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 <linux/textsearch.h>
31 #include <net/checksum.h>
33 #define HAVE_ALLOC_SKB /* For the drivers to know */
34 #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
35 #define SLAB_SKB /* Slabified skbuffs */
37 #define CHECKSUM_NONE 0
39 #define CHECKSUM_UNNECESSARY 2
41 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
42 ~(SMP_CACHE_BYTES - 1))
43 #define SKB_MAX_ORDER(X, ORDER) (((PAGE_SIZE << (ORDER)) - (X) - \
44 sizeof(struct skb_shared_info)) & \
45 ~(SMP_CACHE_BYTES - 1))
46 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
47 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
49 /* A. Checksumming of received packets by device.
51 * NONE: device failed to checksum this packet.
52 * skb->csum is undefined.
54 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
55 * skb->csum is undefined.
56 * It is bad option, but, unfortunately, many of vendors do this.
57 * Apparently with secret goal to sell you new device, when you
58 * will add new protocol to your host. F.e. IPv6. 8)
60 * HW: the most generic way. Device supplied checksum of _all_
61 * the packet as seen by netif_rx in skb->csum.
62 * NOTE: Even if device supports only some protocols, but
63 * is able to produce some skb->csum, it MUST use HW,
66 * B. Checksumming on output.
68 * NONE: skb is checksummed by protocol or csum is not required.
70 * HW: device is required to csum packet as seen by hard_start_xmit
71 * from skb->h.raw to the end and to record the checksum
72 * at skb->h.raw+skb->csum.
74 * Device must show its capabilities in dev->features, set
75 * at device setup time.
76 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
78 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
79 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
80 * TCP/UDP over IPv4. Sigh. Vendors like this
81 * way by an unknown reason. Though, see comment above
82 * about CHECKSUM_UNNECESSARY. 8)
84 * Any questions? No questions, good. --ANK
89 #ifdef CONFIG_NETFILTER
92 void (*destroy
)(struct nf_conntrack
*);
95 #ifdef CONFIG_BRIDGE_NETFILTER
96 struct nf_bridge_info
{
98 struct net_device
*physindev
;
99 struct net_device
*physoutdev
;
100 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
101 struct net_device
*netoutdev
;
104 unsigned long data
[32 / sizeof(unsigned long)];
110 struct sk_buff_head
{
111 /* These two members must be first. */
112 struct sk_buff
*next
;
113 struct sk_buff
*prev
;
121 /* To allow 64K frame to be packed as single skb without frag_list */
122 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
124 typedef struct skb_frag_struct skb_frag_t
;
126 struct skb_frag_struct
{
132 /* This data is invariant across clones and lives at
133 * the end of the header data, ie. at skb->end.
135 struct skb_shared_info
{
137 unsigned int nr_frags
;
138 unsigned short tso_size
;
139 unsigned short tso_segs
;
140 struct sk_buff
*frag_list
;
141 skb_frag_t frags
[MAX_SKB_FRAGS
];
144 /* We divide dataref into two halves. The higher 16 bits hold references
145 * to the payload part of skb->data. The lower 16 bits hold references to
146 * the entire skb->data. It is up to the users of the skb to agree on
147 * where the payload starts.
149 * All users must obey the rule that the skb->data reference count must be
150 * greater than or equal to the payload reference count.
152 * Holding a reference to the payload part means that the user does not
153 * care about modifications to the header part of skb->data.
