2 * Routines having to do with the 'struct sk_buff' memory handlers.
4 * Authors: Alan Cox <iiitac@pyr.swan.ac.uk>
5 * Florian La Roche <rzsfl@rz.uni-sb.de>
7 * Version: $Id: skbuff.c,v 1.90 2001/11/07 05:56:19 davem Exp $
10 * Alan Cox : Fixed the worst of the load
12 * Dave Platt : Interrupt stacking fix.
13 * Richard Kooijman : Timestamp fixes.
14 * Alan Cox : Changed buffer format.
15 * Alan Cox : destructor hook for AF_UNIX etc.
16 * Linus Torvalds : Better skb_clone.
17 * Alan Cox : Added skb_copy.
18 * Alan Cox : Added all the changed routines Linus
19 * only put in the headers
20 * Ray VanTassle : Fixed --skb->lock in free
21 * Alan Cox : skb_copy copy arp field
22 * Andi Kleen : slabified it.
23 * Robert Olsson : Removed skb_head_pool
26 * The __skb_ routines should be called with interrupts
27 * disabled, or you better be *real* sure that the operation is atomic
28 * with respect to whatever list is being frobbed (e.g. via lock_sock()
29 * or via disabling bottom half handlers, etc).
31 * This program is free software; you can redistribute it and/or
32 * modify it under the terms of the GNU General Public License
33 * as published by the Free Software Foundation; either version
34 * 2 of the License, or (at your option) any later version.
38 * The functions in this file will not compile correctly with gcc 2.4.x
41 #include <linux/module.h>
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/sched.h>
46 #include <linux/interrupt.h>
48 #include <linux/inet.h>
49 #include <linux/slab.h>
50 #include <linux/netdevice.h>
51 #ifdef CONFIG_NET_CLS_ACT
52 #include <net/pkt_sched.h>
54 #include <linux/string.h>
55 #include <linux/skbuff.h>
56 #include <linux/cache.h>
57 #include <linux/rtnetlink.h>
58 #include <linux/init.h>
59 #include <linux/highmem.h>
61 #include <net/protocol.h>
64 #include <net/checksum.h>
67 #include <asm/uaccess.h>
68 #include <asm/system.h>
70 static kmem_cache_t
*skbuff_head_cache __read_mostly
;
71 static kmem_cache_t
*skbuff_fclone_cache __read_mostly
;
74 * Keep out-of-line to prevent kernel bloat.
75 * __builtin_return_address is not used because it is not always
80 * skb_over_panic - private function
85 * Out of line support code for skb_put(). Not user callable.
87 void skb_over_panic(struct sk_buff
*skb
, int sz
, void *here
)
89 printk(KERN_EMERG
"skb_over_panic: text:%p len:%d put:%d head:%p "
90 "data:%p tail:%p end:%p dev:%s\n",
91 here
, skb
->len
, sz
, skb
->head
, skb
->data
, skb
->tail
, skb
->end
,
92 skb
->dev
? skb
->dev
->name
: "<NULL>");
97 * skb_under_panic - private function
102 * Out of line support code for skb_push(). Not user callable.
105 void skb_under_panic(struct sk_buff
*skb
, int sz
, void *here
)
107 printk(KERN_EMERG
"skb_under_panic: text:%p len:%d put:%d head:%p "
108 "data:%p tail:%p end:%p dev:%s\n",
109 here
, skb
->len
, sz
, skb
->head
, skb
->data
, skb
->tail
, skb
->end
,
110 skb
->dev
? skb
->dev
->name
: "<NULL>");
114 void skb_truesize_bug(struct sk_buff
*skb
)
116 printk(KERN_ERR
"SKB BUG: Invalid truesize (%u) "
117 "len=%u, sizeof(sk_buff)=%Zd\n",
118 skb
->truesize
, skb
->len
, sizeof(struct sk_buff
));
120 EXPORT_SYMBOL(skb_truesize_bug
);
122 /* Allocate a new skbuff. We do this ourselves so we can fill in a few
123 * 'private' fields and also do memory statistics to find all the
129 * __alloc_skb - allocate a network buffer
130 * @size: size to allocate
131 * @gfp_mask: allocation mask
132 * @fclone: allocate from fclone cache instead of head cache
133 * and allocate a cloned (child) skb
135 * Allocate a new &sk_buff. The returned buffer has no headroom and a
136 * tail room of size bytes. The object has a reference count of one.
137 * The return is the buffer. On a failure the return is %NULL.
139 * Buffers may only be allocated from interrupts using a @gfp_mask of
142 struct sk_buff
*__alloc_skb(unsigned int size
, gfp_t gfp_mask
,
146 struct skb_shared_info
*shinfo
;
150 cache
= fclone
? skbuff_fclone_cache
: skbuff_head_cache
;
153 skb
= kmem_cache_alloc(cache
, gfp_mask
& ~__GFP_DMA
);
157 /* Get the DATA. Size must match skb_add_mtu(). */
158 size
= SKB_DATA_ALIGN(size
);
159 data
= kmalloc_track_caller(size
+ sizeof(struct skb_shared_info
),
164 memset(skb
, 0, offsetof(struct sk_buff
, truesize
));
165 skb
->truesize
= size
+ sizeof(struct sk_buff
);
166 atomic_set(&skb
->users
, 1);
170 skb
->end
= data
+ size
;
171 /* make sure we initialize shinfo sequentially */
172 shinfo
= skb_shinfo(skb
);
173 atomic_set(&shinfo
->dataref
, 1);
174 shinfo
->nr_frags
= 0;
175 shinfo
->gso_size
= 0;
176 shinfo
->gso_segs
= 0;
177 shinfo
->gso_type
= 0;
178 shinfo
->ip6_frag_id
= 0;
179 shinfo
->frag_list
= NULL
;
182 struct sk_buff
*child
= skb
+ 1;
183 atomic_t
*fclone_ref
= (atomic_t
*) (child
+ 1);
185 skb
->fclone
= SKB_FCLONE_ORIG
;
186 atomic_set(fclone_ref
, 1);
188 child
->fclone
= SKB_FCLONE_UNAVAILABLE
;
193 kmem_cache_free(cache
, skb
);
199 * alloc_skb_from_cache - allocate a network buffer
200 * @cp: kmem_cache from which to allocate the data area
201 * (object size must be big enough for @size bytes + skb overheads)
202 * @size: size to allocate
203 * @gfp_mask: allocation mask
205 * Allocate a new &sk_buff. The returned buffer has no headroom and
206 * tail room of size bytes. The object has a reference count of one.
207 * The return is the buffer. On a failure the return is %NULL.
