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 * @iif: ifindex of 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
;
239 /* 4 byte hole on 64 bit*/
244 struct icmphdr
*icmph
;
245 struct igmphdr
*igmph
;
247 struct ipv6hdr
*ipv6h
;
253 struct ipv6hdr
*ipv6h
;
262 struct dst_entry
*dst
;
266 * This is the control buffer. It is free to use for every
267 * layer. Please put your private variables there. If you
268 * want to keep them across layers you have to do a skb_clone()
269 * first. This is owned by whoever has the skb queued ATM.
291 void (*destructor
)(struct sk_buff
*skb
);
292 #ifdef CONFIG_NETFILTER
293 struct nf_conntrack
*nfct
;
294 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
295 struct sk_buff
*nfct_reasm
;
297 #ifdef CONFIG_BRIDGE_NETFILTER
298 struct nf_bridge_info
*nf_bridge
;
300 #endif /* CONFIG_NETFILTER */
301 #ifdef CONFIG_NET_SCHED
302 __u16 tc_index
; /* traffic control index */
303 #ifdef CONFIG_NET_CLS_ACT
304 __u16 tc_verd
; /* traffic control verdict */
307 #ifdef CONFIG_NET_DMA
308 dma_cookie_t dma_cookie
;
310 #ifdef CONFIG_NETWORK_SECMARK
316 /* These elements must be at the end, see alloc_skb() for details. */
317 unsigned int truesize
;
327 * Handling routines are only of interest to the kernel
329 #include <linux/slab.h>
331 #include <asm/system.h>
333 extern void kfree_skb(struct sk_buff
*skb
);
334 extern void __kfree_skb(struct sk_buff
*skb
);
335 extern struct sk_buff
*__alloc_skb(unsigned int size
,
336 gfp_t priority
, int fclone
, int node
);
337 static inline struct sk_buff
*alloc_skb(unsigned int size
,
340 return __alloc_skb(size
, priority
, 0, -1);
343 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
346 return __alloc_skb(size
, priority
, 1, -1);
349 extern void kfree_skbmem(struct sk_buff
*skb
);
350 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
352 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
354 extern struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
356 extern int pskb_expand_head(struct sk_buff
*skb
,
357 int nhead
, int ntail
,
359 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
360 unsigned int headroom
);
361 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
362 int newheadroom
, int newtailroom
,
364 extern int skb_pad(struct sk_buff
*skb
, int pad
);
365 #define dev_kfree_skb(a) kfree_skb(a)
366 extern void skb_over_panic(struct sk_buff
*skb
, int len
,
368 extern void skb_under_panic(struct sk_buff
*skb
, int len
,
370 extern void skb_truesize_bug(struct sk_buff
*skb
);
372 static inline void skb_truesize_check(struct sk_buff
*skb
)
374 if (unlikely((int)skb
->truesize
< sizeof(struct sk_buff
) + skb
->len
))
375 skb_truesize_bug(skb
);
378 extern int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
379 int getfrag(void *from
, char *to
, int offset
,
380 int len
,int odd
, struct sk_buff
*skb
),
381 void *from
, int length
);
388 __u32 stepped_offset
;
389 struct sk_buff
*root_skb
;
390 struct sk_buff
*cur_skb
;
394 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
395 unsigned int from
, unsigned int to
,
396 struct skb_seq_state
*st
);
397 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
398 struct skb_seq_state
*st
);
399 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
401 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
402 unsigned int to
, struct ts_config
*config
,
403 struct ts_state
*state
);
406 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
409 * skb_queue_empty - check if a queue is empty
412 * Returns true if the queue is empty, false otherwise.
414 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
416 return list
->next
== (struct sk_buff
*)list
;
420 * skb_get - reference buffer
421 * @skb: buffer to reference
423 * Makes another reference to a socket buffer and returns a pointer
426 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
428 atomic_inc(&skb
->users
);
433 * If users == 1, we are the only owner and are can avoid redundant
438 * skb_cloned - is the buffer a clone
439 * @skb: buffer to check
441 * Returns true if the buffer was generated with skb_clone() and is
442 * one of multiple shared copies of the buffer. Cloned buffers are
443 * shared data so must not be written to under normal circumstances.
