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/kernel.h>
18 #include <linux/compiler.h>
19 #include <linux/time.h>
20 #include <linux/cache.h>
22 #include <asm/atomic.h>
23 #include <asm/types.h>
24 #include <linux/spinlock.h>
25 #include <linux/net.h>
26 #include <linux/textsearch.h>
27 #include <net/checksum.h>
28 #include <linux/rcupdate.h>
29 #include <linux/dmaengine.h>
31 #define HAVE_ALLOC_SKB /* For the drivers to know */
32 #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
34 #define CHECKSUM_NONE 0
35 #define CHECKSUM_PARTIAL 1
36 #define CHECKSUM_UNNECESSARY 2
37 #define CHECKSUM_COMPLETE 3
39 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
40 ~(SMP_CACHE_BYTES - 1))
41 #define SKB_MAX_ORDER(X, ORDER) (((PAGE_SIZE << (ORDER)) - (X) - \
42 sizeof(struct skb_shared_info)) & \
43 ~(SMP_CACHE_BYTES - 1))
44 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
45 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
47 /* A. Checksumming of received packets by device.
49 * NONE: device failed to checksum this packet.
50 * skb->csum is undefined.
52 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
53 * skb->csum is undefined.
54 * It is bad option, but, unfortunately, many of vendors do this.
55 * Apparently with secret goal to sell you new device, when you
56 * will add new protocol to your host. F.e. IPv6. 8)
58 * COMPLETE: the most generic way. Device supplied checksum of _all_
59 * the packet as seen by netif_rx in skb->csum.
60 * NOTE: Even if device supports only some protocols, but
61 * is able to produce some skb->csum, it MUST use COMPLETE,
64 * B. Checksumming on output.
66 * NONE: skb is checksummed by protocol or csum is not required.
68 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
69 * from skb->h.raw to the end and to record the checksum
70 * at skb->h.raw+skb->csum.
72 * Device must show its capabilities in dev->features, set
73 * at device setup time.
74 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
76 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
77 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
78 * TCP/UDP over IPv4. Sigh. Vendors like this
79 * way by an unknown reason. Though, see comment above
80 * about CHECKSUM_UNNECESSARY. 8)
82 * Any questions? No questions, good. --ANK
87 #ifdef CONFIG_NETFILTER
90 void (*destroy
)(struct nf_conntrack
*);
93 #ifdef CONFIG_BRIDGE_NETFILTER
94 struct nf_bridge_info
{
96 struct net_device
*physindev
;
97 struct net_device
*physoutdev
;
98 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
99 struct net_device
*netoutdev
;
102 unsigned long data
[32 / sizeof(unsigned long)];
108 struct sk_buff_head
{
109 /* These two members must be first. */
110 struct sk_buff
*next
;
111 struct sk_buff
*prev
;
119 /* To allow 64K frame to be packed as single skb without frag_list */
120 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
122 typedef struct skb_frag_struct skb_frag_t
;
124 struct skb_frag_struct
{
130 /* This data is invariant across clones and lives at
131 * the end of the header data, ie. at skb->end.
133 struct skb_shared_info
{
135 unsigned short nr_frags
;
136 unsigned short gso_size
;
137 /* Warning: this field is not always filled in (UFO)! */
138 unsigned short gso_segs
;
139 unsigned short gso_type
;
141 struct sk_buff
*frag_list
;
142 skb_frag_t frags
[MAX_SKB_FRAGS
];
145 /* We divide dataref into two halves. The higher 16 bits hold references
146 * to the payload part of skb->data. The lower 16 bits hold references to
147 * the entire skb->data. It is up to the users of the skb to agree on
148 * where the payload starts.
150 * All users must obey the rule that the skb->data reference count must be
151 * greater than or equal to the payload reference count.
153 * Holding a reference to the payload part means that the user does not
154 * care about modifications to the header part of skb->data.
