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(size
+ sizeof(struct skb_shared_info
), gfp_mask
);
163 memset(skb
, 0, offsetof(struct sk_buff
, truesize
));
164 skb
->truesize
= size
+ sizeof(struct sk_buff
);
165 atomic_set(&skb
->users
, 1);
169 skb
->end
= data
+ size
;
170 /* make sure we initialize shinfo sequentially */
171 shinfo
= skb_shinfo(skb
);
172 atomic_set(&shinfo
->dataref
, 1);
173 shinfo
->nr_frags
= 0;
174 shinfo
->gso_size
= 0;
175 shinfo
->gso_segs
= 0;
176 shinfo
->gso_type
= 0;
177 shinfo
->ip6_frag_id
= 0;
178 shinfo
->frag_list
= NULL
;
181 struct sk_buff
*child
= skb
+ 1;
182 atomic_t
*fclone_ref
= (atomic_t
*) (child
+ 1);
184 skb
->fclone
= SKB_FCLONE_ORIG
;
185 atomic_set(fclone_ref
, 1);
187 child
->fclone
= SKB_FCLONE_UNAVAILABLE
;
192 kmem_cache_free(cache
, skb
);
198 * alloc_skb_from_cache - allocate a network buffer
199 * @cp: kmem_cache from which to allocate the data area
200 * (object size must be big enough for @size bytes + skb overheads)
201 * @size: size to allocate
202 * @gfp_mask: allocation mask
204 * Allocate a new &sk_buff. The returned buffer has no headroom and
205 * tail room of size bytes. The object has a reference count of one.
206 * The return is the buffer. On a failure the return is %NULL.
208 * Buffers may only be allocated from interrupts using a @gfp_mask of
211 struct sk_buff
*alloc_skb_from_cache(kmem_cache_t
*cp
,
219 skb
= kmem_cache_alloc(skbuff_head_cache
,
220 gfp_mask
& ~__GFP_DMA
);
225 size
= SKB_DATA_ALIGN(size
);
226 data
= kmem_cache_alloc(cp
, gfp_mask
);
230 memset(skb
, 0, offsetof(struct sk_buff
, truesize
));
231 skb
->truesize
= size
+ sizeof(struct sk_buff
);
232 atomic_set(&skb
->users
, 1);
236 skb
->end
= data
+ size
;
238 atomic_set(&(skb_shinfo(skb
)->dataref
), 1);
239 skb_shinfo(skb
)->nr_frags
= 0;
240 skb_shinfo(skb
)->gso_size
= 0;
241 skb_shinfo(skb
)->gso_segs
= 0;
242 skb_shinfo(skb
)->gso_type
= 0;
243 skb_shinfo(skb
)->frag_list
= NULL
;
247 kmem_cache_free(skbuff_head_cache
, skb
);
253 * __netdev_alloc_skb - allocate an skbuff for rx on a specific device
254 * @dev: network device to receive on
255 * @length: length to allocate
256 * @gfp_mask: get_free_pages mask, passed to alloc_skb
258 * Allocate a new &sk_buff and assign it a usage count of one. The
259 * buffer has unspecified headroom built in. Users should allocate
260 * the headroom they think they need without accounting for the
261 * built in space. The built in space is used for optimisations.
263 * %NULL is returned if there is no free memory.
265 struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
266 unsigned int length
, gfp_t gfp_mask
)
270 skb
= alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
);
272 skb_reserve(skb
, NET_SKB_PAD
);
276 static void skb_drop_list(struct sk_buff
**listp
)
278 struct sk_buff
*list
= *listp
;
283 struct sk_buff
*this = list
;
289 static inline void skb_drop_fraglist(struct sk_buff
*skb
)
291 skb_drop_list(&skb_shinfo(skb
)->frag_list
);
294 static void skb_clone_fraglist(struct sk_buff
*skb
)
296 struct sk_buff
*list
;
298 for (list
= skb_shinfo(skb
)->frag_list
; list
; list
= list
->next
)
302 static void skb_release_data(struct sk_buff
*skb
)
305 !atomic_sub_return(skb
->nohdr
? (1 << SKB_DATAREF_SHIFT
) + 1 : 1,
306 &skb_shinfo(skb
)->dataref
)) {
307 if (skb_shinfo(skb
)->nr_frags
) {
309 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++)
310 put_page(skb_shinfo(skb
)->frags
[i
].page
);
313 if (skb_shinfo(skb
)->frag_list
)
314 skb_drop_fraglist(skb
);
321 * Free an skbuff by memory without cleaning the state.
323 void kfree_skbmem(struct sk_buff
*skb
)
325 struct sk_buff
*other
;
326 atomic_t
*fclone_ref
;
328 skb_release_data(skb
);
329 switch (skb
->fclone
) {
330 case SKB_FCLONE_UNAVAILABLE
:
331 kmem_cache_free(skbuff_head_cache
, skb
);
334 case SKB_FCLONE_ORIG
:
335 fclone_ref
= (atomic_t
*) (skb
+ 2);
336 if (atomic_dec_and_test(fclone_ref
))
337 kmem_cache_free(skbuff_fclone_cache
, skb
);
340 case SKB_FCLONE_CLONE
:
341 fclone_ref
= (atomic_t
*) (skb
+ 1);
344 /* The clone portion is available for
345 * fast-cloning again.
347 skb
->fclone
= SKB_FCLONE_UNAVAILABLE
;
349 if (atomic_dec_and_test(fclone_ref
))
350 kmem_cache_free(skbuff_fclone_cache
, other
);
356 * __kfree_skb - private function
359 * Free an sk_buff. Release anything attached to the buffer.
360 * Clean the state. This is an internal helper function. Users should
361 * always call kfree_skb
364 void __kfree_skb(struct sk_buff
*skb
)
366 dst_release(skb
->dst
);
368 secpath_put(skb
->sp
);
370 if (skb
->destructor
) {
372 skb
->destructor(skb
);
374 #ifdef CONFIG_NETFILTER
375 nf_conntrack_put(skb
->nfct
);
376 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
377 nf_conntrack_put_reasm(skb
->nfct_reasm
);
379 #ifdef CONFIG_BRIDGE_NETFILTER
380 nf_bridge_put(skb
->nf_bridge
);
383 /* XXX: IS this still necessary? - JHS */
384 #ifdef CONFIG_NET_SCHED
386 #ifdef CONFIG_NET_CLS_ACT
395 * kfree_skb - free an sk_buff
396 * @skb: buffer to free
398 * Drop a reference to the buffer and free it if the usage count has
401 void kfree_skb(struct sk_buff
*skb
)
405 if (likely(atomic_read(&skb
->users
) == 1))
407 else if (likely(!atomic_dec_and_test(&skb
->users
)))
413 * skb_clone - duplicate an sk_buff
414 * @skb: buffer to clone
415 * @gfp_mask: allocation priority
417 * Duplicate an &sk_buff. The new one is not owned by a socket. Both
418 * copies share the same packet data but not structure. The new
419 * buffer has a reference count of 1. If the allocation fails the
420 * function returns %NULL otherwise the new buffer is returned.