445 static inline int skb_cloned(const struct sk_buff
*skb
)
447 return skb
->cloned
&&
448 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
452 * skb_header_cloned - is the header a clone
453 * @skb: buffer to check
455 * Returns true if modifying the header part of the buffer requires
456 * the data to be copied.
458 static inline int skb_header_cloned(const struct sk_buff
*skb
)
465 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
466 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
471 * skb_header_release - release reference to header
472 * @skb: buffer to operate on
474 * Drop a reference to the header part of the buffer. This is done
475 * by acquiring a payload reference. You must not read from the header
476 * part of skb->data after this.
478 static inline void skb_header_release(struct sk_buff
*skb
)
482 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
486 * skb_shared - is the buffer shared
487 * @skb: buffer to check
489 * Returns true if more than one person has a reference to this
492 static inline int skb_shared(const struct sk_buff
*skb
)
494 return atomic_read(&skb
->users
) != 1;
498 * skb_share_check - check if buffer is shared and if so clone it
499 * @skb: buffer to check
500 * @pri: priority for memory allocation
502 * If the buffer is shared the buffer is cloned and the old copy
503 * drops a reference. A new clone with a single reference is returned.
504 * If the buffer is not shared the original buffer is returned. When
505 * being called from interrupt status or with spinlocks held pri must
508 * NULL is returned on a memory allocation failure.
510 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
,
513 might_sleep_if(pri
& __GFP_WAIT
);
514 if (skb_shared(skb
)) {
515 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
523 * Copy shared buffers into a new sk_buff. We effectively do COW on
524 * packets to handle cases where we have a local reader and forward
525 * and a couple of other messy ones. The normal one is tcpdumping
526 * a packet thats being forwarded.
530 * skb_unshare - make a copy of a shared buffer
531 * @skb: buffer to check
532 * @pri: priority for memory allocation
534 * If the socket buffer is a clone then this function creates a new
535 * copy of the data, drops a reference count on the old copy and returns
536 * the new copy with the reference count at 1. If the buffer is not a clone
537 * the original buffer is returned. When called with a spinlock held or
538 * from interrupt state @pri must be %GFP_ATOMIC
540 * %NULL is returned on a memory allocation failure.
542 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
545 might_sleep_if(pri
& __GFP_WAIT
);
546 if (skb_cloned(skb
)) {
547 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
548 kfree_skb(skb
); /* Free our shared copy */
556 * @list_: list to peek at
558 * Peek an &sk_buff. Unlike most other operations you _MUST_
559 * be careful with this one. A peek leaves the buffer on the
560 * list and someone else may run off with it. You must hold
561 * the appropriate locks or have a private queue to do this.
563 * Returns %NULL for an empty list or a pointer to the head element.
564 * The reference count is not incremented and the reference is therefore
565 * volatile. Use with caution.
567 static inline struct sk_buff
*skb_peek(struct sk_buff_head
*list_
)
569 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->next
;
570 if (list
== (struct sk_buff
*)list_
)
577 * @list_: list to peek at
579 * Peek an &sk_buff. Unlike most other operations you _MUST_
580 * be careful with this one. A peek leaves the buffer on the
581 * list and someone else may run off with it. You must hold
582 * the appropriate locks or have a private queue to do this.
584 * Returns %NULL for an empty list or a pointer to the tail element.
585 * The reference count is not incremented and the reference is therefore
586 * volatile. Use with caution.
588 static inline struct sk_buff
*skb_peek_tail(struct sk_buff_head
*list_
)
590 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->prev
;
591 if (list
== (struct sk_buff
*)list_
)
597 * skb_queue_len - get queue length
598 * @list_: list to measure
600 * Return the length of an &sk_buff queue.
