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>
45 #include <linux/interrupt.h>
47 #include <linux/inet.h>
48 #include <linux/slab.h>
49 #include <linux/netdevice.h>
50 #ifdef CONFIG_NET_CLS_ACT
51 #include <net/pkt_sched.h>
53 #include <linux/string.h>
54 #include <linux/skbuff.h>
55 #include <linux/cache.h>
56 #include <linux/rtnetlink.h>
57 #include <linux/init.h>
59 #include <net/protocol.h>
62 #include <net/checksum.h>
65 #include <asm/uaccess.h>
66 #include <asm/system.h>
70 static struct kmem_cache
*skbuff_head_cache __read_mostly
;
71 static struct kmem_cache
*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
134 * @node: numa node to allocate memory on
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
,
144 int fclone
, int node
)
146 struct kmem_cache
*cache
;
147 struct skb_shared_info
*shinfo
;
151 cache
= fclone
? skbuff_fclone_cache
: skbuff_head_cache
;
154 skb
= kmem_cache_alloc_node(cache
, gfp_mask
& ~__GFP_DMA
, node
);
158 /* Get the DATA. Size must match skb_add_mtu(). */
159 size
= SKB_DATA_ALIGN(size
);
160 data
= kmalloc_node_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(struct kmem_cache
*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
)
270 int node
= dev
->dev
.parent
? dev_to_node(dev
->dev
.parent
) : -1;
273 skb
= __alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
, 0, node
);
275 skb_reserve(skb
, NET_SKB_PAD
);
281 static void skb_drop_list(struct sk_buff
**listp
)
283 struct sk_buff
*list
= *listp
;
288 struct sk_buff
*this = list
;
294 static inline void skb_drop_fraglist(struct sk_buff
*skb
)
296 skb_drop_list(&skb_shinfo(skb
)->frag_list
);
299 static void skb_clone_fraglist(struct sk_buff
*skb
)
301 struct sk_buff
*list
;
303 for (list
= skb_shinfo(skb
)->frag_list
; list
; list
= list
->next
)
307 static void skb_release_data(struct sk_buff
*skb
)
310 !atomic_sub_return(skb
->nohdr
? (1 << SKB_DATAREF_SHIFT
) + 1 : 1,
311 &skb_shinfo(skb
)->dataref
)) {
312 if (skb_shinfo(skb
)->nr_frags
) {
314 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++)
315 put_page(skb_shinfo(skb
)->frags
[i
].page
);
318 if (skb_shinfo(skb
)->frag_list
)
319 skb_drop_fraglist(skb
);
326 * Free an skbuff by memory without cleaning the state.
328 void kfree_skbmem(struct sk_buff
*skb
)
330 struct sk_buff
*other
;
331 atomic_t
*fclone_ref
;
333 skb_release_data(skb
);
334 switch (skb
->fclone
) {
335 case SKB_FCLONE_UNAVAILABLE
:
336 kmem_cache_free(skbuff_head_cache
, skb
);
339 case SKB_FCLONE_ORIG
:
340 fclone_ref
= (atomic_t
*) (skb
+ 2);
341 if (atomic_dec_and_test(fclone_ref
))
342 kmem_cache_free(skbuff_fclone_cache
, skb
);
345 case SKB_FCLONE_CLONE
:
346 fclone_ref
= (atomic_t
*) (skb
+ 1);
349 /* The clone portion is available for
350 * fast-cloning again.
352 skb
->fclone
= SKB_FCLONE_UNAVAILABLE
;
354 if (atomic_dec_and_test(fclone_ref
))
355 kmem_cache_free(skbuff_fclone_cache
, other
);
361 * __kfree_skb - private function
364 * Free an sk_buff. Release anything attached to the buffer.
365 * Clean the state. This is an internal helper function. Users should
366 * always call kfree_skb
369 void __kfree_skb(struct sk_buff
*skb
)
371 dst_release(skb
->dst
);
373 secpath_put(skb
->sp
);
375 if (skb
->destructor
) {
377 skb
->destructor(skb
);
379 #ifdef CONFIG_NETFILTER
380 nf_conntrack_put(skb
->nfct
);
381 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
382 nf_conntrack_put_reasm(skb
->nfct_reasm
);
384 #ifdef CONFIG_BRIDGE_NETFILTER
385 nf_bridge_put(skb
->nf_bridge
);
388 /* XXX: IS this still necessary? - JHS */
389 #ifdef CONFIG_NET_SCHED
391 #ifdef CONFIG_NET_CLS_ACT
400 * kfree_skb - free an sk_buff
401 * @skb: buffer to free
403 * Drop a reference to the buffer and free it if the usage count has
406 void kfree_skb(struct sk_buff
*skb
)
410 if (likely(atomic_read(&skb
->users
) == 1))
412 else if (likely(!atomic_dec_and_test(&skb
->users
)))
418 * skb_clone - duplicate an sk_buff
419 * @skb: buffer to clone
420 * @gfp_mask: allocation priority
422 * Duplicate an &sk_buff. The new one is not owned by a socket. Both
423 * copies share the same packet data but not structure. The new
424 * buffer has a reference count of 1. If the allocation fails the
425 * function returns %NULL otherwise the new buffer is returned.
