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)
158 extern struct timeval skb_tv_base
;
166 * struct sk_buff - socket buffer
167 * @next: Next buffer in list
168 * @prev: Previous buffer in list
169 * @list: List we are on
170 * @sk: Socket we are owned by
171 * @tstamp: Time we arrived stored as offset to skb_tv_base
172 * @dev: Device we arrived on/are leaving by
173 * @input_dev: Device we arrived on
174 * @h: Transport layer header
175 * @nh: Network layer header
176 * @mac: Link layer header
177 * @dst: destination entry
178 * @sp: the security path, used for xfrm
179 * @cb: Control buffer. Free for use by every layer. Put private vars here
180 * @len: Length of actual data
181 * @data_len: Data length
182 * @mac_len: Length of link layer header
184 * @local_df: allow local fragmentation
185 * @cloned: Head may be cloned (check refcnt to be sure)
186 * @nohdr: Payload reference only, must not modify header
187 * @pkt_type: Packet class
188 * @ip_summed: Driver fed us an IP checksum
189 * @priority: Packet queueing priority
190 * @users: User count - see {datagram,tcp}.c
191 * @protocol: Packet protocol from driver
192 * @truesize: Buffer size
193 * @head: Head of buffer
194 * @data: Data head pointer
195 * @tail: Tail pointer
197 * @destructor: Destruct function
198 * @nfmark: Can be used for communication between hooks
199 * @nfct: Associated connection, if any
200 * @nfctinfo: Relationship of this skb to the connection
201 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
202 * @tc_index: Traffic control index
203 * @tc_verd: traffic control verdict
207 /* These two members must be first. */
208 struct sk_buff
*next
;
209 struct sk_buff
*prev
;
212 struct skb_timeval tstamp
;
213 struct net_device
*dev
;
214 struct net_device
*input_dev
;
219 struct icmphdr
*icmph
;
220 struct igmphdr
*igmph
;
222 struct ipv6hdr
*ipv6h
;
228 struct ipv6hdr
*ipv6h
;
237 struct dst_entry
*dst
;
241 * This is the control buffer. It is free to use for every
242 * layer. Please put your private variables there. If you
243 * want to keep them across layers you have to do a skb_clone()
244 * first. This is owned by whoever has the skb queued ATM.
261 void (*destructor
)(struct sk_buff
*skb
);
262 #ifdef CONFIG_NETFILTER
264 struct nf_conntrack
*nfct
;
265 #if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
266 __u8 ipvs_property
:1;
268 #ifdef CONFIG_BRIDGE_NETFILTER
269 struct nf_bridge_info
*nf_bridge
;
271 #endif /* CONFIG_NETFILTER */
272 #ifdef CONFIG_NET_SCHED
273 __u16 tc_index
; /* traffic control index */
274 #ifdef CONFIG_NET_CLS_ACT
275 __u16 tc_verd
; /* traffic control verdict */
280 /* These elements must be at the end, see alloc_skb() for details. */
281 unsigned int truesize
;
291 * Handling routines are only of interest to the kernel
293 #include <linux/slab.h>
295 #include <asm/system.h>
297 extern void __kfree_skb(struct sk_buff
*skb
);
298 extern struct sk_buff
*alloc_skb(unsigned int size
,
299 unsigned int __nocast priority
);
300 extern struct sk_buff
*alloc_skb_from_cache(kmem_cache_t
*cp
,
302 unsigned int __nocast priority
);
303 extern void kfree_skbmem(struct sk_buff
*skb
);
304 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
305 unsigned int __nocast priority
);
306 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
307 unsigned int __nocast priority
);
308 extern struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
309 unsigned int __nocast gfp_mask
);
310 extern int pskb_expand_head(struct sk_buff
*skb
,
311 int nhead
, int ntail
,
312 unsigned int __nocast gfp_mask
);
313 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
314 unsigned int headroom
);
315 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
316 int newheadroom
, int newtailroom
,
317 unsigned int __nocast priority
);
318 extern struct sk_buff
* skb_pad(struct sk_buff
*skb
, int pad
);
319 #define dev_kfree_skb(a) kfree_skb(a)
320 extern void skb_over_panic(struct sk_buff
*skb
, int len
,
322 extern void skb_under_panic(struct sk_buff
*skb
, int len
,
330 __u32 stepped_offset
;
331 struct sk_buff
*root_skb
;
332 struct sk_buff
*cur_skb
;
336 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
337 unsigned int from
, unsigned int to
,
338 struct skb_seq_state
*st
);
339 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
340 struct skb_seq_state
*st
);
341 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
343 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
344 unsigned int to
, struct ts_config
*config
,
345 struct ts_state
*state
);
348 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
351 * skb_queue_empty - check if a queue is empty
354 * Returns true if the queue is empty, false otherwise.
