2 * Definitions for the 'struct sk_buff' memory handlers.
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/config.h>
18 #include <linux/kernel.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/cache.h>
23 #include <asm/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
27 #include <linux/highmem.h>
28 #include <linux/poll.h>
29 #include <linux/net.h>
30 #include <net/checksum.h>
32 #define HAVE_ALLOC_SKB /* For the drivers to know */
33 #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
34 #define SLAB_SKB /* Slabified skbuffs */
36 #define CHECKSUM_NONE 0
38 #define CHECKSUM_UNNECESSARY 2
40 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
41 ~(SMP_CACHE_BYTES - 1))
42 #define SKB_MAX_ORDER(X, ORDER) (((PAGE_SIZE << (ORDER)) - (X) - \
43 sizeof(struct skb_shared_info)) & \
44 ~(SMP_CACHE_BYTES - 1))
45 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
46 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
48 /* A. Checksumming of received packets by device.
50 * NONE: device failed to checksum this packet.
51 * skb->csum is undefined.
53 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
54 * skb->csum is undefined.
55 * It is bad option, but, unfortunately, many of vendors do this.
56 * Apparently with secret goal to sell you new device, when you
57 * will add new protocol to your host. F.e. IPv6. 8)
59 * HW: the most generic way. Device supplied checksum of _all_
60 * the packet as seen by netif_rx in skb->csum.
61 * NOTE: Even if device supports only some protocols, but
62 * is able to produce some skb->csum, it MUST use HW,
65 * B. Checksumming on output.
67 * NONE: skb is checksummed by protocol or csum is not required.
69 * HW: device is required to csum packet as seen by hard_start_xmit
70 * from skb->h.raw to the end and to record the checksum
71 * at skb->h.raw+skb->csum.
73 * Device must show its capabilities in dev->features, set
74 * at device setup time.
75 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
77 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
78 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
79 * TCP/UDP over IPv4. Sigh. Vendors like this
80 * way by an unknown reason. Though, see comment above
81 * about CHECKSUM_UNNECESSARY. 8)
83 * Any questions? No questions, good. --ANK
88 #ifdef CONFIG_NETFILTER
91 void (*destroy
)(struct nf_conntrack
*);
94 #ifdef CONFIG_BRIDGE_NETFILTER
95 struct nf_bridge_info
{
97 struct net_device
*physindev
;
98 struct net_device
*physoutdev
;
99 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
100 struct net_device
*netoutdev
;
103 unsigned long data
[32 / sizeof(unsigned long)];
109 struct sk_buff_head
{
110 /* These two members must be first. */
111 struct sk_buff
*next
;
112 struct sk_buff
*prev
;
120 /* To allow 64K frame to be packed as single skb without frag_list */
121 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
123 typedef struct skb_frag_struct skb_frag_t
;
125 struct skb_frag_struct
{
131 /* This data is invariant across clones and lives at
132 * the end of the header data, ie. at skb->end.
134 struct skb_shared_info
{
136 unsigned int nr_frags
;
137 unsigned short tso_size
;
138 unsigned short tso_segs
;
139 struct sk_buff
*frag_list
;
140 skb_frag_t frags
[MAX_SKB_FRAGS
];
143 /* We divide dataref into two halves. The higher 16 bits hold references
144 * to the payload part of skb->data. The lower 16 bits hold references to
145 * the entire skb->data. It is up to the users of the skb to agree on
146 * where the payload starts.
148 * All users must obey the rule that the skb->data reference count must be
149 * greater than or equal to the payload reference count.
151 * Holding a reference to the payload part means that the user does not
152 * care about modifications to the header part of skb->data.