422 * If this function is called from an interrupt gfp_mask() must be
426 struct sk_buff
*skb_clone(struct sk_buff
*skb
, gfp_t gfp_mask
)
431 if (skb
->fclone
== SKB_FCLONE_ORIG
&&
432 n
->fclone
== SKB_FCLONE_UNAVAILABLE
) {
433 atomic_t
*fclone_ref
= (atomic_t
*) (n
+ 1);
434 n
->fclone
= SKB_FCLONE_CLONE
;
435 atomic_inc(fclone_ref
);
437 n
= kmem_cache_alloc(skbuff_head_cache
, gfp_mask
);
440 n
->fclone
= SKB_FCLONE_UNAVAILABLE
;
443 #define C(x) n->x = skb->x
445 n
->next
= n
->prev
= NULL
;
456 secpath_get(skb
->sp
);
458 memcpy(n
->cb
, skb
->cb
, sizeof(skb
->cb
));
468 #if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
472 n
->destructor
= NULL
;
473 #ifdef CONFIG_NETFILTER
476 nf_conntrack_get(skb
->nfct
);
478 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
480 nf_conntrack_get_reasm(skb
->nfct_reasm
);
482 #ifdef CONFIG_BRIDGE_NETFILTER
484 nf_bridge_get(skb
->nf_bridge
);
486 #endif /*CONFIG_NETFILTER*/
487 #ifdef CONFIG_NET_SCHED
489 #ifdef CONFIG_NET_CLS_ACT
490 n
->tc_verd
= SET_TC_VERD(skb
->tc_verd
,0);
491 n
->tc_verd
= CLR_TC_OK2MUNGE(n
->tc_verd
);
492 n
->tc_verd
= CLR_TC_MUNGED(n
->tc_verd
);
495 skb_copy_secmark(n
, skb
);
498 atomic_set(&n
->users
, 1);
504 atomic_inc(&(skb_shinfo(skb
)->dataref
));
510 static void copy_skb_header(struct sk_buff
*new, const struct sk_buff
*old
)
513 * Shift between the two data areas in bytes
515 unsigned long offset
= new->data
- old
->data
;
519 new->priority
= old
->priority
;
520 new->protocol
= old
->protocol
;
521 new->dst
= dst_clone(old
->dst
);
523 new->sp
= secpath_get(old
->sp
);
525 new->h
.raw
= old
->h
.raw
+ offset
;
526 new->nh
.raw
= old
->nh
.raw
+ offset
;
527 new->mac
.raw
= old
->mac
.raw
+ offset
;
528 memcpy(new->cb
, old
->cb
, sizeof(old
->cb
));
529 new->local_df
= old
->local_df
;
530 new->fclone
= SKB_FCLONE_UNAVAILABLE
;
531 new->pkt_type
= old
->pkt_type
;
532 new->tstamp
= old
->tstamp
;
533 new->destructor
= NULL
;
534 #ifdef CONFIG_NETFILTER
535 new->nfmark
= old
->nfmark
;
536 new->nfct
= old
->nfct
;
537 nf_conntrack_get(old
->nfct
);
538 new->nfctinfo
= old
->nfctinfo
;
539 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
540 new->nfct_reasm
= old
->nfct_reasm
;
541 nf_conntrack_get_reasm(old
->nfct_reasm
);
543 #if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
544 new->ipvs_property
= old
->ipvs_property
;
546 #ifdef CONFIG_BRIDGE_NETFILTER
547 new->nf_bridge
= old
->nf_bridge
;
548 nf_bridge_get(old
->nf_bridge
);
551 #ifdef CONFIG_NET_SCHED
552 #ifdef CONFIG_NET_CLS_ACT
553 new->tc_verd
= old
->tc_verd
;
555 new->tc_index
= old
->tc_index
;
557 skb_copy_secmark(new, old
);
558 atomic_set(&new->users
, 1);
559 skb_shinfo(new)->gso_size
= skb_shinfo(old
)->gso_size
;
560 skb_shinfo(new)->gso_segs
= skb_shinfo(old
)->gso_segs
;
561 skb_shinfo(new)->gso_type
= skb_shinfo(old
)->gso_type
;
565 * skb_copy - create private copy of an sk_buff
566 * @skb: buffer to copy
567 * @gfp_mask: allocation priority
569 * Make a copy of both an &sk_buff and its data. This is used when the
570 * caller wishes to modify the data and needs a private copy of the
571 * data to alter. Returns %NULL on failure or the pointer to the buffer
572 * on success. The returned buffer has a reference count of 1.
574 * As by-product this function converts non-linear &sk_buff to linear
575 * one, so that &sk_buff becomes completely private and caller is allowed
576 * to modify all the data of returned buffer. This means that this
577 * function is not recommended for use in circumstances when only
578 * header is going to be modified. Use pskb_copy() instead.
581 struct sk_buff
*skb_copy(const struct sk_buff
*skb
, gfp_t gfp_mask
)
583 int headerlen
= skb
->data
- skb
->head
;
585 * Allocate the copy buffer
587 struct sk_buff
*n
= alloc_skb(skb
->end
- skb
->head
+ skb
->data_len
,
592 /* Set the data pointer */
593 skb_reserve(n
, headerlen
);
594 /* Set the tail pointer and length */
595 skb_put(n
, skb
->len
);
597 n
->ip_summed
= skb
->ip_summed
;
599 if (skb_copy_bits(skb
, -headerlen
, n
->head
, headerlen
+ skb
->len
))
602 copy_skb_header(n
, skb
);
608 * pskb_copy - create copy of an sk_buff with private head.
609 * @skb: buffer to copy
610 * @gfp_mask: allocation priority
612 * Make a copy of both an &sk_buff and part of its data, located
613 * in header. Fragmented data remain shared. This is used when
614 * the caller wishes to modify only header of &sk_buff and needs
615 * private copy of the header to alter. Returns %NULL on failure
616 * or the pointer to the buffer on success.
617 * The returned buffer has a reference count of 1.