156 #define SKB_DATAREF_SHIFT 16
157 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
166 SKB_FCLONE_UNAVAILABLE
,
172 SKB_GSO_TCPV4
= 1 << 0,
173 SKB_GSO_UDP
= 1 << 1,
175 /* This indicates the skb is from an untrusted source. */
176 SKB_GSO_DODGY
= 1 << 2,
178 /* This indicates the tcp segment has CWR set. */
179 SKB_GSO_TCP_ECN
= 1 << 3,
181 SKB_GSO_TCPV6
= 1 << 4,
185 * struct sk_buff - socket buffer
186 * @next: Next buffer in list
187 * @prev: Previous buffer in list
188 * @sk: Socket we are owned by
189 * @tstamp: Time we arrived
190 * @dev: Device we arrived on/are leaving by
191 * @input_dev: Device we arrived on
192 * @h: Transport layer header
193 * @nh: Network layer header
194 * @mac: Link layer header
195 * @dst: destination entry
196 * @sp: the security path, used for xfrm
197 * @cb: Control buffer. Free for use by every layer. Put private vars here
198 * @len: Length of actual data
199 * @data_len: Data length
200 * @mac_len: Length of link layer header
202 * @local_df: allow local fragmentation
203 * @cloned: Head may be cloned (check refcnt to be sure)
204 * @nohdr: Payload reference only, must not modify header
205 * @pkt_type: Packet class
206 * @fclone: skbuff clone status
207 * @ip_summed: Driver fed us an IP checksum
208 * @priority: Packet queueing priority
209 * @users: User count - see {datagram,tcp}.c
210 * @protocol: Packet protocol from driver
211 * @truesize: Buffer size
212 * @head: Head of buffer
213 * @data: Data head pointer
214 * @tail: Tail pointer
216 * @destructor: Destruct function
217 * @mark: Generic packet mark
218 * @nfct: Associated connection, if any
219 * @ipvs_property: skbuff is owned by ipvs
220 * @nfctinfo: Relationship of this skb to the connection
221 * @nfct_reasm: netfilter conntrack re-assembly pointer
222 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
223 * @tc_index: Traffic control index
224 * @tc_verd: traffic control verdict
225 * @dma_cookie: a cookie to one of several possible DMA operations
226 * done by skb DMA functions
227 * @secmark: security marking
231 /* These two members must be first. */
232 struct sk_buff
*next
;
233 struct sk_buff
*prev
;
236 struct skb_timeval tstamp
;
237 struct net_device
*dev
;
238 struct net_device
*input_dev
;
243 struct icmphdr
*icmph
;
244 struct igmphdr
*igmph
;
246 struct ipv6hdr
*ipv6h
;
252 struct ipv6hdr
*ipv6h
;
261 struct dst_entry
*dst
;
265 * This is the control buffer. It is free to use for every
266 * layer. Please put your private variables there. If you
267 * want to keep them across layers you have to do a skb_clone()
268 * first. This is owned by whoever has the skb queued ATM.
290 void (*destructor
)(struct sk_buff
*skb
);
291 #ifdef CONFIG_NETFILTER
292 struct nf_conntrack
*nfct
;
293 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
294 struct sk_buff
*nfct_reasm
;
296 #ifdef CONFIG_BRIDGE_NETFILTER
297 struct nf_bridge_info
*nf_bridge
;
299 #endif /* CONFIG_NETFILTER */
300 #ifdef CONFIG_NET_SCHED
301 __u16 tc_index
; /* traffic control index */
302 #ifdef CONFIG_NET_CLS_ACT
303 __u16 tc_verd
; /* traffic control verdict */
306 #ifdef CONFIG_NET_DMA
307 dma_cookie_t dma_cookie
;
309 #ifdef CONFIG_NETWORK_SECMARK
315 /* These elements must be at the end, see alloc_skb() for details. */
316 unsigned int truesize
;
326 * Handling routines are only of interest to the kernel
328 #include <linux/slab.h>
330 #include <asm/system.h>
332 extern void kfree_skb(struct sk_buff
*skb
);
333 extern void __kfree_skb(struct sk_buff
*skb
);
334 extern struct sk_buff
*__alloc_skb(unsigned int size
,
335 gfp_t priority
, int fclone
, int node
);
336 static inline struct sk_buff
*alloc_skb(unsigned int size
,
339 return __alloc_skb(size
, priority
, 0, -1);
342 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
345 return __alloc_skb(size
, priority
, 1, -1);
348 extern struct sk_buff
*alloc_skb_from_cache(struct kmem_cache
*cp
,
351 extern void kfree_skbmem(struct sk_buff
*skb
);
352 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
354 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
356 extern struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
358 extern int pskb_expand_head(struct sk_buff
*skb
,
359 int nhead
, int ntail
,
361 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
362 unsigned int headroom
);
363 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
364 int newheadroom
, int newtailroom
,
366 extern int skb_pad(struct sk_buff
*skb
, int pad
);
367 #define dev_kfree_skb(a) kfree_skb(a)