356 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
358 return list
->next
== (struct sk_buff
*)list
;
362 * skb_get - reference buffer
363 * @skb: buffer to reference
365 * Makes another reference to a socket buffer and returns a pointer
368 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
370 atomic_inc(&skb
->users
);
375 * If users == 1, we are the only owner and are can avoid redundant
380 * kfree_skb - free an sk_buff
381 * @skb: buffer to free
383 * Drop a reference to the buffer and free it if the usage count has
386 static inline void kfree_skb(struct sk_buff
*skb
)
388 if (likely(atomic_read(&skb
->users
) == 1))
390 else if (likely(!atomic_dec_and_test(&skb
->users
)))
396 * skb_cloned - is the buffer a clone
397 * @skb: buffer to check
399 * Returns true if the buffer was generated with skb_clone() and is
400 * one of multiple shared copies of the buffer. Cloned buffers are
401 * shared data so must not be written to under normal circumstances.
403 static inline int skb_cloned(const struct sk_buff
*skb
)
405 return skb
->cloned
&&
406 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
410 * skb_header_cloned - is the header a clone
411 * @skb: buffer to check
413 * Returns true if modifying the header part of the buffer requires
414 * the data to be copied.
416 static inline int skb_header_cloned(const struct sk_buff
*skb
)
423 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
424 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
429 * skb_header_release - release reference to header
430 * @skb: buffer to operate on
432 * Drop a reference to the header part of the buffer. This is done
433 * by acquiring a payload reference. You must not read from the header
434 * part of skb->data after this.
436 static inline void skb_header_release(struct sk_buff
*skb
)
440 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
444 * skb_shared - is the buffer shared
445 * @skb: buffer to check
447 * Returns true if more than one person has a reference to this
450 static inline int skb_shared(const struct sk_buff
*skb
)
452 return atomic_read(&skb
->users
) != 1;
456 * skb_share_check - check if buffer is shared and if so clone it
457 * @skb: buffer to check
458 * @pri: priority for memory allocation
460 * If the buffer is shared the buffer is cloned and the old copy
461 * drops a reference. A new clone with a single reference is returned.
462 * If the buffer is not shared the original buffer is returned. When
463 * being called from interrupt status or with spinlocks held pri must
466 * NULL is returned on a memory allocation failure.
468 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
,
469 unsigned int __nocast pri
)
471 might_sleep_if(pri
& __GFP_WAIT
);
472 if (skb_shared(skb
)) {
473 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
481 * Copy shared buffers into a new sk_buff. We effectively do COW on
482 * packets to handle cases where we have a local reader and forward
483 * and a couple of other messy ones. The normal one is tcpdumping
484 * a packet thats being forwarded.
488 * skb_unshare - make a copy of a shared buffer
489 * @skb: buffer to check
490 * @pri: priority for memory allocation
492 * If the socket buffer is a clone then this function creates a new
493 * copy of the data, drops a reference count on the old copy and returns
494 * the new copy with the reference count at 1. If the buffer is not a clone
495 * the original buffer is returned. When called with a spinlock held or
496 * from interrupt state @pri must be %GFP_ATOMIC
498 * %NULL is returned on a memory allocation failure.
500 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
501 unsigned int __nocast pri
)
503 might_sleep_if(pri
& __GFP_WAIT
);
504 if (skb_cloned(skb
)) {
505 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
506 kfree_skb(skb
); /* Free our shared copy */
514 * @list_: list to peek at
516 * Peek an &sk_buff. Unlike most other operations you _MUST_
517 * be careful with this one. A peek leaves the buffer on the
518 * list and someone else may run off with it. You must hold
519 * the appropriate locks or have a private queue to do this.
521 * Returns %NULL for an empty list or a pointer to the head element.
522 * The reference count is not incremented and the reference is therefore
523 * volatile. Use with caution.
525 static inline struct sk_buff
*skb_peek(struct sk_buff_head
*list_
)
527 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->next
;
528 if (list
== (struct sk_buff
*)list_
)
535 * @list_: list to peek at
537 * Peek an &sk_buff. Unlike most other operations you _MUST_
538 * be careful with this one. A peek leaves the buffer on the
539 * list and someone else may run off with it. You must hold
540 * the appropriate locks or have a private queue to do this.