620 struct sk_buff
*pskb_copy(struct sk_buff
*skb
, gfp_t gfp_mask
)
623 * Allocate the copy buffer
625 struct sk_buff
*n
= alloc_skb(skb
->end
- skb
->head
, gfp_mask
);
630 /* Set the data pointer */
631 skb_reserve(n
, skb
->data
- skb
->head
);
632 /* Set the tail pointer and length */
633 skb_put(n
, skb_headlen(skb
));
635 memcpy(n
->data
, skb
->data
, n
->len
);
637 n
->ip_summed
= skb
->ip_summed
;
639 n
->data_len
= skb
->data_len
;
642 if (skb_shinfo(skb
)->nr_frags
) {
645 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
646 skb_shinfo(n
)->frags
[i
] = skb_shinfo(skb
)->frags
[i
];
647 get_page(skb_shinfo(n
)->frags
[i
].page
);
649 skb_shinfo(n
)->nr_frags
= i
;
652 if (skb_shinfo(skb
)->frag_list
) {
653 skb_shinfo(n
)->frag_list
= skb_shinfo(skb
)->frag_list
;
654 skb_clone_fraglist(n
);
657 copy_skb_header(n
, skb
);
663 * pskb_expand_head - reallocate header of &sk_buff
664 * @skb: buffer to reallocate
665 * @nhead: room to add at head
666 * @ntail: room to add at tail
667 * @gfp_mask: allocation priority
669 * Expands (or creates identical copy, if &nhead and &ntail are zero)
670 * header of skb. &sk_buff itself is not changed. &sk_buff MUST have
671 * reference count of 1. Returns zero in the case of success or error,
672 * if expansion failed. In the last case, &sk_buff is not changed.
674 * All the pointers pointing into skb header may change and must be
675 * reloaded after call to this function.
678 int pskb_expand_head(struct sk_buff
*skb
, int nhead
, int ntail
,
683 int size
= nhead
+ (skb
->end
- skb
->head
) + ntail
;
689 size
= SKB_DATA_ALIGN(size
);
691 data
= kmalloc(size
+ sizeof(struct skb_shared_info
), gfp_mask
);
695 /* Copy only real data... and, alas, header. This should be
696 * optimized for the cases when header is void. */
697 memcpy(data
+ nhead
, skb
->head
, skb
->tail
- skb
->head
);
698 memcpy(data
+ size
, skb
->end
, sizeof(struct skb_shared_info
));
700 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++)
701 get_page(skb_shinfo(skb
)->frags
[i
].page
);
703 if (skb_shinfo(skb
)->frag_list
)
704 skb_clone_fraglist(skb
);
706 skb_release_data(skb
);
708 off
= (data
+ nhead
) - skb
->head
;
711 skb
->end
= data
+ size
;
719 atomic_set(&skb_shinfo(skb
)->dataref
, 1);
726 /* Make private copy of skb with writable head and some headroom */
728 struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
, unsigned int headroom
)
730 struct sk_buff
*skb2
;
731 int delta
= headroom
- skb_headroom(skb
);
734 skb2
= pskb_copy(skb
, GFP_ATOMIC
);
736 skb2
= skb_clone(skb
, GFP_ATOMIC
);
737 if (skb2
&& pskb_expand_head(skb2
, SKB_DATA_ALIGN(delta
), 0,
748 * skb_copy_expand - copy and expand sk_buff
749 * @skb: buffer to copy
750 * @newheadroom: new free bytes at head
751 * @newtailroom: new free bytes at tail
752 * @gfp_mask: allocation priority
754 * Make a copy of both an &sk_buff and its data and while doing so
755 * allocate additional space.
757 * This is used when the caller wishes to modify the data and needs a
758 * private copy of the data to alter as well as more space for new fields.
759 * Returns %NULL on failure or the pointer to the buffer
760 * on success. The returned buffer has a reference count of 1.
762 * You must pass %GFP_ATOMIC as the allocation priority if this function
763 * is called from an interrupt.
765 * BUG ALERT: ip_summed is not copied. Why does this work? Is it used
766 * only by netfilter in the cases when checksum is recalculated? --ANK
768 struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
769 int newheadroom
, int newtailroom
,
773 * Allocate the copy buffer
775 struct sk_buff
*n
= alloc_skb(newheadroom
+ skb
->len
+ newtailroom
,
777 int head_copy_len
, head_copy_off
;
782 skb_reserve(n
, newheadroom
);
784 /* Set the tail pointer and length */
785 skb_put(n
, skb
->len
);
787 head_copy_len
= skb_headroom(skb
);
789 if (newheadroom
<= head_copy_len
)
790 head_copy_len
= newheadroom
;
792 head_copy_off
= newheadroom
- head_copy_len
;
794 /* Copy the linear header and data. */
795 if (skb_copy_bits(skb
, -head_copy_len
, n
->head
+ head_copy_off
,
796 skb
->len
+ head_copy_len
))
799 copy_skb_header(n
, skb
);
805 * skb_pad - zero pad the tail of an skb
806 * @skb: buffer to pad
809 * Ensure that a buffer is followed by a padding area that is zero
810 * filled. Used by network drivers which may DMA or transfer data
811 * beyond the buffer end onto the wire.
813 * May return error in out of memory cases. The skb is freed on error.
816 int skb_pad(struct sk_buff
*skb
, int pad
)
821 /* If the skbuff is non linear tailroom is always zero.. */
822 if (!skb_cloned(skb
) && skb_tailroom(skb
) >= pad
) {
823 memset(skb
->data
+skb
->len
, 0, pad
);
827 ntail
= skb
->data_len
+ pad
- (skb
->end
- skb
->tail
);
828 if (likely(skb_cloned(skb
) || ntail
> 0)) {
829 err
= pskb_expand_head(skb
, 0, ntail
, GFP_ATOMIC
);
834 /* FIXME: The use of this function with non-linear skb's really needs
837 err
= skb_linearize(skb
);
841 memset(skb
->data
+ skb
->len
, 0, pad
);
849 /* Trims skb to length len. It can change skb pointers.
