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>
30 #include <linux/hrtimer.h>
32 #define HAVE_ALLOC_SKB /* For the drivers to know */
33 #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
35 #define CHECKSUM_NONE 0
36 #define CHECKSUM_PARTIAL 1
37 #define CHECKSUM_UNNECESSARY 2
38 #define CHECKSUM_COMPLETE 3
40 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
41 ~(SMP_CACHE_BYTES - 1))
42 #define SKB_WITH_OVERHEAD(X) \
43 (((X) - sizeof(struct skb_shared_info)) & \
44 ~(SMP_CACHE_BYTES - 1))
45 #define SKB_MAX_ORDER(X, ORDER) \
46 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
47 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
48 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
50 /* A. Checksumming of received packets by device.
52 * NONE: device failed to checksum this packet.
53 * skb->csum is undefined.
55 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
56 * skb->csum is undefined.
57 * It is bad option, but, unfortunately, many of vendors do this.
58 * Apparently with secret goal to sell you new device, when you
59 * will add new protocol to your host. F.e. IPv6. 8)
61 * COMPLETE: the most generic way. Device supplied checksum of _all_
62 * the packet as seen by netif_rx in skb->csum.
63 * NOTE: Even if device supports only some protocols, but
64 * is able to produce some skb->csum, it MUST use COMPLETE,
67 * B. Checksumming on output.
69 * NONE: skb is checksummed by protocol or csum is not required.
71 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
72 * from skb->h.raw to the end and to record the checksum
73 * at skb->h.raw+skb->csum.
75 * Device must show its capabilities in dev->features, set
76 * at device setup time.
77 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
79 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
80 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
81 * TCP/UDP over IPv4. Sigh. Vendors like this
82 * way by an unknown reason. Though, see comment above
83 * about CHECKSUM_UNNECESSARY. 8)
85 * Any questions? No questions, good. --ANK
90 #ifdef CONFIG_NETFILTER
93 void (*destroy
)(struct nf_conntrack
*);
96 #ifdef CONFIG_BRIDGE_NETFILTER
97 struct nf_bridge_info
{
99 struct net_device
*physindev
;
100 struct net_device
*physoutdev
;
101 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
102 struct net_device
*netoutdev
;
105 unsigned long data
[32 / sizeof(unsigned long)];
111 struct sk_buff_head
{
112 /* These two members must be first. */
113 struct sk_buff
*next
;
114 struct sk_buff
*prev
;
122 /* To allow 64K frame to be packed as single skb without frag_list */
123 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
125 typedef struct skb_frag_struct skb_frag_t
;
127 struct skb_frag_struct
{
133 /* This data is invariant across clones and lives at
134 * the end of the header data, ie. at skb->end.
136 struct skb_shared_info
{
138 unsigned short nr_frags
;
139 unsigned short gso_size
;
140 /* Warning: this field is not always filled in (UFO)! */
141 unsigned short gso_segs
;
142 unsigned short gso_type
;
144 struct sk_buff
*frag_list
;
145 skb_frag_t frags
[MAX_SKB_FRAGS
];
148 /* We divide dataref into two halves. The higher 16 bits hold references
149 * to the payload part of skb->data. The lower 16 bits hold references to
150 * the entire skb->data. It is up to the users of the skb to agree on
151 * where the payload starts.
153 * All users must obey the rule that the skb->data reference count must be
154 * greater than or equal to the payload reference count.
156 * Holding a reference to the payload part means that the user does not
157 * care about modifications to the header part of skb->data.