542 * Returns %NULL for an empty list or a pointer to the tail element.
543 * The reference count is not incremented and the reference is therefore
544 * volatile. Use with caution.
546 static inline struct sk_buff
*skb_peek_tail(struct sk_buff_head
*list_
)
548 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->prev
;
549 if (list
== (struct sk_buff
*)list_
)
555 * skb_queue_len - get queue length
556 * @list_: list to measure
558 * Return the length of an &sk_buff queue.
560 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
565 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
567 spin_lock_init(&list
->lock
);
568 list
->prev
= list
->next
= (struct sk_buff
*)list
;
573 * Insert an sk_buff at the start of a list.
575 * The "__skb_xxxx()" functions are the non-atomic ones that
576 * can only be called with interrupts disabled.
580 * __skb_queue_head - queue a buffer at the list head
582 * @newsk: buffer to queue
584 * Queue a buffer at the start 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_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
590 static inline void __skb_queue_head(struct sk_buff_head
*list
,
591 struct sk_buff
*newsk
)
593 struct sk_buff
*prev
, *next
;
596 prev
= (struct sk_buff
*)list
;
600 next
->prev
= prev
->next
= newsk
;
604 * __skb_queue_tail - queue a buffer at the list tail
606 * @newsk: buffer to queue
608 * Queue a buffer at the end of a list. This function takes no locks
609 * and you must therefore hold required locks before calling it.
611 * A buffer cannot be placed on two lists at the same time.
613 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
614 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
615 struct sk_buff
*newsk
)
617 struct sk_buff
*prev
, *next
;
620 next
= (struct sk_buff
*)list
;
624 next
->prev
= prev
->next
= newsk
;
629 * __skb_dequeue - remove from the head of the queue
630 * @list: list to dequeue from
632 * Remove the head of the list. This function does not take any locks
633 * so must be used with appropriate locks held only. The head item is
634 * returned or %NULL if the list is empty.
636 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
637 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
639 struct sk_buff
*next
, *prev
, *result
;
641 prev
= (struct sk_buff
*) list
;
650 result
->next
= result
->prev
= NULL
;
657 * Insert a packet on a list.
659 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
660 static inline void __skb_insert(struct sk_buff
*newsk
,
661 struct sk_buff
*prev
, struct sk_buff
*next
,
662 struct sk_buff_head
*list
)
666 next
->prev
= prev
->next
= newsk
;
671 * Place a packet after a given packet in a list.
673 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
674 static inline void __skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
676 __skb_insert(newsk
, old
, old
->next
, list
);
680 * remove sk_buff from list. _Must_ be called atomically, and with
683 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
684 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
686 struct sk_buff
*next
, *prev
;
691 skb
->next
= skb
->prev
= NULL
;
697 /* XXX: more streamlined implementation */
700 * __skb_dequeue_tail - remove from the tail of the queue
701 * @list: list to dequeue from
703 * Remove the tail of the list. This function does not take any locks
704 * so must be used with appropriate locks held only. The tail item is
705 * returned or %NULL if the list is empty.
707 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
708 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
710 struct sk_buff
*skb
= skb_peek_tail(list
);
712 __skb_unlink(skb
, list
);
717 static inline int skb_is_nonlinear(const struct sk_buff
*skb
)
719 return skb
->data_len
;
722 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
724 return skb
->len
- skb
->data_len
;
727 static inline int skb_pagelen(const struct sk_buff
*skb
)
731 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
732 len
+= skb_shinfo(skb
)->frags
[i
].size
;
733 return len
+ skb_headlen(skb
);
736 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
737 struct page
*page
, int off
, int size
)
739 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
742 frag
->page_offset
= off
;
744 skb_shinfo(skb
)->nr_frags
= i
+ 1;
747 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
748 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
749 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
752 * Add data to an sk_buff
754 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
756 unsigned char *tmp
= skb
->tail
;
757 SKB_LINEAR_ASSERT(skb
);
764 * skb_put - add data to a buffer
765 * @skb: buffer to use
766 * @len: amount of data to add
768 * This function extends the used data area of the buffer. If this would
769 * exceed the total buffer size the kernel will panic. A pointer to the
770 * first byte of the extra data is returned.