427 * If this function is called from an interrupt gfp_mask() must be
431 struct sk_buff
*skb_clone(struct sk_buff
*skb
, gfp_t gfp_mask
)
436 if (skb
->fclone
== SKB_FCLONE_ORIG
&&
437 n
->fclone
== SKB_FCLONE_UNAVAILABLE
) {
438 atomic_t
*fclone_ref
= (atomic_t
*) (n
+ 1);
439 n
->fclone
= SKB_FCLONE_CLONE
;
440 atomic_inc(fclone_ref
);
442 n
= kmem_cache_alloc(skbuff_head_cache
, gfp_mask
);
445 n
->fclone
= SKB_FCLONE_UNAVAILABLE
;
448 #define C(x) n->x = skb->x
450 n
->next
= n
->prev
= NULL
;
461 secpath_get(skb
->sp
);
463 memcpy(n
->cb
, skb
->cb
, sizeof(skb
->cb
));
473 #if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
477 n
->destructor
= NULL
;
479 #ifdef CONFIG_NETFILTER
481 nf_conntrack_get(skb
->nfct
);
483 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
485 nf_conntrack_get_reasm(skb
->nfct_reasm
);
487 #ifdef CONFIG_BRIDGE_NETFILTER
489 nf_bridge_get(skb
->nf_bridge
);
491 #endif /*CONFIG_NETFILTER*/
492 #ifdef CONFIG_NET_SCHED
494 #ifdef CONFIG_NET_CLS_ACT
495 n
->tc_verd
= SET_TC_VERD(skb
->tc_verd
,0);
496 n
->tc_verd
= CLR_TC_OK2MUNGE(n
->tc_verd
);
497 n
->tc_verd
= CLR_TC_MUNGED(n
->tc_verd
);
500 skb_copy_secmark(n
, skb
);
503 atomic_set(&n
->users
, 1);
509 atomic_inc(&(skb_shinfo(skb
)->dataref
));
515 static void copy_skb_header(struct sk_buff
*new, const struct sk_buff
*old
)
518 * Shift between the two data areas in bytes
520 unsigned long offset
= new->data
- old
->data
;
524 new->priority
= old
->priority
;
525 new->protocol
= old
->protocol
;
526 new->dst
= dst_clone(old
->dst
);
528 new->sp
= secpath_get(old
->sp
);
530 new->h
.raw
= old
->h
.raw
+ offset
;
531 new->nh
.raw
= old
->nh
.raw
+ offset
;
532 new->mac
.raw
= old
->mac
.raw
+ offset
;
533 memcpy(new->cb
, old
->cb
, sizeof(old
->cb
));
534 new->local_df
= old
->local_df
;
535 new->fclone
= SKB_FCLONE_UNAVAILABLE
;
536 new->pkt_type
= old
->pkt_type
;
537 new->tstamp
= old
->tstamp
;
538 new->destructor
= NULL
;
539 new->mark
= old
->mark
;
540 #ifdef CONFIG_NETFILTER
541 new->nfct
= old
->nfct
;
542 nf_conntrack_get(old
->nfct
);
543 new->nfctinfo
= old
->nfctinfo
;
544 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
545 new->nfct_reasm
= old
->nfct_reasm
;
546 nf_conntrack_get_reasm(old
->nfct_reasm
);
548 #if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
549 new->ipvs_property
= old
->ipvs_property
;
551 #ifdef CONFIG_BRIDGE_NETFILTER
552 new->nf_bridge
= old
->nf_bridge
;
553 nf_bridge_get(old
->nf_bridge
);
556 #ifdef CONFIG_NET_SCHED
557 #ifdef CONFIG_NET_CLS_ACT
558 new->tc_verd
= old
->tc_verd
;
560 new->tc_index
= old
->tc_index
;
562 skb_copy_secmark(new, old
);
563 atomic_set(&new->users
, 1);
564 skb_shinfo(new)->gso_size
= skb_shinfo(old
)->gso_size
;
565 skb_shinfo(new)->gso_segs
= skb_shinfo(old
)->gso_segs
;
566 skb_shinfo(new)->gso_type
= skb_shinfo(old
)->gso_type
;
570 * skb_copy - create private copy of an sk_buff
571 * @skb: buffer to copy
572 * @gfp_mask: allocation priority
574 * Make a copy of both an &sk_buff and its data. This is used when the
575 * caller wishes to modify the data and needs a private copy of the
576 * data to alter. Returns %NULL on failure or the pointer to the buffer
577 * on success. The returned buffer has a reference count of 1.
579 * As by-product this function converts non-linear &sk_buff to linear
580 * one, so that &sk_buff becomes completely private and caller is allowed
581 * to modify all the data of returned buffer. This means that this
582 * function is not recommended for use in circumstances when only
583 * header is going to be modified. Use pskb_copy() instead.
586 struct sk_buff
*skb_copy(const struct sk_buff
*skb
, gfp_t gfp_mask
)
588 int headerlen
= skb
->data
- skb
->head
;
590 * Allocate the copy buffer
592 struct sk_buff
*n
= alloc_skb(skb
->end
- skb
->head
+ skb
->data_len
,
597 /* Set the data pointer */
598 skb_reserve(n
, headerlen
);
599 /* Set the tail pointer and length */
600 skb_put(n
, skb
->len
);
602 n
->ip_summed
= skb
->ip_summed
;
604 if (skb_copy_bits(skb
, -headerlen
, n
->head
, headerlen
+ skb
->len
))
607 copy_skb_header(n
, skb
);
613 * pskb_copy - create copy of an sk_buff with private head.
614 * @skb: buffer to copy
615 * @gfp_mask: allocation priority
617 * Make a copy of both an &sk_buff and part of its data, located
618 * in header. Fragmented data remain shared. This is used when
619 * the caller wishes to modify only header of &sk_buff and needs
620 * private copy of the header to alter. Returns %NULL on failure
621 * or the pointer to the buffer on success.
622 * The returned buffer has a reference count of 1.
625 struct sk_buff
*pskb_copy(struct sk_buff
*skb
, gfp_t gfp_mask
)
628 * Allocate the copy buffer
630 struct sk_buff
*n
= alloc_skb(skb
->end
- skb
->head
, gfp_mask
);
635 /* Set the data pointer */
636 skb_reserve(n
, skb
->data
- skb
->head
);
637 /* Set the tail pointer and length */
638 skb_put(n
, skb_headlen(skb
));
640 memcpy(n
->data
, skb
->data
, n
->len
);
642 n
->ip_summed
= skb
->ip_summed
;
644 n
->truesize
+= skb
->data_len
;
645 n
->data_len
= skb
->data_len
;
648 if (skb_shinfo(skb
)->nr_frags
) {
651 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
652 skb_shinfo(n
)->frags
[i
] = skb_shinfo(skb
)->frags
[i
];
653 get_page(skb_shinfo(n
)->frags
[i
].page
);
655 skb_shinfo(n
)->nr_frags
= i
;
658 if (skb_shinfo(skb
)->frag_list
) {
659 skb_shinfo(n
)->frag_list
= skb_shinfo(skb
)->frag_list
;
660 skb_clone_fraglist(n
);
663 copy_skb_header(n
, skb
);
669 * pskb_expand_head - reallocate header of &sk_buff
670 * @skb: buffer to reallocate
671 * @nhead: room to add at head
672 * @ntail: room to add at tail
673 * @gfp_mask: allocation priority
675 * Expands (or creates identical copy, if &nhead and &ntail are zero)
676 * header of skb. &sk_buff itself is not changed. &sk_buff MUST have
677 * reference count of 1. Returns zero in the case of success or error,
678 * if expansion failed. In the last case, &sk_buff is not changed.