154 #define SKB_DATAREF_SHIFT 16
155 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
158 * struct sk_buff - socket buffer
159 * @next: Next buffer in list
160 * @prev: Previous buffer in list
161 * @list: List we are on
162 * @sk: Socket we are owned by
163 * @stamp: Time we arrived
164 * @dev: Device we arrived on/are leaving by
165 * @input_dev: Device we arrived on
166 * @real_dev: The real device we are using
167 * @h: Transport layer header
168 * @nh: Network layer header
169 * @mac: Link layer header
170 * @dst: destination entry
171 * @sp: the security path, used for xfrm
172 * @cb: Control buffer. Free for use by every layer. Put private vars here
173 * @len: Length of actual data
174 * @data_len: Data length
175 * @mac_len: Length of link layer header
177 * @local_df: allow local fragmentation
178 * @cloned: Head may be cloned (check refcnt to be sure)
179 * @nohdr: Payload reference only, must not modify header
180 * @pkt_type: Packet class
181 * @ip_summed: Driver fed us an IP checksum
182 * @priority: Packet queueing priority
183 * @users: User count - see {datagram,tcp}.c
184 * @protocol: Packet protocol from driver
185 * @security: Security level of packet
186 * @truesize: Buffer size
187 * @head: Head of buffer
188 * @data: Data head pointer
189 * @tail: Tail pointer
191 * @destructor: Destruct function
192 * @nfmark: Can be used for communication between hooks
193 * @nfcache: Cache info
194 * @nfct: Associated connection, if any
195 * @nfctinfo: Relationship of this skb to the connection
196 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
197 * @private: Data which is private to the HIPPI implementation
198 * @tc_index: Traffic control index
199 * @tc_verd: traffic control verdict
200 * @tc_classid: traffic control classid
204 /* These two members must be first. */
205 struct sk_buff
*next
;
206 struct sk_buff
*prev
;
208 struct sk_buff_head
*list
;
210 struct timeval stamp
;
211 struct net_device
*dev
;
212 struct net_device
*input_dev
;
213 struct net_device
*real_dev
;
218 struct icmphdr
*icmph
;
219 struct igmphdr
*igmph
;
221 struct ipv6hdr
*ipv6h
;
227 struct ipv6hdr
*ipv6h
;
236 struct dst_entry
*dst
;
240 * This is the control buffer. It is free to use for every
241 * layer. Please put your private variables there. If you
242 * want to keep them across layers you have to do a skb_clone()
243 * first. This is owned by whoever has the skb queued ATM.
251 unsigned char local_df
,
257 unsigned short protocol
,
260 void (*destructor
)(struct sk_buff
*skb
);
261 #ifdef CONFIG_NETFILTER
262 unsigned long nfmark
;
265 struct nf_conntrack
*nfct
;
266 #ifdef CONFIG_BRIDGE_NETFILTER
267 struct nf_bridge_info
*nf_bridge
;
269 #endif /* CONFIG_NETFILTER */
270 #if defined(CONFIG_HIPPI)
275 #ifdef CONFIG_NET_SCHED
276 __u32 tc_index
; /* traffic control index */
277 #ifdef CONFIG_NET_CLS_ACT
278 __u32 tc_verd
; /* traffic control verdict */
279 __u32 tc_classid
; /* traffic control classid */
285 /* These elements must be at the end, see alloc_skb() for details. */
286 unsigned int truesize
;
296 * Handling routines are only of interest to the kernel
298 #include <linux/slab.h>
300 #include <asm/system.h>
302 extern void __kfree_skb(struct sk_buff
*skb
);
303 extern struct sk_buff
*alloc_skb(unsigned int size
, int priority
);
304 extern struct sk_buff
*alloc_skb_from_cache(kmem_cache_t
*cp
,
305 unsigned int size
, int priority
);
306 extern void kfree_skbmem(struct sk_buff
*skb
);
307 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
, int priority
);
308 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
, int priority
);
309 extern struct sk_buff
*pskb_copy(struct sk_buff
*skb
, int gfp_mask
);
310 extern int pskb_expand_head(struct sk_buff
*skb
,
311 int nhead
, int ntail
, int gfp_mask
);
312 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
313 unsigned int headroom
);
314 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
315 int newheadroom
, int newtailroom
,
317 extern struct sk_buff
* skb_pad(struct sk_buff
*skb
, int pad
);
318 #define dev_kfree_skb(a) kfree_skb(a)
319 extern void skb_over_panic(struct sk_buff
*skb
, int len
,
321 extern void skb_under_panic(struct sk_buff
*skb
, int len
,
325 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
328 * skb_queue_empty - check if a queue is empty
331 * Returns true if the queue is empty, false otherwise.