852 int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
)
854 struct sk_buff
**fragp
;
855 struct sk_buff
*frag
;
856 int offset
= skb_headlen(skb
);
857 int nfrags
= skb_shinfo(skb
)->nr_frags
;
861 if (skb_cloned(skb
) &&
862 unlikely((err
= pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
))))
869 for (; i
< nfrags
; i
++) {
870 int end
= offset
+ skb_shinfo(skb
)->frags
[i
].size
;
877 skb_shinfo(skb
)->frags
[i
++].size
= len
- offset
;
880 skb_shinfo(skb
)->nr_frags
= i
;
882 for (; i
< nfrags
; i
++)
883 put_page(skb_shinfo(skb
)->frags
[i
].page
);
885 if (skb_shinfo(skb
)->frag_list
)
886 skb_drop_fraglist(skb
);
890 for (fragp
= &skb_shinfo(skb
)->frag_list
; (frag
= *fragp
);
891 fragp
= &frag
->next
) {
892 int end
= offset
+ frag
->len
;
894 if (skb_shared(frag
)) {
895 struct sk_buff
*nfrag
;
897 nfrag
= skb_clone(frag
, GFP_ATOMIC
);
898 if (unlikely(!nfrag
))
901 nfrag
->next
= frag
->next
;
913 unlikely((err
= pskb_trim(frag
, len
- offset
))))
917 skb_drop_list(&frag
->next
);
922 if (len
> skb_headlen(skb
)) {
923 skb
->data_len
-= skb
->len
- len
;
928 skb
->tail
= skb
->data
+ len
;
935 * __pskb_pull_tail - advance tail of skb header
936 * @skb: buffer to reallocate
937 * @delta: number of bytes to advance tail
939 * The function makes a sense only on a fragmented &sk_buff,
940 * it expands header moving its tail forward and copying necessary
941 * data from fragmented part.
943 * &sk_buff MUST have reference count of 1.
945 * Returns %NULL (and &sk_buff does not change) if pull failed
946 * or value of new tail of skb in the case of success.
948 * All the pointers pointing into skb header may change and must be
949 * reloaded after call to this function.
952 /* Moves tail of skb head forward, copying data from fragmented part,
953 * when it is necessary.
954 * 1. It may fail due to malloc failure.
955 * 2. It may change skb pointers.
957 * It is pretty complicated. Luckily, it is called only in exceptional cases.
959 unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
)
961 /* If skb has not enough free space at tail, get new one
962 * plus 128 bytes for future expansions. If we have enough
963 * room at tail, reallocate without expansion only if skb is cloned.
965 int i
, k
, eat
= (skb
->tail
+ delta
) - skb
->end
;
967 if (eat
> 0 || skb_cloned(skb
)) {
968 if (pskb_expand_head(skb
, 0, eat
> 0 ? eat
+ 128 : 0,
973 if (skb_copy_bits(skb
, skb_headlen(skb
), skb
->tail
, delta
))
976 /* Optimization: no fragments, no reasons to preestimate
977 * size of pulled pages. Superb.
979 if (!skb_shinfo(skb
)->frag_list
)
982 /* Estimate size of pulled pages. */
984 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
985 if (skb_shinfo(skb
)->frags
[i
].size
>= eat
)
987 eat
-= skb_shinfo(skb
)->frags
[i
].size
;
990 /* If we need update frag list, we are in troubles.
991 * Certainly, it possible to add an offset to skb data,
992 * but taking into account that pulling is expected to
993 * be very rare operation, it is worth to fight against
994 * further bloating skb head and crucify ourselves here instead.
995 * Pure masohism, indeed. 8)8)
998 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
999 struct sk_buff
*clone
= NULL
;
1000 struct sk_buff
*insp
= NULL
;
1005 if (list
->len
<= eat
) {
1006 /* Eaten as whole. */
1011 /* Eaten partially. */
1013 if (skb_shared(list
)) {
1014 /* Sucks! We need to fork list. :-( */
1015 clone
= skb_clone(list
, GFP_ATOMIC
);
1021 /* This may be pulled without
1025 if (!pskb_pull(list
, eat
)) {
1034 /* Free pulled out fragments. */
1035 while ((list
= skb_shinfo(skb
)->frag_list
) != insp
) {
1036 skb_shinfo(skb
)->frag_list
= list
->next
;
1039 /* And insert new clone at head. */
1042 skb_shinfo(skb
)->frag_list
= clone
;
1045 /* Success! Now we may commit changes to skb data. */
1050 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1051 if (skb_shinfo(skb
)->frags
[i
].size
<= eat
) {
1052 put_page(skb_shinfo(skb
)->frags
[i
].page
);
1053 eat
-= skb_shinfo(skb
)->frags
[i
].size
;
1055 skb_shinfo(skb
)->frags
[k
] = skb_shinfo(skb
)->frags
[i
];
1057 skb_shinfo(skb
)->frags
[k
].page_offset
+= eat
;
1058 skb_shinfo(skb
)->frags
[k
].size
-= eat
;
1064 skb_shinfo(skb
)->nr_frags
= k
;
1067 skb
->data_len
-= delta
;
1072 /* Copy some data bits from skb to kernel buffer. */
1074 int skb_copy_bits(const struct sk_buff
*skb
, int offset
, void *to
, int len
)
1077 int start
= skb_headlen(skb
);
1079 if (offset
> (int)skb
->len
- len
)
1083 if ((copy
= start
- offset
) > 0) {
1086 memcpy(to
, skb
->data
+ offset
, copy
);
1087 if ((len
-= copy
) == 0)
1093 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1096 BUG_TRAP(start
<= offset
+ len
);
1098 end
= start
+ skb_shinfo(skb
)->frags
[i
].size
;
1099 if ((copy
= end
- offset
) > 0) {
1105 vaddr
= kmap_skb_frag(&skb_shinfo(skb
)->frags
[i
]);
1107 vaddr
+ skb_shinfo(skb
)->frags
[i
].page_offset
+
1108 offset
- start
, copy
);
1109 kunmap_skb_frag(vaddr
);
1111 if ((len
-= copy
) == 0)
1119 if (skb_shinfo(skb
)->frag_list
) {
1120 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1122 for (; list
; list
= list
->next
) {
1125 BUG_TRAP(start
<= offset
+ len
);
1127 end
= start
+ list
->len
;
1128 if ((copy
= end
- offset
) > 0) {
1131 if (skb_copy_bits(list
, offset
- start
,
1134 if ((len
-= copy
) == 0)
1150 * skb_store_bits - store bits from kernel buffer to skb
1151 * @skb: destination buffer
1152 * @offset: offset in destination
1153 * @from: source buffer
1154 * @len: number of bytes to copy
1156 * Copy the specified number of bytes from the source buffer to the
1157 * destination skb. This function handles all the messy bits of
1158 * traversing fragment lists and such.