772 static inline unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
)
774 unsigned char *tmp
= skb
->tail
;
775 SKB_LINEAR_ASSERT(skb
);
778 if (unlikely(skb
->tail
>skb
->end
))
779 skb_over_panic(skb
, len
, current_text_addr());
783 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
791 * skb_push - add data to the start of a buffer
792 * @skb: buffer to use
793 * @len: amount of data to add
795 * This function extends the used data area of the buffer at the buffer
796 * start. If this would exceed the total buffer headroom the kernel will
797 * panic. A pointer to the first byte of the extra data is returned.
799 static inline unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
)
803 if (unlikely(skb
->data
<skb
->head
))
804 skb_under_panic(skb
, len
, current_text_addr());
808 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
811 BUG_ON(skb
->len
< skb
->data_len
);
812 return skb
->data
+= len
;
816 * skb_pull - remove data from the start of a buffer
817 * @skb: buffer to use
818 * @len: amount of data to remove
820 * This function removes data from the start of a buffer, returning
821 * the memory to the headroom. A pointer to the next data in the buffer
822 * is returned. Once the data has been pulled future pushes will overwrite
825 static inline unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
)
827 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
830 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
832 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
834 if (len
> skb_headlen(skb
) &&
835 !__pskb_pull_tail(skb
, len
-skb_headlen(skb
)))
838 return skb
->data
+= len
;
841 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
843 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
846 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
848 if (likely(len
<= skb_headlen(skb
)))
850 if (unlikely(len
> skb
->len
))
852 return __pskb_pull_tail(skb
, len
-skb_headlen(skb
)) != NULL
;
856 * skb_headroom - bytes at buffer head
857 * @skb: buffer to check
859 * Return the number of bytes of free space at the head of an &sk_buff.
861 static inline int skb_headroom(const struct sk_buff
*skb
)
863 return skb
->data
- skb
->head
;
867 * skb_tailroom - bytes at buffer end
868 * @skb: buffer to check
870 * Return the number of bytes of free space at the tail of an sk_buff
872 static inline int skb_tailroom(const struct sk_buff
*skb
)
874 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
878 * skb_reserve - adjust headroom
879 * @skb: buffer to alter
880 * @len: bytes to move
882 * Increase the headroom of an empty &sk_buff by reducing the tail
883 * room. This is only allowed for an empty buffer.
885 static inline void skb_reserve(struct sk_buff
*skb
, unsigned int len
)
892 * CPUs often take a performance hit when accessing unaligned memory
893 * locations. The actual performance hit varies, it can be small if the
894 * hardware handles it or large if we have to take an exception and fix it
897 * Since an ethernet header is 14 bytes network drivers often end up with
898 * the IP header at an unaligned offset. The IP header can be aligned by
899 * shifting the start of the packet by 2 bytes. Drivers should do this
902 * skb_reserve(NET_IP_ALIGN);
904 * The downside to this alignment of the IP header is that the DMA is now
905 * unaligned. On some architectures the cost of an unaligned DMA is high
906 * and this cost outweighs the gains made by aligning the IP header.
908 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
912 #define NET_IP_ALIGN 2
915 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
, int realloc
);
917 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
919 if (!skb
->data_len
) {
921 skb
->tail
= skb
->data
+ len
;
923 ___pskb_trim(skb
, len
, 0);
927 * skb_trim - remove end from a buffer
928 * @skb: buffer to alter
931 * Cut the length of a buffer down by removing data from the tail. If
932 * the buffer is already under the length specified it is not modified.
934 static inline void skb_trim(struct sk_buff
*skb
, unsigned int len
)
937 __skb_trim(skb
, len
);
941 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
943 if (!skb
->data_len
) {
945 skb
->tail
= skb
->data
+len
;
948 return ___pskb_trim(skb
, len
, 1);
951 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
953 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
957 * skb_orphan - orphan a buffer
958 * @skb: buffer to orphan
960 * If a buffer currently has an owner then we call the owner's
961 * destructor function and make the @skb unowned. The buffer continues
962 * to exist but is no longer charged to its former owner.
964 static inline void skb_orphan(struct sk_buff
*skb
)
967 skb
->destructor(skb
);
968 skb
->destructor
= NULL
;
973 * __skb_queue_purge - empty a list
974 * @list: list to empty
976 * Delete all buffers on an &sk_buff list. Each buffer is removed from
977 * the list and one reference dropped. This function does not take the
978 * list lock and the caller must hold the relevant locks to use it.