333 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
335 return list
->next
== (struct sk_buff
*)list
;
339 * skb_get - reference buffer
340 * @skb: buffer to reference
342 * Makes another reference to a socket buffer and returns a pointer
345 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
347 atomic_inc(&skb
->users
);
352 * If users == 1, we are the only owner and are can avoid redundant
357 * kfree_skb - free an sk_buff
358 * @skb: buffer to free
360 * Drop a reference to the buffer and free it if the usage count has
363 static inline void kfree_skb(struct sk_buff
*skb
)
365 if (likely(atomic_read(&skb
->users
) == 1))
367 else if (likely(!atomic_dec_and_test(&skb
->users
)))
373 * skb_cloned - is the buffer a clone
374 * @skb: buffer to check
376 * Returns true if the buffer was generated with skb_clone() and is
377 * one of multiple shared copies of the buffer. Cloned buffers are
378 * shared data so must not be written to under normal circumstances.
380 static inline int skb_cloned(const struct sk_buff
*skb
)
382 return skb
->cloned
&&
383 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
387 * skb_header_cloned - is the header a clone
388 * @skb: buffer to check
390 * Returns true if modifying the header part of the buffer requires
391 * the data to be copied.
393 static inline int skb_header_cloned(const struct sk_buff
*skb
)
400 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
401 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
406 * skb_header_release - release reference to header
407 * @skb: buffer to operate on
409 * Drop a reference to the header part of the buffer. This is done
410 * by acquiring a payload reference. You must not read from the header
411 * part of skb->data after this.
413 static inline void skb_header_release(struct sk_buff
*skb
)
417 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
421 * skb_shared - is the buffer shared
422 * @skb: buffer to check
424 * Returns true if more than one person has a reference to this
427 static inline int skb_shared(const struct sk_buff
*skb
)
429 return atomic_read(&skb
->users
) != 1;
433 * skb_share_check - check if buffer is shared and if so clone it
434 * @skb: buffer to check
435 * @pri: priority for memory allocation
437 * If the buffer is shared the buffer is cloned and the old copy
438 * drops a reference. A new clone with a single reference is returned.
439 * If the buffer is not shared the original buffer is returned. When
440 * being called from interrupt status or with spinlocks held pri must
443 * NULL is returned on a memory allocation failure.
445 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
, int pri
)
447 might_sleep_if(pri
& __GFP_WAIT
);
448 if (skb_shared(skb
)) {
449 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
457 * Copy shared buffers into a new sk_buff. We effectively do COW on
458 * packets to handle cases where we have a local reader and forward
459 * and a couple of other messy ones. The normal one is tcpdumping
460 * a packet thats being forwarded.
464 * skb_unshare - make a copy of a shared buffer
465 * @skb: buffer to check
466 * @pri: priority for memory allocation
468 * If the socket buffer is a clone then this function creates a new
469 * copy of the data, drops a reference count on the old copy and returns
470 * the new copy with the reference count at 1. If the buffer is not a clone
471 * the original buffer is returned. When called with a spinlock held or
472 * from interrupt state @pri must be %GFP_ATOMIC
474 * %NULL is returned on a memory allocation failure.
476 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
, int pri
)
478 might_sleep_if(pri
& __GFP_WAIT
);
479 if (skb_cloned(skb
)) {
480 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
481 kfree_skb(skb
); /* Free our shared copy */
489 * @list_: list to peek at
491 * Peek an &sk_buff. Unlike most other operations you _MUST_
492 * be careful with this one. A peek leaves the buffer on the
493 * list and someone else may run off with it. You must hold
494 * the appropriate locks or have a private queue to do this.
496 * Returns %NULL for an empty list or a pointer to the head element.
497 * The reference count is not incremented and the reference is therefore
498 * volatile. Use with caution.