1161 int skb_store_bits(const struct sk_buff
*skb
, int offset
, void *from
, int len
)
1164 int start
= skb_headlen(skb
);
1166 if (offset
> (int)skb
->len
- len
)
1169 if ((copy
= start
- offset
) > 0) {
1172 memcpy(skb
->data
+ offset
, from
, copy
);
1173 if ((len
-= copy
) == 0)
1179 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1180 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1183 BUG_TRAP(start
<= offset
+ len
);
1185 end
= start
+ frag
->size
;
1186 if ((copy
= end
- offset
) > 0) {
1192 vaddr
= kmap_skb_frag(frag
);
1193 memcpy(vaddr
+ frag
->page_offset
+ offset
- start
,
1195 kunmap_skb_frag(vaddr
);
1197 if ((len
-= copy
) == 0)
1205 if (skb_shinfo(skb
)->frag_list
) {
1206 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1208 for (; list
; list
= list
->next
) {
1211 BUG_TRAP(start
<= offset
+ len
);
1213 end
= start
+ list
->len
;
1214 if ((copy
= end
- offset
) > 0) {
1217 if (skb_store_bits(list
, offset
- start
,
1220 if ((len
-= copy
) == 0)
1235 EXPORT_SYMBOL(skb_store_bits
);
1237 /* Checksum skb data. */
1239 unsigned int skb_checksum(const struct sk_buff
*skb
, int offset
,
1240 int len
, unsigned int csum
)
1242 int start
= skb_headlen(skb
);
1243 int i
, copy
= start
- offset
;
1246 /* Checksum header. */
1250 csum
= csum_partial(skb
->data
+ offset
, copy
, csum
);
1251 if ((len
-= copy
) == 0)
1257 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1260 BUG_TRAP(start
<= offset
+ len
);
1262 end
= start
+ skb_shinfo(skb
)->frags
[i
].size
;
1263 if ((copy
= end
- offset
) > 0) {
1266 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1270 vaddr
= kmap_skb_frag(frag
);
1271 csum2
= csum_partial(vaddr
+ frag
->page_offset
+
1272 offset
- start
, copy
, 0);
1273 kunmap_skb_frag(vaddr
);
1274 csum
= csum_block_add(csum
, csum2
, pos
);
1283 if (skb_shinfo(skb
)->frag_list
) {
1284 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1286 for (; list
; list
= list
->next
) {
1289 BUG_TRAP(start
<= offset
+ len
);
1291 end
= start
+ list
->len
;
1292 if ((copy
= end
- offset
) > 0) {
1296 csum2
= skb_checksum(list
, offset
- start
,
1298 csum
= csum_block_add(csum
, csum2
, pos
);
1299 if ((len
-= copy
) == 0)
1312 /* Both of above in one bottle. */
1314 unsigned int skb_copy_and_csum_bits(const struct sk_buff
*skb
, int offset
,
1315 u8
*to
, int len
, unsigned int csum
)
1317 int start
= skb_headlen(skb
);
1318 int i
, copy
= start
- offset
;
1325 csum
= csum_partial_copy_nocheck(skb
->data
+ offset
, to
,
1327 if ((len
-= copy
) == 0)
1334 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1337 BUG_TRAP(start
<= offset
+ len
);
1339 end
= start
+ skb_shinfo(skb
)->frags
[i
].size
;
1340 if ((copy
= end
- offset
) > 0) {
1343 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1347 vaddr
= kmap_skb_frag(frag
);
1348 csum2
= csum_partial_copy_nocheck(vaddr
+
1352 kunmap_skb_frag(vaddr
);
1353 csum
= csum_block_add(csum
, csum2
, pos
);
1363 if (skb_shinfo(skb
)->frag_list
) {
1364 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1366 for (; list
; list
= list
->next
) {
1370 BUG_TRAP(start
<= offset
+ len
);
1372 end
= start
+ list
->len
;
1373 if ((copy
= end
- offset
) > 0) {
1376 csum2
= skb_copy_and_csum_bits(list
,
1379 csum
= csum_block_add(csum
, csum2
, pos
);
1380 if ((len
-= copy
) == 0)
1393 void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
)
1398 if (skb
->ip_summed
== CHECKSUM_HW
)
1399 csstart
= skb
->h
.raw
- skb
->data
;
1401 csstart
= skb_headlen(skb
);
1403 BUG_ON(csstart
> skb_headlen(skb
));
1405 memcpy(to
, skb
->data
, csstart
);
1408 if (csstart
!= skb
->len
)
1409 csum
= skb_copy_and_csum_bits(skb
, csstart
, to
+ csstart
,
1410 skb
->len
- csstart
, 0);
1412 if (skb
->ip_summed
== CHECKSUM_HW
) {
1413 long csstuff
= csstart
+ skb
->csum
;
1415 *((unsigned short *)(to
+ csstuff
)) = csum_fold(csum
);
1420 * skb_dequeue - remove from the head of the queue
1421 * @list: list to dequeue from
1423 * Remove the head of the list. The list lock is taken so the function
1424 * may be used safely with other locking list functions. The head item is
1425 * returned or %NULL if the list is empty.
1428 struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
)
1430 unsigned long flags
;
1431 struct sk_buff
*result
;
1433 spin_lock_irqsave(&list
->lock
, flags
);
1434 result
= __skb_dequeue(list
);
1435 spin_unlock_irqrestore(&list
->lock
, flags
);
1440 * skb_dequeue_tail - remove from the tail of the queue
1441 * @list: list to dequeue from
1443 * Remove the tail of the list. The list lock is taken so the function
1444 * may be used safely with other locking list functions. The tail item is
1445 * returned or %NULL if the list is empty.
1447 struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
)
1449 unsigned long flags
;
1450 struct sk_buff
*result
;
1452 spin_lock_irqsave(&list
->lock
, flags
);
1453 result
= __skb_dequeue_tail(list
);
1454 spin_unlock_irqrestore(&list
->lock
, flags
);
1459 * skb_queue_purge - empty a list
1460 * @list: list to empty
1462 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1463 * the list and one reference dropped. This function takes the list
1464 * lock and is atomic with respect to other list locking functions.
1466 void skb_queue_purge(struct sk_buff_head
*list
)
1468 struct sk_buff
*skb
;
1469 while ((skb
= skb_dequeue(list
)) != NULL
)
1474 * skb_queue_head - queue a buffer at the list head
1475 * @list: list to use
1476 * @newsk: buffer to queue
1478 * Queue a buffer at the start of the list. This function takes the
1479 * list lock and can be used safely with other locking &sk_buff functions
1482 * A buffer cannot be placed on two lists at the same time.
1484 void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
)
1486 unsigned long flags
;
1488 spin_lock_irqsave(&list
->lock
, flags
);
1489 __skb_queue_head(list
, newsk
);
1490 spin_unlock_irqrestore(&list
->lock
, flags
);
1494 * skb_queue_tail - queue a buffer at the list tail
1495 * @list: list to use
1496 * @newsk: buffer to queue
1498 * Queue a buffer at the tail of the list. This function takes the
1499 * list lock and can be used safely with other locking &sk_buff functions
1502 * A buffer cannot be placed on two lists at the same time.
