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>
60 #include <net/protocol.h>
63 #include <net/checksum.h>
66 #include <asm/uaccess.h>
67 #include <asm/system.h>
71 static kmem_cache_t
*skbuff_head_cache __read_mostly
;
72 static kmem_cache_t
*skbuff_fclone_cache __read_mostly
;
75 * Keep out-of-line to prevent kernel bloat.
76 * __builtin_return_address is not used because it is not always
81 * skb_over_panic - private function
86 * Out of line support code for skb_put(). Not user callable.
88 void skb_over_panic(struct sk_buff
*skb
, int sz
, void *here
)
90 printk(KERN_EMERG
"skb_over_panic: text:%p len:%d put:%d head:%p "
91 "data:%p tail:%p end:%p dev:%s\n",
92 here
, skb
->len
, sz
, skb
->head
, skb
->data
, skb
->tail
, skb
->end
,
93 skb
->dev
? skb
->dev
->name
: "<NULL>");
98 * skb_under_panic - private function
103 * Out of line support code for skb_push(). Not user callable.
106 void skb_under_panic(struct sk_buff
*skb
, int sz
, void *here
)
108 printk(KERN_EMERG
"skb_under_panic: text:%p len:%d put:%d head:%p "
109 "data:%p tail:%p end:%p dev:%s\n",
110 here
, skb
->len
, sz
, skb
->head
, skb
->data
, skb
->tail
, skb
->end
,
111 skb
->dev
? skb
->dev
->name
: "<NULL>");
115 void skb_truesize_bug(struct sk_buff
*skb
)
117 printk(KERN_ERR
"SKB BUG: Invalid truesize (%u) "
118 "len=%u, sizeof(sk_buff)=%Zd\n",
119 skb
->truesize
, skb
->len
, sizeof(struct sk_buff
));
121 EXPORT_SYMBOL(skb_truesize_bug
);
123 /* Allocate a new skbuff. We do this ourselves so we can fill in a few
124 * 'private' fields and also do memory statistics to find all the
130 * __alloc_skb - allocate a network buffer
131 * @size: size to allocate
132 * @gfp_mask: allocation mask
133 * @fclone: allocate from fclone cache instead of head cache
134 * and allocate a cloned (child) skb
136 * Allocate a new &sk_buff. The returned buffer has no headroom and a
137 * tail room of size bytes. The object has a reference count of one.
138 * The return is the buffer. On a failure the return is %NULL.
140 * Buffers may only be allocated from interrupts using a @gfp_mask of
143 struct sk_buff
*__alloc_skb(unsigned int size
, gfp_t gfp_mask
,
147 struct skb_shared_info
*shinfo
;
151 cache
= fclone
? skbuff_fclone_cache
: skbuff_head_cache
;
154 skb
= kmem_cache_alloc(cache
, gfp_mask
& ~__GFP_DMA
);
158 /* Get the DATA. Size must match skb_add_mtu(). */
159 size
= SKB_DATA_ALIGN(size
);
160 data
= kmalloc_track_caller(size
+ sizeof(struct skb_shared_info
),
165 memset(skb
, 0, offsetof(struct sk_buff
, truesize
));
166 skb
->truesize
= size
+ sizeof(struct sk_buff
);
167 atomic_set(&skb
->users
, 1);
171 skb
->end
= data
+ size
;
172 /* make sure we initialize shinfo sequentially */
173 shinfo
= skb_shinfo(skb
);
174 atomic_set(&shinfo
->dataref
, 1);
175 shinfo
->nr_frags
= 0;
176 shinfo
->gso_size
= 0;
177 shinfo
->gso_segs
= 0;
178 shinfo
->gso_type
= 0;
179 shinfo
->ip6_frag_id
= 0;
180 shinfo
->frag_list
= NULL
;
183 struct sk_buff
*child
= skb
+ 1;
184 atomic_t
*fclone_ref
= (atomic_t
*) (child
+ 1);
186 skb
->fclone
= SKB_FCLONE_ORIG
;
187 atomic_set(fclone_ref
, 1);
189 child
->fclone
= SKB_FCLONE_UNAVAILABLE
;
194 kmem_cache_free(cache
, skb
);
200 * alloc_skb_from_cache - allocate a network buffer
201 * @cp: kmem_cache from which to allocate the data area
202 * (object size must be big enough for @size bytes + skb overheads)
203 * @size: size to allocate
204 * @gfp_mask: allocation mask
206 * Allocate a new &sk_buff. The returned buffer has no headroom and
207 * tail room of size bytes. The object has a reference count of one.
208 * The return is the buffer. On a failure the return is %NULL.
210 * Buffers may only be allocated from interrupts using a @gfp_mask of
213 struct sk_buff
*alloc_skb_from_cache(kmem_cache_t
*cp
,
221 skb
= kmem_cache_alloc(skbuff_head_cache
,
222 gfp_mask
& ~__GFP_DMA
);
227 size
= SKB_DATA_ALIGN(size
);
228 data
= kmem_cache_alloc(cp
, gfp_mask
);
232 memset(skb
, 0, offsetof(struct sk_buff
, truesize
));
233 skb
->truesize
= size
+ sizeof(struct sk_buff
);
234 atomic_set(&skb
->users
, 1);
238 skb
->end
= data
+ size
;
240 atomic_set(&(skb_shinfo(skb
)->dataref
), 1);
241 skb_shinfo(skb
)->nr_frags
= 0;
242 skb_shinfo(skb
)->gso_size
= 0;
243 skb_shinfo(skb
)->gso_segs
= 0;
244 skb_shinfo(skb
)->gso_type
= 0;
245 skb_shinfo(skb
)->frag_list
= NULL
;
249 kmem_cache_free(skbuff_head_cache
, skb
);
255 * __netdev_alloc_skb - allocate an skbuff for rx on a specific device
256 * @dev: network device to receive on
257 * @length: length to allocate
258 * @gfp_mask: get_free_pages mask, passed to alloc_skb
260 * Allocate a new &sk_buff and assign it a usage count of one. The
261 * buffer has unspecified headroom built in. Users should allocate
262 * the headroom they think they need without accounting for the
263 * built in space. The built in space is used for optimisations.
265 * %NULL is returned if there is no free memory.
267 struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
268 unsigned int length
, gfp_t gfp_mask
)
272 skb
= alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
);
274 skb_reserve(skb
, NET_SKB_PAD
);
280 static void skb_drop_list(struct sk_buff
**listp
)
282 struct sk_buff
*list
= *listp
;
287 struct sk_buff
*this = list
;
293 static inline void skb_drop_fraglist(struct sk_buff
*skb
)
295 skb_drop_list(&skb_shinfo(skb
)->frag_list
);
298 static void skb_clone_fraglist(struct sk_buff
*skb
)
300 struct sk_buff
*list
;
302 for (list
= skb_shinfo(skb
)->frag_list
; list
; list
= list
->next
)
306 static void skb_release_data(struct sk_buff
*skb
)
309 !atomic_sub_return(skb
->nohdr
? (1 << SKB_DATAREF_SHIFT
) + 1 : 1,
310 &skb_shinfo(skb
)->dataref
)) {
311 if (skb_shinfo(skb
)->nr_frags
) {
313 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++)
314 put_page(skb_shinfo(skb
)->frags
[i
].page
);
317 if (skb_shinfo(skb
)->frag_list
)
318 skb_drop_fraglist(skb
);
325 * Free an skbuff by memory without cleaning the state.