500 static inline struct sk_buff
*skb_peek(struct sk_buff_head
*list_
)
502 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->next
;
503 if (list
== (struct sk_buff
*)list_
)
510 * @list_: list to peek at
512 * Peek an &sk_buff. Unlike most other operations you _MUST_
513 * be careful with this one. A peek leaves the buffer on the
514 * list and someone else may run off with it. You must hold
515 * the appropriate locks or have a private queue to do this.
517 * Returns %NULL for an empty list or a pointer to the tail element.
518 * The reference count is not incremented and the reference is therefore
519 * volatile. Use with caution.
521 static inline struct sk_buff
*skb_peek_tail(struct sk_buff_head
*list_
)
523 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->prev
;
524 if (list
== (struct sk_buff
*)list_
)
530 * skb_queue_len - get queue length
531 * @list_: list to measure
533 * Return the length of an &sk_buff queue.
535 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
540 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
542 spin_lock_init(&list
->lock
);
543 list
->prev
= list
->next
= (struct sk_buff
*)list
;
548 * Insert an sk_buff at the start of a list.
550 * The "__skb_xxxx()" functions are the non-atomic ones that
551 * can only be called with interrupts disabled.
555 * __skb_queue_head - queue a buffer at the list head
557 * @newsk: buffer to queue
559 * Queue a buffer at the start of a list. This function takes no locks
560 * and you must therefore hold required locks before calling it.
562 * A buffer cannot be placed on two lists at the same time.
564 extern void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
565 static inline void __skb_queue_head(struct sk_buff_head
*list
,
566 struct sk_buff
*newsk
)
568 struct sk_buff
*prev
, *next
;
572 prev
= (struct sk_buff
*)list
;
576 next
->prev
= prev
->next
= newsk
;
580 * __skb_queue_tail - queue a buffer at the list tail
582 * @newsk: buffer to queue
584 * Queue a buffer at the end of a list. This function takes no locks
585 * and you must therefore hold required locks before calling it.
587 * A buffer cannot be placed on two lists at the same time.
589 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
590 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
591 struct sk_buff
*newsk
)
593 struct sk_buff
*prev
, *next
;
597 next
= (struct sk_buff
*)list
;
601 next
->prev
= prev
->next
= newsk
;
606 * __skb_dequeue - remove from the head of the queue
607 * @list: list to dequeue from
609 * Remove the head of the list. This function does not take any locks
610 * so must be used with appropriate locks held only. The head item is
611 * returned or %NULL if the list is empty.
613 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
614 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
616 struct sk_buff
*next
, *prev
, *result
;
618 prev
= (struct sk_buff
*) list
;
627 result
->next
= result
->prev
= NULL
;
635 * Insert a packet on a list.
637 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
);
638 static inline void __skb_insert(struct sk_buff
*newsk
,
639 struct sk_buff
*prev
, struct sk_buff
*next
,
640 struct sk_buff_head
*list
)
644 next
->prev
= prev
->next
= newsk
;
650 * Place a packet after a given packet in a list.
652 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
);
653 static inline void __skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
)
655 __skb_insert(newsk
, old
, old
->next
, old
->list
);
659 * remove sk_buff from list. _Must_ be called atomically, and with
662 extern void skb_unlink(struct sk_buff
*skb
);
663 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
665 struct sk_buff
*next
, *prev
;
670 skb
->next
= skb
->prev
= NULL
;
677 /* XXX: more streamlined implementation */
680 * __skb_dequeue_tail - remove from the tail of the queue
681 * @list: list to dequeue from
683 * Remove the tail of the list. This function does not take any locks
684 * so must be used with appropriate locks held only. The tail item is
685 * returned or %NULL if the list is empty.