1504 void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
)
1506 unsigned long flags
;
1508 spin_lock_irqsave(&list
->lock
, flags
);
1509 __skb_queue_tail(list
, newsk
);
1510 spin_unlock_irqrestore(&list
->lock
, flags
);
1514 * skb_unlink - remove a buffer from a list
1515 * @skb: buffer to remove
1516 * @list: list to use
1518 * Remove a packet from a list. The list locks are taken and this
1519 * function is atomic with respect to other list locked calls
1521 * You must know what list the SKB is on.
1523 void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1525 unsigned long flags
;
1527 spin_lock_irqsave(&list
->lock
, flags
);
1528 __skb_unlink(skb
, list
);
1529 spin_unlock_irqrestore(&list
->lock
, flags
);
1533 * skb_append - append a buffer
1534 * @old: buffer to insert after
1535 * @newsk: buffer to insert
1536 * @list: list to use
1538 * Place a packet after a given packet in a list. The list locks are taken
1539 * and this function is atomic with respect to other list locked calls.
1540 * A buffer cannot be placed on two lists at the same time.
1542 void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
1544 unsigned long flags
;
1546 spin_lock_irqsave(&list
->lock
, flags
);
1547 __skb_append(old
, newsk
, list
);
1548 spin_unlock_irqrestore(&list
->lock
, flags
);
1553 * skb_insert - insert a buffer
1554 * @old: buffer to insert before
1555 * @newsk: buffer to insert
1556 * @list: list to use
1558 * Place a packet before a given packet in a list. The list locks are
1559 * taken and this function is atomic with respect to other list locked
1562 * A buffer cannot be placed on two lists at the same time.
1564 void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
1566 unsigned long flags
;
1568 spin_lock_irqsave(&list
->lock
, flags
);
1569 __skb_insert(newsk
, old
->prev
, old
, list
);
1570 spin_unlock_irqrestore(&list
->lock
, flags
);
1575 * Tune the memory allocator for a new MTU size.
1577 void skb_add_mtu(int mtu
)
1579 /* Must match allocation in alloc_skb */
1580 mtu
= SKB_DATA_ALIGN(mtu
) + sizeof(struct skb_shared_info
);
1582 kmem_add_cache_size(mtu
);
1586 static inline void skb_split_inside_header(struct sk_buff
*skb
,
1587 struct sk_buff
* skb1
,
1588 const u32 len
, const int pos
)
1592 memcpy(skb_put(skb1
, pos
- len
), skb
->data
+ len
, pos
- len
);
1594 /* And move data appendix as is. */
1595 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++)
1596 skb_shinfo(skb1
)->frags
[i
] = skb_shinfo(skb
)->frags
[i
];
1598 skb_shinfo(skb1
)->nr_frags
= skb_shinfo(skb
)->nr_frags
;
1599 skb_shinfo(skb
)->nr_frags
= 0;
1600 skb1
->data_len
= skb
->data_len
;
1601 skb1
->len
+= skb1
->data_len
;
1604 skb
->tail
= skb
->data
+ len
;
1607 static inline void skb_split_no_header(struct sk_buff
*skb
,
1608 struct sk_buff
* skb1
,
1609 const u32 len
, int pos
)
1612 const int nfrags
= skb_shinfo(skb
)->nr_frags
;
1614 skb_shinfo(skb
)->nr_frags
= 0;
1615 skb1
->len
= skb1
->data_len
= skb
->len
- len
;
1617 skb
->data_len
= len
- pos
;
1619 for (i
= 0; i
< nfrags
; i
++) {
1620 int size
= skb_shinfo(skb
)->frags
[i
].size
;
1622 if (pos
+ size
> len
) {
1623 skb_shinfo(skb1
)->frags
[k
] = skb_shinfo(skb
)->frags
[i
];
1627 * We have two variants in this case:
1628 * 1. Move all the frag to the second
1629 * part, if it is possible. F.e.
1630 * this approach is mandatory for TUX,
1631 * where splitting is expensive.
1632 * 2. Split is accurately. We make this.
1634 get_page(skb_shinfo(skb
)->frags
[i
].page
);
1635 skb_shinfo(skb1
)->frags
[0].page_offset
+= len
- pos
;
1636 skb_shinfo(skb1
)->frags
[0].size
-= len
- pos
;
1637 skb_shinfo(skb
)->frags
[i
].size
= len
- pos
;
1638 skb_shinfo(skb
)->nr_frags
++;
1642 skb_shinfo(skb
)->nr_frags
++;
1645 skb_shinfo(skb1
)->nr_frags
= k
;
1649 * skb_split - Split fragmented skb to two parts at length len.
1650 * @skb: the buffer to split
1651 * @skb1: the buffer to receive the second part
1652 * @len: new length for skb
1654 void skb_split(struct sk_buff
*skb
, struct sk_buff
*skb1
, const u32 len
)
1656 int pos
= skb_headlen(skb
);
1658 if (len
< pos
) /* Split line is inside header. */
1659 skb_split_inside_header(skb
, skb1
, len
, pos
);
1660 else /* Second chunk has no header, nothing to copy. */
1661 skb_split_no_header(skb
, skb1
, len
, pos
);
1665 * skb_prepare_seq_read - Prepare a sequential read of skb data
1666 * @skb: the buffer to read
1667 * @from: lower offset of data to be read
1668 * @to: upper offset of data to be read
1669 * @st: state variable
1671 * Initializes the specified state variable. Must be called before
1672 * invoking skb_seq_read() for the first time.
1674 void skb_prepare_seq_read(struct sk_buff
*skb
, unsigned int from
,
1675 unsigned int to
, struct skb_seq_state
*st
)
1677 st
->lower_offset
= from
;
1678 st
->upper_offset
= to
;
1679 st
->root_skb
= st
->cur_skb
= skb
;
1680 st
->frag_idx
= st
->stepped_offset
= 0;
1681 st
->frag_data
= NULL
;
1685 * skb_seq_read - Sequentially read skb data
1686 * @consumed: number of bytes consumed by the caller so far
1687 * @data: destination pointer for data to be returned
1688 * @st: state variable
1690 * Reads a block of skb data at &consumed relative to the
1691 * lower offset specified to skb_prepare_seq_read(). Assigns
1692 * the head of the data block to &data and returns the length
1693 * of the block or 0 if the end of the skb data or the upper
1694 * offset has been reached.
1696 * The caller is not required to consume all of the data
1697 * returned, i.e. &consumed is typically set to the number
1698 * of bytes already consumed and the next call to
1699 * skb_seq_read() will return the remaining part of the block.