155 #define SKB_DATAREF_SHIFT 16
156 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
159 * struct sk_buff - socket buffer
160 * @next: Next buffer in list
161 * @prev: Previous buffer in list
162 * @list: List we are on
163 * @sk: Socket we are owned by
164 * @stamp: Time we arrived
165 * @dev: Device we arrived on/are leaving by
166 * @input_dev: Device we arrived on
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 * @truesize: Buffer size
186 * @head: Head of buffer
187 * @data: Data head pointer
188 * @tail: Tail pointer
190 * @destructor: Destruct function
191 * @nfmark: Can be used for communication between hooks
192 * @nfct: Associated connection, if any
193 * @nfctinfo: Relationship of this skb to the connection
194 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
195 * @tc_index: Traffic control index
196 * @tc_verd: traffic control verdict
200 /* These two members must be first. */
201 struct sk_buff
*next
;
202 struct sk_buff
*prev
;
205 struct timeval stamp
;
206 struct net_device
*dev
;
207 struct net_device
*input_dev
;
212 struct icmphdr
*icmph
;
213 struct igmphdr
*igmph
;
215 struct ipv6hdr
*ipv6h
;
221 struct ipv6hdr
*ipv6h
;
230 struct dst_entry
*dst
;
234 * This is the control buffer. It is free to use for every
235 * layer. Please put your private variables there. If you
236 * want to keep them across layers you have to do a skb_clone()
237 * first. This is owned by whoever has the skb queued ATM.
254 void (*destructor
)(struct sk_buff
*skb
);
255 #ifdef CONFIG_NETFILTER
257 struct nf_conntrack
*nfct
;
258 #if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
259 __u8 ipvs_property
:1;
261 #ifdef CONFIG_BRIDGE_NETFILTER
262 struct nf_bridge_info
*nf_bridge
;
264 #endif /* CONFIG_NETFILTER */
265 #ifdef CONFIG_NET_SCHED
266 __u16 tc_index
; /* traffic control index */
267 #ifdef CONFIG_NET_CLS_ACT
268 __u16 tc_verd
; /* traffic control verdict */
273 /* These elements must be at the end, see alloc_skb() for details. */
274 unsigned int truesize
;
284 * Handling routines are only of interest to the kernel
286 #include <linux/slab.h>
288 #include <asm/system.h>
290 extern void __kfree_skb(struct sk_buff
*skb
);
291 extern struct sk_buff
*alloc_skb(unsigned int size
,
292 unsigned int __nocast priority
);
293 extern struct sk_buff
*alloc_skb_from_cache(kmem_cache_t
*cp
,
295 unsigned int __nocast priority
);
296 extern void kfree_skbmem(struct sk_buff
*skb
);
297 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
298 unsigned int __nocast priority
);
299 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
300 unsigned int __nocast priority
);
301 extern struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
302 unsigned int __nocast gfp_mask
);
303 extern int pskb_expand_head(struct sk_buff
*skb
,
304 int nhead
, int ntail
,
305 unsigned int __nocast gfp_mask
);
306 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
307 unsigned int headroom
);
308 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
309 int newheadroom
, int newtailroom
,
310 unsigned int __nocast priority
);
311 extern struct sk_buff
* skb_pad(struct sk_buff
*skb
, int pad
);
312 #define dev_kfree_skb(a) kfree_skb(a)
313 extern void skb_over_panic(struct sk_buff
*skb
, int len
,
315 extern void skb_under_panic(struct sk_buff
*skb
, int len
,
323 __u32 stepped_offset
;
324 struct sk_buff
*root_skb
;
325 struct sk_buff
*cur_skb
;
329 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
330 unsigned int from
, unsigned int to
,
331 struct skb_seq_state
*st
);
332 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
333 struct skb_seq_state
*st
);
334 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
336 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
337 unsigned int to
, struct ts_config
*config
,
338 struct ts_state
*state
);
341 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
344 * skb_queue_empty - check if a queue is empty
347 * Returns true if the queue is empty, false otherwise.
349 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
351 return list
->next
== (struct sk_buff
*)list
;
355 * skb_get - reference buffer
356 * @skb: buffer to reference
358 * Makes another reference to a socket buffer and returns a pointer
361 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
363 atomic_inc(&skb
->users
);
368 * If users == 1, we are the only owner and are can avoid redundant
373 * kfree_skb - free an sk_buff
374 * @skb: buffer to free
376 * Drop a reference to the buffer and free it if the usage count has
379 static inline void kfree_skb(struct sk_buff
*skb
)
381 if (likely(atomic_read(&skb
->users
) == 1))
383 else if (likely(!atomic_dec_and_test(&skb
->users
)))
389 * skb_cloned - is the buffer a clone
390 * @skb: buffer to check
392 * Returns true if the buffer was generated with skb_clone() and is
393 * one of multiple shared copies of the buffer. Cloned buffers are
394 * shared data so must not be written to under normal circumstances.