209 * Buffers may only be allocated from interrupts using a @gfp_mask of
212 struct sk_buff
*alloc_skb_from_cache(kmem_cache_t
*cp
,
220 skb
= kmem_cache_alloc(skbuff_head_cache
,
221 gfp_mask
& ~__GFP_DMA
);
226 size
= SKB_DATA_ALIGN(size
);
227 data
= kmem_cache_alloc(cp
, gfp_mask
);
231 memset(skb
, 0, offsetof(struct sk_buff
, truesize
));
232 skb
->truesize
= size
+ sizeof(struct sk_buff
);
233 atomic_set(&skb
->users
, 1);
237 skb
->end
= data
+ size
;
239 atomic_set(&(skb_shinfo(skb
)->dataref
), 1);
240 skb_shinfo(skb
)->nr_frags
= 0;
241 skb_shinfo(skb
)->gso_size
= 0;
242 skb_shinfo(skb
)->gso_segs
= 0;
243 skb_shinfo(skb
)->gso_type
= 0;
244 skb_shinfo(skb
)->frag_list
= NULL
;
248 kmem_cache_free(skbuff_head_cache
, skb
);
254 * __netdev_alloc_skb - allocate an skbuff for rx on a specific device
255 * @dev: network device to receive on
256 * @length: length to allocate
257 * @gfp_mask: get_free_pages mask, passed to alloc_skb
259 * Allocate a new &sk_buff and assign it a usage count of one. The
260 * buffer has unspecified headroom built in. Users should allocate
261 * the headroom they think they need without accounting for the
262 * built in space. The built in space is used for optimisations.
264 * %NULL is returned if there is no free memory.
266 struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
267 unsigned int length
, gfp_t gfp_mask
)
271 skb
= alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
);
273 skb_reserve(skb
, NET_SKB_PAD
);
279 static void skb_drop_list(struct sk_buff
**listp
)
281 struct sk_buff
*list
= *listp
;
286 struct sk_buff
*this = list
;
292 static inline void skb_drop_fraglist(struct sk_buff
*skb
)
294 skb_drop_list(&skb_shinfo(skb
)->frag_list
);
297 static void skb_clone_fraglist(struct sk_buff
*skb
)
299 struct sk_buff
*list
;
301 for (list
= skb_shinfo(skb
)->frag_list
; list
; list
= list
->next
)
305 static void skb_release_data(struct sk_buff
*skb
)
308 !atomic_sub_return(skb
->nohdr
? (1 << SKB_DATAREF_SHIFT
) + 1 : 1,
309 &skb_shinfo(skb
)->dataref
)) {
310 if (skb_shinfo(skb
)->nr_frags
) {
312 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++)
313 put_page(skb_shinfo(skb
)->frags
[i
].page
);
316 if (skb_shinfo(skb
)->frag_list
)
317 skb_drop_fraglist(skb
);
324 * Free an skbuff by memory without cleaning the state.
326 void kfree_skbmem(struct sk_buff
*skb
)
328 struct sk_buff
*other
;
329 atomic_t
*fclone_ref
;
331 skb_release_data(skb
);
332 switch (skb
->fclone
) {
333 case SKB_FCLONE_UNAVAILABLE
:
334 kmem_cache_free(skbuff_head_cache
, skb
);
337 case SKB_FCLONE_ORIG
:
338 fclone_ref
= (atomic_t
*) (skb
+ 2);
339 if (atomic_dec_and_test(fclone_ref
))
340 kmem_cache_free(skbuff_fclone_cache
, skb
);
343 case SKB_FCLONE_CLONE
:
344 fclone_ref
= (atomic_t
*) (skb
+ 1);
347 /* The clone portion is available for
348 * fast-cloning again.
350 skb
->fclone
= SKB_FCLONE_UNAVAILABLE
;
352 if (atomic_dec_and_test(fclone_ref
))
353 kmem_cache_free(skbuff_fclone_cache
, other
);
359 * __kfree_skb - private function
362 * Free an sk_buff. Release anything attached to the buffer.
363 * Clean the state. This is an internal helper function. Users should
364 * always call kfree_skb
367 void __kfree_skb(struct sk_buff
*skb
)
369 dst_release(skb
->dst
);
371 secpath_put(skb
->sp
);
373 if (skb
->destructor
) {
375 skb
->destructor(skb
);
377 #ifdef CONFIG_NETFILTER
378 nf_conntrack_put(skb
->nfct
);
379 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
380 nf_conntrack_put_reasm(skb
->nfct_reasm
);
382 #ifdef CONFIG_BRIDGE_NETFILTER
383 nf_bridge_put(skb
->nf_bridge
);
386 /* XXX: IS this still necessary? - JHS */
387 #ifdef CONFIG_NET_SCHED
389 #ifdef CONFIG_NET_CLS_ACT
398 * kfree_skb - free an sk_buff
399 * @skb: buffer to free
401 * Drop a reference to the buffer and free it if the usage count has
404 void kfree_skb(struct sk_buff
*skb
)
408 if (likely(atomic_read(&skb
->users
) == 1))
410 else if (likely(!atomic_dec_and_test(&skb
->users
)))
416 * skb_clone - duplicate an sk_buff
417 * @skb: buffer to clone
418 * @gfp_mask: allocation priority
420 * Duplicate an &sk_buff. The new one is not owned by a socket. Both
421 * copies share the same packet data but not structure. The new
422 * buffer has a reference count of 1. If the allocation fails the
423 * function returns %NULL otherwise the new buffer is returned.
425 * If this function is called from an interrupt gfp_mask() must be
429 struct sk_buff
*skb_clone(struct sk_buff
*skb
, gfp_t gfp_mask
)
434 if (skb
->fclone
== SKB_FCLONE_ORIG
&&
435 n
->fclone
== SKB_FCLONE_UNAVAILABLE
) {
436 atomic_t
*fclone_ref
= (atomic_t
*) (n
+ 1);
437 n
->fclone
= SKB_FCLONE_CLONE
;
438 atomic_inc(fclone_ref
);
440 n
= kmem_cache_alloc(skbuff_head_cache
, gfp_mask
);
443 n
->fclone
= SKB_FCLONE_UNAVAILABLE
;
446 #define C(x) n->x = skb->x
448 n
->next
= n
->prev
= NULL
;
459 secpath_get(skb
->sp
);
461 memcpy(n
->cb
, skb
->cb
, sizeof(skb
->cb
));
471 #if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
475 n
->destructor
= NULL
;
476 #ifdef CONFIG_NETFILTER
479 nf_conntrack_get(skb
->nfct
);
481 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
483 nf_conntrack_get_reasm(skb
->nfct_reasm
);
485 #ifdef CONFIG_BRIDGE_NETFILTER
487 nf_bridge_get(skb
->nf_bridge
);
489 #endif /*CONFIG_NETFILTER*/
490 #ifdef CONFIG_NET_SCHED
492 #ifdef CONFIG_NET_CLS_ACT
493 n
->tc_verd
= SET_TC_VERD(skb
->tc_verd
,0);
494 n
->tc_verd
= CLR_TC_OK2MUNGE(n
->tc_verd
);
495 n
->tc_verd
= CLR_TC_MUNGED(n
->tc_verd
);
498 skb_copy_secmark(n
, skb
);
501 atomic_set(&n
->users
, 1);
507 atomic_inc(&(skb_shinfo(skb
)->dataref
));
513 static void copy_skb_header(struct sk_buff
*new, const struct sk_buff
*old
)
516 * Shift between the two data areas in bytes
518 unsigned long offset
= new->data
- old
->data
;
522 new->priority
= old
->priority
;
523 new->protocol
= old
->protocol
;
524 new->dst
= dst_clone(old
->dst
);
526 new->sp
= secpath_get(old
->sp
);
528 new->h
.raw
= old
->h
.raw
+ offset
;
529 new->nh
.raw
= old
->nh
.raw
+ offset
;
530 new->mac
.raw
= old
->mac
.raw
+ offset
;
531 memcpy(new->cb
, old
->cb
, sizeof(old
->cb
));
532 new->local_df
= old
->local_df
;
533 new->fclone
= SKB_FCLONE_UNAVAILABLE
;
534 new->pkt_type
= old
->pkt_type
;
535 new->tstamp
= old
->tstamp
;
536 new->destructor
= NULL
;
537 #ifdef CONFIG_NETFILTER
538 new->nfmark
= old
->nfmark
;
539 new->nfct
= old
->nfct
;
540 nf_conntrack_get(old
->nfct
);
541 new->nfctinfo
= old
->nfctinfo
;
542 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
543 new->nfct_reasm
= old
->nfct_reasm
;
544 nf_conntrack_get_reasm(old
->nfct_reasm
);
546 #if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
547 new->ipvs_property
= old
->ipvs_property
;
549 #ifdef CONFIG_BRIDGE_NETFILTER
550 new->nf_bridge
= old
->nf_bridge
;
551 nf_bridge_get(old
->nf_bridge
);
554 #ifdef CONFIG_NET_SCHED
555 #ifdef CONFIG_NET_CLS_ACT
556 new->tc_verd
= old
->tc_verd
;
558 new->tc_index
= old
->tc_index
;
560 skb_copy_secmark(new, old
);
561 atomic_set(&new->users
, 1);
562 skb_shinfo(new)->gso_size
= skb_shinfo(old
)->gso_size
;
563 skb_shinfo(new)->gso_segs
= skb_shinfo(old
)->gso_segs
;
564 skb_shinfo(new)->gso_type
= skb_shinfo(old
)->gso_type
;
568 * skb_copy - create private copy of an sk_buff
569 * @skb: buffer to copy
570 * @gfp_mask: allocation priority
572 * Make a copy of both an &sk_buff and its data. This is used when the
573 * caller wishes to modify the data and needs a private copy of the
574 * data to alter. Returns %NULL on failure or the pointer to the buffer
575 * on success. The returned buffer has a reference count of 1.