602 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
608 * This function creates a split out lock class for each invocation;
609 * this is needed for now since a whole lot of users of the skb-queue
610 * infrastructure in drivers have different locking usage (in hardirq)
611 * than the networking core (in softirq only). In the long run either the
612 * network layer or drivers should need annotation to consolidate the
613 * main types of usage into 3 classes.
615 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
617 spin_lock_init(&list
->lock
);
618 list
->prev
= list
->next
= (struct sk_buff
*)list
;
622 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
623 struct lock_class_key
*class)
625 skb_queue_head_init(list
);
626 lockdep_set_class(&list
->lock
, class);
630 * Insert an sk_buff at the start of a list.
632 * The "__skb_xxxx()" functions are the non-atomic ones that
633 * can only be called with interrupts disabled.
637 * __skb_queue_after - queue a buffer at the list head
639 * @prev: place after this buffer
640 * @newsk: buffer to queue
642 * Queue a buffer int the middle of a list. This function takes no locks
643 * and you must therefore hold required locks before calling it.
645 * A buffer cannot be placed on two lists at the same time.
647 static inline void __skb_queue_after(struct sk_buff_head
*list
,
648 struct sk_buff
*prev
,
649 struct sk_buff
*newsk
)
651 struct sk_buff
*next
;
657 next
->prev
= prev
->next
= newsk
;
661 * __skb_queue_head - queue a buffer at the list head
663 * @newsk: buffer to queue
665 * Queue a buffer at the start of a list. This function takes no locks
666 * and you must therefore hold required locks before calling it.
668 * A buffer cannot be placed on two lists at the same time.
670 extern void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
671 static inline void __skb_queue_head(struct sk_buff_head
*list
,
672 struct sk_buff
*newsk
)
674 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
678 * __skb_queue_tail - queue a buffer at the list tail
680 * @newsk: buffer to queue
682 * Queue a buffer at the end of a list. This function takes no locks
683 * and you must therefore hold required locks before calling it.
685 * A buffer cannot be placed on two lists at the same time.
687 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
688 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
689 struct sk_buff
*newsk
)
691 struct sk_buff
*prev
, *next
;
694 next
= (struct sk_buff
*)list
;
698 next
->prev
= prev
->next
= newsk
;
703 * __skb_dequeue - remove from the head of the queue
704 * @list: list to dequeue from
706 * Remove the head of the list. This function does not take any locks
707 * so must be used with appropriate locks held only. The head item is
708 * returned or %NULL if the list is empty.
710 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
711 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
713 struct sk_buff
*next
, *prev
, *result
;
715 prev
= (struct sk_buff
*) list
;
724 result
->next
= result
->prev
= NULL
;
731 * Insert a packet on a list.
733 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
734 static inline void __skb_insert(struct sk_buff
*newsk
,
735 struct sk_buff
*prev
, struct sk_buff
*next
,
736 struct sk_buff_head
*list
)
740 next
->prev
= prev
->next
= newsk
;
745 * Place a packet after a given packet in a list.
747 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
748 static inline void __skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
750 __skb_insert(newsk
, old
, old
->next
, list
);
754 * remove sk_buff from list. _Must_ be called atomically, and with
757 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
758 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
760 struct sk_buff
*next
, *prev
;
765 skb
->next
= skb
->prev
= NULL
;
771 /* XXX: more streamlined implementation */
774 * __skb_dequeue_tail - remove from the tail of the queue
775 * @list: list to dequeue from
777 * Remove the tail of the list. This function does not take any locks
778 * so must be used with appropriate locks held only. The tail item is
779 * returned or %NULL if the list is empty.
781 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
782 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
784 struct sk_buff
*skb
= skb_peek_tail(list
);
786 __skb_unlink(skb
, list
);
791 static inline int skb_is_nonlinear(const struct sk_buff
*skb
)
793 return skb
->data_len
;
796 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
798 return skb
->len
- skb
->data_len
;
801 static inline int skb_pagelen(const struct sk_buff
*skb
)
805 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
806 len
+= skb_shinfo(skb
)->frags
[i
].size
;
807 return len
+ skb_headlen(skb
);
810 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
811 struct page
*page
, int off
, int size
)
813 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
816 frag
->page_offset
= off
;
818 skb_shinfo(skb
)->nr_frags
= i
+ 1;
821 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
822 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
823 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
826 * Add data to an sk_buff
828 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
830 unsigned char *tmp
= skb
->tail
;
831 SKB_LINEAR_ASSERT(skb
);
838 * skb_put - add data to a buffer
839 * @skb: buffer to use
840 * @len: amount of data to add
842 * This function extends the used data area of the buffer. If this would
843 * exceed the total buffer size the kernel will panic. A pointer to the
844 * first byte of the extra data is returned.