368 extern void skb_over_panic(struct sk_buff
*skb
, int len
,
370 extern void skb_under_panic(struct sk_buff
*skb
, int len
,
372 extern void skb_truesize_bug(struct sk_buff
*skb
);
374 static inline void skb_truesize_check(struct sk_buff
*skb
)
376 if (unlikely((int)skb
->truesize
< sizeof(struct sk_buff
) + skb
->len
))
377 skb_truesize_bug(skb
);
380 extern int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
381 int getfrag(void *from
, char *to
, int offset
,
382 int len
,int odd
, struct sk_buff
*skb
),
383 void *from
, int length
);
390 __u32 stepped_offset
;
391 struct sk_buff
*root_skb
;
392 struct sk_buff
*cur_skb
;
396 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
397 unsigned int from
, unsigned int to
,
398 struct skb_seq_state
*st
);
399 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
400 struct skb_seq_state
*st
);
401 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
403 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
404 unsigned int to
, struct ts_config
*config
,
405 struct ts_state
*state
);
408 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
411 * skb_queue_empty - check if a queue is empty
414 * Returns true if the queue is empty, false otherwise.
416 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
418 return list
->next
== (struct sk_buff
*)list
;
422 * skb_get - reference buffer
423 * @skb: buffer to reference
425 * Makes another reference to a socket buffer and returns a pointer
428 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
430 atomic_inc(&skb
->users
);
435 * If users == 1, we are the only owner and are can avoid redundant
440 * skb_cloned - is the buffer a clone
441 * @skb: buffer to check
443 * Returns true if the buffer was generated with skb_clone() and is
444 * one of multiple shared copies of the buffer. Cloned buffers are
445 * shared data so must not be written to under normal circumstances.
447 static inline int skb_cloned(const struct sk_buff
*skb
)
449 return skb
->cloned
&&
450 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
454 * skb_header_cloned - is the header a clone
455 * @skb: buffer to check
457 * Returns true if modifying the header part of the buffer requires
458 * the data to be copied.
460 static inline int skb_header_cloned(const struct sk_buff
*skb
)
467 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
468 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
473 * skb_header_release - release reference to header
474 * @skb: buffer to operate on
476 * Drop a reference to the header part of the buffer. This is done
477 * by acquiring a payload reference. You must not read from the header
478 * part of skb->data after this.
480 static inline void skb_header_release(struct sk_buff
*skb
)
484 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
488 * skb_shared - is the buffer shared
489 * @skb: buffer to check
491 * Returns true if more than one person has a reference to this
494 static inline int skb_shared(const struct sk_buff
*skb
)
496 return atomic_read(&skb
->users
) != 1;
500 * skb_share_check - check if buffer is shared and if so clone it
501 * @skb: buffer to check
502 * @pri: priority for memory allocation
504 * If the buffer is shared the buffer is cloned and the old copy
505 * drops a reference. A new clone with a single reference is returned.
506 * If the buffer is not shared the original buffer is returned. When
507 * being called from interrupt status or with spinlocks held pri must
510 * NULL is returned on a memory allocation failure.
512 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
,
515 might_sleep_if(pri
& __GFP_WAIT
);
516 if (skb_shared(skb
)) {
517 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
525 * Copy shared buffers into a new sk_buff. We effectively do COW on
526 * packets to handle cases where we have a local reader and forward
527 * and a couple of other messy ones. The normal one is tcpdumping
528 * a packet thats being forwarded.
532 * skb_unshare - make a copy of a shared buffer
533 * @skb: buffer to check
534 * @pri: priority for memory allocation
536 * If the socket buffer is a clone then this function creates a new
537 * copy of the data, drops a reference count on the old copy and returns
538 * the new copy with the reference count at 1. If the buffer is not a clone
539 * the original buffer is returned. When called with a spinlock held or
540 * from interrupt state @pri must be %GFP_ATOMIC
542 * %NULL is returned on a memory allocation failure.