577 * As by-product this function converts non-linear &sk_buff to linear
578 * one, so that &sk_buff becomes completely private and caller is allowed
579 * to modify all the data of returned buffer. This means that this
580 * function is not recommended for use in circumstances when only
581 * header is going to be modified. Use pskb_copy() instead.
584 struct sk_buff
*skb_copy(const struct sk_buff
*skb
, gfp_t gfp_mask
)
586 int headerlen
= skb
->data
- skb
->head
;
588 * Allocate the copy buffer
590 struct sk_buff
*n
= alloc_skb(skb
->end
- skb
->head
+ skb
->data_len
,
595 /* Set the data pointer */
596 skb_reserve(n
, headerlen
);
597 /* Set the tail pointer and length */
598 skb_put(n
, skb
->len
);
600 n
->ip_summed
= skb
->ip_summed
;
602 if (skb_copy_bits(skb
, -headerlen
, n
->head
, headerlen
+ skb
->len
))
605 copy_skb_header(n
, skb
);
611 * pskb_copy - create copy of an sk_buff with private head.
612 * @skb: buffer to copy
613 * @gfp_mask: allocation priority
615 * Make a copy of both an &sk_buff and part of its data, located
616 * in header. Fragmented data remain shared. This is used when
617 * the caller wishes to modify only header of &sk_buff and needs
618 * private copy of the header to alter. Returns %NULL on failure
619 * or the pointer to the buffer on success.
620 * The returned buffer has a reference count of 1.
623 struct sk_buff
*pskb_copy(struct sk_buff
*skb
, gfp_t gfp_mask
)
626 * Allocate the copy buffer
628 struct sk_buff
*n
= alloc_skb(skb
->end
- skb
->head
, gfp_mask
);
633 /* Set the data pointer */
634 skb_reserve(n
, skb
->data
- skb
->head
);
635 /* Set the tail pointer and length */
636 skb_put(n
, skb_headlen(skb
));
638 memcpy(n
->data
, skb
->data
, n
->len
);
640 n
->ip_summed
= skb
->ip_summed
;
642 n
->data_len
= skb
->data_len
;
645 if (skb_shinfo(skb
)->nr_frags
) {
648 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
649 skb_shinfo(n
)->frags
[i
] = skb_shinfo(skb
)->frags
[i
];
650 get_page(skb_shinfo(n
)->frags
[i
].page
);
652 skb_shinfo(n
)->nr_frags
= i
;
655 if (skb_shinfo(skb
)->frag_list
) {
656 skb_shinfo(n
)->frag_list
= skb_shinfo(skb
)->frag_list
;
657 skb_clone_fraglist(n
);
660 copy_skb_header(n
, skb
);
666 * pskb_expand_head - reallocate header of &sk_buff
667 * @skb: buffer to reallocate
668 * @nhead: room to add at head
669 * @ntail: room to add at tail
670 * @gfp_mask: allocation priority
672 * Expands (or creates identical copy, if &nhead and &ntail are zero)
673 * header of skb. &sk_buff itself is not changed. &sk_buff MUST have
674 * reference count of 1. Returns zero in the case of success or error,
675 * if expansion failed. In the last case, &sk_buff is not changed.
677 * All the pointers pointing into skb header may change and must be
678 * reloaded after call to this function.
681 int pskb_expand_head(struct sk_buff
*skb
, int nhead
, int ntail
,
686 int size
= nhead
+ (skb
->end
- skb
->head
) + ntail
;
692 size
= SKB_DATA_ALIGN(size
);
694 data
= kmalloc(size
+ sizeof(struct skb_shared_info
), gfp_mask
);
698 /* Copy only real data... and, alas, header. This should be
699 * optimized for the cases when header is void. */
700 memcpy(data
+ nhead
, skb
->head
, skb
->tail
- skb
->head
);
701 memcpy(data
+ size
, skb
->end
, sizeof(struct skb_shared_info
));
703 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++)
704 get_page(skb_shinfo(skb
)->frags
[i
].page
);
706 if (skb_shinfo(skb
)->frag_list
)
707 skb_clone_fraglist(skb
);
709 skb_release_data(skb
);
711 off
= (data
+ nhead
) - skb
->head
;
714 skb
->end
= data
+ size
;
722 atomic_set(&skb_shinfo(skb
)->dataref
, 1);
729 /* Make private copy of skb with writable head and some headroom */
731 struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
, unsigned int headroom
)
733 struct sk_buff
*skb2
;
734 int delta
= headroom
- skb_headroom(skb
);
737 skb2
= pskb_copy(skb
, GFP_ATOMIC
);
739 skb2
= skb_clone(skb
, GFP_ATOMIC
);
740 if (skb2
&& pskb_expand_head(skb2
, SKB_DATA_ALIGN(delta
), 0,
751 * skb_copy_expand - copy and expand sk_buff
752 * @skb: buffer to copy
753 * @newheadroom: new free bytes at head
754 * @newtailroom: new free bytes at tail
755 * @gfp_mask: allocation priority
757 * Make a copy of both an &sk_buff and its data and while doing so
758 * allocate additional space.