680 * All the pointers pointing into skb header may change and must be
681 * reloaded after call to this function.
684 int pskb_expand_head(struct sk_buff
*skb
, int nhead
, int ntail
,
689 int size
= nhead
+ (skb
->end
- skb
->head
) + ntail
;
695 size
= SKB_DATA_ALIGN(size
);
697 data
= kmalloc(size
+ sizeof(struct skb_shared_info
), gfp_mask
);
701 /* Copy only real data... and, alas, header. This should be
702 * optimized for the cases when header is void. */
703 memcpy(data
+ nhead
, skb
->head
, skb
->tail
- skb
->head
);
704 memcpy(data
+ size
, skb
->end
, sizeof(struct skb_shared_info
));
706 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++)
707 get_page(skb_shinfo(skb
)->frags
[i
].page
);
709 if (skb_shinfo(skb
)->frag_list
)
710 skb_clone_fraglist(skb
);
712 skb_release_data(skb
);
714 off
= (data
+ nhead
) - skb
->head
;
717 skb
->end
= data
+ size
;
725 atomic_set(&skb_shinfo(skb
)->dataref
, 1);
732 /* Make private copy of skb with writable head and some headroom */
734 struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
, unsigned int headroom
)
736 struct sk_buff
*skb2
;
737 int delta
= headroom
- skb_headroom(skb
);
740 skb2
= pskb_copy(skb
, GFP_ATOMIC
);
742 skb2
= skb_clone(skb
, GFP_ATOMIC
);
743 if (skb2
&& pskb_expand_head(skb2
, SKB_DATA_ALIGN(delta
), 0,
754 * skb_copy_expand - copy and expand sk_buff
755 * @skb: buffer to copy
756 * @newheadroom: new free bytes at head
757 * @newtailroom: new free bytes at tail
758 * @gfp_mask: allocation priority
760 * Make a copy of both an &sk_buff and its data and while doing so
761 * allocate additional space.
763 * This is used when the caller wishes to modify the data and needs a
764 * private copy of the data to alter as well as more space for new fields.
765 * Returns %NULL on failure or the pointer to the buffer
766 * on success. The returned buffer has a reference count of 1.
768 * You must pass %GFP_ATOMIC as the allocation priority if this function
769 * is called from an interrupt.
771 * BUG ALERT: ip_summed is not copied. Why does this work? Is it used
772 * only by netfilter in the cases when checksum is recalculated? --ANK
774 struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
775 int newheadroom
, int newtailroom
,
779 * Allocate the copy buffer
781 struct sk_buff
*n
= alloc_skb(newheadroom
+ skb
->len
+ newtailroom
,
783 int head_copy_len
, head_copy_off
;
788 skb_reserve(n
, newheadroom
);
790 /* Set the tail pointer and length */
791 skb_put(n
, skb
->len
);
793 head_copy_len
= skb_headroom(skb
);
795 if (newheadroom
<= head_copy_len
)
796 head_copy_len
= newheadroom
;
798 head_copy_off
= newheadroom
- head_copy_len
;
800 /* Copy the linear header and data. */
801 if (skb_copy_bits(skb
, -head_copy_len
, n
->head
+ head_copy_off
,
802 skb
->len
+ head_copy_len
))
805 copy_skb_header(n
, skb
);
811 * skb_pad - zero pad the tail of an skb
812 * @skb: buffer to pad
815 * Ensure that a buffer is followed by a padding area that is zero
816 * filled. Used by network drivers which may DMA or transfer data
817 * beyond the buffer end onto the wire.
819 * May return error in out of memory cases. The skb is freed on error.
822 int skb_pad(struct sk_buff
*skb
, int pad
)
827 /* If the skbuff is non linear tailroom is always zero.. */
828 if (!skb_cloned(skb
) && skb_tailroom(skb
) >= pad
) {
829 memset(skb
->data
+skb
->len
, 0, pad
);
833 ntail
= skb
->data_len
+ pad
- (skb
->end
- skb
->tail
);
834 if (likely(skb_cloned(skb
) || ntail
> 0)) {
835 err
= pskb_expand_head(skb
, 0, ntail
, GFP_ATOMIC
);
840 /* FIXME: The use of this function with non-linear skb's really needs
843 err
= skb_linearize(skb
);
847 memset(skb
->data
+ skb
->len
, 0, pad
);
855 /* Trims skb to length len. It can change skb pointers.
858 int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
)
860 struct sk_buff
**fragp
;
861 struct sk_buff
*frag
;
862 int offset
= skb_headlen(skb
);
863 int nfrags
= skb_shinfo(skb
)->nr_frags
;
867 if (skb_cloned(skb
) &&
868 unlikely((err
= pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
))))
875 for (; i
< nfrags
; i
++) {
876 int end
= offset
+ skb_shinfo(skb
)->frags
[i
].size
;
883 skb_shinfo(skb
)->frags
[i
++].size
= len
- offset
;
886 skb_shinfo(skb
)->nr_frags
= i
;
888 for (; i
< nfrags
; i
++)
889 put_page(skb_shinfo(skb
)->frags
[i
].page
);
891 if (skb_shinfo(skb
)->frag_list
)
892 skb_drop_fraglist(skb
);
896 for (fragp
= &skb_shinfo(skb
)->frag_list
; (frag
= *fragp
);
897 fragp
= &frag
->next
) {
898 int end
= offset
+ frag
->len
;
900 if (skb_shared(frag
)) {
901 struct sk_buff
*nfrag
;
903 nfrag
= skb_clone(frag
, GFP_ATOMIC
);
904 if (unlikely(!nfrag
))
907 nfrag
->next
= frag
->next
;
919 unlikely((err
= pskb_trim(frag
, len
- offset
))))
923 skb_drop_list(&frag
->next
);
928 if (len
> skb_headlen(skb
)) {
929 skb
->data_len
-= skb
->len
- len
;
934 skb
->tail
= skb
->data
+ len
;
941 * __pskb_pull_tail - advance tail of skb header
942 * @skb: buffer to reallocate
943 * @delta: number of bytes to advance tail
945 * The function makes a sense only on a fragmented &sk_buff,
946 * it expands header moving its tail forward and copying necessary
947 * data from fragmented part.