1701 * Note: The size of each block of data returned can be arbitary,
1702 * this limitation is the cost for zerocopy seqeuental
1703 * reads of potentially non linear data.
1705 * Note: Fragment lists within fragments are not implemented
1706 * at the moment, state->root_skb could be replaced with
1707 * a stack for this purpose.
1709 unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
1710 struct skb_seq_state
*st
)
1712 unsigned int block_limit
, abs_offset
= consumed
+ st
->lower_offset
;
1715 if (unlikely(abs_offset
>= st
->upper_offset
))
1719 block_limit
= skb_headlen(st
->cur_skb
);
1721 if (abs_offset
< block_limit
) {
1722 *data
= st
->cur_skb
->data
+ abs_offset
;
1723 return block_limit
- abs_offset
;
1726 if (st
->frag_idx
== 0 && !st
->frag_data
)
1727 st
->stepped_offset
+= skb_headlen(st
->cur_skb
);
1729 while (st
->frag_idx
< skb_shinfo(st
->cur_skb
)->nr_frags
) {
1730 frag
= &skb_shinfo(st
->cur_skb
)->frags
[st
->frag_idx
];
1731 block_limit
= frag
->size
+ st
->stepped_offset
;
1733 if (abs_offset
< block_limit
) {
1735 st
->frag_data
= kmap_skb_frag(frag
);
1737 *data
= (u8
*) st
->frag_data
+ frag
->page_offset
+
1738 (abs_offset
- st
->stepped_offset
);
1740 return block_limit
- abs_offset
;
1743 if (st
->frag_data
) {
1744 kunmap_skb_frag(st
->frag_data
);
1745 st
->frag_data
= NULL
;
1749 st
->stepped_offset
+= frag
->size
;
1752 if (st
->cur_skb
->next
) {
1753 st
->cur_skb
= st
->cur_skb
->next
;
1756 } else if (st
->root_skb
== st
->cur_skb
&&
1757 skb_shinfo(st
->root_skb
)->frag_list
) {
1758 st
->cur_skb
= skb_shinfo(st
->root_skb
)->frag_list
;
1766 * skb_abort_seq_read - Abort a sequential read of skb data
1767 * @st: state variable
1769 * Must be called if skb_seq_read() was not called until it
1772 void skb_abort_seq_read(struct skb_seq_state
*st
)
1775 kunmap_skb_frag(st
->frag_data
);
1778 #define TS_SKB_CB(state) ((struct skb_seq_state *) &((state)->cb))
1780 static unsigned int skb_ts_get_next_block(unsigned int offset
, const u8
**text
,
1781 struct ts_config
*conf
,
1782 struct ts_state
*state
)
1784 return skb_seq_read(offset
, text
, TS_SKB_CB(state
));
1787 static void skb_ts_finish(struct ts_config
*conf
, struct ts_state
*state
)
1789 skb_abort_seq_read(TS_SKB_CB(state
));
1793 * skb_find_text - Find a text pattern in skb data
1794 * @skb: the buffer to look in
1795 * @from: search offset
1797 * @config: textsearch configuration
1798 * @state: uninitialized textsearch state variable
1800 * Finds a pattern in the skb data according to the specified
1801 * textsearch configuration. Use textsearch_next() to retrieve
1802 * subsequent occurrences of the pattern. Returns the offset
1803 * to the first occurrence or UINT_MAX if no match was found.
1805 unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
1806 unsigned int to
, struct ts_config
*config
,
1807 struct ts_state
*state
)
1811 config
->get_next_block
= skb_ts_get_next_block
;
1812 config
->finish
= skb_ts_finish
;
1814 skb_prepare_seq_read(skb
, from
, to
, TS_SKB_CB(state
));
1816 ret
= textsearch_find(config
, state
);
1817 return (ret
<= to
- from
? ret
: UINT_MAX
);
1821 * skb_append_datato_frags: - append the user data to a skb
1822 * @sk: sock structure
1823 * @skb: skb structure to be appened with user data.
1824 * @getfrag: call back function to be used for getting the user data
1825 * @from: pointer to user message iov
1826 * @length: length of the iov message
1828 * Description: This procedure append the user data in the fragment part
1829 * of the skb if any page alloc fails user this procedure returns -ENOMEM
1831 int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
1832 int (*getfrag
)(void *from
, char *to
, int offset
,
1833 int len
, int odd
, struct sk_buff
*skb
),
1834 void *from
, int length
)
1837 skb_frag_t
*frag
= NULL
;
1838 struct page
*page
= NULL
;
1844 /* Return error if we don't have space for new frag */
1845 frg_cnt
= skb_shinfo(skb
)->nr_frags
;
1846 if (frg_cnt
>= MAX_SKB_FRAGS
)
1849 /* allocate a new page for next frag */
1850 page
= alloc_pages(sk
->sk_allocation
, 0);
1852 /* If alloc_page fails just return failure and caller will
1853 * free previous allocated pages by doing kfree_skb()
1858 /* initialize the next frag */
1859 sk
->sk_sndmsg_page
= page
;
1860 sk
->sk_sndmsg_off
= 0;
1861 skb_fill_page_desc(skb
, frg_cnt
, page
, 0, 0);
1862 skb
->truesize
+= PAGE_SIZE
;
1863 atomic_add(PAGE_SIZE
, &sk
->sk_wmem_alloc
);
1865 /* get the new initialized frag */
1866 frg_cnt
= skb_shinfo(skb
)->nr_frags
;
1867 frag
= &skb_shinfo(skb
)->frags
[frg_cnt
- 1];
1869 /* copy the user data to page */
1870 left
= PAGE_SIZE
- frag
->page_offset
;
1871 copy
= (length
> left
)? left
: length
;
1873 ret
= getfrag(from
, (page_address(frag
->page
) +
1874 frag
->page_offset
+ frag
->size
),
1875 offset
, copy
, 0, skb
);
1879 /* copy was successful so update the size parameters */
1880 sk
->sk_sndmsg_off
+= copy
;
1883 skb
->data_len
+= copy
;
1887 } while (length
> 0);
1893 * skb_pull_rcsum - pull skb and update receive checksum
1894 * @skb: buffer to update
1895 * @start: start of data before pull
1896 * @len: length of data pulled
1898 * This function performs an skb_pull on the packet and updates
1899 * update the CHECKSUM_HW checksum. It should be used on receive
1900 * path processing instead of skb_pull unless you know that the
1901 * checksum difference is zero (e.g., a valid IP header) or you
1902 * are setting ip_summed to CHECKSUM_NONE.