760 * This is used when the caller wishes to modify the data and needs a
761 * private copy of the data to alter as well as more space for new fields.
762 * Returns %NULL on failure or the pointer to the buffer
763 * on success. The returned buffer has a reference count of 1.
765 * You must pass %GFP_ATOMIC as the allocation priority if this function
766 * is called from an interrupt.
768 * BUG ALERT: ip_summed is not copied. Why does this work? Is it used
769 * only by netfilter in the cases when checksum is recalculated? --ANK
771 struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
772 int newheadroom
, int newtailroom
,
776 * Allocate the copy buffer
778 struct sk_buff
*n
= alloc_skb(newheadroom
+ skb
->len
+ newtailroom
,
780 int head_copy_len
, head_copy_off
;
785 skb_reserve(n
, newheadroom
);
787 /* Set the tail pointer and length */
788 skb_put(n
, skb
->len
);
790 head_copy_len
= skb_headroom(skb
);
792 if (newheadroom
<= head_copy_len
)
793 head_copy_len
= newheadroom
;
795 head_copy_off
= newheadroom
- head_copy_len
;
797 /* Copy the linear header and data. */
798 if (skb_copy_bits(skb
, -head_copy_len
, n
->head
+ head_copy_off
,
799 skb
->len
+ head_copy_len
))
802 copy_skb_header(n
, skb
);
808 * skb_pad - zero pad the tail of an skb
809 * @skb: buffer to pad
812 * Ensure that a buffer is followed by a padding area that is zero
813 * filled. Used by network drivers which may DMA or transfer data
814 * beyond the buffer end onto the wire.
816 * May return error in out of memory cases. The skb is freed on error.
819 int skb_pad(struct sk_buff
*skb
, int pad
)
824 /* If the skbuff is non linear tailroom is always zero.. */
825 if (!skb_cloned(skb
) && skb_tailroom(skb
) >= pad
) {
826 memset(skb
->data
+skb
->len
, 0, pad
);
830 ntail
= skb
->data_len
+ pad
- (skb
->end
- skb
->tail
);
831 if (likely(skb_cloned(skb
) || ntail
> 0)) {
832 err
= pskb_expand_head(skb
, 0, ntail
, GFP_ATOMIC
);
837 /* FIXME: The use of this function with non-linear skb's really needs
840 err
= skb_linearize(skb
);
844 memset(skb
->data
+ skb
->len
, 0, pad
);
852 /* Trims skb to length len. It can change skb pointers.
855 int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
)
857 struct sk_buff
**fragp
;
858 struct sk_buff
*frag
;
859 int offset
= skb_headlen(skb
);
860 int nfrags
= skb_shinfo(skb
)->nr_frags
;
864 if (skb_cloned(skb
) &&
865 unlikely((err
= pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
))))
872 for (; i
< nfrags
; i
++) {
873 int end
= offset
+ skb_shinfo(skb
)->frags
[i
].size
;
880 skb_shinfo(skb
)->frags
[i
++].size
= len
- offset
;
883 skb_shinfo(skb
)->nr_frags
= i
;
885 for (; i
< nfrags
; i
++)
886 put_page(skb_shinfo(skb
)->frags
[i
].page
);
888 if (skb_shinfo(skb
)->frag_list
)
889 skb_drop_fraglist(skb
);
893 for (fragp
= &skb_shinfo(skb
)->frag_list
; (frag
= *fragp
);
894 fragp
= &frag
->next
) {
895 int end
= offset
+ frag
->len
;
897 if (skb_shared(frag
)) {
898 struct sk_buff
*nfrag
;
900 nfrag
= skb_clone(frag
, GFP_ATOMIC
);
901 if (unlikely(!nfrag
))
904 nfrag
->next
= frag
->next
;
916 unlikely((err
= pskb_trim(frag
, len
- offset
))))
920 skb_drop_list(&frag
->next
);
925 if (len
> skb_headlen(skb
)) {
926 skb
->data_len
-= skb
->len
- len
;
931 skb
->tail
= skb
->data
+ len
;
938 * __pskb_pull_tail - advance tail of skb header
939 * @skb: buffer to reallocate
940 * @delta: number of bytes to advance tail
942 * The function makes a sense only on a fragmented &sk_buff,
943 * it expands header moving its tail forward and copying necessary
944 * data from fragmented part.
946 * &sk_buff MUST have reference count of 1.
948 * Returns %NULL (and &sk_buff does not change) if pull failed
949 * or value of new tail of skb in the case of success.
951 * All the pointers pointing into skb header may change and must be
952 * reloaded after call to this function.
955 /* Moves tail of skb head forward, copying data from fragmented part,
956 * when it is necessary.
957 * 1. It may fail due to malloc failure.
958 * 2. It may change skb pointers.
960 * It is pretty complicated. Luckily, it is called only in exceptional cases.
962 unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
)
964 /* If skb has not enough free space at tail, get new one
965 * plus 128 bytes for future expansions. If we have enough
966 * room at tail, reallocate without expansion only if skb is cloned.
968 int i
, k
, eat
= (skb
->tail
+ delta
) - skb
->end
;
970 if (eat
> 0 || skb_cloned(skb
)) {
971 if (pskb_expand_head(skb
, 0, eat
> 0 ? eat
+ 128 : 0,
976 if (skb_copy_bits(skb
, skb_headlen(skb
), skb
->tail
, delta
))
979 /* Optimization: no fragments, no reasons to preestimate
980 * size of pulled pages. Superb.
982 if (!skb_shinfo(skb
)->frag_list
)
985 /* Estimate size of pulled pages. */
987 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
988 if (skb_shinfo(skb
)->frags
[i
].size
>= eat
)
990 eat
-= skb_shinfo(skb
)->frags
[i
].size
;
993 /* If we need update frag list, we are in troubles.
994 * Certainly, it possible to add an offset to skb data,
995 * but taking into account that pulling is expected to
996 * be very rare operation, it is worth to fight against
997 * further bloating skb head and crucify ourselves here instead.