544 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
547 might_sleep_if(pri
& __GFP_WAIT
);
548 if (skb_cloned(skb
)) {
549 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
550 kfree_skb(skb
); /* Free our shared copy */
558 * @list_: list to peek at
560 * Peek an &sk_buff. Unlike most other operations you _MUST_
561 * be careful with this one. A peek leaves the buffer on the
562 * list and someone else may run off with it. You must hold
563 * the appropriate locks or have a private queue to do this.
565 * Returns %NULL for an empty list or a pointer to the head element.
566 * The reference count is not incremented and the reference is therefore
567 * volatile. Use with caution.
569 static inline struct sk_buff
*skb_peek(struct sk_buff_head
*list_
)
571 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->next
;
572 if (list
== (struct sk_buff
*)list_
)
579 * @list_: list to peek at
581 * Peek an &sk_buff. Unlike most other operations you _MUST_
582 * be careful with this one. A peek leaves the buffer on the
583 * list and someone else may run off with it. You must hold
584 * the appropriate locks or have a private queue to do this.
586 * Returns %NULL for an empty list or a pointer to the tail element.
587 * The reference count is not incremented and the reference is therefore
588 * volatile. Use with caution.
590 static inline struct sk_buff
*skb_peek_tail(struct sk_buff_head
*list_
)
592 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->prev
;
593 if (list
== (struct sk_buff
*)list_
)
599 * skb_queue_len - get queue length
600 * @list_: list to measure
602 * Return the length of an &sk_buff queue.
604 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
610 * This function creates a split out lock class for each invocation;
611 * this is needed for now since a whole lot of users of the skb-queue
612 * infrastructure in drivers have different locking usage (in hardirq)
613 * than the networking core (in softirq only). In the long run either the
614 * network layer or drivers should need annotation to consolidate the
615 * main types of usage into 3 classes.
617 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
619 spin_lock_init(&list
->lock
);
620 list
->prev
= list
->next
= (struct sk_buff
*)list
;
625 * Insert an sk_buff at the start of a list.
627 * The "__skb_xxxx()" functions are the non-atomic ones that
628 * can only be called with interrupts disabled.
632 * __skb_queue_after - queue a buffer at the list head
634 * @prev: place after this buffer
635 * @newsk: buffer to queue
637 * Queue a buffer int the middle of a list. This function takes no locks
638 * and you must therefore hold required locks before calling it.
640 * A buffer cannot be placed on two lists at the same time.
642 static inline void __skb_queue_after(struct sk_buff_head
*list
,
643 struct sk_buff
*prev
,
644 struct sk_buff
*newsk
)
646 struct sk_buff
*next
;
652 next
->prev
= prev
->next
= newsk
;
656 * __skb_queue_head - queue a buffer at the list head
658 * @newsk: buffer to queue
660 * Queue a buffer at the start of a list. This function takes no locks
661 * and you must therefore hold required locks before calling it.
663 * A buffer cannot be placed on two lists at the same time.
665 extern void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
666 static inline void __skb_queue_head(struct sk_buff_head
*list
,
667 struct sk_buff
*newsk
)
669 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
673 * __skb_queue_tail - queue a buffer at the list tail
675 * @newsk: buffer to queue
677 * Queue a buffer at the end of a list. This function takes no locks
678 * and you must therefore hold required locks before calling it.
680 * A buffer cannot be placed on two lists at the same time.
682 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
683 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
684 struct sk_buff
*newsk
)
686 struct sk_buff
*prev
, *next
;
689 next
= (struct sk_buff
*)list
;
693 next
->prev
= prev
->next
= newsk
;
698 * __skb_dequeue - remove from the head of the queue
699 * @list: list to dequeue from
701 * Remove the head of the list. This function does not take any locks
702 * so must be used with appropriate locks held only. The head item is
703 * returned or %NULL if the list is empty.
705 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
706 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
708 struct sk_buff
*next
, *prev
, *result
;
710 prev
= (struct sk_buff
*) list
;
719 result
->next
= result
->prev
= NULL
;
726 * Insert a packet on a list.
728 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
729 static inline void __skb_insert(struct sk_buff
*newsk
,
730 struct sk_buff
*prev
, struct sk_buff
*next
,
731 struct sk_buff_head
*list
)
735 next
->prev
= prev
->next
= newsk
;
740 * Place a packet after a given packet in a list.