327 void kfree_skbmem(struct sk_buff
*skb
)
329 struct sk_buff
*other
;
330 atomic_t
*fclone_ref
;
332 skb_release_data(skb
);
333 switch (skb
->fclone
) {
334 case SKB_FCLONE_UNAVAILABLE
:
335 kmem_cache_free(skbuff_head_cache
, skb
);
338 case SKB_FCLONE_ORIG
:
339 fclone_ref
= (atomic_t
*) (skb
+ 2);
340 if (atomic_dec_and_test(fclone_ref
))
341 kmem_cache_free(skbuff_fclone_cache
, skb
);
344 case SKB_FCLONE_CLONE
:
345 fclone_ref
= (atomic_t
*) (skb
+ 1);
348 /* The clone portion is available for
349 * fast-cloning again.
351 skb
->fclone
= SKB_FCLONE_UNAVAILABLE
;
353 if (atomic_dec_and_test(fclone_ref
))
354 kmem_cache_free(skbuff_fclone_cache
, other
);
360 * __kfree_skb - private function
363 * Free an sk_buff. Release anything attached to the buffer.
364 * Clean the state. This is an internal helper function. Users should
365 * always call kfree_skb
368 void __kfree_skb(struct sk_buff
*skb
)
370 dst_release(skb
->dst
);
372 secpath_put(skb
->sp
);
374 if (skb
->destructor
) {
376 skb
->destructor(skb
);
378 #ifdef CONFIG_NETFILTER
379 nf_conntrack_put(skb
->nfct
);
380 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
381 nf_conntrack_put_reasm(skb
->nfct_reasm
);
383 #ifdef CONFIG_BRIDGE_NETFILTER
384 nf_bridge_put(skb
->nf_bridge
);
387 /* XXX: IS this still necessary? - JHS */
388 #ifdef CONFIG_NET_SCHED
390 #ifdef CONFIG_NET_CLS_ACT
399 * kfree_skb - free an sk_buff
400 * @skb: buffer to free
402 * Drop a reference to the buffer and free it if the usage count has
405 void kfree_skb(struct sk_buff
*skb
)
409 if (likely(atomic_read(&skb
->users
) == 1))
411 else if (likely(!atomic_dec_and_test(&skb
->users
)))
417 * skb_clone - duplicate an sk_buff
418 * @skb: buffer to clone
419 * @gfp_mask: allocation priority
421 * Duplicate an &sk_buff. The new one is not owned by a socket. Both
422 * copies share the same packet data but not structure. The new
423 * buffer has a reference count of 1. If the allocation fails the
424 * function returns %NULL otherwise the new buffer is returned.
426 * If this function is called from an interrupt gfp_mask() must be
430 struct sk_buff
*skb_clone(struct sk_buff
*skb
, gfp_t gfp_mask
)
435 if (skb
->fclone
== SKB_FCLONE_ORIG
&&
436 n
->fclone
== SKB_FCLONE_UNAVAILABLE
) {
437 atomic_t
*fclone_ref
= (atomic_t
*) (n
+ 1);
438 n
->fclone
= SKB_FCLONE_CLONE
;
439 atomic_inc(fclone_ref
);
441 n
= kmem_cache_alloc(skbuff_head_cache
, gfp_mask
);
444 n
->fclone
= SKB_FCLONE_UNAVAILABLE
;
447 #define C(x) n->x = skb->x
449 n
->next
= n
->prev
= NULL
;
460 secpath_get(skb
->sp
);
462 memcpy(n
->cb
, skb
->cb
, sizeof(skb
->cb
));
472 #if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
476 n
->destructor
= NULL
;
478 #ifdef CONFIG_NETFILTER
480 nf_conntrack_get(skb
->nfct
);
482 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
484 nf_conntrack_get_reasm(skb
->nfct_reasm
);
486 #ifdef CONFIG_BRIDGE_NETFILTER
488 nf_bridge_get(skb
->nf_bridge
);
490 #endif /*CONFIG_NETFILTER*/
491 #ifdef CONFIG_NET_SCHED
493 #ifdef CONFIG_NET_CLS_ACT
494 n
->tc_verd
= SET_TC_VERD(skb
->tc_verd
,0);
495 n
->tc_verd
= CLR_TC_OK2MUNGE(n
->tc_verd
);
496 n
->tc_verd
= CLR_TC_MUNGED(n
->tc_verd
);
499 skb_copy_secmark(n
, skb
);
502 atomic_set(&n
->users
, 1);
508 atomic_inc(&(skb_shinfo(skb
)->dataref
));
514 static void copy_skb_header(struct sk_buff
*new, const struct sk_buff
*old
)
517 * Shift between the two data areas in bytes
519 unsigned long offset
= new->data
- old
->data
;
523 new->priority
= old
->priority
;
524 new->protocol
= old
->protocol
;
525 new->dst
= dst_clone(old
->dst
);
527 new->sp
= secpath_get(old
->sp
);
529 new->h
.raw
= old
->h
.raw
+ offset
;
530 new->nh
.raw
= old
->nh
.raw
+ offset
;
531 new->mac
.raw
= old
->mac
.raw
+ offset
;
532 memcpy(new->cb
, old
->cb
, sizeof(old
->cb
));
533 new->local_df
= old
->local_df
;
534 new->fclone
= SKB_FCLONE_UNAVAILABLE
;
535 new->pkt_type
= old
->pkt_type
;
536 new->tstamp
= old
->tstamp
;
537 new->destructor
= NULL
;
538 new->mark
= old
->mark
;
539 #ifdef CONFIG_NETFILTER
540 new->nfct
= old
->nfct
;
541 nf_conntrack_get(old
->nfct
);
542 new->nfctinfo
= old
->nfctinfo
;
543 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
544 new->nfct_reasm
= old
->nfct_reasm
;
545 nf_conntrack_get_reasm(old
->nfct_reasm
);
547 #if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
548 new->ipvs_property
= old
->ipvs_property
;
550 #ifdef CONFIG_BRIDGE_NETFILTER
551 new->nf_bridge
= old
->nf_bridge
;
552 nf_bridge_get(old
->nf_bridge
);
555 #ifdef CONFIG_NET_SCHED
556 #ifdef CONFIG_NET_CLS_ACT
557 new->tc_verd
= old
->tc_verd
;
559 new->tc_index
= old
->tc_index
;
561 skb_copy_secmark(new, old
);
562 atomic_set(&new->users
, 1);
563 skb_shinfo(new)->gso_size
= skb_shinfo(old
)->gso_size
;
564 skb_shinfo(new)->gso_segs
= skb_shinfo(old
)->gso_segs
;
565 skb_shinfo(new)->gso_type
= skb_shinfo(old
)->gso_type
;
569 * skb_copy - create private copy of an sk_buff
570 * @skb: buffer to copy
571 * @gfp_mask: allocation priority
573 * Make a copy of both an &sk_buff and its data. This is used when the
574 * caller wishes to modify the data and needs a private copy of the
575 * data to alter. Returns %NULL on failure or the pointer to the buffer
576 * on success. The returned buffer has a reference count of 1.
