2 * Definitions for the 'struct sk_buff' memory handlers.
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/config.h>
18 #include <linux/kernel.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/cache.h>
23 #include <asm/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
27 #include <linux/highmem.h>
28 #include <linux/poll.h>
29 #include <linux/net.h>
30 #include <linux/textsearch.h>
31 #include <net/checksum.h>
33 #define HAVE_ALLOC_SKB /* For the drivers to know */
34 #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
35 #define SLAB_SKB /* Slabified skbuffs */
37 #define CHECKSUM_NONE 0
39 #define CHECKSUM_UNNECESSARY 2
41 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
42 ~(SMP_CACHE_BYTES - 1))
43 #define SKB_MAX_ORDER(X, ORDER) (((PAGE_SIZE << (ORDER)) - (X) - \
44 sizeof(struct skb_shared_info)) & \
45 ~(SMP_CACHE_BYTES - 1))
46 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
47 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
49 /* A. Checksumming of received packets by device.
51 * NONE: device failed to checksum this packet.
52 * skb->csum is undefined.
54 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
55 * skb->csum is undefined.
56 * It is bad option, but, unfortunately, many of vendors do this.
57 * Apparently with secret goal to sell you new device, when you
58 * will add new protocol to your host. F.e. IPv6. 8)
60 * HW: the most generic way. Device supplied checksum of _all_
61 * the packet as seen by netif_rx in skb->csum.
62 * NOTE: Even if device supports only some protocols, but
63 * is able to produce some skb->csum, it MUST use HW,
66 * B. Checksumming on output.
68 * NONE: skb is checksummed by protocol or csum is not required.
70 * HW: device is required to csum packet as seen by hard_start_xmit
71 * from skb->h.raw to the end and to record the checksum
72 * at skb->h.raw+skb->csum.
74 * Device must show its capabilities in dev->features, set
75 * at device setup time.
76 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
78 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
79 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
80 * TCP/UDP over IPv4. Sigh. Vendors like this
81 * way by an unknown reason. Though, see comment above
82 * about CHECKSUM_UNNECESSARY. 8)
84 * Any questions? No questions, good. --ANK
89 #ifdef CONFIG_NETFILTER
92 void (*destroy)(struct nf_conntrack *);
95 #ifdef CONFIG_BRIDGE_NETFILTER
96 struct nf_bridge_info {
98 struct net_device *physindev;
99 struct net_device *physoutdev;
100 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
101 struct net_device *netoutdev;
104 unsigned long data[32 / sizeof(unsigned long)];
110 struct sk_buff_head {
111 /* These two members must be first. */
112 struct sk_buff *next;
113 struct sk_buff *prev;
121 /* To allow 64K frame to be packed as single skb without frag_list */
122 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
124 typedef struct skb_frag_struct skb_frag_t;
126 struct skb_frag_struct {
132 /* This data is invariant across clones and lives at
133 * the end of the header data, ie. at skb->end.
135 struct skb_shared_info {
137 unsigned int nr_frags;
138 unsigned short tso_size;
139 unsigned short tso_segs;
140 unsigned short ufo_size;
141 unsigned int ip6_frag_id;
142 struct sk_buff *frag_list;
143 skb_frag_t frags[MAX_SKB_FRAGS];
146 /* We divide dataref into two halves. The higher 16 bits hold references
147 * to the payload part of skb->data. The lower 16 bits hold references to
148 * the entire skb->data. It is up to the users of the skb to agree on
149 * where the payload starts.
151 * All users must obey the rule that the skb->data reference count must be
152 * greater than or equal to the payload reference count.
154 * Holding a reference to the payload part means that the user does not
155 * care about modifications to the header part of skb->data.
157 #define SKB_DATAREF_SHIFT 16
158 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
167 SKB_FCLONE_UNAVAILABLE,
173 * struct sk_buff - socket buffer
174 * @next: Next buffer in list
175 * @prev: Previous buffer in list
176 * @sk: Socket we are owned by
177 * @tstamp: Time we arrived
178 * @dev: Device we arrived on/are leaving by
179 * @input_dev: Device we arrived on
180 * @h: Transport layer header
181 * @nh: Network layer header
182 * @mac: Link layer header
183 * @dst: destination entry
184 * @sp: the security path, used for xfrm
185 * @cb: Control buffer. Free for use by every layer. Put private vars here
186 * @len: Length of actual data
187 * @data_len: Data length
188 * @mac_len: Length of link layer header
190 * @local_df: allow local fragmentation
191 * @cloned: Head may be cloned (check refcnt to be sure)
192 * @nohdr: Payload reference only, must not modify header
193 * @pkt_type: Packet class
194 * @fclone: skbuff clone status
195 * @ip_summed: Driver fed us an IP checksum
196 * @priority: Packet queueing priority
197 * @users: User count - see {datagram,tcp}.c
198 * @protocol: Packet protocol from driver
199 * @truesize: Buffer size
200 * @head: Head of buffer
201 * @data: Data head pointer
202 * @tail: Tail pointer
204 * @destructor: Destruct function
205 * @nfmark: Can be used for communication between hooks
206 * @nfct: Associated connection, if any
207 * @ipvs_property: skbuff is owned by ipvs
208 * @nfctinfo: Relationship of this skb to the connection
209 * @nfct_reasm: netfilter conntrack re-assembly pointer
210 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
211 * @tc_index: Traffic control index
212 * @tc_verd: traffic control verdict
216 /* These two members must be first. */
217 struct sk_buff *next;
218 struct sk_buff *prev;
221 struct skb_timeval tstamp;
222 struct net_device *dev;
223 struct net_device *input_dev;
228 struct icmphdr *icmph;
229 struct igmphdr *igmph;
231 struct ipv6hdr *ipv6h;
237 struct ipv6hdr *ipv6h;
246 struct dst_entry *dst;
250 * This is the control buffer. It is free to use for every
251 * layer. Please put your private variables there. If you
252 * want to keep them across layers you have to do a skb_clone()
253 * first. This is owned by whoever has the skb queued ATM.