742 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
743 static inline void __skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
745 __skb_insert(newsk
, old
, old
->next
, list
);
749 * remove sk_buff from list. _Must_ be called atomically, and with
752 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
753 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
755 struct sk_buff
*next
, *prev
;
760 skb
->next
= skb
->prev
= NULL
;
766 /* XXX: more streamlined implementation */
769 * __skb_dequeue_tail - remove from the tail of the queue
770 * @list: list to dequeue from
772 * Remove the tail of the list. This function does not take any locks
773 * so must be used with appropriate locks held only. The tail item is
774 * returned or %NULL if the list is empty.
776 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
777 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
779 struct sk_buff
*skb
= skb_peek_tail(list
);
781 __skb_unlink(skb
, list
);
786 static inline int skb_is_nonlinear(const struct sk_buff
*skb
)
788 return skb
->data_len
;
791 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
793 return skb
->len
- skb
->data_len
;
796 static inline int skb_pagelen(const struct sk_buff
*skb
)
800 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
801 len
+= skb_shinfo(skb
)->frags
[i
].size
;
802 return len
+ skb_headlen(skb
);
805 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
806 struct page
*page
, int off
, int size
)
808 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
811 frag
->page_offset
= off
;
813 skb_shinfo(skb
)->nr_frags
= i
+ 1;
816 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
817 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
818 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
821 * Add data to an sk_buff
823 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
825 unsigned char *tmp
= skb
->tail
;
826 SKB_LINEAR_ASSERT(skb
);
833 * skb_put - add data to a buffer
834 * @skb: buffer to use
835 * @len: amount of data to add
837 * This function extends the used data area of the buffer. If this would
838 * exceed the total buffer size the kernel will panic. A pointer to the
839 * first byte of the extra data is returned.
841 static inline unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
)
843 unsigned char *tmp
= skb
->tail
;
844 SKB_LINEAR_ASSERT(skb
);
847 if (unlikely(skb
->tail
>skb
->end
))
848 skb_over_panic(skb
, len
, current_text_addr());
852 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
860 * skb_push - add data to the start of a buffer
861 * @skb: buffer to use
862 * @len: amount of data to add
864 * This function extends the used data area of the buffer at the buffer
865 * start. If this would exceed the total buffer headroom the kernel will
866 * panic. A pointer to the first byte of the extra data is returned.
868 static inline unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
)
872 if (unlikely(skb
->data
<skb
->head
))
873 skb_under_panic(skb
, len
, current_text_addr());
877 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
880 BUG_ON(skb
->len
< skb
->data_len
);
881 return skb
->data
+= len
;
885 * skb_pull - remove data from the start of a buffer
886 * @skb: buffer to use
887 * @len: amount of data to remove
889 * This function removes data from the start of a buffer, returning
890 * the memory to the headroom. A pointer to the next data in the buffer
891 * is returned. Once the data has been pulled future pushes will overwrite
894 static inline unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
)
896 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
899 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
901 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
903 if (len
> skb_headlen(skb
) &&
904 !__pskb_pull_tail(skb
, len
-skb_headlen(skb
)))
907 return skb
->data
+= len
;
910 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
912 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
915 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
917 if (likely(len
<= skb_headlen(skb
)))
919 if (unlikely(len
> skb
->len
))
921 return __pskb_pull_tail(skb
, len
-skb_headlen(skb
)) != NULL
;
925 * skb_headroom - bytes at buffer head
926 * @skb: buffer to check
928 * Return the number of bytes of free space at the head of an &sk_buff.
930 static inline int skb_headroom(const struct sk_buff
*skb
)
932 return skb
->data
- skb
->head
;
936 * skb_tailroom - bytes at buffer end
937 * @skb: buffer to check
939 * Return the number of bytes of free space at the tail of an sk_buff
941 static inline int skb_tailroom(const struct sk_buff
*skb
)
943 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
947 * skb_reserve - adjust headroom
948 * @skb: buffer to alter
949 * @len: bytes to move
951 * Increase the headroom of an empty &sk_buff by reducing the tail
952 * room. This is only allowed for an empty buffer.