949 * &sk_buff MUST have reference count of 1.
951 * Returns %NULL (and &sk_buff does not change) if pull failed
952 * or value of new tail of skb in the case of success.
954 * All the pointers pointing into skb header may change and must be
955 * reloaded after call to this function.
958 /* Moves tail of skb head forward, copying data from fragmented part,
959 * when it is necessary.
960 * 1. It may fail due to malloc failure.
961 * 2. It may change skb pointers.
963 * It is pretty complicated. Luckily, it is called only in exceptional cases.
965 unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
)
967 /* If skb has not enough free space at tail, get new one
968 * plus 128 bytes for future expansions. If we have enough
969 * room at tail, reallocate without expansion only if skb is cloned.
971 int i
, k
, eat
= (skb
->tail
+ delta
) - skb
->end
;
973 if (eat
> 0 || skb_cloned(skb
)) {
974 if (pskb_expand_head(skb
, 0, eat
> 0 ? eat
+ 128 : 0,
979 if (skb_copy_bits(skb
, skb_headlen(skb
), skb
->tail
, delta
))
982 /* Optimization: no fragments, no reasons to preestimate
983 * size of pulled pages. Superb.
985 if (!skb_shinfo(skb
)->frag_list
)
988 /* Estimate size of pulled pages. */
990 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
991 if (skb_shinfo(skb
)->frags
[i
].size
>= eat
)
993 eat
-= skb_shinfo(skb
)->frags
[i
].size
;
996 /* If we need update frag list, we are in troubles.
997 * Certainly, it possible to add an offset to skb data,
998 * but taking into account that pulling is expected to
999 * be very rare operation, it is worth to fight against
1000 * further bloating skb head and crucify ourselves here instead.
1001 * Pure masohism, indeed. 8)8)
1004 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1005 struct sk_buff
*clone
= NULL
;
1006 struct sk_buff
*insp
= NULL
;
1011 if (list
->len
<= eat
) {
1012 /* Eaten as whole. */
1017 /* Eaten partially. */
1019 if (skb_shared(list
)) {
1020 /* Sucks! We need to fork list. :-( */
1021 clone
= skb_clone(list
, GFP_ATOMIC
);
1027 /* This may be pulled without
1031 if (!pskb_pull(list
, eat
)) {
1040 /* Free pulled out fragments. */
1041 while ((list
= skb_shinfo(skb
)->frag_list
) != insp
) {
1042 skb_shinfo(skb
)->frag_list
= list
->next
;
1045 /* And insert new clone at head. */
1048 skb_shinfo(skb
)->frag_list
= clone
;
1051 /* Success! Now we may commit changes to skb data. */
1056 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1057 if (skb_shinfo(skb
)->frags
[i
].size
<= eat
) {
1058 put_page(skb_shinfo(skb
)->frags
[i
].page
);
1059 eat
-= skb_shinfo(skb
)->frags
[i
].size
;
1061 skb_shinfo(skb
)->frags
[k
] = skb_shinfo(skb
)->frags
[i
];
1063 skb_shinfo(skb
)->frags
[k
].page_offset
+= eat
;
1064 skb_shinfo(skb
)->frags
[k
].size
-= eat
;
1070 skb_shinfo(skb
)->nr_frags
= k
;
1073 skb
->data_len
-= delta
;
1078 /* Copy some data bits from skb to kernel buffer. */
1080 int skb_copy_bits(const struct sk_buff
*skb
, int offset
, void *to
, int len
)
1083 int start
= skb_headlen(skb
);
1085 if (offset
> (int)skb
->len
- len
)
1089 if ((copy
= start
- offset
) > 0) {
1092 memcpy(to
, skb
->data
+ offset
, copy
);
1093 if ((len
-= copy
) == 0)
1099 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1102 BUG_TRAP(start
<= offset
+ len
);
1104 end
= start
+ skb_shinfo(skb
)->frags
[i
].size
;
1105 if ((copy
= end
- offset
) > 0) {
1111 vaddr
= kmap_skb_frag(&skb_shinfo(skb
)->frags
[i
]);
1113 vaddr
+ skb_shinfo(skb
)->frags
[i
].page_offset
+
1114 offset
- start
, copy
);
1115 kunmap_skb_frag(vaddr
);
1117 if ((len
-= copy
) == 0)
1125 if (skb_shinfo(skb
)->frag_list
) {
1126 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1128 for (; list
; list
= list
->next
) {
1131 BUG_TRAP(start
<= offset
+ len
);
1133 end
= start
+ list
->len
;
1134 if ((copy
= end
- offset
) > 0) {
1137 if (skb_copy_bits(list
, offset
- start
,
1140 if ((len
-= copy
) == 0)
1156 * skb_store_bits - store bits from kernel buffer to skb
1157 * @skb: destination buffer
1158 * @offset: offset in destination
1159 * @from: source buffer
1160 * @len: number of bytes to copy
1162 * Copy the specified number of bytes from the source buffer to the
1163 * destination skb. This function handles all the messy bits of
1164 * traversing fragment lists and such.