272 void (*destructor)(struct sk_buff *skb);
273 #ifdef CONFIG_NETFILTER
275 struct nf_conntrack *nfct;
276 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
277 struct sk_buff *nfct_reasm;
279 #ifdef CONFIG_BRIDGE_NETFILTER
280 struct nf_bridge_info *nf_bridge;
282 #endif /* CONFIG_NETFILTER */
283 #ifdef CONFIG_NET_SCHED
284 __u16 tc_index; /* traffic control index */
285 #ifdef CONFIG_NET_CLS_ACT
286 __u16 tc_verd; /* traffic control verdict */
291 /* These elements must be at the end, see alloc_skb() for details. */
292 unsigned int truesize;
302 * Handling routines are only of interest to the kernel
304 #include <linux/slab.h>
306 #include <asm/system.h>
308 extern void __kfree_skb(struct sk_buff *skb);
309 extern struct sk_buff *__alloc_skb(unsigned int size,
310 gfp_t priority, int fclone);
311 static inline struct sk_buff *alloc_skb(unsigned int size,
314 return __alloc_skb(size, priority, 0);
317 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
320 return __alloc_skb(size, priority, 1);
323 extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
326 extern void kfree_skbmem(struct sk_buff *skb);
327 extern struct sk_buff *skb_clone(struct sk_buff *skb,
329 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
331 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
333 extern int pskb_expand_head(struct sk_buff *skb,
334 int nhead, int ntail,
336 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
337 unsigned int headroom);
338 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
339 int newheadroom, int newtailroom,
341 extern struct sk_buff * skb_pad(struct sk_buff *skb, int pad);
342 #define dev_kfree_skb(a) kfree_skb(a)
343 extern void skb_over_panic(struct sk_buff *skb, int len,
345 extern void skb_under_panic(struct sk_buff *skb, int len,
348 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
349 int getfrag(void *from, char *to, int offset,
350 int len,int odd, struct sk_buff *skb),
351 void *from, int length);
358 __u32 stepped_offset;
359 struct sk_buff *root_skb;
360 struct sk_buff *cur_skb;
364 extern void skb_prepare_seq_read(struct sk_buff *skb,
365 unsigned int from, unsigned int to,
366 struct skb_seq_state *st);
367 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
368 struct skb_seq_state *st);
369 extern void skb_abort_seq_read(struct skb_seq_state *st);
371 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
372 unsigned int to, struct ts_config *config,
373 struct ts_state *state);
376 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
379 * skb_queue_empty - check if a queue is empty
382 * Returns true if the queue is empty, false otherwise.
384 static inline int skb_queue_empty(const struct sk_buff_head *list)
386 return list->next == (struct sk_buff *)list;
390 * skb_get - reference buffer
391 * @skb: buffer to reference
393 * Makes another reference to a socket buffer and returns a pointer
396 static inline struct sk_buff *skb_get(struct sk_buff *skb)
398 atomic_inc(&skb->users);
403 * If users == 1, we are the only owner and are can avoid redundant
408 * kfree_skb - free an sk_buff
409 * @skb: buffer to free
411 * Drop a reference to the buffer and free it if the usage count has
414 static inline void kfree_skb(struct sk_buff *skb)
416 if (likely(atomic_read(&skb->users) == 1))
418 else if (likely(!atomic_dec_and_test(&skb->users)))
424 * skb_cloned - is the buffer a clone
425 * @skb: buffer to check
427 * Returns true if the buffer was generated with skb_clone() and is
428 * one of multiple shared copies of the buffer. Cloned buffers are
429 * shared data so must not be written to under normal circumstances.
431 static inline int skb_cloned(const struct sk_buff *skb)
433 return skb->cloned &&
434 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
438 * skb_header_cloned - is the header a clone
439 * @skb: buffer to check
441 * Returns true if modifying the header part of the buffer requires
442 * the data to be copied.
444 static inline int skb_header_cloned(const struct sk_buff *skb)
451 dataref = atomic_read(&skb_shinfo(skb)->dataref);
452 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
457 * skb_header_release - release reference to header
458 * @skb: buffer to operate on
460 * Drop a reference to the header part of the buffer. This is done
461 * by acquiring a payload reference. You must not read from the header
462 * part of skb->data after this.