159 #define SKB_DATAREF_SHIFT 16
160 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
164 SKB_FCLONE_UNAVAILABLE
,
170 SKB_GSO_TCPV4
= 1 << 0,
171 SKB_GSO_UDP
= 1 << 1,
173 /* This indicates the skb is from an untrusted source. */
174 SKB_GSO_DODGY
= 1 << 2,
176 /* This indicates the tcp segment has CWR set. */
177 SKB_GSO_TCP_ECN
= 1 << 3,
179 SKB_GSO_TCPV6
= 1 << 4,
183 * struct sk_buff - socket buffer
184 * @next: Next buffer in list
185 * @prev: Previous buffer in list
186 * @sk: Socket we are owned by
187 * @tstamp: Time we arrived
188 * @dev: Device we arrived on/are leaving by
189 * @iif: ifindex of device we arrived on
190 * @h: Transport layer header
191 * @nh: Network layer header
192 * @mac: Link layer header
193 * @dst: destination entry
194 * @sp: the security path, used for xfrm
195 * @cb: Control buffer. Free for use by every layer. Put private vars here
196 * @len: Length of actual data
197 * @data_len: Data length
198 * @mac_len: Length of link layer header
200 * @local_df: allow local fragmentation
201 * @cloned: Head may be cloned (check refcnt to be sure)
202 * @nohdr: Payload reference only, must not modify header
203 * @pkt_type: Packet class
204 * @fclone: skbuff clone status
205 * @ip_summed: Driver fed us an IP checksum
206 * @priority: Packet queueing priority
207 * @users: User count - see {datagram,tcp}.c
208 * @protocol: Packet protocol from driver
209 * @truesize: Buffer size
210 * @head: Head of buffer
211 * @data: Data head pointer
212 * @tail: Tail pointer
214 * @destructor: Destruct function
215 * @mark: Generic packet mark
216 * @nfct: Associated connection, if any
217 * @ipvs_property: skbuff is owned by ipvs
218 * @nfctinfo: Relationship of this skb to the connection
219 * @nfct_reasm: netfilter conntrack re-assembly pointer
220 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
221 * @tc_index: Traffic control index
222 * @tc_verd: traffic control verdict
223 * @dma_cookie: a cookie to one of several possible DMA operations
224 * done by skb DMA functions
225 * @secmark: security marking
229 /* These two members must be first. */
230 struct sk_buff
*next
;
231 struct sk_buff
*prev
;
235 struct net_device
*dev
;
237 /* 4 byte hole on 64 bit*/
242 struct icmphdr
*icmph
;
244 struct ipv6hdr
*ipv6h
;
256 struct dst_entry
*dst
;
260 * This is the control buffer. It is free to use for every
261 * layer. Please put your private variables there. If you
262 * want to keep them across layers you have to do a skb_clone()
263 * first. This is owned by whoever has the skb queued ATM.
285 void (*destructor
)(struct sk_buff
*skb
);
286 #ifdef CONFIG_NETFILTER
287 struct nf_conntrack
*nfct
;
288 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
289 struct sk_buff
*nfct_reasm
;
291 #ifdef CONFIG_BRIDGE_NETFILTER
292 struct nf_bridge_info
*nf_bridge
;
294 #endif /* CONFIG_NETFILTER */
295 #ifdef CONFIG_NET_SCHED
296 __u16 tc_index
; /* traffic control index */
297 #ifdef CONFIG_NET_CLS_ACT
298 __u16 tc_verd
; /* traffic control verdict */
301 #ifdef CONFIG_NET_DMA
302 dma_cookie_t dma_cookie
;
304 #ifdef CONFIG_NETWORK_SECMARK
310 /* These elements must be at the end, see alloc_skb() for details. */
311 unsigned int truesize
;
321 * Handling routines are only of interest to the kernel
323 #include <linux/slab.h>
325 #include <asm/system.h>
327 extern void kfree_skb(struct sk_buff
*skb
);
328 extern void __kfree_skb(struct sk_buff
*skb
);
329 extern struct sk_buff
*__alloc_skb(unsigned int size
,
330 gfp_t priority
, int fclone
, int node
);
331 static inline struct sk_buff
*alloc_skb(unsigned int size
,
334 return __alloc_skb(size
, priority
, 0, -1);
337 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
340 return __alloc_skb(size
, priority
, 1, -1);
343 extern void kfree_skbmem(struct sk_buff
*skb
);
344 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
346 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
348 extern struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
350 extern int pskb_expand_head(struct sk_buff
*skb
,
351 int nhead
, int ntail
,
353 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
354 unsigned int headroom
);
355 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
356 int newheadroom
, int newtailroom
,
358 extern int skb_pad(struct sk_buff
*skb
, int pad
);
359 #define dev_kfree_skb(a) kfree_skb(a)
360 extern void skb_over_panic(struct sk_buff
*skb
, int len
,
362 extern void skb_under_panic(struct sk_buff
*skb
, int len
,
364 extern void skb_truesize_bug(struct sk_buff
*skb
);
366 static inline void skb_truesize_check(struct sk_buff
*skb
)
368 if (unlikely((int)skb
->truesize
< sizeof(struct sk_buff
) + skb
->len
))
369 skb_truesize_bug(skb
);
372 extern int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
373 int getfrag(void *from
, char *to
, int offset
,
374 int len
,int odd
, struct sk_buff
*skb
),
375 void *from
, int length
);
382 __u32 stepped_offset
;
383 struct sk_buff
*root_skb
;
384 struct sk_buff
*cur_skb
;
388 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
389 unsigned int from
, unsigned int to
,
390 struct skb_seq_state
*st
);
391 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
392 struct skb_seq_state
*st
);
393 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
395 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
396 unsigned int to
, struct ts_config
*config
,
397 struct ts_state
*state
);
400 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
403 * skb_queue_empty - check if a queue is empty
406 * Returns true if the queue is empty, false otherwise.