1167 int skb_store_bits(const struct sk_buff
*skb
, int offset
, void *from
, int len
)
1170 int start
= skb_headlen(skb
);
1172 if (offset
> (int)skb
->len
- len
)
1175 if ((copy
= start
- offset
) > 0) {
1178 memcpy(skb
->data
+ offset
, from
, copy
);
1179 if ((len
-= copy
) == 0)
1185 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1186 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1189 BUG_TRAP(start
<= offset
+ len
);
1191 end
= start
+ frag
->size
;
1192 if ((copy
= end
- offset
) > 0) {
1198 vaddr
= kmap_skb_frag(frag
);
1199 memcpy(vaddr
+ frag
->page_offset
+ offset
- start
,
1201 kunmap_skb_frag(vaddr
);
1203 if ((len
-= copy
) == 0)
1211 if (skb_shinfo(skb
)->frag_list
) {
1212 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1214 for (; list
; list
= list
->next
) {
1217 BUG_TRAP(start
<= offset
+ len
);
1219 end
= start
+ list
->len
;
1220 if ((copy
= end
- offset
) > 0) {
1223 if (skb_store_bits(list
, offset
- start
,
1226 if ((len
-= copy
) == 0)
1241 EXPORT_SYMBOL(skb_store_bits
);
1243 /* Checksum skb data. */
1245 __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
,
1246 int len
, __wsum csum
)
1248 int start
= skb_headlen(skb
);
1249 int i
, copy
= start
- offset
;
1252 /* Checksum header. */
1256 csum
= csum_partial(skb
->data
+ offset
, copy
, csum
);
1257 if ((len
-= copy
) == 0)
1263 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1266 BUG_TRAP(start
<= offset
+ len
);
1268 end
= start
+ skb_shinfo(skb
)->frags
[i
].size
;
1269 if ((copy
= end
- offset
) > 0) {
1272 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1276 vaddr
= kmap_skb_frag(frag
);
1277 csum2
= csum_partial(vaddr
+ frag
->page_offset
+
1278 offset
- start
, copy
, 0);
1279 kunmap_skb_frag(vaddr
);
1280 csum
= csum_block_add(csum
, csum2
, pos
);
1289 if (skb_shinfo(skb
)->frag_list
) {
1290 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1292 for (; list
; list
= list
->next
) {
1295 BUG_TRAP(start
<= offset
+ len
);
1297 end
= start
+ list
->len
;
1298 if ((copy
= end
- offset
) > 0) {
1302 csum2
= skb_checksum(list
, offset
- start
,
1304 csum
= csum_block_add(csum
, csum2
, pos
);
1305 if ((len
-= copy
) == 0)
1318 /* Both of above in one bottle. */
1320 __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
, int offset
,
1321 u8
*to
, int len
, __wsum csum
)
1323 int start
= skb_headlen(skb
);
1324 int i
, copy
= start
- offset
;
1331 csum
= csum_partial_copy_nocheck(skb
->data
+ offset
, to
,
1333 if ((len
-= copy
) == 0)
1340 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++) {
1343 BUG_TRAP(start
<= offset
+ len
);
1345 end
= start
+ skb_shinfo(skb
)->frags
[i
].size
;
1346 if ((copy
= end
- offset
) > 0) {
1349 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1353 vaddr
= kmap_skb_frag(frag
);
1354 csum2
= csum_partial_copy_nocheck(vaddr
+
1358 kunmap_skb_frag(vaddr
);
1359 csum
= csum_block_add(csum
, csum2
, pos
);
1369 if (skb_shinfo(skb
)->frag_list
) {
1370 struct sk_buff
*list
= skb_shinfo(skb
)->frag_list
;
1372 for (; list
; list
= list
->next
) {
1376 BUG_TRAP(start
<= offset
+ len
);
1378 end
= start
+ list
->len
;
1379 if ((copy
= end
- offset
) > 0) {
1382 csum2
= skb_copy_and_csum_bits(list
,
1385 csum
= csum_block_add(csum
, csum2
, pos
);
1386 if ((len
-= copy
) == 0)
1399 void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
)
1404 if (skb
->ip_summed
== CHECKSUM_PARTIAL
)
1405 csstart
= skb
->h
.raw
- skb
->data
;
1407 csstart
= skb_headlen(skb
);
1409 BUG_ON(csstart
> skb_headlen(skb
));
1411 memcpy(to
, skb
->data
, csstart
);
1414 if (csstart
!= skb
->len
)
1415 csum
= skb_copy_and_csum_bits(skb
, csstart
, to
+ csstart
,
1416 skb
->len
- csstart
, 0);
1418 if (skb
->ip_summed
== CHECKSUM_PARTIAL
) {
1419 long csstuff
= csstart
+ skb
->csum_offset
;
1421 *((__sum16
*)(to
+ csstuff
)) = csum_fold(csum
);
1426 * skb_dequeue - remove from the head of the queue
1427 * @list: list to dequeue from
1429 * Remove the head of the list. The list lock is taken so the function
1430 * may be used safely with other locking list functions. The head item is
1431 * returned or %NULL if the list is empty.
1434 struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
)
1436 unsigned long flags
;
1437 struct sk_buff
*result
;
1439 spin_lock_irqsave(&list
->lock
, flags
);
1440 result
= __skb_dequeue(list
);
1441 spin_unlock_irqrestore(&list
->lock
, flags
);
1446 * skb_dequeue_tail - remove from the tail of the queue
1447 * @list: list to dequeue from
1449 * Remove the tail of the list. The list lock is taken so the function
1450 * may be used safely with other locking list functions. The tail item is
1451 * returned or %NULL if the list is empty.
1453 struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
)
1455 unsigned long flags
;
1456 struct sk_buff
*result
;
1458 spin_lock_irqsave(&list
->lock
, flags
);
1459 result
= __skb_dequeue_tail(list
);
1460 spin_unlock_irqrestore(&list
->lock
, flags
);
1465 * skb_queue_purge - empty a list
1466 * @list: list to empty
1468 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1469 * the list and one reference dropped. This function takes the list
1470 * lock and is atomic with respect to other list locking functions.