578 * As by-product this function converts non-linear &sk_buff to linear
579 * one, so that &sk_buff becomes completely private and caller is allowed
580 * to modify all the data of returned buffer. This means that this
581 * function is not recommended for use in circumstances when only
582 * header is going to be modified. Use pskb_copy() instead.
585 struct sk_buff
*skb_copy(const struct sk_buff
*skb
, gfp_t gfp_mask
)
587 int headerlen
= skb
->data
- skb
->head
;
589 * Allocate the copy buffer
591 struct sk_buff
*n
= alloc_skb(skb
->end
- skb
->head
+ skb
->data_len
,
596 /* Set the data pointer */
597 skb_reserve(n
, headerlen
);
598 /* Set the tail pointer and length */
599 skb_put(n
, skb
->len
);
601 n
->ip_summed
= skb
->ip_summed
;
603 if (skb_copy_bits(skb
, -headerlen
, n
->head
, headerlen
+ skb
->len
))
606 copy_skb_header(n
, skb
);
612 * pskb_copy - create copy of an sk_buff with private head.
613 * @skb: buffer to copy
614 * @gfp_mask: allocation priority
616 * Make a copy of both an &sk_buff and part of its data, located
617 * in header. Fragmented data remain shared. This is used when
618 * the caller wishes to modify only header of &sk_buff and needs
619 * private copy of the header to alter. Returns %NULL on failure
620 * or the pointer to the buffer on success.
621 * The returned buffer has a reference count of 1.
624 struct sk_buff
*pskb_copy(struct sk_buff
*skb
, gfp_t gfp_mask
)
627 * Allocate the copy buffer
629 struct sk_buff
*n
= alloc_skb(skb
->end
- skb
->head
, gfp_mask
);
634 /* Set the data pointer */
635 skb_reserve(n
, skb
->data
- skb
->head
);
636 /* Set the tail pointer and length */
637 skb_put(n
, skb_headlen(skb
));
639 memcpy(n
->data
, skb
->data
, n
->len
);
641 n
->ip_summed
= skb
->ip_summed
;
643 n
->truesize
+= skb
->data_len
;
644 n
->data_len
= skb
->data_len
;
647 if (skb_shinfo(skb
)->nr_frags
) {
650 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
651 skb_shinfo(n
)->frags
[i
] = skb_shinfo(skb
)->frags
[i
];
652 get_page(skb_shinfo(n
)->frags
[i
].page
);
654 skb_shinfo(n
)->nr_frags
= i
;
657 if (skb_shinfo(skb
)->frag_list
) {
658 skb_shinfo(n
)->frag_list
= skb_shinfo(skb
)->frag_list
;
659 skb_clone_fraglist(n
);
662 copy_skb_header(n
, skb
);
668 * pskb_expand_head - reallocate header of &sk_buff
669 * @skb: buffer to reallocate
670 * @nhead: room to add at head
671 * @ntail: room to add at tail
672 * @gfp_mask: allocation priority
674 * Expands (or creates identical copy, if &nhead and &ntail are zero)
675 * header of skb. &sk_buff itself is not changed. &sk_buff MUST have
676 * reference count of 1. Returns zero in the case of success or error,
677 * if expansion failed. In the last case, &sk_buff is not changed.
679 * All the pointers pointing into skb header may change and must be
680 * reloaded after call to this function.
683 int pskb_expand_head(struct sk_buff
*skb
, int nhead
, int ntail
,
688 int size
= nhead
+ (skb
->end
- skb
->head
) + ntail
;
694 size
= SKB_DATA_ALIGN(size
);
696 data
= kmalloc(size
+ sizeof(struct skb_shared_info
), gfp_mask
);
700 /* Copy only real data... and, alas, header. This should be
701 * optimized for the cases when header is void. */
702 memcpy(data
+ nhead
, skb
->head
, skb
->tail
- skb
->head
);
703 memcpy(data
+ size
, skb
->end
, sizeof(struct skb_shared_info
));
705 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++)
706 get_page(skb_shinfo(skb
)->frags
[i
].page
);
708 if (skb_shinfo(skb
)->frag_list
)
709 skb_clone_fraglist(skb
);
711 skb_release_data(skb
);
713 off
= (data
+ nhead
) - skb
->head
;
716 skb
->end
= data
+ size
;
724 atomic_set(&skb_shinfo(skb
)->dataref
, 1);
731 /* Make private copy of skb with writable head and some headroom */
733 struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
, unsigned int headroom
)
735 struct sk_buff
*skb2
;
736 int delta
= headroom
- skb_headroom(skb
);
739 skb2
= pskb_copy(skb
, GFP_ATOMIC
);
741 skb2
= skb_clone(skb
, GFP_ATOMIC
);
742 if (skb2
&& pskb_expand_head(skb2
, SKB_DATA_ALIGN(delta
), 0,
753 * skb_copy_expand - copy and expand sk_buff
754 * @skb: buffer to copy
755 * @newheadroom: new free bytes at head
756 * @newtailroom: new free bytes at tail
757 * @gfp_mask: allocation priority
759 * Make a copy of both an &sk_buff and its data and while doing so
760 * allocate additional space.
762 * This is used when the caller wishes to modify the data and needs a
763 * private copy of the data to alter as well as more space for new fields.
764 * Returns %NULL on failure or the pointer to the buffer
765 * on success. The returned buffer has a reference count of 1.
767 * You must pass %GFP_ATOMIC as the allocation priority if this function
768 * is called from an interrupt.
770 * BUG ALERT: ip_summed is not copied. Why does this work? Is it used
771 * only by netfilter in the cases when checksum is recalculated? --ANK
773 struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
774 int newheadroom
, int newtailroom
,
778 * Allocate the copy buffer
780 struct sk_buff
*n
= alloc_skb(newheadroom
+ skb
->len
+ newtailroom
,
782 int head_copy_len
, head_copy_off
;
787 skb_reserve(n
, newheadroom
);
789 /* Set the tail pointer and length */
790 skb_put(n
, skb
->len
);
792 head_copy_len
= skb_headroom(skb
);
794 if (newheadroom
<= head_copy_len
)
795 head_copy_len
= newheadroom
;
797 head_copy_off
= newheadroom
- head_copy_len
;
799 /* Copy the linear header and data. */
800 if (skb_copy_bits(skb
, -head_copy_len
, n
->head
+ head_copy_off
,
801 skb
->len
+ head_copy_len
))
804 copy_skb_header(n
, skb
);
810 * skb_pad - zero pad the tail of an skb
811 * @skb: buffer to pad
814 * Ensure that a buffer is followed by a padding area that is zero
815 * filled. Used by network drivers which may DMA or transfer data
816 * beyond the buffer end onto the wire.