396 static inline int skb_cloned(const struct sk_buff
*skb
)
398 return skb
->cloned
&&
399 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
403 * skb_header_cloned - is the header a clone
404 * @skb: buffer to check
406 * Returns true if modifying the header part of the buffer requires
407 * the data to be copied.
409 static inline int skb_header_cloned(const struct sk_buff
*skb
)
416 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
417 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
422 * skb_header_release - release reference to header
423 * @skb: buffer to operate on
425 * Drop a reference to the header part of the buffer. This is done
426 * by acquiring a payload reference. You must not read from the header
427 * part of skb->data after this.
429 static inline void skb_header_release(struct sk_buff
*skb
)
433 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
437 * skb_shared - is the buffer shared
438 * @skb: buffer to check
440 * Returns true if more than one person has a reference to this
443 static inline int skb_shared(const struct sk_buff
*skb
)
445 return atomic_read(&skb
->users
) != 1;
449 * skb_share_check - check if buffer is shared and if so clone it
450 * @skb: buffer to check
451 * @pri: priority for memory allocation
453 * If the buffer is shared the buffer is cloned and the old copy
454 * drops a reference. A new clone with a single reference is returned.
455 * If the buffer is not shared the original buffer is returned. When
456 * being called from interrupt status or with spinlocks held pri must
459 * NULL is returned on a memory allocation failure.
461 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
,
462 unsigned int __nocast pri
)
464 might_sleep_if(pri
& __GFP_WAIT
);
465 if (skb_shared(skb
)) {
466 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
474 * Copy shared buffers into a new sk_buff. We effectively do COW on
475 * packets to handle cases where we have a local reader and forward
476 * and a couple of other messy ones. The normal one is tcpdumping
477 * a packet thats being forwarded.
481 * skb_unshare - make a copy of a shared buffer
482 * @skb: buffer to check
483 * @pri: priority for memory allocation
485 * If the socket buffer is a clone then this function creates a new
486 * copy of the data, drops a reference count on the old copy and returns
487 * the new copy with the reference count at 1. If the buffer is not a clone
488 * the original buffer is returned. When called with a spinlock held or
489 * from interrupt state @pri must be %GFP_ATOMIC
491 * %NULL is returned on a memory allocation failure.
493 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
494 unsigned int __nocast pri
)
496 might_sleep_if(pri
& __GFP_WAIT
);
497 if (skb_cloned(skb
)) {
498 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
499 kfree_skb(skb
); /* Free our shared copy */
507 * @list_: list to peek at
509 * Peek an &sk_buff. Unlike most other operations you _MUST_
510 * be careful with this one. A peek leaves the buffer on the
511 * list and someone else may run off with it. You must hold
512 * the appropriate locks or have a private queue to do this.
514 * Returns %NULL for an empty list or a pointer to the head element.
515 * The reference count is not incremented and the reference is therefore
516 * volatile. Use with caution.
518 static inline struct sk_buff
*skb_peek(struct sk_buff_head
*list_
)
520 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->next
;
521 if (list
== (struct sk_buff
*)list_
)
528 * @list_: list to peek at
530 * Peek an &sk_buff. Unlike most other operations you _MUST_
531 * be careful with this one. A peek leaves the buffer on the
532 * list and someone else may run off with it. You must hold
533 * the appropriate locks or have a private queue to do this.
535 * Returns %NULL for an empty list or a pointer to the tail element.
536 * The reference count is not incremented and the reference is therefore
537 * volatile. Use with caution.