846 static inline unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
)
848 unsigned char *tmp
= skb
->tail
;
849 SKB_LINEAR_ASSERT(skb
);
852 if (unlikely(skb
->tail
>skb
->end
))
853 skb_over_panic(skb
, len
, current_text_addr());
857 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
865 * skb_push - add data to the start of a buffer
866 * @skb: buffer to use
867 * @len: amount of data to add
869 * This function extends the used data area of the buffer at the buffer
870 * start. If this would exceed the total buffer headroom the kernel will
871 * panic. A pointer to the first byte of the extra data is returned.
873 static inline unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
)
877 if (unlikely(skb
->data
<skb
->head
))
878 skb_under_panic(skb
, len
, current_text_addr());
882 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
885 BUG_ON(skb
->len
< skb
->data_len
);
886 return skb
->data
+= len
;
890 * skb_pull - remove data from the start of a buffer
891 * @skb: buffer to use
892 * @len: amount of data to remove
894 * This function removes data from the start of a buffer, returning
895 * the memory to the headroom. A pointer to the next data in the buffer
896 * is returned. Once the data has been pulled future pushes will overwrite
899 static inline unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
)
901 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
904 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
906 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
908 if (len
> skb_headlen(skb
) &&
909 !__pskb_pull_tail(skb
, len
-skb_headlen(skb
)))
912 return skb
->data
+= len
;
915 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
917 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
920 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
922 if (likely(len
<= skb_headlen(skb
)))
924 if (unlikely(len
> skb
->len
))
926 return __pskb_pull_tail(skb
, len
-skb_headlen(skb
)) != NULL
;
930 * skb_headroom - bytes at buffer head
931 * @skb: buffer to check
933 * Return the number of bytes of free space at the head of an &sk_buff.
935 static inline int skb_headroom(const struct sk_buff
*skb
)
937 return skb
->data
- skb
->head
;
941 * skb_tailroom - bytes at buffer end
942 * @skb: buffer to check
944 * Return the number of bytes of free space at the tail of an sk_buff
946 static inline int skb_tailroom(const struct sk_buff
*skb
)
948 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
952 * skb_reserve - adjust headroom
953 * @skb: buffer to alter
954 * @len: bytes to move
956 * Increase the headroom of an empty &sk_buff by reducing the tail
957 * room. This is only allowed for an empty buffer.
959 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
966 * CPUs often take a performance hit when accessing unaligned memory
967 * locations. The actual performance hit varies, it can be small if the
968 * hardware handles it or large if we have to take an exception and fix it
971 * Since an ethernet header is 14 bytes network drivers often end up with
972 * the IP header at an unaligned offset. The IP header can be aligned by
973 * shifting the start of the packet by 2 bytes. Drivers should do this
976 * skb_reserve(NET_IP_ALIGN);
978 * The downside to this alignment of the IP header is that the DMA is now
979 * unaligned. On some architectures the cost of an unaligned DMA is high
980 * and this cost outweighs the gains made by aligning the IP header.
982 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
986 #define NET_IP_ALIGN 2
990 * The networking layer reserves some headroom in skb data (via
991 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
992 * the header has to grow. In the default case, if the header has to grow
993 * 16 bytes or less we avoid the reallocation.
995 * Unfortunately this headroom changes the DMA alignment of the resulting
996 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
997 * on some architectures. An architecture can override this value,
998 * perhaps setting it to a cacheline in size (since that will maintain
999 * cacheline alignment of the DMA). It must be a power of 2.