980 extern void skb_queue_purge(struct sk_buff_head
*list
);
981 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
984 while ((skb
= __skb_dequeue(list
)) != NULL
)
988 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
990 * __dev_alloc_skb - allocate an skbuff for sending
991 * @length: length to allocate
992 * @gfp_mask: get_free_pages mask, passed to alloc_skb
994 * Allocate a new &sk_buff and assign it a usage count of one. The
995 * buffer has unspecified headroom built in. Users should allocate
996 * the headroom they think they need without accounting for the
997 * built in space. The built in space is used for optimisations.
999 * %NULL is returned in there is no free memory.
1001 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
1002 unsigned int __nocast gfp_mask
)
1004 struct sk_buff
*skb
= alloc_skb(length
+ 16, gfp_mask
);
1006 skb_reserve(skb
, 16);
1010 extern struct sk_buff
*__dev_alloc_skb(unsigned int length
, int gfp_mask
);
1014 * dev_alloc_skb - allocate an skbuff for sending
1015 * @length: length to allocate
1017 * Allocate a new &sk_buff and assign it a usage count of one. The
1018 * buffer has unspecified headroom built in. Users should allocate
1019 * the headroom they think they need without accounting for the
1020 * built in space. The built in space is used for optimisations.
1022 * %NULL is returned in there is no free memory. Although this function
1023 * allocates memory it can be called from an interrupt.
1025 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
1027 return __dev_alloc_skb(length
, GFP_ATOMIC
);
1031 * skb_cow - copy header of skb when it is required
1032 * @skb: buffer to cow
1033 * @headroom: needed headroom
1035 * If the skb passed lacks sufficient headroom or its data part
1036 * is shared, data is reallocated. If reallocation fails, an error
1037 * is returned and original skb is not changed.
1039 * The result is skb with writable area skb->head...skb->tail
1040 * and at least @headroom of space at head.
1042 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
1044 int delta
= (headroom
> 16 ? headroom
: 16) - skb_headroom(skb
);
1049 if (delta
|| skb_cloned(skb
))
1050 return pskb_expand_head(skb
, (delta
+ 15) & ~15, 0, GFP_ATOMIC
);
1055 * skb_padto - pad an skbuff up to a minimal size
1056 * @skb: buffer to pad
1057 * @len: minimal length
1059 * Pads up a buffer to ensure the trailing bytes exist and are
1060 * blanked. If the buffer already contains sufficient data it
1061 * is untouched. Returns the buffer, which may be a replacement
1062 * for the original, or NULL for out of memory - in which case
1063 * the original buffer is still freed.
1066 static inline struct sk_buff
*skb_padto(struct sk_buff
*skb
, unsigned int len
)
1068 unsigned int size
= skb
->len
;
1069 if (likely(size
>= len
))
1071 return skb_pad(skb
, len
-size
);
1074 static inline int skb_add_data(struct sk_buff
*skb
,
1075 char __user
*from
, int copy
)
1077 const int off
= skb
->len
;
1079 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1081 unsigned int csum
= csum_and_copy_from_user(from
,
1085 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
1088 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
1091 __skb_trim(skb
, off
);
1095 static inline int skb_can_coalesce(struct sk_buff
*skb
, int i
,
1096 struct page
*page
, int off
)
1099 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
1101 return page
== frag
->page
&&
1102 off
== frag
->page_offset
+ frag
->size
;
1108 * skb_linearize - convert paged skb to linear one
1109 * @skb: buffer to linarize
1110 * @gfp: allocation mode
1112 * If there is no free memory -ENOMEM is returned, otherwise zero
1113 * is returned and the old skb data released.
1115 extern int __skb_linearize(struct sk_buff
*skb
, unsigned int __nocast gfp
);
1116 static inline int skb_linearize(struct sk_buff
*skb
, unsigned int __nocast gfp
)
1118 return __skb_linearize(skb
, gfp
);
1122 * skb_postpull_rcsum - update checksum for received skb after pull
1123 * @skb: buffer to update
1124 * @start: start of data before pull
1125 * @len: length of data pulled
1127 * After doing a pull on a received packet, you need to call this to
1128 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1129 * so that it can be recomputed from scratch.
1132 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
1133 const void *start
, int len
)
1135 if (skb
->ip_summed
== CHECKSUM_HW
)
1136 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
1140 * pskb_trim_rcsum - trim received skb and update checksum
1141 * @skb: buffer to trim
1144 * This is exactly the same as pskb_trim except that it ensures the
1145 * checksum of received packets are still valid after the operation.