1472 void skb_queue_purge(struct sk_buff_head
*list
)
1474 struct sk_buff
*skb
;
1475 while ((skb
= skb_dequeue(list
)) != NULL
)
1480 * skb_queue_head - queue a buffer at the list head
1481 * @list: list to use
1482 * @newsk: buffer to queue
1484 * Queue a buffer at the start of the list. This function takes the
1485 * list lock and can be used safely with other locking &sk_buff functions
1488 * A buffer cannot be placed on two lists at the same time.
1490 void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
)
1492 unsigned long flags
;
1494 spin_lock_irqsave(&list
->lock
, flags
);
1495 __skb_queue_head(list
, newsk
);
1496 spin_unlock_irqrestore(&list
->lock
, flags
);
1500 * skb_queue_tail - queue a buffer at the list tail
1501 * @list: list to use
1502 * @newsk: buffer to queue
1504 * Queue a buffer at the tail of the list. This function takes the
1505 * list lock and can be used safely with other locking &sk_buff functions
1508 * A buffer cannot be placed on two lists at the same time.
1510 void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
)
1512 unsigned long flags
;
1514 spin_lock_irqsave(&list
->lock
, flags
);
1515 __skb_queue_tail(list
, newsk
);
1516 spin_unlock_irqrestore(&list
->lock
, flags
);
1520 * skb_unlink - remove a buffer from a list
1521 * @skb: buffer to remove
1522 * @list: list to use
1524 * Remove a packet from a list. The list locks are taken and this
1525 * function is atomic with respect to other list locked calls
1527 * You must know what list the SKB is on.
1529 void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1531 unsigned long flags
;
1533 spin_lock_irqsave(&list
->lock
, flags
);
1534 __skb_unlink(skb
, list
);
1535 spin_unlock_irqrestore(&list
->lock
, flags
);
1539 * skb_append - append a buffer
1540 * @old: buffer to insert after
1541 * @newsk: buffer to insert
1542 * @list: list to use
1544 * Place a packet after a given packet in a list. The list locks are taken
1545 * and this function is atomic with respect to other list locked calls.
1546 * A buffer cannot be placed on two lists at the same time.
1548 void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
1550 unsigned long flags
;
1552 spin_lock_irqsave(&list
->lock
, flags
);
1553 __skb_append(old
, newsk
, list
);
1554 spin_unlock_irqrestore(&list
->lock
, flags
);
1559 * skb_insert - insert a buffer
1560 * @old: buffer to insert before
1561 * @newsk: buffer to insert
1562 * @list: list to use
1564 * Place a packet before a given packet in a list. The list locks are
1565 * taken and this function is atomic with respect to other list locked
1568 * A buffer cannot be placed on two lists at the same time.
1570 void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
1572 unsigned long flags
;
1574 spin_lock_irqsave(&list
->lock
, flags
);
1575 __skb_insert(newsk
, old
->prev
, old
, list
);
1576 spin_unlock_irqrestore(&list
->lock
, flags
);
1581 * Tune the memory allocator for a new MTU size.
1583 void skb_add_mtu(int mtu
)
1585 /* Must match allocation in alloc_skb */
1586 mtu
= SKB_DATA_ALIGN(mtu
) + sizeof(struct skb_shared_info
);
1588 kmem_add_cache_size(mtu
);
1592 static inline void skb_split_inside_header(struct sk_buff
*skb
,
1593 struct sk_buff
* skb1
,
1594 const u32 len
, const int pos
)
1598 memcpy(skb_put(skb1
, pos
- len
), skb
->data
+ len
, pos
- len
);
1600 /* And move data appendix as is. */
1601 for (i
= 0; i
< skb_shinfo(skb
)->nr_frags
; i
++)
1602 skb_shinfo(skb1
)->frags
[i
] = skb_shinfo(skb
)->frags
[i
];
1604 skb_shinfo(skb1
)->nr_frags
= skb_shinfo(skb
)->nr_frags
;
1605 skb_shinfo(skb
)->nr_frags
= 0;
1606 skb1
->data_len
= skb
->data_len
;
1607 skb1
->len
+= skb1
->data_len
;
1610 skb
->tail
= skb
->data
+ len
;
1613 static inline void skb_split_no_header(struct sk_buff
*skb
,
1614 struct sk_buff
* skb1
,
1615 const u32 len
, int pos
)
1618 const int nfrags
= skb_shinfo(skb
)->nr_frags
;
1620 skb_shinfo(skb
)->nr_frags
= 0;
1621 skb1
->len
= skb1
->data_len
= skb
->len
- len
;
1623 skb
->data_len
= len
- pos
;
1625 for (i
= 0; i
< nfrags
; i
++) {
1626 int size
= skb_shinfo(skb
)->frags
[i
].size
;
1628 if (pos
+ size
> len
) {
1629 skb_shinfo(skb1
)->frags
[k
] = skb_shinfo(skb
)->frags
[i
];
1633 * We have two variants in this case:
1634 * 1. Move all the frag to the second
1635 * part, if it is possible. F.e.
1636 * this approach is mandatory for TUX,
1637 * where splitting is expensive.
1638 * 2. Split is accurately. We make this.
1640 get_page(skb_shinfo(skb
)->frags
[i
].page
);
1641 skb_shinfo(skb1
)->frags
[0].page_offset
+= len
- pos
;
1642 skb_shinfo(skb1
)->frags
[0].size
-= len
- pos
;
1643 skb_shinfo(skb
)->frags
[i
].size
= len
- pos
;
1644 skb_shinfo(skb
)->nr_frags
++;
1648 skb_shinfo(skb
)->nr_frags
++;
1651 skb_shinfo(skb1
)->nr_frags
= k
;
1655 * skb_split - Split fragmented skb to two parts at length len.
1656 * @skb: the buffer to split
1657 * @skb1: the buffer to receive the second part
1658 * @len: new length for skb
1660 void skb_split(struct sk_buff
*skb
, struct sk_buff
*skb1
, const u32 len
)
1662 int pos
= skb_headlen(skb
);
1664 if (len
< pos
) /* Split line is inside header. */
1665 skb_split_inside_header(skb
, skb1
, len
, pos
);
1666 else /* Second chunk has no header, nothing to copy. */
1667 skb_split_no_header(skb
, skb1
, len
, pos
);
1671 * skb_prepare_seq_read - Prepare a sequential read of skb data
1672 * @skb: the buffer to read
1673 * @from: lower offset of data to be read
1674 * @to: upper offset of data to be read
1675 * @st: state variable
1677 * Initializes the specified state variable. Must be called before
1678 * invoking skb_seq_read() for the first time.