998 * Pure masohism, indeed. 8)8)
1001 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1002 struct sk_buff
*clone
= NULL
;
1003 struct sk_buff
*insp
= NULL
;
1008 if (list
->len
<= eat
) {
1009 /* Eaten as whole. */
1014 /* Eaten partially. */
1016 if (skb_shared(list
)) {
1017 /* Sucks! We need to fork list. :-( */
1018 clone
= skb_clone(list
, GFP_ATOMIC
);
1024 /* This may be pulled without
1028 if (!pskb_pull(list
, eat
)) {
1037 /* Free pulled out fragments. */
1038 while ((list
= skb_shinfo(skb
)->frag_list
) != insp
) {
1039 skb_shinfo(skb
)->frag_list
= list
->next
;
1042 /* And insert new clone at head. */
1045 skb_shinfo(skb
)->frag_list
= clone
;
1048 /* Success! Now we may commit changes to skb data. */
1053 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1054 if (skb_shinfo(skb
)->frags
[i
].size
<= eat
) {
1055 put_page(skb_shinfo(skb
)->frags
[i
].page
);
1056 eat
-= skb_shinfo(skb
)->frags
[i
].size
;
1058 skb_shinfo(skb
)->frags
[k
] = skb_shinfo(skb
)->frags
[i
];
1060 skb_shinfo(skb
)->frags
[k
].page_offset
+= eat
;
1061 skb_shinfo(skb
)->frags
[k
].size
-= eat
;
1067 skb_shinfo(skb
)->nr_frags
= k
;
1070 skb
->data_len
-= delta
;
1075 /* Copy some data bits from skb to kernel buffer. */
1077 int skb_copy_bits(const struct sk_buff
*skb
, int offset
, void *to
, int len
)
1080 int start
= skb_headlen(skb
);
1082 if (offset
> (int)skb
->len
- len
)
1086 if ((copy
= start
- offset
) > 0) {
1089 memcpy(to
, skb
->data
+ offset
, copy
);
1090 if ((len
-= copy
) == 0)
1096 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1099 BUG_TRAP(start
<= offset
+ len
);
1101 end
= start
+ skb_shinfo(skb
)->frags
[i
].size
;
1102 if ((copy
= end
- offset
) > 0) {
1108 vaddr
= kmap_skb_frag(&skb_shinfo(skb
)->frags
[i
]);
1110 vaddr
+ skb_shinfo(skb
)->frags
[i
].page_offset
+
1111 offset
- start
, copy
);
1112 kunmap_skb_frag(vaddr
);
1114 if ((len
-= copy
) == 0)
1122 if (skb_shinfo(skb
)->frag_list
) {
1123 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1125 for (; list
; list
= list
->next
) {
1128 BUG_TRAP(start
<= offset
+ len
);
1130 end
= start
+ list
->len
;
1131 if ((copy
= end
- offset
) > 0) {
1134 if (skb_copy_bits(list
, offset
- start
,
1137 if ((len
-= copy
) == 0)
1153 * skb_store_bits - store bits from kernel buffer to skb
1154 * @skb: destination buffer
1155 * @offset: offset in destination
1156 * @from: source buffer
1157 * @len: number of bytes to copy
1159 * Copy the specified number of bytes from the source buffer to the
1160 * destination skb. This function handles all the messy bits of
1161 * traversing fragment lists and such.
1164 int skb_store_bits(const struct sk_buff
*skb
, int offset
, void *from
, int len
)
1167 int start
= skb_headlen(skb
);
1169 if (offset
> (int)skb
->len
- len
)
1172 if ((copy
= start
- offset
) > 0) {
1175 memcpy(skb
->data
+ offset
, from
, copy
);
1176 if ((len
-= copy
) == 0)
1182 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1183 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1186 BUG_TRAP(start
<= offset
+ len
);
1188 end
= start
+ frag
->size
;
1189 if ((copy
= end
- offset
) > 0) {
1195 vaddr
= kmap_skb_frag(frag
);
1196 memcpy(vaddr
+ frag
->page_offset
+ offset
- start
,
1198 kunmap_skb_frag(vaddr
);
1200 if ((len
-= copy
) == 0)
1208 if (skb_shinfo(skb
)->frag_list
) {
1209 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1211 for (; list
; list
= list
->next
) {
1214 BUG_TRAP(start
<= offset
+ len
);
1216 end
= start
+ list
->len
;
1217 if ((copy
= end
- offset
) > 0) {
1220 if (skb_store_bits(list
, offset
- start
,
1223 if ((len
-= copy
) == 0)
1238 EXPORT_SYMBOL(skb_store_bits
);
1240 /* Checksum skb data. */
1242 unsigned int skb_checksum(const struct sk_buff
*skb
, int offset
,
1243 int len
, unsigned int csum
)
1245 int start
= skb_headlen(skb
);
1246 int i
, copy
= start
- offset
;
1249 /* Checksum header. */
1253 csum
= csum_partial(skb
->data
+ offset
, copy
, csum
);
1254 if ((len
-= copy
) == 0)
1260 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1263 BUG_TRAP(start
<= offset
+ len
);
1265 end
= start
+ skb_shinfo(skb
)->frags
[i
].size
;
1266 if ((copy
= end
- offset
) > 0) {
1269 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1273 vaddr
= kmap_skb_frag(frag
);
1274 csum2
= csum_partial(vaddr
+ frag
->page_offset
+
1275 offset
- start
, copy
, 0);
1276 kunmap_skb_frag(vaddr
);
1277 csum
= csum_block_add(csum
, csum2
, pos
);
1286 if (skb_shinfo(skb
)->frag_list
) {
1287 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1289 for (; list
; list
= list
->next
) {
1292 BUG_TRAP(start
<= offset
+ len
);
1294 end
= start
+ list
->len
;
1295 if ((copy
= end
- offset
) > 0) {
1299 csum2
= skb_checksum(list
, offset
- start
,
1301 csum
= csum_block_add(csum
, csum2
, pos
);
1302 if ((len
-= copy
) == 0)
1315 /* Both of above in one bottle. */
1317 unsigned int skb_copy_and_csum_bits(const struct sk_buff
*skb
, int offset
,
1318 u8
*to
, int len
, unsigned int csum
)
1320 int start
= skb_headlen(skb
);
1321 int i
, copy
= start
- offset
;
1328 csum
= csum_partial_copy_nocheck(skb
->data
+ offset
, to
,
1330 if ((len
-= copy
) == 0)
1337 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1340 BUG_TRAP(start
<= offset
+ len
);
1342 end
= start
+ skb_shinfo(skb
)->frags
[i
].size
;
1343 if ((copy
= end
- offset
) > 0) {
1346 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1350 vaddr
= kmap_skb_frag(frag
);
1351 csum2
= csum_partial_copy_nocheck(vaddr
+
1355 kunmap_skb_frag(vaddr
);
1356 csum
= csum_block_add(csum
, csum2
, pos
);
1366 if (skb_shinfo(skb
)->frag_list
) {
1367 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1369 for (; list
; list
= list
->next
) {
1373 BUG_TRAP(start
<= offset
+ len
);
1375 end
= start
+ list
->len
;
1376 if ((copy
= end
- offset
) > 0) {
1379 csum2
= skb_copy_and_csum_bits(list
,
1382 csum
= csum_block_add(csum
, csum2
, pos
);
1383 if ((len
-= copy
) == 0)
1396 void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
)
1401 if (skb
->ip_summed
== CHECKSUM_PARTIAL
)
1402 csstart
= skb
->h
.raw
- skb
->data
;
1404 csstart
= skb_headlen(skb
);
1406 BUG_ON(csstart
> skb_headlen(skb
));
1408 memcpy(to
, skb
->data
, csstart
);
1411 if (csstart
!= skb
->len
)
1412 csum
= skb_copy_and_csum_bits(skb
, csstart
, to
+ csstart
,
1413 skb
->len
- csstart
, 0);
1415 if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1416 long csstuff
= csstart
+ skb
->csum
;
1418 *((unsigned short *)(to
+ csstuff
)) = csum_fold(csum
);
1423 * skb_dequeue - remove from the head of the queue
1424 * @list: list to dequeue from
1426 * Remove the head of the list. The list lock is taken so the function
1427 * may be used safely with other locking list functions. The head item is
1428 * returned or %NULL if the list is empty.