687 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
688 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
690 struct sk_buff
*skb
= skb_peek_tail(list
);
692 __skb_unlink(skb
, list
);
697 static inline int skb_is_nonlinear(const struct sk_buff
*skb
)
699 return skb
->data_len
;
702 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
704 return skb
->len
- skb
->data_len
;
707 static inline int skb_pagelen(const struct sk_buff
*skb
)
711 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
712 len
+= skb_shinfo(skb
)->frags
[i
].size
;
713 return len
+ skb_headlen(skb
);
716 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
717 struct page
*page
, int off
, int size
)
719 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
722 frag
->page_offset
= off
;
724 skb_shinfo(skb
)->nr_frags
= i
+ 1;
727 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
728 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
729 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
732 * Add data to an sk_buff
734 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
736 unsigned char *tmp
= skb
->tail
;
737 SKB_LINEAR_ASSERT(skb
);
744 * skb_put - add data to a buffer
745 * @skb: buffer to use
746 * @len: amount of data to add
748 * This function extends the used data area of the buffer. If this would
749 * exceed the total buffer size the kernel will panic. A pointer to the
750 * first byte of the extra data is returned.
752 static inline unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
)
754 unsigned char *tmp
= skb
->tail
;
755 SKB_LINEAR_ASSERT(skb
);
758 if (unlikely(skb
->tail
>skb
->end
))
759 skb_over_panic(skb
, len
, current_text_addr());
763 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
771 * skb_push - add data to the start of a buffer
772 * @skb: buffer to use
773 * @len: amount of data to add
775 * This function extends the used data area of the buffer at the buffer
776 * start. If this would exceed the total buffer headroom the kernel will
777 * panic. A pointer to the first byte of the extra data is returned.
779 static inline unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
)
783 if (unlikely(skb
->data
<skb
->head
))
784 skb_under_panic(skb
, len
, current_text_addr());
788 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
791 BUG_ON(skb
->len
< skb
->data_len
);
792 return skb
->data
+= len
;
796 * skb_pull - remove data from the start of a buffer
797 * @skb: buffer to use
798 * @len: amount of data to remove
800 * This function removes data from the start of a buffer, returning
801 * the memory to the headroom. A pointer to the next data in the buffer
802 * is returned. Once the data has been pulled future pushes will overwrite
805 static inline unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
)
807 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
810 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
812 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
814 if (len
> skb_headlen(skb
) &&
815 !__pskb_pull_tail(skb
, len
-skb_headlen(skb
)))
818 return skb
->data
+= len
;
821 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
823 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
826 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
828 if (likely(len
<= skb_headlen(skb
)))
830 if (unlikely(len
> skb
->len
))
832 return __pskb_pull_tail(skb
, len
-skb_headlen(skb
)) != NULL
;
836 * skb_headroom - bytes at buffer head
837 * @skb: buffer to check
839 * Return the number of bytes of free space at the head of an &sk_buff.
841 static inline int skb_headroom(const struct sk_buff
*skb
)
843 return skb
->data
- skb
->head
;
847 * skb_tailroom - bytes at buffer end
848 * @skb: buffer to check
850 * Return the number of bytes of free space at the tail of an sk_buff
852 static inline int skb_tailroom(const struct sk_buff
*skb
)
854 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
858 * skb_reserve - adjust headroom
859 * @skb: buffer to alter
860 * @len: bytes to move
862 * Increase the headroom of an empty &sk_buff by reducing the tail
863 * room. This is only allowed for an empty buffer.
865 static inline void skb_reserve(struct sk_buff
*skb
, unsigned int len
)
872 * CPUs often take a performance hit when accessing unaligned memory
873 * locations. The actual performance hit varies, it can be small if the
874 * hardware handles it or large if we have to take an exception and fix it
877 * Since an ethernet header is 14 bytes network drivers often end up with
878 * the IP header at an unaligned offset. The IP header can be aligned by
879 * shifting the start of the packet by 2 bytes. Drivers should do this
882 * skb_reserve(NET_IP_ALIGN);
884 * The downside to this alignment of the IP header is that the DMA is now
885 * unaligned. On some architectures the cost of an unaligned DMA is high
886 * and this cost outweighs the gains made by aligning the IP header.
888 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
892 #define NET_IP_ALIGN 2
895 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
, int realloc
);
897 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
899 if (!skb
->data_len
) {
901 skb
->tail
= skb
->data
+ len
;
903 ___pskb_trim(skb
, len
, 0);
907 * skb_trim - remove end from a buffer
908 * @skb: buffer to alter
911 * Cut the length of a buffer down by removing data from the tail. If
912 * the buffer is already under the length specified it is not modified.