1001 * Various parts of the networking layer expect at least 16 bytes of
1002 * headroom, you should not reduce this.
1005 #define NET_SKB_PAD 16
1008 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
1010 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
1012 if (unlikely(skb
->data_len
)) {
1017 skb
->tail
= skb
->data
+ len
;
1021 * skb_trim - remove end from a buffer
1022 * @skb: buffer to alter
1025 * Cut the length of a buffer down by removing data from the tail. If
1026 * the buffer is already under the length specified it is not modified.
1027 * The skb must be linear.
1029 static inline void skb_trim(struct sk_buff
*skb
, unsigned int len
)
1032 __skb_trim(skb
, len
);
1036 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1039 return ___pskb_trim(skb
, len
);
1040 __skb_trim(skb
, len
);
1044 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1046 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
1050 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1051 * @skb: buffer to alter
1054 * This is identical to pskb_trim except that the caller knows that
1055 * the skb is not cloned so we should never get an error due to out-
1058 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
1060 int err
= pskb_trim(skb
, len
);
1065 * skb_orphan - orphan a buffer
1066 * @skb: buffer to orphan
1068 * If a buffer currently has an owner then we call the owner's
1069 * destructor function and make the @skb unowned. The buffer continues
1070 * to exist but is no longer charged to its former owner.
1072 static inline void skb_orphan(struct sk_buff
*skb
)
1074 if (skb
->destructor
)
1075 skb
->destructor(skb
);
1076 skb
->destructor
= NULL
;
1081 * __skb_queue_purge - empty a list
1082 * @list: list to empty
1084 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1085 * the list and one reference dropped. This function does not take the
1086 * list lock and the caller must hold the relevant locks to use it.
1088 extern void skb_queue_purge(struct sk_buff_head
*list
);
1089 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
1091 struct sk_buff
*skb
;
1092 while ((skb
= __skb_dequeue(list
)) != NULL
)
1097 * __dev_alloc_skb - allocate an skbuff for receiving
1098 * @length: length to allocate
1099 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1101 * Allocate a new &sk_buff and assign it a usage count of one. The
1102 * buffer has unspecified headroom built in. Users should allocate
1103 * the headroom they think they need without accounting for the
1104 * built in space. The built in space is used for optimisations.
1106 * %NULL is returned if there is no free memory.
1108 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
1111 struct sk_buff
*skb
= alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
);
1113 skb_reserve(skb
, NET_SKB_PAD
);
1118 * dev_alloc_skb - allocate an skbuff for receiving
1119 * @length: length to allocate
1121 * Allocate a new &sk_buff and assign it a usage count of one. The
1122 * buffer has unspecified headroom built in. Users should allocate
1123 * the headroom they think they need without accounting for the
1124 * built in space. The built in space is used for optimisations.
1126 * %NULL is returned if there is no free memory. Although this function
1127 * allocates memory it can be called from an interrupt.
1129 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
1131 return __dev_alloc_skb(length
, GFP_ATOMIC
);
1134 extern struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
1135 unsigned int length
, gfp_t gfp_mask
);
1138 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1139 * @dev: network device to receive on
1140 * @length: length to allocate
1142 * Allocate a new &sk_buff and assign it a usage count of one. The
1143 * buffer has unspecified headroom built in. Users should allocate
1144 * the headroom they think they need without accounting for the
1145 * built in space. The built in space is used for optimisations.
1147 * %NULL is returned if there is no free memory. Although this function
1148 * allocates memory it can be called from an interrupt.
1150 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
1151 unsigned int length
)
1153 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
1157 * skb_cow - copy header of skb when it is required
1158 * @skb: buffer to cow
1159 * @headroom: needed headroom
1161 * If the skb passed lacks sufficient headroom or its data part
1162 * is shared, data is reallocated. If reallocation fails, an error
1163 * is returned and original skb is not changed.
1165 * The result is skb with writable area skb->head...skb->tail
1166 * and at least @headroom of space at head.