818 * May return error in out of memory cases. The skb is freed on error.
821 int skb_pad(struct sk_buff
*skb
, int pad
)
826 /* If the skbuff is non linear tailroom is always zero.. */
827 if (!skb_cloned(skb
) && skb_tailroom(skb
) >= pad
) {
828 memset(skb
->data
+skb
->len
, 0, pad
);
832 ntail
= skb
->data_len
+ pad
- (skb
->end
- skb
->tail
);
833 if (likely(skb_cloned(skb
) || ntail
> 0)) {
834 err
= pskb_expand_head(skb
, 0, ntail
, GFP_ATOMIC
);
839 /* FIXME: The use of this function with non-linear skb's really needs
842 err
= skb_linearize(skb
);
846 memset(skb
->data
+ skb
->len
, 0, pad
);
854 /* Trims skb to length len. It can change skb pointers.
857 int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
)
859 struct sk_buff
**fragp
;
860 struct sk_buff
*frag
;
861 int offset
= skb_headlen(skb
);
862 int nfrags
= skb_shinfo(skb
)->nr_frags
;
866 if (skb_cloned(skb
) &&
867 unlikely((err
= pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
))))
874 for (; i
< nfrags
; i
++) {
875 int end
= offset
+ skb_shinfo(skb
)->frags
[i
].size
;
882 skb_shinfo(skb
)->frags
[i
++].size
= len
- offset
;
885 skb_shinfo(skb
)->nr_frags
= i
;
887 for (; i
< nfrags
; i
++)
888 put_page(skb_shinfo(skb
)->frags
[i
].page
);
890 if (skb_shinfo(skb
)->frag_list
)
891 skb_drop_fraglist(skb
);
895 for (fragp
= &skb_shinfo(skb
)->frag_list
; (frag
= *fragp
);
896 fragp
= &frag
->next
) {
897 int end
= offset
+ frag
->len
;
899 if (skb_shared(frag
)) {
900 struct sk_buff
*nfrag
;
902 nfrag
= skb_clone(frag
, GFP_ATOMIC
);
903 if (unlikely(!nfrag
))
906 nfrag
->next
= frag
->next
;
918 unlikely((err
= pskb_trim(frag
, len
- offset
))))
922 skb_drop_list(&frag
->next
);
927 if (len
> skb_headlen(skb
)) {
928 skb
->data_len
-= skb
->len
- len
;
933 skb
->tail
= skb
->data
+ len
;
940 * __pskb_pull_tail - advance tail of skb header
941 * @skb: buffer to reallocate
942 * @delta: number of bytes to advance tail
944 * The function makes a sense only on a fragmented &sk_buff,
945 * it expands header moving its tail forward and copying necessary
946 * data from fragmented part.
948 * &sk_buff MUST have reference count of 1.
950 * Returns %NULL (and &sk_buff does not change) if pull failed
951 * or value of new tail of skb in the case of success.
953 * All the pointers pointing into skb header may change and must be
954 * reloaded after call to this function.
957 /* Moves tail of skb head forward, copying data from fragmented part,
958 * when it is necessary.
959 * 1. It may fail due to malloc failure.
960 * 2. It may change skb pointers.
962 * It is pretty complicated. Luckily, it is called only in exceptional cases.
964 unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
)
966 /* If skb has not enough free space at tail, get new one
967 * plus 128 bytes for future expansions. If we have enough
968 * room at tail, reallocate without expansion only if skb is cloned.
970 int i
, k
, eat
= (skb
->tail
+ delta
) - skb
->end
;
972 if (eat
> 0 || skb_cloned(skb
)) {
973 if (pskb_expand_head(skb
, 0, eat
> 0 ? eat
+ 128 : 0,
978 if (skb_copy_bits(skb
, skb_headlen(skb
), skb
->tail
, delta
))
981 /* Optimization: no fragments, no reasons to preestimate
982 * size of pulled pages. Superb.
984 if (!skb_shinfo(skb
)->frag_list
)
987 /* Estimate size of pulled pages. */
989 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
990 if (skb_shinfo(skb
)->frags
[i
].size
>= eat
)
992 eat
-= skb_shinfo(skb
)->frags
[i
].size
;
995 /* If we need update frag list, we are in troubles.
996 * Certainly, it possible to add an offset to skb data,
997 * but taking into account that pulling is expected to
998 * be very rare operation, it is worth to fight against
999 * further bloating skb head and crucify ourselves here instead.
1000 * Pure masohism, indeed. 8)8)
1003 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1004 struct sk_buff
*clone
= NULL
;
1005 struct sk_buff
*insp
= NULL
;
1010 if (list
->len
<= eat
) {
1011 /* Eaten as whole. */
1016 /* Eaten partially. */
1018 if (skb_shared(list
)) {
1019 /* Sucks! We need to fork list. :-( */
1020 clone
= skb_clone(list
, GFP_ATOMIC
);
1026 /* This may be pulled without
1030 if (!pskb_pull(list
, eat
)) {
1039 /* Free pulled out fragments. */
1040 while ((list
= skb_shinfo(skb
)->frag_list
) != insp
) {
1041 skb_shinfo(skb
)->frag_list
= list
->next
;
1044 /* And insert new clone at head. */
1047 skb_shinfo(skb
)->frag_list
= clone
;
1050 /* Success! Now we may commit changes to skb data. */
1055 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1056 if (skb_shinfo(skb
)->frags
[i
].size
<= eat
) {
1057 put_page(skb_shinfo(skb
)->frags
[i
].page
);
1058 eat
-= skb_shinfo(skb
)->frags
[i
].size
;
1060 skb_shinfo(skb
)->frags
[k
] = skb_shinfo(skb
)->frags
[i
];
1062 skb_shinfo(skb
)->frags
[k
].page_offset
+= eat
;
1063 skb_shinfo(skb
)->frags
[k
].size
-= eat
;
1069 skb_shinfo(skb
)->nr_frags
= k
;
1072 skb
->data_len
-= delta
;
1077 /* Copy some data bits from skb to kernel buffer. */
1079 int skb_copy_bits(const struct sk_buff
*skb
, int offset
, void *to
, int len
)
1082 int start
= skb_headlen(skb
);
1084 if (offset
> (int)skb
->len
- len
)
1088 if ((copy
= start
- offset
) > 0) {
1091 memcpy(to
, skb
->data
+ offset
, copy
);
1092 if ((len
-= copy
) == 0)
1098 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1101 BUG_TRAP(start
<= offset
+ len
);
1103 end
= start
+ skb_shinfo(skb
)->frags
[i
].size
;
1104 if ((copy
= end
- offset
) > 0) {
1110 vaddr
= kmap_skb_frag(&skb_shinfo(skb
)->frags
[i
]);
1112 vaddr
+ skb_shinfo(skb
)->frags
[i
].page_offset
+
1113 offset
- start
, copy
);
1114 kunmap_skb_frag(vaddr
);
1116 if ((len
-= copy
) == 0)
1124 if (skb_shinfo(skb
)->frag_list
) {
1125 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1127 for (; list
; list
= list
->next
) {
1130 BUG_TRAP(start
<= offset
+ len
);
1132 end
= start
+ list
->len
;
1133 if ((copy
= end
- offset
) > 0) {
1136 if (skb_copy_bits(list
, offset
- start
,
1139 if ((len
-= copy
) == 0)
1155 * skb_store_bits - store bits from kernel buffer to skb
1156 * @skb: destination buffer
1157 * @offset: offset in destination
1158 * @from: source buffer
1159 * @len: number of bytes to copy
1161 * Copy the specified number of bytes from the source buffer to the
1162 * destination skb. This function handles all the messy bits of
1163 * traversing fragment lists and such.