914 static inline void skb_trim(struct sk_buff
*skb
, unsigned int len
)
917 __skb_trim(skb
, len
);
921 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
923 if (!skb
->data_len
) {
925 skb
->tail
= skb
->data
+len
;
928 return ___pskb_trim(skb
, len
, 1);
931 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
933 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
937 * skb_orphan - orphan a buffer
938 * @skb: buffer to orphan
940 * If a buffer currently has an owner then we call the owner's
941 * destructor function and make the @skb unowned. The buffer continues
942 * to exist but is no longer charged to its former owner.
944 static inline void skb_orphan(struct sk_buff
*skb
)
947 skb
->destructor(skb
);
948 skb
->destructor
= NULL
;
953 * __skb_queue_purge - empty a list
954 * @list: list to empty
956 * Delete all buffers on an &sk_buff list. Each buffer is removed from
957 * the list and one reference dropped. This function does not take the
958 * list lock and the caller must hold the relevant locks to use it.
960 extern void skb_queue_purge(struct sk_buff_head
*list
);
961 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
964 while ((skb
= __skb_dequeue(list
)) != NULL
)
968 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
970 * __dev_alloc_skb - allocate an skbuff for sending
971 * @length: length to allocate
972 * @gfp_mask: get_free_pages mask, passed to alloc_skb
974 * Allocate a new &sk_buff and assign it a usage count of one. The
975 * buffer has unspecified headroom built in. Users should allocate
976 * the headroom they think they need without accounting for the
977 * built in space. The built in space is used for optimisations.
979 * %NULL is returned in there is no free memory.
981 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
984 struct sk_buff
*skb
= alloc_skb(length
+ 16, gfp_mask
);
986 skb_reserve(skb
, 16);
990 extern struct sk_buff
*__dev_alloc_skb(unsigned int length
, int gfp_mask
);
994 * dev_alloc_skb - allocate an skbuff for sending
995 * @length: length to allocate
997 * Allocate a new &sk_buff and assign it a usage count of one. The
998 * buffer has unspecified headroom built in. Users should allocate
999 * the headroom they think they need without accounting for the
1000 * built in space. The built in space is used for optimisations.
1002 * %NULL is returned in there is no free memory. Although this function
1003 * allocates memory it can be called from an interrupt.
1005 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
1007 return __dev_alloc_skb(length
, GFP_ATOMIC
);
1011 * skb_cow - copy header of skb when it is required
1012 * @skb: buffer to cow
1013 * @headroom: needed headroom
1015 * If the skb passed lacks sufficient headroom or its data part
1016 * is shared, data is reallocated. If reallocation fails, an error
1017 * is returned and original skb is not changed.
1019 * The result is skb with writable area skb->head...skb->tail
1020 * and at least @headroom of space at head.
1022 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
1024 int delta
= (headroom
> 16 ? headroom
: 16) - skb_headroom(skb
);
1029 if (delta
|| skb_cloned(skb
))
1030 return pskb_expand_head(skb
, (delta
+ 15) & ~15, 0, GFP_ATOMIC
);
1035 * skb_padto - pad an skbuff up to a minimal size
1036 * @skb: buffer to pad
1037 * @len: minimal length
1039 * Pads up a buffer to ensure the trailing bytes exist and are
1040 * blanked. If the buffer already contains sufficient data it
1041 * is untouched. Returns the buffer, which may be a replacement
1042 * for the original, or NULL for out of memory - in which case
1043 * the original buffer is still freed.
1046 static inline struct sk_buff
*skb_padto(struct sk_buff
*skb
, unsigned int len
)
1048 unsigned int size
= skb
->len
;
1049 if (likely(size
>= len
))
1051 return skb_pad(skb
, len
-size
);
1054 static inline int skb_add_data(struct sk_buff
*skb
,
1055 char __user
*from
, int copy
)
1057 const int off
= skb
->len
;
1059 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1061 unsigned int csum
= csum_and_copy_from_user(from
,
1065 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
1068 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
1071 __skb_trim(skb
, off
);
1075 static inline int skb_can_coalesce(struct sk_buff
*skb
, int i
,
1076 struct page
*page
, int off
)
1079 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
1081 return page
== frag
->page
&&
1082 off
== frag
->page_offset
+ frag
->size
;
1088 * skb_linearize - convert paged skb to linear one
1089 * @skb: buffer to linarize
1090 * @gfp: allocation mode
1092 * If there is no free memory -ENOMEM is returned, otherwise zero
1093 * is returned and the old skb data released.