1904 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
)
1906 BUG_ON(len
> skb
->len
);
1908 BUG_ON(skb
->len
< skb
->data_len
);
1909 skb_postpull_rcsum(skb
, skb
->data
, len
);
1910 return skb
->data
+= len
;
1913 EXPORT_SYMBOL_GPL(skb_pull_rcsum
);
1916 * skb_segment - Perform protocol segmentation on skb.
1917 * @skb: buffer to segment
1918 * @features: features for the output path (see dev->features)
1920 * This function performs segmentation on the given skb. It returns
1921 * the segment at the given position. It returns NULL if there are
1922 * no more segments to generate, or when an error is encountered.
1924 struct sk_buff
*skb_segment(struct sk_buff
*skb
, int features
)
1926 struct sk_buff
*segs
= NULL
;
1927 struct sk_buff
*tail
= NULL
;
1928 unsigned int mss
= skb_shinfo(skb
)->gso_size
;
1929 unsigned int doffset
= skb
->data
- skb
->mac
.raw
;
1930 unsigned int offset
= doffset
;
1931 unsigned int headroom
;
1933 int sg
= features
& NETIF_F_SG
;
1934 int nfrags
= skb_shinfo(skb
)->nr_frags
;
1939 __skb_push(skb
, doffset
);
1940 headroom
= skb_headroom(skb
);
1941 pos
= skb_headlen(skb
);
1944 struct sk_buff
*nskb
;
1950 len
= skb
->len
- offset
;
1954 hsize
= skb_headlen(skb
) - offset
;
1957 nsize
= hsize
+ doffset
;
1958 if (nsize
> len
+ doffset
|| !sg
)
1959 nsize
= len
+ doffset
;
1961 nskb
= alloc_skb(nsize
+ headroom
, GFP_ATOMIC
);
1962 if (unlikely(!nskb
))
1971 nskb
->dev
= skb
->dev
;
1972 nskb
->priority
= skb
->priority
;
1973 nskb
->protocol
= skb
->protocol
;
1974 nskb
->dst
= dst_clone(skb
->dst
);
1975 memcpy(nskb
->cb
, skb
->cb
, sizeof(skb
->cb
));
1976 nskb
->pkt_type
= skb
->pkt_type
;
1977 nskb
->mac_len
= skb
->mac_len
;
1979 skb_reserve(nskb
, headroom
);
1980 nskb
->mac
.raw
= nskb
->data
;
1981 nskb
->nh
.raw
= nskb
->data
+ skb
->mac_len
;
1982 nskb
->h
.raw
= nskb
->nh
.raw
+ (skb
->h
.raw
- skb
->nh
.raw
);
1983 memcpy(skb_put(nskb
, doffset
), skb
->data
, doffset
);
1986 nskb
->csum
= skb_copy_and_csum_bits(skb
, offset
,
1992 frag
= skb_shinfo(nskb
)->frags
;
1995 nskb
->ip_summed
= CHECKSUM_HW
;
1996 nskb
->csum
= skb
->csum
;
1997 memcpy(skb_put(nskb
, hsize
), skb
->data
+ offset
, hsize
);
1999 while (pos
< offset
+ len
) {
2000 BUG_ON(i
>= nfrags
);
2002 *frag
= skb_shinfo(skb
)->frags
[i
];
2003 get_page(frag
->page
);
2007 frag
->page_offset
+= offset
- pos
;
2008 frag
->size
-= offset
- pos
;
2013 if (pos
+ size
<= offset
+ len
) {
2017 frag
->size
-= pos
+ size
- (offset
+ len
);
2024 skb_shinfo(nskb
)->nr_frags
= k
;
2025 nskb
->data_len
= len
- hsize
;
2026 nskb
->len
+= nskb
->data_len
;
2027 nskb
->truesize
+= nskb
->data_len
;
2028 } while ((offset
+= len
) < skb
->len
);
2033 while ((skb
= segs
)) {
2037 return ERR_PTR(err
);
2040 EXPORT_SYMBOL_GPL(skb_segment
);
2042 void __init
skb_init(void)
2044 skbuff_head_cache
= kmem_cache_create("skbuff_head_cache",
2045 sizeof(struct sk_buff
),
2049 if (!skbuff_head_cache
)
2050 panic("cannot create skbuff cache");
2052 skbuff_fclone_cache
= kmem_cache_create("skbuff_fclone_cache",
2053 (2*sizeof(struct sk_buff
)) +
2058 if (!skbuff_fclone_cache
)
2059 panic("cannot create skbuff cache");
2062 EXPORT_SYMBOL(___pskb_trim
);
2063 EXPORT_SYMBOL(__kfree_skb
);
2064 EXPORT_SYMBOL(kfree_skb
);
2065 EXPORT_SYMBOL(__pskb_pull_tail
);
2066 EXPORT_SYMBOL(__alloc_skb
);
2067 EXPORT_SYMBOL(__netdev_alloc_skb
);
2068 EXPORT_SYMBOL(pskb_copy
);
2069 EXPORT_SYMBOL(pskb_expand_head
);
2070 EXPORT_SYMBOL(skb_checksum
);
2071 EXPORT_SYMBOL(skb_clone
);
2072 EXPORT_SYMBOL(skb_clone_fraglist
);
2073 EXPORT_SYMBOL(skb_copy
);
2074 EXPORT_SYMBOL(skb_copy_and_csum_bits
);
2075 EXPORT_SYMBOL(skb_copy_and_csum_dev
);
2076 EXPORT_SYMBOL(skb_copy_bits
);
2077 EXPORT_SYMBOL(skb_copy_expand
);
2078 EXPORT_SYMBOL(skb_over_panic
);
2079 EXPORT_SYMBOL(skb_pad
);
2080 EXPORT_SYMBOL(skb_realloc_headroom
);
2081 EXPORT_SYMBOL(skb_under_panic
);
2082 EXPORT_SYMBOL(skb_dequeue
);
2083 EXPORT_SYMBOL(skb_dequeue_tail
);
2084 EXPORT_SYMBOL(skb_insert
);
2085 EXPORT_SYMBOL(skb_queue_purge
);
2086 EXPORT_SYMBOL(skb_queue_head
);
2087 EXPORT_SYMBOL(skb_queue_tail
);
2088 EXPORT_SYMBOL(skb_unlink
);
2089 EXPORT_SYMBOL(skb_append
);
2090 EXPORT_SYMBOL(skb_split
);
2091 EXPORT_SYMBOL(skb_prepare_seq_read
);
2092 EXPORT_SYMBOL(skb_seq_read
);
2093 EXPORT_SYMBOL(skb_abort_seq_read
);
2094 EXPORT_SYMBOL(skb_find_text
);
2095 EXPORT_SYMBOL(skb_append_datato_frags
);