1680 void skb_prepare_seq_read(struct sk_buff
*skb
, unsigned int from
,
1681 unsigned int to
, struct skb_seq_state
*st
)
1683 st
->lower_offset
= from
;
1684 st
->upper_offset
= to
;
1685 st
->root_skb
= st
->cur_skb
= skb
;
1686 st
->frag_idx
= st
->stepped_offset
= 0;
1687 st
->frag_data
= NULL
;
1691 * skb_seq_read - Sequentially read skb data
1692 * @consumed: number of bytes consumed by the caller so far
1693 * @data: destination pointer for data to be returned
1694 * @st: state variable
1696 * Reads a block of skb data at &consumed relative to the
1697 * lower offset specified to skb_prepare_seq_read(). Assigns
1698 * the head of the data block to &data and returns the length
1699 * of the block or 0 if the end of the skb data or the upper
1700 * offset has been reached.
1702 * The caller is not required to consume all of the data
1703 * returned, i.e. &consumed is typically set to the number
1704 * of bytes already consumed and the next call to
1705 * skb_seq_read() will return the remaining part of the block.
1707 * Note: The size of each block of data returned can be arbitary,
1708 * this limitation is the cost for zerocopy seqeuental
1709 * reads of potentially non linear data.
1711 * Note: Fragment lists within fragments are not implemented
1712 * at the moment, state->root_skb could be replaced with
1713 * a stack for this purpose.
1715 unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
1716 struct skb_seq_state
*st
)
1718 unsigned int block_limit
, abs_offset
= consumed
+ st
->lower_offset
;
1721 if (unlikely(abs_offset
>= st
->upper_offset
))
1725 block_limit
= skb_headlen(st
->cur_skb
);
1727 if (abs_offset
< block_limit
) {
1728 *data
= st
->cur_skb
->data
+ abs_offset
;
1729 return block_limit
- abs_offset
;
1732 if (st
->frag_idx
== 0 && !st
->frag_data
)
1733 st
->stepped_offset
+= skb_headlen(st
->cur_skb
);
1735 while (st
->frag_idx
< skb_shinfo(st
->cur_skb
)->nr_frags
) {
1736 frag
= &skb_shinfo(st
->cur_skb
)->frags
[st
->frag_idx
];
1737 block_limit
= frag
->size
+ st
->stepped_offset
;
1739 if (abs_offset
< block_limit
) {
1741 st
->frag_data
= kmap_skb_frag(frag
);
1743 *data
= (u8
*) st
->frag_data
+ frag
->page_offset
+
1744 (abs_offset
- st
->stepped_offset
);
1746 return block_limit
- abs_offset
;
1749 if (st
->frag_data
) {
1750 kunmap_skb_frag(st
->frag_data
);
1751 st
->frag_data
= NULL
;
1755 st
->stepped_offset
+= frag
->size
;
1758 if (st
->cur_skb
->next
) {
1759 st
->cur_skb
= st
->cur_skb
->next
;
1762 } else if (st
->root_skb
== st
->cur_skb
&&
1763 skb_shinfo(st
->root_skb
)->frag_list
) {
1764 st
->cur_skb
= skb_shinfo(st
->root_skb
)->frag_list
;
1772 * skb_abort_seq_read - Abort a sequential read of skb data
1773 * @st: state variable
1775 * Must be called if skb_seq_read() was not called until it
1778 void skb_abort_seq_read(struct skb_seq_state
*st
)
1781 kunmap_skb_frag(st
->frag_data
);
1784 #define TS_SKB_CB(state) ((struct skb_seq_state *) &((state)->cb))
1786 static unsigned int skb_ts_get_next_block(unsigned int offset
, const u8
**text
,
1787 struct ts_config
*conf
,
1788 struct ts_state
*state
)
1790 return skb_seq_read(offset
, text
, TS_SKB_CB(state
));
1793 static void skb_ts_finish(struct ts_config
*conf
, struct ts_state
*state
)
1795 skb_abort_seq_read(TS_SKB_CB(state
));
1799 * skb_find_text - Find a text pattern in skb data
1800 * @skb: the buffer to look in
1801 * @from: search offset
1803 * @config: textsearch configuration
1804 * @state: uninitialized textsearch state variable
1806 * Finds a pattern in the skb data according to the specified
1807 * textsearch configuration. Use textsearch_next() to retrieve
1808 * subsequent occurrences of the pattern. Returns the offset
1809 * to the first occurrence or UINT_MAX if no match was found.
1811 unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
1812 unsigned int to
, struct ts_config
*config
,
1813 struct ts_state
*state
)
1817 config
->get_next_block
= skb_ts_get_next_block
;
1818 config
->finish
= skb_ts_finish
;
1820 skb_prepare_seq_read(skb
, from
, to
, TS_SKB_CB(state
));
1822 ret
= textsearch_find(config
, state
);
1823 return (ret
<= to
- from
? ret
: UINT_MAX
);
1827 * skb_append_datato_frags: - append the user data to a skb
1828 * @sk: sock structure
1829 * @skb: skb structure to be appened with user data.