1095 extern int __skb_linearize(struct sk_buff
*skb
, int gfp
);
1096 static inline int skb_linearize(struct sk_buff
*skb
, int gfp
)
1098 return __skb_linearize(skb
, gfp
);
1102 * skb_postpull_rcsum - update checksum for received skb after pull
1103 * @skb: buffer to update
1104 * @start: start of data before pull
1105 * @len: length of data pulled
1107 * After doing a pull on a received packet, you need to call this to
1108 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1109 * so that it can be recomputed from scratch.
1112 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
1113 const void *start
, int len
)
1115 if (skb
->ip_summed
== CHECKSUM_HW
)
1116 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
1120 * pskb_trim_rcsum - trim received skb and update checksum
1121 * @skb: buffer to trim
1124 * This is exactly the same as pskb_trim except that it ensures the
1125 * checksum of received packets are still valid after the operation.
1128 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
1130 if (len
>= skb
->len
)
1132 if (skb
->ip_summed
== CHECKSUM_HW
)
1133 skb
->ip_summed
= CHECKSUM_NONE
;
1134 return __pskb_trim(skb
, len
);
1137 static inline void *kmap_skb_frag(const skb_frag_t
*frag
)
1139 #ifdef CONFIG_HIGHMEM
1144 return kmap_atomic(frag
->page
, KM_SKB_DATA_SOFTIRQ
);
1147 static inline void kunmap_skb_frag(void *vaddr
)
1149 kunmap_atomic(vaddr
, KM_SKB_DATA_SOFTIRQ
);
1150 #ifdef CONFIG_HIGHMEM
1155 #define skb_queue_walk(queue, skb) \
1156 for (skb = (queue)->next; \
1157 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1161 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1162 int noblock
, int *err
);
1163 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
1164 struct poll_table_struct
*wait
);
1165 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
1166 int offset
, struct iovec
*to
,
1168 extern int skb_copy_and_csum_datagram_iovec(const
1169 struct sk_buff
*skb
,
1172 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
1173 extern unsigned int skb_checksum(const struct sk_buff
*skb
, int offset
,
1174 int len
, unsigned int csum
);
1175 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
1177 extern int skb_store_bits(const struct sk_buff
*skb
, int offset
,
1178 void *from
, int len
);
1179 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff
*skb
,
1180 int offset
, u8
*to
, int len
,
1182 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
1183 extern void skb_split(struct sk_buff
*skb
,
1184 struct sk_buff
*skb1
, const u32 len
);
1186 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
1187 int len
, void *buffer
)
1189 int hlen
= skb_headlen(skb
);
1191 if (offset
+ len
<= hlen
)
1192 return skb
->data
+ offset
;
1194 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
1200 extern void skb_init(void);
1201 extern void skb_add_mtu(int mtu
);
1203 #ifdef CONFIG_NETFILTER
1204 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
1206 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
1207 nfct
->destroy(nfct
);
1209 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
1212 atomic_inc(&nfct
->use
);
1214 static inline void nf_reset(struct sk_buff
*skb
)
1216 nf_conntrack_put(skb
->nfct
);
1220 #ifdef CONFIG_BRIDGE_NETFILTER
1221 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
1223 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
1226 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
1229 atomic_inc(&nf_bridge
->use
);
1231 #endif /* CONFIG_BRIDGE_NETFILTER */
1232 #else /* CONFIG_NETFILTER */
1233 static inline void nf_reset(struct sk_buff
*skb
) {}
1234 #endif /* CONFIG_NETFILTER */
1236 #endif /* __KERNEL__ */
1237 #endif /* _LINUX_SKBUFF_H */