1168 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
1170 int delta
= (headroom
> NET_SKB_PAD
? headroom
: NET_SKB_PAD
) -
1176 if (delta
|| skb_cloned(skb
))
1177 return pskb_expand_head(skb
, (delta
+ (NET_SKB_PAD
-1)) &
1178 ~(NET_SKB_PAD
-1), 0, GFP_ATOMIC
);
1183 * skb_padto - pad an skbuff up to a minimal size
1184 * @skb: buffer to pad
1185 * @len: minimal length
1187 * Pads up a buffer to ensure the trailing bytes exist and are
1188 * blanked. If the buffer already contains sufficient data it
1189 * is untouched. Otherwise it is extended. Returns zero on
1190 * success. The skb is freed on error.
1193 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
1195 unsigned int size
= skb
->len
;
1196 if (likely(size
>= len
))
1198 return skb_pad(skb
, len
-size
);
1201 static inline int skb_add_data(struct sk_buff
*skb
,
1202 char __user
*from
, int copy
)
1204 const int off
= skb
->len
;
1206 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1208 __wsum csum
= csum_and_copy_from_user(from
, skb_put(skb
, copy
),
1211 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
1214 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
1217 __skb_trim(skb
, off
);
1221 static inline int skb_can_coalesce(struct sk_buff
*skb
, int i
,
1222 struct page
*page
, int off
)
1225 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
1227 return page
== frag
->page
&&
1228 off
== frag
->page_offset
+ frag
->size
;
1233 static inline int __skb_linearize(struct sk_buff
*skb
)
1235 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
1239 * skb_linearize - convert paged skb to linear one
1240 * @skb: buffer to linarize
1242 * If there is no free memory -ENOMEM is returned, otherwise zero
1243 * is returned and the old skb data released.
1245 static inline int skb_linearize(struct sk_buff
*skb
)
1247 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
1251 * skb_linearize_cow - make sure skb is linear and writable
1252 * @skb: buffer to process
1254 * If there is no free memory -ENOMEM is returned, otherwise zero
1255 * is returned and the old skb data released.
1257 static inline int skb_linearize_cow(struct sk_buff
*skb
)
1259 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
1260 __skb_linearize(skb
) : 0;
1264 * skb_postpull_rcsum - update checksum for received skb after pull
1265 * @skb: buffer to update
1266 * @start: start of data before pull
1267 * @len: length of data pulled
1269 * After doing a pull on a received packet, you need to call this to
1270 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1271 * CHECKSUM_NONE so that it can be recomputed from scratch.
1274 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
1275 const void *start
, unsigned int len
)
1277 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1278 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
1281 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
1284 * pskb_trim_rcsum - trim received skb and update checksum
1285 * @skb: buffer to trim
1288 * This is exactly the same as pskb_trim except that it ensures the
1289 * checksum of received packets are still valid after the operation.
1292 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
1294 if (likely(len
>= skb
->len
))
1296 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1297 skb
->ip_summed
= CHECKSUM_NONE
;
1298 return __pskb_trim(skb
, len
);
1301 #define skb_queue_walk(queue, skb) \
1302 for (skb = (queue)->next; \
1303 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1306 #define skb_queue_reverse_walk(queue, skb) \
1307 for (skb = (queue)->prev; \
1308 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1312 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1313 int noblock
, int *err
);
1314 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
1315 struct poll_table_struct
*wait
);
1316 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
1317 int offset
, struct iovec
*to
,
1319 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
,
1322 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
1323 extern void skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
,
1324 unsigned int flags
);
1325 extern __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
,
1326 int len
, __wsum csum
);
1327 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
1329 extern int skb_store_bits(const struct sk_buff
*skb
, int offset
,
1330 void *from
, int len
);
1331 extern __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
,
1332 int offset
, u8
*to
, int len
,
1334 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
1335 extern void skb_split(struct sk_buff
*skb
,
1336 struct sk_buff
*skb1
, const u32 len
);
1338 extern struct sk_buff
*skb_segment(struct sk_buff
*skb
, int features
);
1340 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
1341 int len
, void *buffer
)
1343 int hlen
= skb_headlen(skb
);
1345 if (hlen
- offset
>= len
)
1346 return skb
->data
+ offset
;
1348 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
1354 extern void skb_init(void);
1355 extern void skb_add_mtu(int mtu
);
1358 * skb_get_timestamp - get timestamp from a skb
1359 * @skb: skb to get stamp from
1360 * @stamp: pointer to struct timeval to store stamp in
1362 * Timestamps are stored in the skb as offsets to a base timestamp.