1166 int skb_store_bits(const struct sk_buff
*skb
, int offset
, void *from
, int len
)
1169 int start
= skb_headlen(skb
);
1171 if (offset
> (int)skb
->len
- len
)
1174 if ((copy
= start
- offset
) > 0) {
1177 memcpy(skb
->data
+ offset
, from
, copy
);
1178 if ((len
-= copy
) == 0)
1184 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1185 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1188 BUG_TRAP(start
<= offset
+ len
);
1190 end
= start
+ frag
->size
;
1191 if ((copy
= end
- offset
) > 0) {
1197 vaddr
= kmap_skb_frag(frag
);
1198 memcpy(vaddr
+ frag
->page_offset
+ offset
- start
,
1200 kunmap_skb_frag(vaddr
);
1202 if ((len
-= copy
) == 0)
1210 if (skb_shinfo(skb
)->frag_list
) {
1211 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1213 for (; list
; list
= list
->next
) {
1216 BUG_TRAP(start
<= offset
+ len
);
1218 end
= start
+ list
->len
;
1219 if ((copy
= end
- offset
) > 0) {
1222 if (skb_store_bits(list
, offset
- start
,
1225 if ((len
-= copy
) == 0)
1240 EXPORT_SYMBOL(skb_store_bits
);
1242 /* Checksum skb data. */
1244 __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
,
1245 int len
, __wsum csum
)
1247 int start
= skb_headlen(skb
);
1248 int i
, copy
= start
- offset
;
1251 /* Checksum header. */
1255 csum
= csum_partial(skb
->data
+ offset
, copy
, csum
);
1256 if ((len
-= copy
) == 0)
1262 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1265 BUG_TRAP(start
<= offset
+ len
);
1267 end
= start
+ skb_shinfo(skb
)->frags
[i
].size
;
1268 if ((copy
= end
- offset
) > 0) {
1271 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1275 vaddr
= kmap_skb_frag(frag
);
1276 csum2
= csum_partial(vaddr
+ frag
->page_offset
+
1277 offset
- start
, copy
, 0);
1278 kunmap_skb_frag(vaddr
);
1279 csum
= csum_block_add(csum
, csum2
, pos
);
1288 if (skb_shinfo(skb
)->frag_list
) {
1289 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1291 for (; list
; list
= list
->next
) {
1294 BUG_TRAP(start
<= offset
+ len
);
1296 end
= start
+ list
->len
;
1297 if ((copy
= end
- offset
) > 0) {
1301 csum2
= skb_checksum(list
, offset
- start
,
1303 csum
= csum_block_add(csum
, csum2
, pos
);
1304 if ((len
-= copy
) == 0)
1317 /* Both of above in one bottle. */
1319 __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
, int offset
,
1320 u8
*to
, int len
, __wsum csum
)
1322 int start
= skb_headlen(skb
);
1323 int i
, copy
= start
- offset
;
1330 csum
= csum_partial_copy_nocheck(skb
->data
+ offset
, to
,
1332 if ((len
-= copy
) == 0)
1339 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1342 BUG_TRAP(start
<= offset
+ len
);
1344 end
= start
+ skb_shinfo(skb
)->frags
[i
].size
;
1345 if ((copy
= end
- offset
) > 0) {
1348 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1352 vaddr
= kmap_skb_frag(frag
);
1353 csum2
= csum_partial_copy_nocheck(vaddr
+
1357 kunmap_skb_frag(vaddr
);
1358 csum
= csum_block_add(csum
, csum2
, pos
);
1368 if (skb_shinfo(skb
)->frag_list
) {
1369 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1371 for (; list
; list
= list
->next
) {
1375 BUG_TRAP(start
<= offset
+ len
);
1377 end
= start
+ list
->len
;
1378 if ((copy
= end
- offset
) > 0) {
1381 csum2
= skb_copy_and_csum_bits(list
,
1384 csum
= csum_block_add(csum
, csum2
, pos
);
1385 if ((len
-= copy
) == 0)
1398 void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
)
1403 if (skb
->ip_summed
== CHECKSUM_PARTIAL
)
1404 csstart
= skb
->h
.raw
- skb
->data
;
1406 csstart
= skb_headlen(skb
);
1408 BUG_ON(csstart
> skb_headlen(skb
));
1410 memcpy(to
, skb
->data
, csstart
);
1413 if (csstart
!= skb
->len
)
1414 csum
= skb_copy_and_csum_bits(skb
, csstart
, to
+ csstart
,
1415 skb
->len
- csstart
, 0);
1417 if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1418 long csstuff
= csstart
+ skb
->csum_offset
;
1420 *((__sum16
*)(to
+ csstuff
)) = csum_fold(csum
);
1425 * skb_dequeue - remove from the head of the queue
1426 * @list: list to dequeue from
1428 * Remove the head of the list. The list lock is taken so the function
1429 * may be used safely with other locking list functions. The head item is
1430 * returned or %NULL if the list is empty.
1433 struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
)
1435 unsigned long flags
;
1436 struct sk_buff
*result
;
1438 spin_lock_irqsave(&list
->lock
, flags
);
1439 result
= __skb_dequeue(list
);
1440 spin_unlock_irqrestore(&list
->lock
, flags
);
1445 * skb_dequeue_tail - remove from the tail of the queue
1446 * @list: list to dequeue from
1448 * Remove the tail of the list. The list lock is taken so the function
1449 * may be used safely with other locking list functions. The tail item is
1450 * returned or %NULL if the list is empty.
1452 struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
)
1454 unsigned long flags
;
1455 struct sk_buff
*result
;
1457 spin_lock_irqsave(&list
->lock
, flags
);
1458 result
= __skb_dequeue_tail(list
);
1459 spin_unlock_irqrestore(&list
->lock
, flags
);
1464 * skb_queue_purge - empty a list
1465 * @list: list to empty
1467 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1468 * the list and one reference dropped. This function takes the list
1469 * lock and is atomic with respect to other list locking functions.