1431 struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
)
1433 unsigned long flags
;
1434 struct sk_buff
*result
;
1436 spin_lock_irqsave(&list
->lock
, flags
);
1437 result
= __skb_dequeue(list
);
1438 spin_unlock_irqrestore(&list
->lock
, flags
);
1443 * skb_dequeue_tail - remove from the tail of the queue
1444 * @list: list to dequeue from
1446 * Remove the tail of the list. The list lock is taken so the function
1447 * may be used safely with other locking list functions. The tail item is
1448 * returned or %NULL if the list is empty.
1450 struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
)
1452 unsigned long flags
;
1453 struct sk_buff
*result
;
1455 spin_lock_irqsave(&list
->lock
, flags
);
1456 result
= __skb_dequeue_tail(list
);
1457 spin_unlock_irqrestore(&list
->lock
, flags
);
1462 * skb_queue_purge - empty a list
1463 * @list: list to empty
1465 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1466 * the list and one reference dropped. This function takes the list
1467 * lock and is atomic with respect to other list locking functions.
1469 void skb_queue_purge(struct sk_buff_head
*list
)
1471 struct sk_buff
*skb
;
1472 while ((skb
= skb_dequeue(list
)) != NULL
)
1477 * skb_queue_head - queue a buffer at the list head
1478 * @list: list to use
1479 * @newsk: buffer to queue
1481 * Queue a buffer at the start of the list. This function takes the
1482 * list lock and can be used safely with other locking &sk_buff functions
1485 * A buffer cannot be placed on two lists at the same time.
1487 void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
)
1489 unsigned long flags
;
1491 spin_lock_irqsave(&list
->lock
, flags
);
1492 __skb_queue_head(list
, newsk
);
1493 spin_unlock_irqrestore(&list
->lock
, flags
);
1497 * skb_queue_tail - queue a buffer at the list tail
1498 * @list: list to use
1499 * @newsk: buffer to queue
1501 * Queue a buffer at the tail of the list. This function takes the
1502 * list lock and can be used safely with other locking &sk_buff functions
1505 * A buffer cannot be placed on two lists at the same time.
1507 void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
)
1509 unsigned long flags
;
1511 spin_lock_irqsave(&list
->lock
, flags
);
1512 __skb_queue_tail(list
, newsk
);
1513 spin_unlock_irqrestore(&list
->lock
, flags
);
1517 * skb_unlink - remove a buffer from a list
1518 * @skb: buffer to remove
1519 * @list: list to use
1521 * Remove a packet from a list. The list locks are taken and this
1522 * function is atomic with respect to other list locked calls
1524 * You must know what list the SKB is on.
1526 void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1528 unsigned long flags
;
1530 spin_lock_irqsave(&list
->lock
, flags
);
1531 __skb_unlink(skb
, list
);
1532 spin_unlock_irqrestore(&list
->lock
, flags
);
1536 * skb_append - append a buffer
1537 * @old: buffer to insert after
1538 * @newsk: buffer to insert
1539 * @list: list to use
1541 * Place a packet after a given packet in a list. The list locks are taken
1542 * and this function is atomic with respect to other list locked calls.
1543 * A buffer cannot be placed on two lists at the same time.
1545 void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
1547 unsigned long flags
;
1549 spin_lock_irqsave(&list
->lock
, flags
);
1550 __skb_append(old
, newsk
, list
);
1551 spin_unlock_irqrestore(&list
->lock
, flags
);
1556 * skb_insert - insert a buffer
1557 * @old: buffer to insert before
1558 * @newsk: buffer to insert
1559 * @list: list to use
1561 * Place a packet before a given packet in a list. The list locks are
1562 * taken and this function is atomic with respect to other list locked
1565 * A buffer cannot be placed on two lists at the same time.
1567 void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
1569 unsigned long flags
;
1571 spin_lock_irqsave(&list
->lock
, flags
);
1572 __skb_insert(newsk
, old
->prev
, old
, list
);
1573 spin_unlock_irqrestore(&list
->lock
, flags
);
1578 * Tune the memory allocator for a new MTU size.
1580 void skb_add_mtu(int mtu
)
1582 /* Must match allocation in alloc_skb */
1583 mtu
= SKB_DATA_ALIGN(mtu
) + sizeof(struct skb_shared_info
);
1585 kmem_add_cache_size(mtu
);
1589 static inline void skb_split_inside_header(struct sk_buff
*skb
,
1590 struct sk_buff
* skb1
,
1591 const u32 len
, const int pos
)
1595 memcpy(skb_put(skb1
, pos
- len
), skb
->data
+ len
, pos
- len
);
1597 /* And move data appendix as is. */
1598 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++)
1599 skb_shinfo(skb1
)->frags
[i
] = skb_shinfo(skb
)->frags
[i
];
1601 skb_shinfo(skb1
)->nr_frags
= skb_shinfo(skb
)->nr_frags
;
1602 skb_shinfo(skb
)->nr_frags
= 0;
1603 skb1
->data_len
= skb
->data_len
;
1604 skb1
->len
+= skb1
->data_len
;
1607 skb
->tail
= skb
->data
+ len
;
1610 static inline void skb_split_no_header(struct sk_buff
*skb
,
1611 struct sk_buff
* skb1
,
1612 const u32 len
, int pos
)
1615 const int nfrags
= skb_shinfo(skb
)->nr_frags
;
1617 skb_shinfo(skb
)->nr_frags
= 0;
1618 skb1
->len
= skb1
->data_len
= skb
->len
- len
;
1620 skb
->data_len
= len
- pos
;
1622 for (i
= 0; i
< nfrags
; i
++) {
1623 int size
= skb_shinfo(skb
)->frags
[i
].size
;
1625 if (pos
+ size
> len
) {
1626 skb_shinfo(skb1
)->frags
[k
] = skb_shinfo(skb
)->frags
[i
];
1630 * We have two variants in this case:
1631 * 1. Move all the frag to the second
1632 * part, if it is possible. F.e.
1633 * this approach is mandatory for TUX,
1634 * where splitting is expensive.
1635 * 2. Split is accurately. We make this.
1637 get_page(skb_shinfo(skb
)->frags
[i
].page
);
1638 skb_shinfo(skb1
)->frags
[0].page_offset
+= len
- pos
;
1639 skb_shinfo(skb1
)->frags
[0].size
-= len
- pos
;
1640 skb_shinfo(skb
)->frags
[i
].size
= len
- pos
;
1641 skb_shinfo(skb
)->nr_frags
++;
1645 skb_shinfo(skb
)->nr_frags
++;
1648 skb_shinfo(skb1
)->nr_frags
= k
;
1652 * skb_split - Split fragmented skb to two parts at length len.
1653 * @skb: the buffer to split
1654 * @skb1: the buffer to receive the second part
1655 * @len: new length for skb
1657 void skb_split(struct sk_buff
*skb
, struct sk_buff
*skb1
, const u32 len
)
1659 int pos
= skb_headlen(skb
);
1661 if (len
< pos
) /* Split line is inside header. */
1662 skb_split_inside_header(skb
, skb1
, len
, pos
);
1663 else /* Second chunk has no header, nothing to copy. */
1664 skb_split_no_header(skb
, skb1
, len
, pos
);
1668 * skb_prepare_seq_read - Prepare a sequential read of skb data
1669 * @skb: the buffer to read
1670 * @from: lower offset of data to be read
1671 * @to: upper offset of data to be read
1672 * @st: state variable
1674 * Initializes the specified state variable. Must be called before
1675 * invoking skb_seq_read() for the first time.