1363 * This function converts the offset back to a struct timeval and stores
1366 static inline void skb_get_timestamp(const struct sk_buff
*skb
, struct timeval
*stamp
)
1368 stamp
->tv_sec
= skb
->tstamp
.off_sec
;
1369 stamp
->tv_usec
= skb
->tstamp
.off_usec
;
1373 * skb_set_timestamp - set timestamp of a skb
1374 * @skb: skb to set stamp of
1375 * @stamp: pointer to struct timeval to get stamp from
1377 * Timestamps are stored in the skb as offsets to a base timestamp.
1378 * This function converts a struct timeval to an offset and stores
1381 static inline void skb_set_timestamp(struct sk_buff
*skb
, const struct timeval
*stamp
)
1383 skb
->tstamp
.off_sec
= stamp
->tv_sec
;
1384 skb
->tstamp
.off_usec
= stamp
->tv_usec
;
1387 extern void __net_timestamp(struct sk_buff
*skb
);
1389 extern __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
1392 * skb_checksum_complete - Calculate checksum of an entire packet
1393 * @skb: packet to process
1395 * This function calculates the checksum over the entire packet plus
1396 * the value of skb->csum. The latter can be used to supply the
1397 * checksum of a pseudo header as used by TCP/UDP. It returns the
1400 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1401 * this function can be used to verify that checksum on received
1402 * packets. In that case the function should return zero if the
1403 * checksum is correct. In particular, this function will return zero
1404 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1405 * hardware has already verified the correctness of the checksum.
1407 static inline unsigned int skb_checksum_complete(struct sk_buff
*skb
)
1409 return skb
->ip_summed
!= CHECKSUM_UNNECESSARY
&&
1410 __skb_checksum_complete(skb
);
1413 #ifdef CONFIG_NETFILTER
1414 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
1416 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
1417 nfct
->destroy(nfct
);
1419 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
1422 atomic_inc(&nfct
->use
);
1424 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1425 static inline void nf_conntrack_get_reasm(struct sk_buff
*skb
)
1428 atomic_inc(&skb
->users
);
1430 static inline void nf_conntrack_put_reasm(struct sk_buff
*skb
)
1436 #ifdef CONFIG_BRIDGE_NETFILTER
1437 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
1439 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
1442 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
1445 atomic_inc(&nf_bridge
->use
);
1447 #endif /* CONFIG_BRIDGE_NETFILTER */
1448 static inline void nf_reset(struct sk_buff
*skb
)
1450 nf_conntrack_put(skb
->nfct
);
1452 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1453 nf_conntrack_put_reasm(skb
->nfct_reasm
);
1454 skb
->nfct_reasm
= NULL
;
1456 #ifdef CONFIG_BRIDGE_NETFILTER
1457 nf_bridge_put(skb
->nf_bridge
);
1458 skb
->nf_bridge
= NULL
;
1462 #else /* CONFIG_NETFILTER */
1463 static inline void nf_reset(struct sk_buff
*skb
) {}
1464 #endif /* CONFIG_NETFILTER */
1466 #ifdef CONFIG_NETWORK_SECMARK
1467 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
1469 to
->secmark
= from
->secmark
;
1472 static inline void skb_init_secmark(struct sk_buff
*skb
)
1477 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
1480 static inline void skb_init_secmark(struct sk_buff
*skb
)
1484 static inline int skb_is_gso(const struct sk_buff
*skb
)
1486 return skb_shinfo(skb
)->gso_size
;
1489 #endif /* __KERNEL__ */
1490 #endif /* _LINUX_SKBUFF_H */