954 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
961 * CPUs often take a performance hit when accessing unaligned memory
962 * locations. The actual performance hit varies, it can be small if the
963 * hardware handles it or large if we have to take an exception and fix it
966 * Since an ethernet header is 14 bytes network drivers often end up with
967 * the IP header at an unaligned offset. The IP header can be aligned by
968 * shifting the start of the packet by 2 bytes. Drivers should do this
971 * skb_reserve(NET_IP_ALIGN);
973 * The downside to this alignment of the IP header is that the DMA is now
974 * unaligned. On some architectures the cost of an unaligned DMA is high
975 * and this cost outweighs the gains made by aligning the IP header.
977 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
981 #define NET_IP_ALIGN 2
985 * The networking layer reserves some headroom in skb data (via
986 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
987 * the header has to grow. In the default case, if the header has to grow
988 * 16 bytes or less we avoid the reallocation.
990 * Unfortunately this headroom changes the DMA alignment of the resulting
991 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
992 * on some architectures. An architecture can override this value,
993 * perhaps setting it to a cacheline in size (since that will maintain
994 * cacheline alignment of the DMA). It must be a power of 2.
996 * Various parts of the networking layer expect at least 16 bytes of
997 * headroom, you should not reduce this.
1000 #define NET_SKB_PAD 16
1003 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
1005 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
1007 if (unlikely(skb
->data_len
)) {
1012 skb
->tail
= skb
->data
+ len
;
1016 * skb_trim - remove end from a buffer
1017 * @skb: buffer to alter
1020 * Cut the length of a buffer down by removing data from the tail. If
1021 * the buffer is already under the length specified it is not modified.
1022 * The skb must be linear.
1024 static inline void skb_trim(struct sk_buff
*skb
, unsigned int len
)
1027 __skb_trim(skb
, len
);
1031 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1034 return ___pskb_trim(skb
, len
);
1035 __skb_trim(skb
, len
);
1039 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1041 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
1045 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1046 * @skb: buffer to alter
1049 * This is identical to pskb_trim except that the caller knows that
1050 * the skb is not cloned so we should never get an error due to out-
1053 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
1055 int err
= pskb_trim(skb
, len
);
1060 * skb_orphan - orphan a buffer
1061 * @skb: buffer to orphan
1063 * If a buffer currently has an owner then we call the owner's
1064 * destructor function and make the @skb unowned. The buffer continues
1065 * to exist but is no longer charged to its former owner.
1067 static inline void skb_orphan(struct sk_buff
*skb
)
1069 if (skb
->destructor
)
1070 skb
->destructor(skb
);
1071 skb
->destructor
= NULL
;
1076 * __skb_queue_purge - empty a list
1077 * @list: list to empty
1079 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1080 * the list and one reference dropped. This function does not take the
1081 * list lock and the caller must hold the relevant locks to use it.
1083 extern void skb_queue_purge(struct sk_buff_head
*list
);
1084 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
1086 struct sk_buff
*skb
;
1087 while ((skb
= __skb_dequeue(list
)) != NULL
)
1092 * __dev_alloc_skb - allocate an skbuff for receiving
1093 * @length: length to allocate
1094 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1096 * Allocate a new &sk_buff and assign it a usage count of one. The
1097 * buffer has unspecified headroom built in. Users should allocate
1098 * the headroom they think they need without accounting for the
1099 * built in space. The built in space is used for optimisations.
1101 * %NULL is returned if there is no free memory.
1103 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
1106 struct sk_buff
*skb
= alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
);
1108 skb_reserve(skb
, NET_SKB_PAD
);
1113 * dev_alloc_skb - allocate an skbuff for receiving
1114 * @length: length to allocate
1116 * Allocate a new &sk_buff and assign it a usage count of one. The
1117 * buffer has unspecified headroom built in. Users should allocate
1118 * the headroom they think they need without accounting for the
1119 * built in space. The built in space is used for optimisations.
1121 * %NULL is returned if there is no free memory. Although this function
1122 * allocates memory it can be called from an interrupt.
1124 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
1126 return __dev_alloc_skb(length
, GFP_ATOMIC
);
1129 extern struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
1130 unsigned int length
, gfp_t gfp_mask
);
1133 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1134 * @dev: network device to receive on
1135 * @length: length to allocate
1137 * Allocate a new &sk_buff and assign it a usage count of one. The
1138 * buffer has unspecified headroom built in. Users should allocate
1139 * the headroom they think they need without accounting for the
1140 * built in space. The built in space is used for optimisations.
1142 * %NULL is returned if there is no free memory. Although this function
1143 * allocates memory it can be called from an interrupt.