1830 * @getfrag: call back function to be used for getting the user data
1831 * @from: pointer to user message iov
1832 * @length: length of the iov message
1834 * Description: This procedure append the user data in the fragment part
1835 * of the skb if any page alloc fails user this procedure returns -ENOMEM
1837 int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
1838 int (*getfrag
)(void *from
, char *to
, int offset
,
1839 int len
, int odd
, struct sk_buff
*skb
),
1840 void *from
, int length
)
1843 skb_frag_t
*frag
= NULL
;
1844 struct page
*page
= NULL
;
1850 /* Return error if we don't have space for new frag */
1851 frg_cnt
= skb_shinfo(skb
)->nr_frags
;
1852 if (frg_cnt
>= MAX_SKB_FRAGS
)
1855 /* allocate a new page for next frag */
1856 page
= alloc_pages(sk
->sk_allocation
, 0);
1858 /* If alloc_page fails just return failure and caller will
1859 * free previous allocated pages by doing kfree_skb()
1864 /* initialize the next frag */
1865 sk
->sk_sndmsg_page
= page
;
1866 sk
->sk_sndmsg_off
= 0;
1867 skb_fill_page_desc(skb
, frg_cnt
, page
, 0, 0);
1868 skb
->truesize
+= PAGE_SIZE
;
1869 atomic_add(PAGE_SIZE
, &sk
->sk_wmem_alloc
);
1871 /* get the new initialized frag */
1872 frg_cnt
= skb_shinfo(skb
)->nr_frags
;
1873 frag
= &skb_shinfo(skb
)->frags
[frg_cnt
- 1];
1875 /* copy the user data to page */
1876 left
= PAGE_SIZE
- frag
->page_offset
;
1877 copy
= (length
> left
)? left
: length
;
1879 ret
= getfrag(from
, (page_address(frag
->page
) +
1880 frag
->page_offset
+ frag
->size
),
1881 offset
, copy
, 0, skb
);
1885 /* copy was successful so update the size parameters */
1886 sk
->sk_sndmsg_off
+= copy
;
1889 skb
->data_len
+= copy
;
1893 } while (length
> 0);
1899 * skb_pull_rcsum - pull skb and update receive checksum
1900 * @skb: buffer to update
1901 * @start: start of data before pull
1902 * @len: length of data pulled
1904 * This function performs an skb_pull on the packet and updates
1905 * update the CHECKSUM_COMPLETE checksum. It should be used on
1906 * receive path processing instead of skb_pull unless you know
1907 * that the checksum difference is zero (e.g., a valid IP header)
1908 * or you are setting ip_summed to CHECKSUM_NONE.
1910 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
)
1912 BUG_ON(len
> skb
->len
);
1914 BUG_ON(skb
->len
< skb
->data_len
);
1915 skb_postpull_rcsum(skb
, skb
->data
, len
);
1916 return skb
->data
+= len
;
1919 EXPORT_SYMBOL_GPL(skb_pull_rcsum
);
1922 * skb_segment - Perform protocol segmentation on skb.
1923 * @skb: buffer to segment
1924 * @features: features for the output path (see dev->features)
1926 * This function performs segmentation on the given skb. It returns
1927 * the segment at the given position. It returns NULL if there are
1928 * no more segments to generate, or when an error is encountered.
1930 struct sk_buff
*skb_segment(struct sk_buff
*skb
, int features
)
1932 struct sk_buff
*segs
= NULL
;
1933 struct sk_buff
*tail
= NULL
;
1934 unsigned int mss
= skb_shinfo(skb
)->gso_size
;
1935 unsigned int doffset
= skb
->data
- skb
->mac
.raw
;
1936 unsigned int offset
= doffset
;
1937 unsigned int headroom
;
1939 int sg
= features
& NETIF_F_SG
;
1940 int nfrags
= skb_shinfo(skb
)->nr_frags
;
1945 __skb_push(skb
, doffset
);
1946 headroom
= skb_headroom(skb
);
1947 pos
= skb_headlen(skb
);
1950 struct sk_buff
*nskb
;
1956 len
= skb
->len
- offset
;
1960 hsize
= skb_headlen(skb
) - offset
;
1963 if (hsize
> len
|| !sg
)
1966 nskb
= alloc_skb(hsize
+ doffset
+ headroom
, GFP_ATOMIC
);
1967 if (unlikely(!nskb
))
1976 nskb
->dev
= skb
->dev
;
1977 nskb
->priority
= skb
->priority
;
1978 nskb
->protocol
= skb
->protocol
;
1979 nskb
->dst
= dst_clone(skb
->dst
);
1980 memcpy(nskb
->cb
, skb
->cb
, sizeof(skb
->cb
));
1981 nskb
->pkt_type
= skb
->pkt_type
;
1982 nskb
->mac_len
= skb
->mac_len
;
1984 skb_reserve(nskb
, headroom
);
1985 nskb
->mac
.raw
= nskb
->data
;
1986 nskb
->nh
.raw
= nskb
->data
+ skb
->mac_len
;
1987 nskb
->h
.raw
= nskb
->nh
.raw
+ (skb
->h
.raw
- skb
->nh
.raw
);
1988 memcpy(skb_put(nskb
, doffset
), skb
->data
, doffset
);
1991 nskb
->csum
= skb_copy_and_csum_bits(skb
, offset
,
1997 frag
= skb_shinfo(nskb
)->frags
;
2000 nskb
->ip_summed
= CHECKSUM_PARTIAL
;
2001 nskb
->csum
= skb
->csum
;
2002 memcpy(skb_put(nskb
, hsize
), skb
->data
+ offset
, hsize
);
2004 while (pos
< offset
+ len
) {
2005 BUG_ON(i
>= nfrags
);
2007 *frag
= skb_shinfo(skb
)->frags
[i
];
2008 get_page(frag
->page
);
2012 frag
->page_offset
+= offset
- pos
;
2013 frag
->size
-= offset
- pos
;
2018 if (pos
+ size
<= offset
+ len
) {
2022 frag
->size
-= pos
+ size
- (offset
+ len
);
2029 skb_shinfo(nskb
)->nr_frags
= k
;
2030 nskb
->data_len
= len
- hsize
;
2031 nskb
->len
+= nskb
->data_len
;
2032 nskb
->truesize
+= nskb
->data_len
;
2033 } while ((offset
+= len
) < skb
->len
);
2038 while ((skb
= segs
)) {
2042 return ERR_PTR(err
);
2045 EXPORT_SYMBOL_GPL(skb_segment
);
2047 void __init
skb_init(void)
2049 skbuff_head_cache
= kmem_cache_create("skbuff_head_cache",
2050 sizeof(struct sk_buff
),
2052 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
2054 skbuff_fclone_cache
= kmem_cache_create("skbuff_fclone_cache",
2055 (2*sizeof(struct sk_buff
)) +
2058 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,
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
);