539 static inline struct sk_buff
*skb_peek_tail(struct sk_buff_head
*list_
)
541 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->prev
;
542 if (list
== (struct sk_buff
*)list_
)
548 * skb_queue_len - get queue length
549 * @list_: list to measure
551 * Return the length of an &sk_buff queue.
553 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
558 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
560 spin_lock_init(&list
->lock
);
561 list
->prev
= list
->next
= (struct sk_buff
*)list
;
566 * Insert an sk_buff at the start of a list.
568 * The "__skb_xxxx()" functions are the non-atomic ones that
569 * can only be called with interrupts disabled.
573 * __skb_queue_head - queue a buffer at the list head
575 * @newsk: buffer to queue
577 * Queue a buffer at the start of a list. This function takes no locks
578 * and you must therefore hold required locks before calling it.
580 * A buffer cannot be placed on two lists at the same time.
582 extern void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
583 static inline void __skb_queue_head(struct sk_buff_head
*list
,
584 struct sk_buff
*newsk
)
586 struct sk_buff
*prev
, *next
;
589 prev
= (struct sk_buff
*)list
;
593 next
->prev
= prev
->next
= newsk
;
597 * __skb_queue_tail - queue a buffer at the list tail
599 * @newsk: buffer to queue
601 * Queue a buffer at the end of a list. This function takes no locks
602 * and you must therefore hold required locks before calling it.
604 * A buffer cannot be placed on two lists at the same time.
606 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
607 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
608 struct sk_buff
*newsk
)
610 struct sk_buff
*prev
, *next
;
613 next
= (struct sk_buff
*)list
;
617 next
->prev
= prev
->next
= newsk
;
622 * __skb_dequeue - remove from the head of the queue
623 * @list: list to dequeue from
625 * Remove the head of the list. This function does not take any locks
626 * so must be used with appropriate locks held only. The head item is
627 * returned or %NULL if the list is empty.
629 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
630 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
632 struct sk_buff
*next
, *prev
, *result
;
634 prev
= (struct sk_buff
*) list
;
643 result
->next
= result
->prev
= NULL
;
650 * Insert a packet on a list.
652 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
653 static inline void __skb_insert(struct sk_buff
*newsk
,
654 struct sk_buff
*prev
, struct sk_buff
*next
,
655 struct sk_buff_head
*list
)
659 next
->prev
= prev
->next
= newsk
;
664 * Place a packet after a given packet in a list.
666 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
667 static inline void __skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
669 __skb_insert(newsk
, old
, old
->next
, list
);
673 * remove sk_buff from list. _Must_ be called atomically, and with
676 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
677 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
679 struct sk_buff
*next
, *prev
;
684 skb
->next
= skb
->prev
= NULL
;
690 /* XXX: more streamlined implementation */
693 * __skb_dequeue_tail - remove from the tail of the queue
694 * @list: list to dequeue from
696 * Remove the tail of the list. This function does not take any locks
697 * so must be used with appropriate locks held only. The tail item is
698 * returned or %NULL if the list is empty.
700 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
701 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
703 struct sk_buff
*skb
= skb_peek_tail(list
);
705 __skb_unlink(skb
, list
);
710 static inline int skb_is_nonlinear(const struct sk_buff
*skb
)
712 return skb
->data_len
;
715 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
717 return skb
->len
- skb
->data_len
;
720 static inline int skb_pagelen(const struct sk_buff
*skb
)
724 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
725 len
+= skb_shinfo(skb
)->frags
[i
].size
;
726 return len
+ skb_headlen(skb
);
729 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
730 struct page
*page
, int off
, int size
)
732 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
735 frag
->page_offset
= off
;
737 skb_shinfo(skb
)->nr_frags
= i
+ 1;
740 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
741 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
742 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
745 * Add data to an sk_buff
747 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
749 unsigned char *tmp
= skb
->tail
;
750 SKB_LINEAR_ASSERT(skb
);
757 * skb_put - add data to a buffer
758 * @skb: buffer to use
759 * @len: amount of data to add
761 * This function extends the used data area of the buffer. If this would
762 * exceed the total buffer size the kernel will panic. A pointer to the
763 * first byte of the extra data is returned.