1145 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
1146 unsigned int length
)
1148 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
1152 * skb_cow - copy header of skb when it is required
1153 * @skb: buffer to cow
1154 * @headroom: needed headroom
1156 * If the skb passed lacks sufficient headroom or its data part
1157 * is shared, data is reallocated. If reallocation fails, an error
1158 * is returned and original skb is not changed.
1160 * The result is skb with writable area skb->head...skb->tail
1161 * and at least @headroom of space at head.
1163 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
1165 int delta
= (headroom
> NET_SKB_PAD
? headroom
: NET_SKB_PAD
) -
1171 if (delta
|| skb_cloned(skb
))
1172 return pskb_expand_head(skb
, (delta
+ (NET_SKB_PAD
-1)) &
1173 ~(NET_SKB_PAD
-1), 0, GFP_ATOMIC
);
1178 * skb_padto - pad an skbuff up to a minimal size
1179 * @skb: buffer to pad
1180 * @len: minimal length
1182 * Pads up a buffer to ensure the trailing bytes exist and are
1183 * blanked. If the buffer already contains sufficient data it
1184 * is untouched. Otherwise it is extended. Returns zero on
1185 * success. The skb is freed on error.
1188 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
1190 unsigned int size
= skb
->len
;
1191 if (likely(size
>= len
))
1193 return skb_pad(skb
, len
-size
);
1196 static inline int skb_add_data(struct sk_buff
*skb
,
1197 char __user
*from
, int copy
)
1199 const int off
= skb
->len
;
1201 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1203 __wsum csum
= csum_and_copy_from_user(from
, skb_put(skb
, copy
),
1206 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
1209 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
1212 __skb_trim(skb
, off
);
1216 static inline int skb_can_coalesce(struct sk_buff
*skb
, int i
,
1217 struct page
*page
, int off
)
1220 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
1222 return page
== frag
->page
&&
1223 off
== frag
->page_offset
+ frag
->size
;
1228 static inline int __skb_linearize(struct sk_buff
*skb
)
1230 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
1234 * skb_linearize - convert paged skb to linear one
1235 * @skb: buffer to linarize
1237 * If there is no free memory -ENOMEM is returned, otherwise zero
1238 * is returned and the old skb data released.
1240 static inline int skb_linearize(struct sk_buff
*skb
)
1242 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
1246 * skb_linearize_cow - make sure skb is linear and writable
1247 * @skb: buffer to process
1249 * If there is no free memory -ENOMEM is returned, otherwise zero
1250 * is returned and the old skb data released.
1252 static inline int skb_linearize_cow(struct sk_buff
*skb
)
1254 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
1255 __skb_linearize(skb
) : 0;
1259 * skb_postpull_rcsum - update checksum for received skb after pull
1260 * @skb: buffer to update
1261 * @start: start of data before pull
1262 * @len: length of data pulled
1264 * After doing a pull on a received packet, you need to call this to
1265 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1266 * CHECKSUM_NONE so that it can be recomputed from scratch.
1269 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
1270 const void *start
, unsigned int len
)
1272 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1273 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
1276 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
1279 * pskb_trim_rcsum - trim received skb and update checksum
1280 * @skb: buffer to trim
1283 * This is exactly the same as pskb_trim except that it ensures the
1284 * checksum of received packets are still valid after the operation.
1287 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
1289 if (likely(len
>= skb
->len
))
1291 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1292 skb
->ip_summed
= CHECKSUM_NONE
;
1293 return __pskb_trim(skb
, len
);
1296 #define skb_queue_walk(queue, skb) \
1297 for (skb = (queue)->next; \
1298 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1301 #define skb_queue_reverse_walk(queue, skb) \
1302 for (skb = (queue)->prev; \
1303 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1307 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1308 int noblock
, int *err
);
1309 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
1310 struct poll_table_struct
*wait
);
1311 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
1312 int offset
, struct iovec
*to
,
1314 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
,
1317 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
1318 extern void skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
,
1319 unsigned int flags
);
1320 extern __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
,
1321 int len
, __wsum csum
);
1322 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
1324 extern int skb_store_bits(const struct sk_buff
*skb
, int offset
,
1325 void *from
, int len
);
1326 extern __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
,
1327 int offset
, u8
*to
, int len
,
1329 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
1330 extern void skb_split(struct sk_buff
*skb
,
1331 struct sk_buff
*skb1
, const u32 len
);
1333 extern struct sk_buff
*skb_segment(struct sk_buff
*skb
, int features
);
1335 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
1336 int len
, void *buffer
)
1338 int hlen
= skb_headlen(skb
);
1340 if (hlen
- offset
>= len
)
1341 return skb
->data
+ offset
;
1343 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
1349 extern void skb_init(void);
1350 extern void skb_add_mtu(int mtu
);
1353 * skb_get_timestamp - get timestamp from a skb
1354 * @skb: skb to get stamp from
1355 * @stamp: pointer to struct timeval to store stamp in
1357 * Timestamps are stored in the skb as offsets to a base timestamp.