408 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
410 return list
->next
== (struct sk_buff
*)list
;
414 * skb_get - reference buffer
415 * @skb: buffer to reference
417 * Makes another reference to a socket buffer and returns a pointer
420 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
422 atomic_inc(&skb
->users
);
427 * If users == 1, we are the only owner and are can avoid redundant
432 * skb_cloned - is the buffer a clone
433 * @skb: buffer to check
435 * Returns true if the buffer was generated with skb_clone() and is
436 * one of multiple shared copies of the buffer. Cloned buffers are
437 * shared data so must not be written to under normal circumstances.
439 static inline int skb_cloned(const struct sk_buff
*skb
)
441 return skb
->cloned
&&
442 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
446 * skb_header_cloned - is the header a clone
447 * @skb: buffer to check
449 * Returns true if modifying the header part of the buffer requires
450 * the data to be copied.
452 static inline int skb_header_cloned(const struct sk_buff
*skb
)
459 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
460 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
465 * skb_header_release - release reference to header
466 * @skb: buffer to operate on
468 * Drop a reference to the header part of the buffer. This is done
469 * by acquiring a payload reference. You must not read from the header
470 * part of skb->data after this.
472 static inline void skb_header_release(struct sk_buff
*skb
)
476 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
480 * skb_shared - is the buffer shared
481 * @skb: buffer to check
483 * Returns true if more than one person has a reference to this
486 static inline int skb_shared(const struct sk_buff
*skb
)
488 return atomic_read(&skb
->users
) != 1;
492 * skb_share_check - check if buffer is shared and if so clone it
493 * @skb: buffer to check
494 * @pri: priority for memory allocation
496 * If the buffer is shared the buffer is cloned and the old copy
497 * drops a reference. A new clone with a single reference is returned.
498 * If the buffer is not shared the original buffer is returned. When
499 * being called from interrupt status or with spinlocks held pri must
502 * NULL is returned on a memory allocation failure.
504 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
,
507 might_sleep_if(pri
& __GFP_WAIT
);
508 if (skb_shared(skb
)) {
509 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
517 * Copy shared buffers into a new sk_buff. We effectively do COW on
518 * packets to handle cases where we have a local reader and forward
519 * and a couple of other messy ones. The normal one is tcpdumping
520 * a packet thats being forwarded.
524 * skb_unshare - make a copy of a shared buffer
525 * @skb: buffer to check
526 * @pri: priority for memory allocation
528 * If the socket buffer is a clone then this function creates a new
529 * copy of the data, drops a reference count on the old copy and returns
530 * the new copy with the reference count at 1. If the buffer is not a clone
531 * the original buffer is returned. When called with a spinlock held or
532 * from interrupt state @pri must be %GFP_ATOMIC
534 * %NULL is returned on a memory allocation failure.
536 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
539 might_sleep_if(pri
& __GFP_WAIT
);
540 if (skb_cloned(skb
)) {
541 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
542 kfree_skb(skb
); /* Free our shared copy */
550 * @list_: list to peek at
552 * Peek an &sk_buff. Unlike most other operations you _MUST_
553 * be careful with this one. A peek leaves the buffer on the
554 * list and someone else may run off with it. You must hold
555 * the appropriate locks or have a private queue to do this.
557 * Returns %NULL for an empty list or a pointer to the head element.
558 * The reference count is not incremented and the reference is therefore
559 * volatile. Use with caution.