765 static inline unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
)
767 unsigned char *tmp
= skb
->tail
;
768 SKB_LINEAR_ASSERT(skb
);
771 if (unlikely(skb
->tail
>skb
->end
))
772 skb_over_panic(skb
, len
, current_text_addr());
776 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
784 * skb_push - add data to the start of a buffer
785 * @skb: buffer to use
786 * @len: amount of data to add
788 * This function extends the used data area of the buffer at the buffer
789 * start. If this would exceed the total buffer headroom the kernel will
790 * panic. A pointer to the first byte of the extra data is returned.
792 static inline unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
)
796 if (unlikely(skb
->data
<skb
->head
))
797 skb_under_panic(skb
, len
, current_text_addr());
801 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
804 BUG_ON(skb
->len
< skb
->data_len
);
805 return skb
->data
+= len
;
809 * skb_pull - remove data from the start of a buffer
810 * @skb: buffer to use
811 * @len: amount of data to remove
813 * This function removes data from the start of a buffer, returning
814 * the memory to the headroom. A pointer to the next data in the buffer
815 * is returned. Once the data has been pulled future pushes will overwrite
818 static inline unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
)
820 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
823 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
825 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
827 if (len
> skb_headlen(skb
) &&
828 !__pskb_pull_tail(skb
, len
-skb_headlen(skb
)))
831 return skb
->data
+= len
;
834 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
836 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
839 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
841 if (likely(len
<= skb_headlen(skb
)))
843 if (unlikely(len
> skb
->len
))
845 return __pskb_pull_tail(skb
, len
-skb_headlen(skb
)) != NULL
;
849 * skb_headroom - bytes at buffer head
850 * @skb: buffer to check
852 * Return the number of bytes of free space at the head of an &sk_buff.
854 static inline int skb_headroom(const struct sk_buff
*skb
)
856 return skb
->data
- skb
->head
;
860 * skb_tailroom - bytes at buffer end
861 * @skb: buffer to check
863 * Return the number of bytes of free space at the tail of an sk_buff
865 static inline int skb_tailroom(const struct sk_buff
*skb
)
867 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
871 * skb_reserve - adjust headroom
872 * @skb: buffer to alter
873 * @len: bytes to move
875 * Increase the headroom of an empty &sk_buff by reducing the tail
876 * room. This is only allowed for an empty buffer.
878 static inline void skb_reserve(struct sk_buff
*skb
, unsigned int len
)
885 * CPUs often take a performance hit when accessing unaligned memory
886 * locations. The actual performance hit varies, it can be small if the
887 * hardware handles it or large if we have to take an exception and fix it
890 * Since an ethernet header is 14 bytes network drivers often end up with
891 * the IP header at an unaligned offset. The IP header can be aligned by
892 * shifting the start of the packet by 2 bytes. Drivers should do this
895 * skb_reserve(NET_IP_ALIGN);
897 * The downside to this alignment of the IP header is that the DMA is now
898 * unaligned. On some architectures the cost of an unaligned DMA is high
899 * and this cost outweighs the gains made by aligning the IP header.
901 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
905 #define NET_IP_ALIGN 2
908 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
, int realloc
);
910 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
912 if (!skb
->data_len
) {
914 skb
->tail
= skb
->data
+ len
;
916 ___pskb_trim(skb
, len
, 0);
920 * skb_trim - remove end from a buffer
921 * @skb: buffer to alter
924 * Cut the length of a buffer down by removing data from the tail. If
925 * the buffer is already under the length specified it is not modified.
927 static inline void skb_trim(struct sk_buff
*skb
, unsigned int len
)
930 __skb_trim(skb
, len
);
934 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
936 if (!skb
->data_len
) {
938 skb
->tail
= skb
->data
+len
;
941 return ___pskb_trim(skb
, len
, 1);
944 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
946 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
950 * skb_orphan - orphan a buffer
951 * @skb: buffer to orphan
953 * If a buffer currently has an owner then we call the owner's
954 * destructor function and make the @skb unowned. The buffer continues
955 * to exist but is no longer charged to its former owner.