1358 * This function converts the offset back to a struct timeval and stores
1361 static inline void skb_get_timestamp(const struct sk_buff
*skb
, struct timeval
*stamp
)
1363 stamp
->tv_sec
= skb
->tstamp
.off_sec
;
1364 stamp
->tv_usec
= skb
->tstamp
.off_usec
;
1368 * skb_set_timestamp - set timestamp of a skb
1369 * @skb: skb to set stamp of
1370 * @stamp: pointer to struct timeval to get stamp from
1372 * Timestamps are stored in the skb as offsets to a base timestamp.
1373 * This function converts a struct timeval to an offset and stores
1376 static inline void skb_set_timestamp(struct sk_buff
*skb
, const struct timeval
*stamp
)
1378 skb
->tstamp
.off_sec
= stamp
->tv_sec
;
1379 skb
->tstamp
.off_usec
= stamp
->tv_usec
;
1382 extern void __net_timestamp(struct sk_buff
*skb
);
1384 extern __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
1387 * skb_checksum_complete - Calculate checksum of an entire packet
1388 * @skb: packet to process
1390 * This function calculates the checksum over the entire packet plus
1391 * the value of skb->csum. The latter can be used to supply the
1392 * checksum of a pseudo header as used by TCP/UDP. It returns the
1395 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1396 * this function can be used to verify that checksum on received
1397 * packets. In that case the function should return zero if the
1398 * checksum is correct. In particular, this function will return zero
1399 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1400 * hardware has already verified the correctness of the checksum.
1402 static inline unsigned int skb_checksum_complete(struct sk_buff
*skb
)
1404 return skb
->ip_summed
!= CHECKSUM_UNNECESSARY
&&
1405 __skb_checksum_complete(skb
);
1408 #ifdef CONFIG_NETFILTER
1409 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
1411 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
1412 nfct
->destroy(nfct
);
1414 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
1417 atomic_inc(&nfct
->use
);
1419 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1420 static inline void nf_conntrack_get_reasm(struct sk_buff
*skb
)
1423 atomic_inc(&skb
->users
);
1425 static inline void nf_conntrack_put_reasm(struct sk_buff
*skb
)
1431 #ifdef CONFIG_BRIDGE_NETFILTER
1432 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
1434 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
1437 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
1440 atomic_inc(&nf_bridge
->use
);
1442 #endif /* CONFIG_BRIDGE_NETFILTER */
1443 static inline void nf_reset(struct sk_buff
*skb
)
1445 nf_conntrack_put(skb
->nfct
);
1447 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1448 nf_conntrack_put_reasm(skb
->nfct_reasm
);
1449 skb
->nfct_reasm
= NULL
;
1451 #ifdef CONFIG_BRIDGE_NETFILTER
1452 nf_bridge_put(skb
->nf_bridge
);
1453 skb
->nf_bridge
= NULL
;
1457 #else /* CONFIG_NETFILTER */
1458 static inline void nf_reset(struct sk_buff
*skb
) {}
1459 #endif /* CONFIG_NETFILTER */
1461 #ifdef CONFIG_NETWORK_SECMARK
1462 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
1464 to
->secmark
= from
->secmark
;
1467 static inline void skb_init_secmark(struct sk_buff
*skb
)
1472 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
1475 static inline void skb_init_secmark(struct sk_buff
*skb
)
1479 static inline int skb_is_gso(const struct sk_buff
*skb
)
1481 return skb_shinfo(skb
)->gso_size
;
1484 #endif /* __KERNEL__ */
1485 #endif /* _LINUX_SKBUFF_H */