1148 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
1150 if (len
>= skb
->len
)
1152 if (skb
->ip_summed
== CHECKSUM_HW
)
1153 skb
->ip_summed
= CHECKSUM_NONE
;
1154 return __pskb_trim(skb
, len
);
1157 static inline void *kmap_skb_frag(const skb_frag_t
*frag
)
1159 #ifdef CONFIG_HIGHMEM
1164 return kmap_atomic(frag
->page
, KM_SKB_DATA_SOFTIRQ
);
1167 static inline void kunmap_skb_frag(void *vaddr
)
1169 kunmap_atomic(vaddr
, KM_SKB_DATA_SOFTIRQ
);
1170 #ifdef CONFIG_HIGHMEM
1175 #define skb_queue_walk(queue, skb) \
1176 for (skb = (queue)->next; \
1177 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1181 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1182 int noblock
, int *err
);
1183 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
1184 struct poll_table_struct
*wait
);
1185 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
1186 int offset
, struct iovec
*to
,
1188 extern int skb_copy_and_csum_datagram_iovec(const
1189 struct sk_buff
*skb
,
1192 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
1193 extern unsigned int skb_checksum(const struct sk_buff
*skb
, int offset
,
1194 int len
, unsigned int csum
);
1195 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
1197 extern int skb_store_bits(const struct sk_buff
*skb
, int offset
,
1198 void *from
, int len
);
1199 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff
*skb
,
1200 int offset
, u8
*to
, int len
,
1202 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
1203 extern void skb_split(struct sk_buff
*skb
,
1204 struct sk_buff
*skb1
, const u32 len
);
1206 extern void skb_release_data(struct sk_buff
*skb
);
1208 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
1209 int len
, void *buffer
)
1211 int hlen
= skb_headlen(skb
);
1213 if (hlen
- offset
>= len
)
1214 return skb
->data
+ offset
;
1216 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
1222 extern void skb_init(void);
1223 extern void skb_add_mtu(int mtu
);
1226 * skb_get_timestamp - get timestamp from a skb
1227 * @skb: skb to get stamp from
1228 * @stamp: pointer to struct timeval to store stamp in
1230 * Timestamps are stored in the skb as offsets to a base timestamp.
1231 * This function converts the offset back to a struct timeval and stores
1234 static inline void skb_get_timestamp(struct sk_buff
*skb
, struct timeval
*stamp
)
1236 stamp
->tv_sec
= skb
->tstamp
.off_sec
;
1237 stamp
->tv_usec
= skb
->tstamp
.off_usec
;
1238 if (skb
->tstamp
.off_sec
) {
1239 stamp
->tv_sec
+= skb_tv_base
.tv_sec
;
1240 stamp
->tv_usec
+= skb_tv_base
.tv_usec
;
1245 * skb_set_timestamp - set timestamp of a skb
1246 * @skb: skb to set stamp of
1247 * @stamp: pointer to struct timeval to get stamp from
1249 * Timestamps are stored in the skb as offsets to a base timestamp.
1250 * This function converts a struct timeval to an offset and stores
1253 static inline void skb_set_timestamp(struct sk_buff
*skb
, struct timeval
*stamp
)
1255 skb
->tstamp
.off_sec
= stamp
->tv_sec
- skb_tv_base
.tv_sec
;
1256 skb
->tstamp
.off_usec
= stamp
->tv_usec
- skb_tv_base
.tv_usec
;
1259 extern void __net_timestamp(struct sk_buff
*skb
);
1261 #ifdef CONFIG_NETFILTER
1262 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
1264 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
1265 nfct
->destroy(nfct
);
1267 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
1270 atomic_inc(&nfct
->use
);
1272 static inline void nf_reset(struct sk_buff
*skb
)
1274 nf_conntrack_put(skb
->nfct
);
1278 #ifdef CONFIG_BRIDGE_NETFILTER
1279 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
1281 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
1284 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
1287 atomic_inc(&nf_bridge
->use
);
1289 #endif /* CONFIG_BRIDGE_NETFILTER */
1290 #else /* CONFIG_NETFILTER */
1291 static inline void nf_reset(struct sk_buff
*skb
) {}
1292 #endif /* CONFIG_NETFILTER */
1294 #endif /* __KERNEL__ */
1295 #endif /* _LINUX_SKBUFF_H */