957 static inline void skb_orphan(struct sk_buff
*skb
)
960 skb
->destructor(skb
);
961 skb
->destructor
= NULL
;
966 * __skb_queue_purge - empty a list
967 * @list: list to empty
969 * Delete all buffers on an &sk_buff list. Each buffer is removed from
970 * the list and one reference dropped. This function does not take the
971 * list lock and the caller must hold the relevant locks to use it.
973 extern void skb_queue_purge(struct sk_buff_head
*list
);
974 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
977 while ((skb
= __skb_dequeue(list
)) != NULL
)
981 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
983 * __dev_alloc_skb - allocate an skbuff for sending
984 * @length: length to allocate
985 * @gfp_mask: get_free_pages mask, passed to alloc_skb
987 * Allocate a new &sk_buff and assign it a usage count of one. The
988 * buffer has unspecified headroom built in. Users should allocate
989 * the headroom they think they need without accounting for the
990 * built in space. The built in space is used for optimisations.
992 * %NULL is returned in there is no free memory.
994 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
995 unsigned int __nocast gfp_mask
)
997 struct sk_buff
*skb
= alloc_skb(length
+ 16, gfp_mask
);
999 skb_reserve(skb
, 16);
1003 extern struct sk_buff
*__dev_alloc_skb(unsigned int length
, int gfp_mask
);
1007 * dev_alloc_skb - allocate an skbuff for sending
1008 * @length: length to allocate
1010 * Allocate a new &sk_buff and assign it a usage count of one. The
1011 * buffer has unspecified headroom built in. Users should allocate
1012 * the headroom they think they need without accounting for the
1013 * built in space. The built in space is used for optimisations.
1015 * %NULL is returned in there is no free memory. Although this function
1016 * allocates memory it can be called from an interrupt.
1018 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
1020 return __dev_alloc_skb(length
, GFP_ATOMIC
);
1024 * skb_cow - copy header of skb when it is required
1025 * @skb: buffer to cow
1026 * @headroom: needed headroom
1028 * If the skb passed lacks sufficient headroom or its data part
1029 * is shared, data is reallocated. If reallocation fails, an error
1030 * is returned and original skb is not changed.
1032 * The result is skb with writable area skb->head...skb->tail
1033 * and at least @headroom of space at head.
1035 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
1037 int delta
= (headroom
> 16 ? headroom
: 16) - skb_headroom(skb
);
1042 if (delta
|| skb_cloned(skb
))
1043 return pskb_expand_head(skb
, (delta
+ 15) & ~15, 0, GFP_ATOMIC
);
1048 * skb_padto - pad an skbuff up to a minimal size
1049 * @skb: buffer to pad
1050 * @len: minimal length
1052 * Pads up a buffer to ensure the trailing bytes exist and are
1053 * blanked. If the buffer already contains sufficient data it
1054 * is untouched. Returns the buffer, which may be a replacement
1055 * for the original, or NULL for out of memory - in which case
1056 * the original buffer is still freed.
1059 static inline struct sk_buff
*skb_padto(struct sk_buff
*skb
, unsigned int len
)
1061 unsigned int size
= skb
->len
;
1062 if (likely(size
>= len
))
1064 return skb_pad(skb
, len
-size
);
1067 static inline int skb_add_data(struct sk_buff
*skb
,
1068 char __user
*from
, int copy
)
1070 const int off
= skb
->len
;
1072 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1074 unsigned int csum
= csum_and_copy_from_user(from
,
1078 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
1081 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
1084 __skb_trim(skb
, off
);
1088 static inline int skb_can_coalesce(struct sk_buff
*skb
, int i
,
1089 struct page
*page
, int off
)
1092 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
1094 return page
== frag
->page
&&
1095 off
== frag
->page_offset
+ frag
->size
;
1101 * skb_linearize - convert paged skb to linear one
1102 * @skb: buffer to linarize
1103 * @gfp: allocation mode
1105 * If there is no free memory -ENOMEM is returned, otherwise zero
1106 * is returned and the old skb data released.