561 static inline struct sk_buff
*skb_peek(struct sk_buff_head
*list_
)
563 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->next
;
564 if (list
== (struct sk_buff
*)list_
)
571 * @list_: list to peek at
573 * Peek an &sk_buff. Unlike most other operations you _MUST_
574 * be careful with this one. A peek leaves the buffer on the
575 * list and someone else may run off with it. You must hold
576 * the appropriate locks or have a private queue to do this.
578 * Returns %NULL for an empty list or a pointer to the tail element.
579 * The reference count is not incremented and the reference is therefore
580 * volatile. Use with caution.
582 static inline struct sk_buff
*skb_peek_tail(struct sk_buff_head
*list_
)
584 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->prev
;
585 if (list
== (struct sk_buff
*)list_
)
591 * skb_queue_len - get queue length
592 * @list_: list to measure
594 * Return the length of an &sk_buff queue.
596 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
602 * This function creates a split out lock class for each invocation;
603 * this is needed for now since a whole lot of users of the skb-queue
604 * infrastructure in drivers have different locking usage (in hardirq)
605 * than the networking core (in softirq only). In the long run either the
606 * network layer or drivers should need annotation to consolidate the
607 * main types of usage into 3 classes.
609 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
611 spin_lock_init(&list
->lock
);
612 list
->prev
= list
->next
= (struct sk_buff
*)list
;
616 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
617 struct lock_class_key
*class)
619 skb_queue_head_init(list
);
620 lockdep_set_class(&list
->lock
, class);
624 * Insert an sk_buff at the start of a list.
626 * The "__skb_xxxx()" functions are the non-atomic ones that
627 * can only be called with interrupts disabled.
631 * __skb_queue_after - queue a buffer at the list head
633 * @prev: place after this buffer
634 * @newsk: buffer to queue
636 * Queue a buffer int the middle of a list. This function takes no locks
637 * and you must therefore hold required locks before calling it.
639 * A buffer cannot be placed on two lists at the same time.
641 static inline void __skb_queue_after(struct sk_buff_head
*list
,
642 struct sk_buff
*prev
,
643 struct sk_buff
*newsk
)
645 struct sk_buff
*next
;
651 next
->prev
= prev
->next
= newsk
;
655 * __skb_queue_head - queue a buffer at the list head
657 * @newsk: buffer to queue
659 * Queue a buffer at the start of a list. This function takes no locks
660 * and you must therefore hold required locks before calling it.
662 * A buffer cannot be placed on two lists at the same time.
664 extern void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
665 static inline void __skb_queue_head(struct sk_buff_head
*list
,
666 struct sk_buff
*newsk
)
668 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
672 * __skb_queue_tail - queue a buffer at the list tail
674 * @newsk: buffer to queue
676 * Queue a buffer at the end of a list. This function takes no locks
677 * and you must therefore hold required locks before calling it.
679 * A buffer cannot be placed on two lists at the same time.
681 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
682 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
683 struct sk_buff
*newsk
)
685 struct sk_buff
*prev
, *next
;
688 next
= (struct sk_buff
*)list
;
692 next
->prev
= prev
->next
= newsk
;
697 * __skb_dequeue - remove from the head of the queue
698 * @list: list to dequeue from
700 * Remove the head of the list. This function does not take any locks
701 * so must be used with appropriate locks held only. The head item is
702 * returned or %NULL if the list is empty.
704 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
705 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
707 struct sk_buff
*next
, *prev
, *result
;
709 prev
= (struct sk_buff
*) list
;
718 result
->next
= result
->prev
= NULL
;
725 * Insert a packet on a list.
727 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
728 static inline void __skb_insert(struct sk_buff
*newsk
,
729 struct sk_buff
*prev
, struct sk_buff
*next
,
730 struct sk_buff_head
*list
)
734 next
->prev
= prev
->next
= newsk
;
739 * Place a packet after a given packet in a list.
741 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
742 static inline void __skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
744 __skb_insert(newsk
, old
, old
->next
, list
);
748 * remove sk_buff from list. _Must_ be called atomically, and with
751 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
752 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
754 struct sk_buff
*next
, *prev
;
759 skb
->next
= skb
->prev
= NULL
;
765 /* XXX: more streamlined implementation */
768 * __skb_dequeue_tail - remove from the tail of the queue
769 * @list: list to dequeue from
771 * Remove the tail of the list. This function does not take any locks
772 * so must be used with appropriate locks held only. The tail item is
773 * returned or %NULL if the list is empty.