1471 void skb_queue_purge(struct sk_buff_head
*list
)
1473 struct sk_buff
*skb
;
1474 while ((skb
= skb_dequeue(list
)) != NULL
)
1479 * skb_queue_head - queue a buffer at the list head
1480 * @list: list to use
1481 * @newsk: buffer to queue
1483 * Queue a buffer at the start of the list. This function takes the
1484 * list lock and can be used safely with other locking &sk_buff functions
1487 * A buffer cannot be placed on two lists at the same time.
1489 void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
)
1491 unsigned long flags
;
1493 spin_lock_irqsave(&list
->lock
, flags
);
1494 __skb_queue_head(list
, newsk
);
1495 spin_unlock_irqrestore(&list
->lock
, flags
);
1499 * skb_queue_tail - queue a buffer at the list tail
1500 * @list: list to use
1501 * @newsk: buffer to queue
1503 * Queue a buffer at the tail of the list. This function takes the
1504 * list lock and can be used safely with other locking &sk_buff functions
1507 * A buffer cannot be placed on two lists at the same time.
1509 void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
)
1511 unsigned long flags
;
1513 spin_lock_irqsave(&list
->lock
, flags
);
1514 __skb_queue_tail(list
, newsk
);
1515 spin_unlock_irqrestore(&list
->lock
, flags
);
1519 * skb_unlink - remove a buffer from a list
1520 * @skb: buffer to remove
1521 * @list: list to use
1523 * Remove a packet from a list. The list locks are taken and this
1524 * function is atomic with respect to other list locked calls
1526 * You must know what list the SKB is on.
1528 void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1530 unsigned long flags
;
1532 spin_lock_irqsave(&list
->lock
, flags
);
1533 __skb_unlink(skb
, list
);
1534 spin_unlock_irqrestore(&list
->lock
, flags
);
1538 * skb_append - append a buffer
1539 * @old: buffer to insert after
1540 * @newsk: buffer to insert
1541 * @list: list to use
1543 * Place a packet after a given packet in a list. The list locks are taken
1544 * and this function is atomic with respect to other list locked calls.
1545 * A buffer cannot be placed on two lists at the same time.
1547 void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
1549 unsigned long flags
;
1551 spin_lock_irqsave(&list
->lock
, flags
);
1552 __skb_append(old
, newsk
, list
);
1553 spin_unlock_irqrestore(&list
->lock
, flags
);
1558 * skb_insert - insert a buffer
1559 * @old: buffer to insert before
1560 * @newsk: buffer to insert
1561 * @list: list to use
1563 * Place a packet before a given packet in a list. The list locks are
1564 * taken and this function is atomic with respect to other list locked
1567 * A buffer cannot be placed on two lists at the same time.
1569 void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
1571 unsigned long flags
;
1573 spin_lock_irqsave(&list
->lock
, flags
);
1574 __skb_insert(newsk
, old
->prev
, old
, list
);
1575 spin_unlock_irqrestore(&list
->lock
, flags
);
1580 * Tune the memory allocator for a new MTU size.
1582 void skb_add_mtu(int mtu
)
1584 /* Must match allocation in alloc_skb */
1585 mtu
= SKB_DATA_ALIGN(mtu
) + sizeof(struct skb_shared_info
);
1587 kmem_add_cache_size(mtu
);
1591 static inline void skb_split_inside_header(struct sk_buff
*skb
,
1592 struct sk_buff
* skb1
,
1593 const u32 len
, const int pos
)
1597 memcpy(skb_put(skb1
, pos
- len
), skb
->data
+ len
, pos
- len
);
1599 /* And move data appendix as is. */
1600 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++)
1601 skb_shinfo(skb1
)->frags
[i
] = skb_shinfo(skb
)->frags
[i
];
1603 skb_shinfo(skb1
)->nr_frags
= skb_shinfo(skb
)->nr_frags
;
1604 skb_shinfo(skb
)->nr_frags
= 0;
1605 skb1
->data_len
= skb
->data_len
;
1606 skb1
->len
+= skb1
->data_len
;
1609 skb
->tail
= skb
->data
+ len
;
1612 static inline void skb_split_no_header(struct sk_buff
*skb
,
1613 struct sk_buff
* skb1
,
1614 const u32 len
, int pos
)
1617 const int nfrags
= skb_shinfo(skb
)->nr_frags
;
1619 skb_shinfo(skb
)->nr_frags
= 0;
1620 skb1
->len
= skb1
->data_len
= skb
->len
- len
;
1622 skb
->data_len
= len
- pos
;
1624 for (i
= 0; i
< nfrags
; i
++) {
1625 int size
= skb_shinfo(skb
)->frags
[i
].size
;
1627 if (pos
+ size
> len
) {
1628 skb_shinfo(skb1
)->frags
[k
] = skb_shinfo(skb
)->frags
[i
];
1632 * We have two variants in this case:
1633 * 1. Move all the frag to the second
1634 * part, if it is possible. F.e.
1635 * this approach is mandatory for TUX,
1636 * where splitting is expensive.
1637 * 2. Split is accurately. We make this.
1639 get_page(skb_shinfo(skb
)->frags
[i
].page
);
1640 skb_shinfo(skb1
)->frags
[0].page_offset
+= len
- pos
;
1641 skb_shinfo(skb1
)->frags
[0].size
-= len
- pos
;
1642 skb_shinfo(skb
)->frags
[i
].size
= len
- pos
;
1643 skb_shinfo(skb
)->nr_frags
++;
1647 skb_shinfo(skb
)->nr_frags
++;
1650 skb_shinfo(skb1
)->nr_frags
= k
;
1654 * skb_split - Split fragmented skb to two parts at length len.
1655 * @skb: the buffer to split
1656 * @skb1: the buffer to receive the second part
1657 * @len: new length for skb
1659 void skb_split(struct sk_buff
*skb
, struct sk_buff
*skb1
, const u32 len
)
1661 int pos
= skb_headlen(skb
);
1663 if (len
< pos
) /* Split line is inside header. */
1664 skb_split_inside_header(skb
, skb1
, len
, pos
);
1665 else /* Second chunk has no header, nothing to copy. */
1666 skb_split_no_header(skb
, skb1
, len
, pos
);
1670 * skb_prepare_seq_read - Prepare a sequential read of skb data
1671 * @skb: the buffer to read
1672 * @from: lower offset of data to be read
1673 * @to: upper offset of data to be read
1674 * @st: state variable
1676 * Initializes the specified state variable. Must be called before
1677 * invoking skb_seq_read() for the first time.