1108 extern int __skb_linearize(struct sk_buff
*skb
, unsigned int __nocast gfp
);
1109 static inline int skb_linearize(struct sk_buff
*skb
, unsigned int __nocast gfp
)
1111 return __skb_linearize(skb
, gfp
);
1115 * skb_postpull_rcsum - update checksum for received skb after pull
1116 * @skb: buffer to update
1117 * @start: start of data before pull
1118 * @len: length of data pulled
1120 * After doing a pull on a received packet, you need to call this to
1121 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1122 * so that it can be recomputed from scratch.
1125 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
1126 const void *start
, int len
)
1128 if (skb
->ip_summed
== CHECKSUM_HW
)
1129 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
1133 * pskb_trim_rcsum - trim received skb and update checksum
1134 * @skb: buffer to trim
1137 * This is exactly the same as pskb_trim except that it ensures the
1138 * checksum of received packets are still valid after the operation.
1141 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
1143 if (len
>= skb
->len
)
1145 if (skb
->ip_summed
== CHECKSUM_HW
)
1146 skb
->ip_summed
= CHECKSUM_NONE
;
1147 return __pskb_trim(skb
, len
);
1150 static inline void *kmap_skb_frag(const skb_frag_t
*frag
)
1152 #ifdef CONFIG_HIGHMEM
1157 return kmap_atomic(frag
->page
, KM_SKB_DATA_SOFTIRQ
);
1160 static inline void kunmap_skb_frag(void *vaddr
)
1162 kunmap_atomic(vaddr
, KM_SKB_DATA_SOFTIRQ
);
1163 #ifdef CONFIG_HIGHMEM
1168 #define skb_queue_walk(queue, skb) \
1169 for (skb = (queue)->next; \
1170 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1174 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1175 int noblock
, int *err
);
1176 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
1177 struct poll_table_struct
*wait
);
1178 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
1179 int offset
, struct iovec
*to
,
1181 extern int skb_copy_and_csum_datagram_iovec(const
1182 struct sk_buff
*skb
,
1185 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
1186 extern unsigned int skb_checksum(const struct sk_buff
*skb
, int offset
,
1187 int len
, unsigned int csum
);
1188 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
1190 extern int skb_store_bits(const struct sk_buff
*skb
, int offset
,
1191 void *from
, int len
);
1192 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff
*skb
,
1193 int offset
, u8
*to
, int len
,
1195 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
1196 extern void skb_split(struct sk_buff
*skb
,
1197 struct sk_buff
*skb1
, const u32 len
);
1199 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
1200 int len
, void *buffer
)
1202 int hlen
= skb_headlen(skb
);
1204 if (hlen
- offset
>= len
)
1205 return skb
->data
+ offset
;
1207 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
1213 extern void skb_init(void);
1214 extern void skb_add_mtu(int mtu
);
1216 #ifdef CONFIG_NETFILTER
1217 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
1219 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
1220 nfct
->destroy(nfct
);
1222 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
1225 atomic_inc(&nfct
->use
);
1227 static inline void nf_reset(struct sk_buff
*skb
)
1229 nf_conntrack_put(skb
->nfct
);
1233 #ifdef CONFIG_BRIDGE_NETFILTER
1234 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
1236 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
1239 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
1242 atomic_inc(&nf_bridge
->use
);
1244 #endif /* CONFIG_BRIDGE_NETFILTER */
1245 #else /* CONFIG_NETFILTER */
1246 static inline void nf_reset(struct sk_buff
*skb
) {}
1247 #endif /* CONFIG_NETFILTER */
1249 #endif /* __KERNEL__ */
1250 #endif /* _LINUX_SKBUFF_H */