775 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
776 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
778 struct sk_buff
*skb
= skb_peek_tail(list
);
780 __skb_unlink(skb
, list
);
785 static inline int skb_is_nonlinear(const struct sk_buff
*skb
)
787 return skb
->data_len
;
790 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
792 return skb
->len
- skb
->data_len
;
795 static inline int skb_pagelen(const struct sk_buff
*skb
)
799 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
800 len
+= skb_shinfo(skb
)->frags
[i
].size
;
801 return len
+ skb_headlen(skb
);
804 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
805 struct page
*page
, int off
, int size
)
807 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
810 frag
->page_offset
= off
;
812 skb_shinfo(skb
)->nr_frags
= i
+ 1;
815 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
816 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
817 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
820 * Add data to an sk_buff
822 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
824 unsigned char *tmp
= skb
->tail
;
825 SKB_LINEAR_ASSERT(skb
);
832 * skb_put - add data to a buffer
833 * @skb: buffer to use
834 * @len: amount of data to add
836 * This function extends the used data area of the buffer. If this would
837 * exceed the total buffer size the kernel will panic. A pointer to the
838 * first byte of the extra data is returned.
840 static inline unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
)
842 unsigned char *tmp
= skb
->tail
;
843 SKB_LINEAR_ASSERT(skb
);
846 if (unlikely(skb
->tail
>skb
->end
))
847 skb_over_panic(skb
, len
, current_text_addr());
851 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
859 * skb_push - add data to the start of a buffer
860 * @skb: buffer to use
861 * @len: amount of data to add
863 * This function extends the used data area of the buffer at the buffer
864 * start. If this would exceed the total buffer headroom the kernel will
865 * panic. A pointer to the first byte of the extra data is returned.
867 static inline unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
)
871 if (unlikely(skb
->data
<skb
->head
))
872 skb_under_panic(skb
, len
, current_text_addr());
876 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
879 BUG_ON(skb
->len
< skb
->data_len
);
880 return skb
->data
+= len
;
884 * skb_pull - remove data from the start of a buffer
885 * @skb: buffer to use
886 * @len: amount of data to remove
888 * This function removes data from the start of a buffer, returning
889 * the memory to the headroom. A pointer to the next data in the buffer
890 * is returned. Once the data has been pulled future pushes will overwrite
893 static inline unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
)
895 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
898 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
900 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
902 if (len
> skb_headlen(skb
) &&
903 !__pskb_pull_tail(skb
, len
-skb_headlen(skb
)))
906 return skb
->data
+= len
;
909 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
911 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
914 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
916 if (likely(len
<= skb_headlen(skb
)))
918 if (unlikely(len
> skb
->len
))
920 return __pskb_pull_tail(skb
, len
-skb_headlen(skb
)) != NULL
;
924 * skb_headroom - bytes at buffer head
925 * @skb: buffer to check
927 * Return the number of bytes of free space at the head of an &sk_buff.
929 static inline int skb_headroom(const struct sk_buff
*skb
)
931 return skb
->data
- skb
->head
;
935 * skb_tailroom - bytes at buffer end
936 * @skb: buffer to check
938 * Return the number of bytes of free space at the tail of an sk_buff
940 static inline int skb_tailroom(const struct sk_buff
*skb
)
942 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
946 * skb_reserve - adjust headroom
947 * @skb: buffer to alter
948 * @len: bytes to move
950 * Increase the headroom of an empty &sk_buff by reducing the tail
951 * room. This is only allowed for an empty buffer.