1677 void skb_prepare_seq_read(struct sk_buff
*skb
, unsigned int from
,
1678 unsigned int to
, struct skb_seq_state
*st
)
1680 st
->lower_offset
= from
;
1681 st
->upper_offset
= to
;
1682 st
->root_skb
= st
->cur_skb
= skb
;
1683 st
->frag_idx
= st
->stepped_offset
= 0;
1684 st
->frag_data
= NULL
;
1688 * skb_seq_read - Sequentially read skb data
1689 * @consumed: number of bytes consumed by the caller so far
1690 * @data: destination pointer for data to be returned
1691 * @st: state variable
1693 * Reads a block of skb data at &consumed relative to the
1694 * lower offset specified to skb_prepare_seq_read(). Assigns
1695 * the head of the data block to &data and returns the length
1696 * of the block or 0 if the end of the skb data or the upper
1697 * offset has been reached.
1699 * The caller is not required to consume all of the data
1700 * returned, i.e. &consumed is typically set to the number
1701 * of bytes already consumed and the next call to
1702 * skb_seq_read() will return the remaining part of the block.
1704 * Note: The size of each block of data returned can be arbitary,
1705 * this limitation is the cost for zerocopy seqeuental
1706 * reads of potentially non linear data.
1708 * Note: Fragment lists within fragments are not implemented
1709 * at the moment, state->root_skb could be replaced with
1710 * a stack for this purpose.
1712 unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
1713 struct skb_seq_state
*st
)
1715 unsigned int block_limit
, abs_offset
= consumed
+ st
->lower_offset
;
1718 if (unlikely(abs_offset
>= st
->upper_offset
))
1722 block_limit
= skb_headlen(st
->cur_skb
);
1724 if (abs_offset
< block_limit
) {
1725 *data
= st
->cur_skb
->data
+ abs_offset
;
1726 return block_limit
- abs_offset
;
1729 if (st
->frag_idx
== 0 && !st
->frag_data
)
1730 st
->stepped_offset
+= skb_headlen(st
->cur_skb
);
1732 while (st
->frag_idx
< skb_shinfo(st
->cur_skb
)->nr_frags
) {
1733 frag
= &skb_shinfo(st
->cur_skb
)->frags
[st
->frag_idx
];
1734 block_limit
= frag
->size
+ st
->stepped_offset
;
1736 if (abs_offset
< block_limit
) {
1738 st
->frag_data
= kmap_skb_frag(frag
);
1740 *data
= (u8
*) st
->frag_data
+ frag
->page_offset
+
1741 (abs_offset
- st
->stepped_offset
);
1743 return block_limit
- abs_offset
;
1746 if (st
->frag_data
) {
1747 kunmap_skb_frag(st
->frag_data
);
1748 st
->frag_data
= NULL
;
1752 st
->stepped_offset
+= frag
->size
;
1755 if (st
->cur_skb
->next
) {
1756 st
->cur_skb
= st
->cur_skb
->next
;
1759 } else if (st
->root_skb
== st
->cur_skb
&&
1760 skb_shinfo(st
->root_skb
)->frag_list
) {
1761 st
->cur_skb
= skb_shinfo(st
->root_skb
)->frag_list
;
1769 * skb_abort_seq_read - Abort a sequential read of skb data
1770 * @st: state variable
1772 * Must be called if skb_seq_read() was not called until it
1775 void skb_abort_seq_read(struct skb_seq_state
*st
)
1778 kunmap_skb_frag(st
->frag_data
);
1781 #define TS_SKB_CB(state) ((struct skb_seq_state *) &((state)->cb))
1783 static unsigned int skb_ts_get_next_block(unsigned int offset
, const u8
**text
,
1784 struct ts_config
*conf
,
1785 struct ts_state
*state
)
1787 return skb_seq_read(offset
, text
, TS_SKB_CB(state
));
1790 static void skb_ts_finish(struct ts_config
*conf
, struct ts_state
*state
)
1792 skb_abort_seq_read(TS_SKB_CB(state
));
1796 * skb_find_text - Find a text pattern in skb data
1797 * @skb: the buffer to look in
1798 * @from: search offset
1800 * @config: textsearch configuration
1801 * @state: uninitialized textsearch state variable
1803 * Finds a pattern in the skb data according to the specified
1804 * textsearch configuration. Use textsearch_next() to retrieve
1805 * subsequent occurrences of the pattern. Returns the offset
1806 * to the first occurrence or UINT_MAX if no match was found.
1808 unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
1809 unsigned int to
, struct ts_config
*config
,
1810 struct ts_state
*state
)
1814 config
->get_next_block
= skb_ts_get_next_block
;
1815 config
->finish
= skb_ts_finish
;
1817 skb_prepare_seq_read(skb
, from
, to
, TS_SKB_CB(state
));
1819 ret
= textsearch_find(config
, state
);
1820 return (ret
<= to
- from
? ret
: UINT_MAX
);
1824 * skb_append_datato_frags: - append the user data to a skb
1825 * @sk: sock structure
1826 * @skb: skb structure to be appened with user data.
1827 * @getfrag: call back function to be used for getting the user data
1828 * @from: pointer to user message iov
1829 * @length: length of the iov message
1831 * Description: This procedure append the user data in the fragment part
1832 * of the skb if any page alloc fails user this procedure returns -ENOMEM
1834 int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
1835 int (*getfrag
)(void *from
, char *to
, int offset
,
1836 int len
, int odd
, struct sk_buff
*skb
),
1837 void *from
, int length
)
1840 skb_frag_t
*frag
= NULL
;
1841 struct page
*page
= NULL
;
1847 /* Return error if we don't have space for new frag */
1848 frg_cnt
= skb_shinfo(skb
)->nr_frags
;
1849 if (frg_cnt
>= MAX_SKB_FRAGS
)
1852 /* allocate a new page for next frag */
1853 page
= alloc_pages(sk
->sk_allocation
, 0);
1855 /* If alloc_page fails just return failure and caller will
1856 * free previous allocated pages by doing kfree_skb()
1861 /* initialize the next frag */
1862 sk
->sk_sndmsg_page
= page
;
1863 sk
->sk_sndmsg_off
= 0;
1864 skb_fill_page_desc(skb
, frg_cnt
, page
, 0, 0);
1865 skb
->truesize
+= PAGE_SIZE
;
1866 atomic_add(PAGE_SIZE
, &sk
->sk_wmem_alloc
);
1868 /* get the new initialized frag */
1869 frg_cnt
= skb_shinfo(skb
)->nr_frags
;
1870 frag
= &skb_shinfo(skb
)->frags
[frg_cnt
- 1];
1872 /* copy the user data to page */
1873 left
= PAGE_SIZE
- frag
->page_offset
;
1874 copy
= (length
> left
)? left
: length
;
1876 ret
= getfrag(from
, (page_address(frag
->page
) +
1877 frag
->page_offset
+ frag
->size
),
1878 offset
, copy
, 0, skb
);
1882 /* copy was successful so update the size parameters */
1883 sk
->sk_sndmsg_off
+= copy
;
1886 skb
->data_len
+= copy
;
1890 } while (length
> 0);
1896 * skb_pull_rcsum - pull skb and update receive checksum
1897 * @skb: buffer to update
1898 * @start: start of data before pull
1899 * @len: length of data pulled
1901 * This function performs an skb_pull on the packet and updates
1902 * update the CHECKSUM_COMPLETE checksum. It should be used on
1903 * receive path processing instead of skb_pull unless you know
1904 * that the checksum difference is zero (e.g., a valid IP header)
1905 * or you are setting ip_summed to CHECKSUM_NONE.