1679 void skb_prepare_seq_read(struct sk_buff
*skb
, unsigned int from
,
1680 unsigned int to
, struct skb_seq_state
*st
)
1682 st
->lower_offset
= from
;
1683 st
->upper_offset
= to
;
1684 st
->root_skb
= st
->cur_skb
= skb
;
1685 st
->frag_idx
= st
->stepped_offset
= 0;
1686 st
->frag_data
= NULL
;
1690 * skb_seq_read - Sequentially read skb data
1691 * @consumed: number of bytes consumed by the caller so far
1692 * @data: destination pointer for data to be returned
1693 * @st: state variable
1695 * Reads a block of skb data at &consumed relative to the
1696 * lower offset specified to skb_prepare_seq_read(). Assigns
1697 * the head of the data block to &data and returns the length
1698 * of the block or 0 if the end of the skb data or the upper
1699 * offset has been reached.
1701 * The caller is not required to consume all of the data
1702 * returned, i.e. &consumed is typically set to the number
1703 * of bytes already consumed and the next call to
1704 * skb_seq_read() will return the remaining part of the block.
1706 * Note: The size of each block of data returned can be arbitary,
1707 * this limitation is the cost for zerocopy seqeuental
1708 * reads of potentially non linear data.
1710 * Note: Fragment lists within fragments are not implemented
1711 * at the moment, state->root_skb could be replaced with
1712 * a stack for this purpose.
1714 unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
1715 struct skb_seq_state
*st
)
1717 unsigned int block_limit
, abs_offset
= consumed
+ st
->lower_offset
;
1720 if (unlikely(abs_offset
>= st
->upper_offset
))
1724 block_limit
= skb_headlen(st
->cur_skb
);
1726 if (abs_offset
< block_limit
) {
1727 *data
= st
->cur_skb
->data
+ abs_offset
;
1728 return block_limit
- abs_offset
;
1731 if (st
->frag_idx
== 0 && !st
->frag_data
)
1732 st
->stepped_offset
+= skb_headlen(st
->cur_skb
);
1734 while (st
->frag_idx
< skb_shinfo(st
->cur_skb
)->nr_frags
) {
1735 frag
= &skb_shinfo(st
->cur_skb
)->frags
[st
->frag_idx
];
1736 block_limit
= frag
->size
+ st
->stepped_offset
;
1738 if (abs_offset
< block_limit
) {
1740 st
->frag_data
= kmap_skb_frag(frag
);
1742 *data
= (u8
*) st
->frag_data
+ frag
->page_offset
+
1743 (abs_offset
- st
->stepped_offset
);
1745 return block_limit
- abs_offset
;
1748 if (st
->frag_data
) {
1749 kunmap_skb_frag(st
->frag_data
);
1750 st
->frag_data
= NULL
;
1754 st
->stepped_offset
+= frag
->size
;
1757 if (st
->cur_skb
->next
) {
1758 st
->cur_skb
= st
->cur_skb
->next
;
1761 } else if (st
->root_skb
== st
->cur_skb
&&
1762 skb_shinfo(st
->root_skb
)->frag_list
) {
1763 st
->cur_skb
= skb_shinfo(st
->root_skb
)->frag_list
;
1771 * skb_abort_seq_read - Abort a sequential read of skb data
1772 * @st: state variable
1774 * Must be called if skb_seq_read() was not called until it
1777 void skb_abort_seq_read(struct skb_seq_state
*st
)
1780 kunmap_skb_frag(st
->frag_data
);
1783 #define TS_SKB_CB(state) ((struct skb_seq_state *) &((state)->cb))
1785 static unsigned int skb_ts_get_next_block(unsigned int offset
, const u8
**text
,
1786 struct ts_config
*conf
,
1787 struct ts_state
*state
)
1789 return skb_seq_read(offset
, text
, TS_SKB_CB(state
));
1792 static void skb_ts_finish(struct ts_config
*conf
, struct ts_state
*state
)
1794 skb_abort_seq_read(TS_SKB_CB(state
));
1798 * skb_find_text - Find a text pattern in skb data
1799 * @skb: the buffer to look in
1800 * @from: search offset
1802 * @config: textsearch configuration
1803 * @state: uninitialized textsearch state variable
1805 * Finds a pattern in the skb data according to the specified
1806 * textsearch configuration. Use textsearch_next() to retrieve
1807 * subsequent occurrences of the pattern. Returns the offset
1808 * to the first occurrence or UINT_MAX if no match was found.
1810 unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
1811 unsigned int to
, struct ts_config
*config
,
1812 struct ts_state
*state
)
1816 config
->get_next_block
= skb_ts_get_next_block
;
1817 config
->finish
= skb_ts_finish
;
1819 skb_prepare_seq_read(skb
, from
, to
, TS_SKB_CB(state
));
1821 ret
= textsearch_find(config
, state
);
1822 return (ret
<= to
- from
? ret
: UINT_MAX
);
1826 * skb_append_datato_frags: - append the user data to a skb
1827 * @sk: sock structure
1828 * @skb: skb structure to be appened with user data.
1829 * @getfrag: call back function to be used for getting the user data
1830 * @from: pointer to user message iov
1831 * @length: length of the iov message
1833 * Description: This procedure append the user data in the fragment part
1834 * of the skb if any page alloc fails user this procedure returns -ENOMEM
1836 int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
1837 int (*getfrag
)(void *from
, char *to
, int offset
,
1838 int len
, int odd
, struct sk_buff
*skb
),
1839 void *from
, int length
)
1842 skb_frag_t
*frag
= NULL
;
1843 struct page
*page
= NULL
;
1849 /* Return error if we don't have space for new frag */
1850 frg_cnt
= skb_shinfo(skb
)->nr_frags
;
1851 if (frg_cnt
>= MAX_SKB_FRAGS
)
1854 /* allocate a new page for next frag */
1855 page
= alloc_pages(sk
->sk_allocation
, 0);
1857 /* If alloc_page fails just return failure and caller will
1858 * free previous allocated pages by doing kfree_skb()
1863 /* initialize the next frag */
1864 sk
->sk_sndmsg_page
= page
;
1865 sk
->sk_sndmsg_off
= 0;
1866 skb_fill_page_desc(skb
, frg_cnt
, page
, 0, 0);
1867 skb
->truesize
+= PAGE_SIZE
;
1868 atomic_add(PAGE_SIZE
, &sk
->sk_wmem_alloc
);
1870 /* get the new initialized frag */
1871 frg_cnt
= skb_shinfo(skb
)->nr_frags
;
1872 frag
= &skb_shinfo(skb
)->frags
[frg_cnt
- 1];
1874 /* copy the user data to page */
1875 left
= PAGE_SIZE
- frag
->page_offset
;
1876 copy
= (length
> left
)? left
: length
;
1878 ret
= getfrag(from
, (page_address(frag
->page
) +
1879 frag
->page_offset
+ frag
->size
),
1880 offset
, copy
, 0, skb
);
1884 /* copy was successful so update the size parameters */
1885 sk
->sk_sndmsg_off
+= copy
;
1888 skb
->data_len
+= copy
;
1892 } while (length
> 0);
1898 * skb_pull_rcsum - pull skb and update receive checksum
1899 * @skb: buffer to update
1900 * @start: start of data before pull
1901 * @len: length of data pulled
1903 * This function performs an skb_pull on the packet and updates
1904 * update the CHECKSUM_COMPLETE checksum. It should be used on
1905 * receive path processing instead of skb_pull unless you know
1906 * that the checksum difference is zero (e.g., a valid IP header)
1907 * or you are setting ip_summed to CHECKSUM_NONE.