953 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
959 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
961 skb
->h
.raw
= skb
->data
;
964 static inline void skb_set_transport_header(struct sk_buff
*skb
,
967 skb
->h
.raw
= skb
->data
+ offset
;
970 static inline int skb_transport_offset(const struct sk_buff
*skb
)
972 return skb
->h
.raw
- skb
->data
;
975 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
980 static inline void skb_reset_network_header(struct sk_buff
*skb
)
982 skb
->nh
.raw
= skb
->data
;
985 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
987 skb
->nh
.raw
= skb
->data
+ offset
;
990 static inline int skb_network_offset(const struct sk_buff
*skb
)
992 return skb
->nh
.raw
- skb
->data
;
995 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1000 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1002 return skb
->mac
.raw
!= NULL
;
1005 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1007 skb
->mac
.raw
= skb
->data
;
1010 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1012 skb
->mac
.raw
= skb
->data
+ offset
;
1016 * CPUs often take a performance hit when accessing unaligned memory
1017 * locations. The actual performance hit varies, it can be small if the
1018 * hardware handles it or large if we have to take an exception and fix it
1021 * Since an ethernet header is 14 bytes network drivers often end up with
1022 * the IP header at an unaligned offset. The IP header can be aligned by
1023 * shifting the start of the packet by 2 bytes. Drivers should do this
1026 * skb_reserve(NET_IP_ALIGN);
1028 * The downside to this alignment of the IP header is that the DMA is now
1029 * unaligned. On some architectures the cost of an unaligned DMA is high
1030 * and this cost outweighs the gains made by aligning the IP header.
1032 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1035 #ifndef NET_IP_ALIGN
1036 #define NET_IP_ALIGN 2
1040 * The networking layer reserves some headroom in skb data (via
1041 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1042 * the header has to grow. In the default case, if the header has to grow
1043 * 16 bytes or less we avoid the reallocation.
1045 * Unfortunately this headroom changes the DMA alignment of the resulting
1046 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1047 * on some architectures. An architecture can override this value,
1048 * perhaps setting it to a cacheline in size (since that will maintain
1049 * cacheline alignment of the DMA). It must be a power of 2.
1051 * Various parts of the networking layer expect at least 16 bytes of
1052 * headroom, you should not reduce this.
1055 #define NET_SKB_PAD 16
1058 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
1060 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
1062 if (unlikely(skb
->data_len
)) {
1067 skb
->tail
= skb
->data
+ len
;
1071 * skb_trim - remove end from a buffer
1072 * @skb: buffer to alter
1075 * Cut the length of a buffer down by removing data from the tail. If
1076 * the buffer is already under the length specified it is not modified.
1077 * The skb must be linear.
1079 static inline void skb_trim(struct sk_buff
*skb
, unsigned int len
)
1082 __skb_trim(skb
, len
);
1086 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1089 return ___pskb_trim(skb
, len
);
1090 __skb_trim(skb
, len
);
1094 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1096 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
1100 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1101 * @skb: buffer to alter
1104 * This is identical to pskb_trim except that the caller knows that
1105 * the skb is not cloned so we should never get an error due to out-
1108 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
1110 int err
= pskb_trim(skb
, len
);
1115 * skb_orphan - orphan a buffer
1116 * @skb: buffer to orphan
1118 * If a buffer currently has an owner then we call the owner's
1119 * destructor function and make the @skb unowned. The buffer continues
1120 * to exist but is no longer charged to its former owner.
1122 static inline void skb_orphan(struct sk_buff
*skb
)
1124 if (skb
->destructor
)
1125 skb
->destructor(skb
);
1126 skb
->destructor
= NULL
;
1131 * __skb_queue_purge - empty a list
1132 * @list: list to empty
1134 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1135 * the list and one reference dropped. This function does not take the
1136 * list lock and the caller must hold the relevant locks to use it.
1138 extern void skb_queue_purge(struct sk_buff_head
*list
);
1139 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
1141 struct sk_buff
*skb
;
1142 while ((skb
= __skb_dequeue(list
)) != NULL
)
1147 * __dev_alloc_skb - allocate an skbuff for receiving
1148 * @length: length to allocate
1149 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1151 * Allocate a new &sk_buff and assign it a usage count of one. The
1152 * buffer has unspecified headroom built in. Users should allocate
1153 * the headroom they think they need without accounting for the
1154 * built in space. The built in space is used for optimisations.
1156 * %NULL is returned if there is no free memory.
1158 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
1161 struct sk_buff
*skb
= alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
);
1163 skb_reserve(skb
, NET_SKB_PAD
);
1168 * dev_alloc_skb - allocate an skbuff for receiving
1169 * @length: length to allocate
1171 * Allocate a new &sk_buff and assign it a usage count of one. The
1172 * buffer has unspecified headroom built in. Users should allocate
1173 * the headroom they think they need without accounting for the
1174 * built in space. The built in space is used for optimisations.
1176 * %NULL is returned if there is no free memory. Although this function
1177 * allocates memory it can be called from an interrupt.