1907 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
)
1909 BUG_ON(len
> skb
->len
);
1911 BUG_ON(skb
->len
< skb
->data_len
);
1912 skb_postpull_rcsum(skb
, skb
->data
, len
);
1913 return skb
->data
+= len
;
1916 EXPORT_SYMBOL_GPL(skb_pull_rcsum
);
1919 * skb_segment - Perform protocol segmentation on skb.
1920 * @skb: buffer to segment
1921 * @features: features for the output path (see dev->features)
1923 * This function performs segmentation on the given skb. It returns
1924 * the segment at the given position. It returns NULL if there are
1925 * no more segments to generate, or when an error is encountered.
1927 struct sk_buff
*skb_segment(struct sk_buff
*skb
, int features
)
1929 struct sk_buff
*segs
= NULL
;
1930 struct sk_buff
*tail
= NULL
;
1931 unsigned int mss
= skb_shinfo(skb
)->gso_size
;
1932 unsigned int doffset
= skb
->data
- skb
->mac
.raw
;
1933 unsigned int offset
= doffset
;
1934 unsigned int headroom
;
1936 int sg
= features
& NETIF_F_SG
;
1937 int nfrags
= skb_shinfo(skb
)->nr_frags
;
1942 __skb_push(skb
, doffset
);
1943 headroom
= skb_headroom(skb
);
1944 pos
= skb_headlen(skb
);
1947 struct sk_buff
*nskb
;
1953 len
= skb
->len
- offset
;
1957 hsize
= skb_headlen(skb
) - offset
;
1960 nsize
= hsize
+ doffset
;
1961 if (nsize
> len
+ doffset
|| !sg
)
1962 nsize
= len
+ doffset
;
1964 nskb
= alloc_skb(nsize
+ headroom
, GFP_ATOMIC
);
1965 if (unlikely(!nskb
))
1974 nskb
->dev
= skb
->dev
;
1975 nskb
->priority
= skb
->priority
;
1976 nskb
->protocol
= skb
->protocol
;
1977 nskb
->dst
= dst_clone(skb
->dst
);
1978 memcpy(nskb
->cb
, skb
->cb
, sizeof(skb
->cb
));
1979 nskb
->pkt_type
= skb
->pkt_type
;
1980 nskb
->mac_len
= skb
->mac_len
;
1982 skb_reserve(nskb
, headroom
);
1983 nskb
->mac
.raw
= nskb
->data
;
1984 nskb
->nh
.raw
= nskb
->data
+ skb
->mac_len
;
1985 nskb
->h
.raw
= nskb
->nh
.raw
+ (skb
->h
.raw
- skb
->nh
.raw
);
1986 memcpy(skb_put(nskb
, doffset
), skb
->data
, doffset
);
1989 nskb
->csum
= skb_copy_and_csum_bits(skb
, offset
,
1995 frag
= skb_shinfo(nskb
)->frags
;
1998 nskb
->ip_summed
= CHECKSUM_PARTIAL
;
1999 nskb
->csum
= skb
->csum
;
2000 memcpy(skb_put(nskb
, hsize
), skb
->data
+ offset
, hsize
);
2002 while (pos
< offset
+ len
) {
2003 BUG_ON(i
>= nfrags
);
2005 *frag
= skb_shinfo(skb
)->frags
[i
];
2006 get_page(frag
->page
);
2010 frag
->page_offset
+= offset
- pos
;
2011 frag
->size
-= offset
- pos
;
2016 if (pos
+ size
<= offset
+ len
) {
2020 frag
->size
-= pos
+ size
- (offset
+ len
);
2027 skb_shinfo(nskb
)->nr_frags
= k
;
2028 nskb
->data_len
= len
- hsize
;
2029 nskb
->len
+= nskb
->data_len
;
2030 nskb
->truesize
+= nskb
->data_len
;
2031 } while ((offset
+= len
) < skb
->len
);
2036 while ((skb
= segs
)) {
2040 return ERR_PTR(err
);
2043 EXPORT_SYMBOL_GPL(skb_segment
);
2045 void __init
skb_init(void)
2047 skbuff_head_cache
= kmem_cache_create("skbuff_head_cache",
2048 sizeof(struct sk_buff
),
2050 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2052 skbuff_fclone_cache
= kmem_cache_create("skbuff_fclone_cache",
2053 (2*sizeof(struct sk_buff
)) +
2056 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2060 EXPORT_SYMBOL(___pskb_trim
);
2061 EXPORT_SYMBOL(__kfree_skb
);
2062 EXPORT_SYMBOL(kfree_skb
);
2063 EXPORT_SYMBOL(__pskb_pull_tail
);
2064 EXPORT_SYMBOL(__alloc_skb
);
2065 EXPORT_SYMBOL(__netdev_alloc_skb
);
2066 EXPORT_SYMBOL(pskb_copy
);
2067 EXPORT_SYMBOL(pskb_expand_head
);
2068 EXPORT_SYMBOL(skb_checksum
);
2069 EXPORT_SYMBOL(skb_clone
);
2070 EXPORT_SYMBOL(skb_clone_fraglist
);
2071 EXPORT_SYMBOL(skb_copy
);
2072 EXPORT_SYMBOL(skb_copy_and_csum_bits
);
2073 EXPORT_SYMBOL(skb_copy_and_csum_dev
);
2074 EXPORT_SYMBOL(skb_copy_bits
);
2075 EXPORT_SYMBOL(skb_copy_expand
);
2076 EXPORT_SYMBOL(skb_over_panic
);
2077 EXPORT_SYMBOL(skb_pad
);
2078 EXPORT_SYMBOL(skb_realloc_headroom
);
2079 EXPORT_SYMBOL(skb_under_panic
);
2080 EXPORT_SYMBOL(skb_dequeue
);
2081 EXPORT_SYMBOL(skb_dequeue_tail
);
2082 EXPORT_SYMBOL(skb_insert
);
2083 EXPORT_SYMBOL(skb_queue_purge
);
2084 EXPORT_SYMBOL(skb_queue_head
);
2085 EXPORT_SYMBOL(skb_queue_tail
);
2086 EXPORT_SYMBOL(skb_unlink
);
2087 EXPORT_SYMBOL(skb_append
);
2088 EXPORT_SYMBOL(skb_split
);
2089 EXPORT_SYMBOL(skb_prepare_seq_read
);
2090 EXPORT_SYMBOL(skb_seq_read
);
2091 EXPORT_SYMBOL(skb_abort_seq_read
);
2092 EXPORT_SYMBOL(skb_find_text
);
2093 EXPORT_SYMBOL(skb_append_datato_frags
);