1909 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
)
1911 BUG_ON(len
> skb
->len
);
1913 BUG_ON(skb
->len
< skb
->data_len
);
1914 skb_postpull_rcsum(skb
, skb
->data
, len
);
1915 return skb
->data
+= len
;
1918 EXPORT_SYMBOL_GPL(skb_pull_rcsum
);
1921 * skb_segment - Perform protocol segmentation on skb.
1922 * @skb: buffer to segment
1923 * @features: features for the output path (see dev->features)
1925 * This function performs segmentation on the given skb. It returns
1926 * the segment at the given position. It returns NULL if there are
1927 * no more segments to generate, or when an error is encountered.
1929 struct sk_buff
*skb_segment(struct sk_buff
*skb
, int features
)
1931 struct sk_buff
*segs
= NULL
;
1932 struct sk_buff
*tail
= NULL
;
1933 unsigned int mss
= skb_shinfo(skb
)->gso_size
;
1934 unsigned int doffset
= skb
->data
- skb
->mac
.raw
;
1935 unsigned int offset
= doffset
;
1936 unsigned int headroom
;
1938 int sg
= features
& NETIF_F_SG
;
1939 int nfrags
= skb_shinfo(skb
)->nr_frags
;
1944 __skb_push(skb
, doffset
);
1945 headroom
= skb_headroom(skb
);
1946 pos
= skb_headlen(skb
);
1949 struct sk_buff
*nskb
;
1955 len
= skb
->len
- offset
;
1959 hsize
= skb_headlen(skb
) - offset
;
1962 if (hsize
> len
|| !sg
)
1965 nskb
= alloc_skb(hsize
+ doffset
+ headroom
, GFP_ATOMIC
);
1966 if (unlikely(!nskb
))
1975 nskb
->dev
= skb
->dev
;
1976 nskb
->priority
= skb
->priority
;
1977 nskb
->protocol
= skb
->protocol
;
1978 nskb
->dst
= dst_clone(skb
->dst
);
1979 memcpy(nskb
->cb
, skb
->cb
, sizeof(skb
->cb
));
1980 nskb
->pkt_type
= skb
->pkt_type
;
1981 nskb
->mac_len
= skb
->mac_len
;
1983 skb_reserve(nskb
, headroom
);
1984 nskb
->mac
.raw
= nskb
->data
;
1985 nskb
->nh
.raw
= nskb
->data
+ skb
->mac_len
;
1986 nskb
->h
.raw
= nskb
->nh
.raw
+ (skb
->h
.raw
- skb
->nh
.raw
);
1987 memcpy(skb_put(nskb
, doffset
), skb
->data
, doffset
);
1990 nskb
->csum
= skb_copy_and_csum_bits(skb
, offset
,
1996 frag
= skb_shinfo(nskb
)->frags
;
1999 nskb
->ip_summed
= CHECKSUM_PARTIAL
;
2000 nskb
->csum
= skb
->csum
;
2001 memcpy(skb_put(nskb
, hsize
), skb
->data
+ offset
, hsize
);
2003 while (pos
< offset
+ len
) {
2004 BUG_ON(i
>= nfrags
);
2006 *frag
= skb_shinfo(skb
)->frags
[i
];
2007 get_page(frag
->page
);
2011 frag
->page_offset
+= offset
- pos
;
2012 frag
->size
-= offset
- pos
;
2017 if (pos
+ size
<= offset
+ len
) {
2021 frag
->size
-= pos
+ size
- (offset
+ len
);
2028 skb_shinfo(nskb
)->nr_frags
= k
;
2029 nskb
->data_len
= len
- hsize
;
2030 nskb
->len
+= nskb
->data_len
;
2031 nskb
->truesize
+= nskb
->data_len
;
2032 } while ((offset
+= len
) < skb
->len
);
2037 while ((skb
= segs
)) {
2041 return ERR_PTR(err
);
2044 EXPORT_SYMBOL_GPL(skb_segment
);
2046 void __init
skb_init(void)
2048 skbuff_head_cache
= kmem_cache_create("skbuff_head_cache",
2049 sizeof(struct sk_buff
),
2051 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2053 skbuff_fclone_cache
= kmem_cache_create("skbuff_fclone_cache",
2054 (2*sizeof(struct sk_buff
)) +
2057 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2061 EXPORT_SYMBOL(___pskb_trim
);
2062 EXPORT_SYMBOL(__kfree_skb
);
2063 EXPORT_SYMBOL(kfree_skb
);
2064 EXPORT_SYMBOL(__pskb_pull_tail
);
2065 EXPORT_SYMBOL(__alloc_skb
);
2066 EXPORT_SYMBOL(__netdev_alloc_skb
);
2067 EXPORT_SYMBOL(pskb_copy
);
2068 EXPORT_SYMBOL(pskb_expand_head
);
2069 EXPORT_SYMBOL(skb_checksum
);
2070 EXPORT_SYMBOL(skb_clone
);
2071 EXPORT_SYMBOL(skb_clone_fraglist
);
2072 EXPORT_SYMBOL(skb_copy
);
2073 EXPORT_SYMBOL(skb_copy_and_csum_bits
);
2074 EXPORT_SYMBOL(skb_copy_and_csum_dev
);
2075 EXPORT_SYMBOL(skb_copy_bits
);
2076 EXPORT_SYMBOL(skb_copy_expand
);
2077 EXPORT_SYMBOL(skb_over_panic
);
2078 EXPORT_SYMBOL(skb_pad
);
2079 EXPORT_SYMBOL(skb_realloc_headroom
);
2080 EXPORT_SYMBOL(skb_under_panic
);
2081 EXPORT_SYMBOL(skb_dequeue
);
2082 EXPORT_SYMBOL(skb_dequeue_tail
);
2083 EXPORT_SYMBOL(skb_insert
);
2084 EXPORT_SYMBOL(skb_queue_purge
);
2085 EXPORT_SYMBOL(skb_queue_head
);
2086 EXPORT_SYMBOL(skb_queue_tail
);
2087 EXPORT_SYMBOL(skb_unlink
);
2088 EXPORT_SYMBOL(skb_append
);
2089 EXPORT_SYMBOL(skb_split
);
2090 EXPORT_SYMBOL(skb_prepare_seq_read
);
2091 EXPORT_SYMBOL(skb_seq_read
);
2092 EXPORT_SYMBOL(skb_abort_seq_read
);
2093 EXPORT_SYMBOL(skb_find_text
);
2094 EXPORT_SYMBOL(skb_append_datato_frags
);