1179 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
1181 return __dev_alloc_skb(length
, GFP_ATOMIC
);
1184 extern struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
1185 unsigned int length
, gfp_t gfp_mask
);
1188 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1189 * @dev: network device to receive on
1190 * @length: length to allocate
1192 * Allocate a new &sk_buff and assign it a usage count of one. The
1193 * buffer has unspecified headroom built in. Users should allocate
1194 * the headroom they think they need without accounting for the
1195 * built in space. The built in space is used for optimisations.
1197 * %NULL is returned if there is no free memory. Although this function
1198 * allocates memory it can be called from an interrupt.
1200 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
1201 unsigned int length
)
1203 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
1207 * skb_cow - copy header of skb when it is required
1208 * @skb: buffer to cow
1209 * @headroom: needed headroom
1211 * If the skb passed lacks sufficient headroom or its data part
1212 * is shared, data is reallocated. If reallocation fails, an error
1213 * is returned and original skb is not changed.
1215 * The result is skb with writable area skb->head...skb->tail
1216 * and at least @headroom of space at head.
1218 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
1220 int delta
= (headroom
> NET_SKB_PAD
? headroom
: NET_SKB_PAD
) -
1226 if (delta
|| skb_cloned(skb
))
1227 return pskb_expand_head(skb
, (delta
+ (NET_SKB_PAD
-1)) &
1228 ~(NET_SKB_PAD
-1), 0, GFP_ATOMIC
);
1233 * skb_padto - pad an skbuff up to a minimal size
1234 * @skb: buffer to pad
1235 * @len: minimal length
1237 * Pads up a buffer to ensure the trailing bytes exist and are
1238 * blanked. If the buffer already contains sufficient data it
1239 * is untouched. Otherwise it is extended. Returns zero on
1240 * success. The skb is freed on error.
1243 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
1245 unsigned int size
= skb
->len
;
1246 if (likely(size
>= len
))
1248 return skb_pad(skb
, len
-size
);
1251 static inline int skb_add_data(struct sk_buff
*skb
,
1252 char __user
*from
, int copy
)
1254 const int off
= skb
->len
;
1256 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1258 __wsum csum
= csum_and_copy_from_user(from
, skb_put(skb
, copy
),
1261 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
1264 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
1267 __skb_trim(skb
, off
);
1271 static inline int skb_can_coalesce(struct sk_buff
*skb
, int i
,
1272 struct page
*page
, int off
)
1275 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
1277 return page
== frag
->page
&&
1278 off
== frag
->page_offset
+ frag
->size
;
1283 static inline int __skb_linearize(struct sk_buff
*skb
)
1285 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
1289 * skb_linearize - convert paged skb to linear one
1290 * @skb: buffer to linarize
1292 * If there is no free memory -ENOMEM is returned, otherwise zero
1293 * is returned and the old skb data released.
1295 static inline int skb_linearize(struct sk_buff
*skb
)
1297 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
1301 * skb_linearize_cow - make sure skb is linear and writable
1302 * @skb: buffer to process
1304 * If there is no free memory -ENOMEM is returned, otherwise zero
1305 * is returned and the old skb data released.
1307 static inline int skb_linearize_cow(struct sk_buff
*skb
)
1309 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
1310 __skb_linearize(skb
) : 0;
1314 * skb_postpull_rcsum - update checksum for received skb after pull
1315 * @skb: buffer to update
1316 * @start: start of data before pull
1317 * @len: length of data pulled
1319 * After doing a pull on a received packet, you need to call this to
1320 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1321 * CHECKSUM_NONE so that it can be recomputed from scratch.
1324 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
1325 const void *start
, unsigned int len
)
1327 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1328 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
1331 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
1334 * pskb_trim_rcsum - trim received skb and update checksum
1335 * @skb: buffer to trim
1338 * This is exactly the same as pskb_trim except that it ensures the
1339 * checksum of received packets are still valid after the operation.
1342 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
1344 if (likely(len
>= skb
->len
))
1346 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1347 skb
->ip_summed
= CHECKSUM_NONE
;
1348 return __pskb_trim(skb
, len
);
1351 #define skb_queue_walk(queue, skb) \
1352 for (skb = (queue)->next; \
1353 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1356 #define skb_queue_reverse_walk(queue, skb) \
1357 for (skb = (queue)->prev; \
1358 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1362 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1363 int noblock
, int *err
);
1364 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
1365 struct poll_table_struct
*wait
);
1366 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
1367 int offset
, struct iovec
*to
,
1369 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
,
1372 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
1373 extern void skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
,
1374 unsigned int flags
);
1375 extern __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
,
1376 int len
, __wsum csum
);
1377 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
1379 extern int skb_store_bits(const struct sk_buff
*skb
, int offset
,
1380 void *from
, int len
);
1381 extern __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
,
1382 int offset
, u8
*to
, int len
,
1384 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
1385 extern void skb_split(struct sk_buff
*skb
,
1386 struct sk_buff
*skb1
, const u32 len
);
1388 extern struct sk_buff
*skb_segment(struct sk_buff
*skb
, int features
);
1390 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
1391 int len
, void *buffer
)
1393 int hlen
= skb_headlen(skb
);
1395 if (hlen
- offset
>= len
)
1396 return skb
->data
+ offset
;
1398 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
1404 extern void skb_init(void);
1405 extern void skb_add_mtu(int mtu
);
1408 * skb_get_timestamp - get timestamp from a skb
1409 * @skb: skb to get stamp from
1410 * @stamp: pointer to struct timeval to store stamp in
1412 * Timestamps are stored in the skb as offsets to a base timestamp.
1413 * This function converts the offset back to a struct timeval and stores
1416 static inline void skb_get_timestamp(const struct sk_buff
*skb
, struct timeval
*stamp
)
1418 *stamp
= ktime_to_timeval(skb
->tstamp
);
1421 static inline void __net_timestamp(struct sk_buff
*skb
)
1423 skb
->tstamp
= ktime_get_real();
1427 extern __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
1428 extern __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
1431 * skb_checksum_complete - Calculate checksum of an entire packet
1432 * @skb: packet to process
1434 * This function calculates the checksum over the entire packet plus
1435 * the value of skb->csum. The latter can be used to supply the
1436 * checksum of a pseudo header as used by TCP/UDP. It returns the
1439 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1440 * this function can be used to verify that checksum on received
1441 * packets. In that case the function should return zero if the
1442 * checksum is correct. In particular, this function will return zero
1443 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1444 * hardware has already verified the correctness of the checksum.
1446 static inline unsigned int skb_checksum_complete(struct sk_buff
*skb
)
1448 return skb
->ip_summed
!= CHECKSUM_UNNECESSARY
&&
1449 __skb_checksum_complete(skb
);
1452 #ifdef CONFIG_NETFILTER
1453 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
1455 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
1456 nfct
->destroy(nfct
);
1458 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
1461 atomic_inc(&nfct
->use
);
1463 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1464 static inline void nf_conntrack_get_reasm(struct sk_buff
*skb
)
1467 atomic_inc(&skb
->users
);
1469 static inline void nf_conntrack_put_reasm(struct sk_buff
*skb
)
1475 #ifdef CONFIG_BRIDGE_NETFILTER
1476 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
1478 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
1481 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
1484 atomic_inc(&nf_bridge
->use
);
1486 #endif /* CONFIG_BRIDGE_NETFILTER */
1487 static inline void nf_reset(struct sk_buff
*skb
)
1489 nf_conntrack_put(skb
->nfct
);
1491 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1492 nf_conntrack_put_reasm(skb
->nfct_reasm
);
1493 skb
->nfct_reasm
= NULL
;
1495 #ifdef CONFIG_BRIDGE_NETFILTER
1496 nf_bridge_put(skb
->nf_bridge
);
1497 skb
->nf_bridge
= NULL
;
1501 #else /* CONFIG_NETFILTER */
1502 static inline void nf_reset(struct sk_buff
*skb
) {}
1503 #endif /* CONFIG_NETFILTER */
1505 #ifdef CONFIG_NETWORK_SECMARK
1506 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
1508 to
->secmark
= from
->secmark
;
1511 static inline void skb_init_secmark(struct sk_buff
*skb
)
1516 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
1519 static inline void skb_init_secmark(struct sk_buff
*skb
)
1523 static inline int skb_is_gso(const struct sk_buff
*skb
)
1525 return skb_shinfo(skb
)->gso_size
;
1528 #endif /* __KERNEL__ */
1529 #endif /* _LINUX_SKBUFF_H */