464 static inline void skb_header_release(struct sk_buff *skb)
468 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
472 * skb_shared - is the buffer shared
473 * @skb: buffer to check
475 * Returns true if more than one person has a reference to this
478 static inline int skb_shared(const struct sk_buff *skb)
480 return atomic_read(&skb->users) != 1;
484 * skb_share_check - check if buffer is shared and if so clone it
485 * @skb: buffer to check
486 * @pri: priority for memory allocation
488 * If the buffer is shared the buffer is cloned and the old copy
489 * drops a reference. A new clone with a single reference is returned.
490 * If the buffer is not shared the original buffer is returned. When
491 * being called from interrupt status or with spinlocks held pri must
494 * NULL is returned on a memory allocation failure.
496 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
499 might_sleep_if(pri & __GFP_WAIT);
500 if (skb_shared(skb)) {
501 struct sk_buff *nskb = skb_clone(skb, pri);
509 * Copy shared buffers into a new sk_buff. We effectively do COW on
510 * packets to handle cases where we have a local reader and forward
511 * and a couple of other messy ones. The normal one is tcpdumping
512 * a packet thats being forwarded.
516 * skb_unshare - make a copy of a shared buffer
517 * @skb: buffer to check
518 * @pri: priority for memory allocation
520 * If the socket buffer is a clone then this function creates a new
521 * copy of the data, drops a reference count on the old copy and returns
522 * the new copy with the reference count at 1. If the buffer is not a clone
523 * the original buffer is returned. When called with a spinlock held or
524 * from interrupt state @pri must be %GFP_ATOMIC
526 * %NULL is returned on a memory allocation failure.
528 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
531 might_sleep_if(pri & __GFP_WAIT);
532 if (skb_cloned(skb)) {
533 struct sk_buff *nskb = skb_copy(skb, pri);
534 kfree_skb(skb); /* Free our shared copy */
542 * @list_: list to peek at
544 * Peek an &sk_buff. Unlike most other operations you _MUST_
545 * be careful with this one. A peek leaves the buffer on the
546 * list and someone else may run off with it. You must hold
547 * the appropriate locks or have a private queue to do this.
549 * Returns %NULL for an empty list or a pointer to the head element.
550 * The reference count is not incremented and the reference is therefore
551 * volatile. Use with caution.
553 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
555 struct sk_buff *list = ((struct sk_buff *)list_)->next;
556 if (list == (struct sk_buff *)list_)
563 * @list_: list to peek at
565 * Peek an &sk_buff. Unlike most other operations you _MUST_
566 * be careful with this one. A peek leaves the buffer on the
567 * list and someone else may run off with it. You must hold
568 * the appropriate locks or have a private queue to do this.
570 * Returns %NULL for an empty list or a pointer to the tail element.
571 * The reference count is not incremented and the reference is therefore
572 * volatile. Use with caution.
574 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
576 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
577 if (list == (struct sk_buff *)list_)
583 * skb_queue_len - get queue length
584 * @list_: list to measure
586 * Return the length of an &sk_buff queue.
588 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
593 static inline void skb_queue_head_init(struct sk_buff_head *list)
595 spin_lock_init(&list->lock);
596 list->prev = list->next = (struct sk_buff *)list;
601 * Insert an sk_buff at the start of a list.
603 * The "__skb_xxxx()" functions are the non-atomic ones that
604 * can only be called with interrupts disabled.
608 * __skb_queue_after - queue a buffer at the list head
610 * @prev: place after this buffer
611 * @newsk: buffer to queue
613 * Queue a buffer int the middle of a list. This function takes no locks
614 * and you must therefore hold required locks before calling it.
616 * A buffer cannot be placed on two lists at the same time.
618 static inline void __skb_queue_after(struct sk_buff_head *list,
619 struct sk_buff *prev,
620 struct sk_buff *newsk)
622 struct sk_buff *next;
628 next->prev = prev->next = newsk;
632 * __skb_queue_head - queue a buffer at the list head
634 * @newsk: buffer to queue
636 * Queue a buffer at the start 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 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
642 static inline void __skb_queue_head(struct sk_buff_head *list,
643 struct sk_buff *newsk)
645 __skb_queue_after(list, (struct sk_buff *)list, newsk);
649 * __skb_queue_tail - queue a buffer at the list tail
651 * @newsk: buffer to queue
653 * Queue a buffer at the end of a list. This function takes no locks
654 * and you must therefore hold required locks before calling it.
656 * A buffer cannot be placed on two lists at the same time.
658 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
659 static inline void __skb_queue_tail(struct sk_buff_head *list,
660 struct sk_buff *newsk)
662 struct sk_buff *prev, *next;
665 next = (struct sk_buff *)list;
669 next->prev = prev->next = newsk;
674 * __skb_dequeue - remove from the head of the queue
675 * @list: list to dequeue from
677 * Remove the head of the list. This function does not take any locks
678 * so must be used with appropriate locks held only. The head item is
679 * returned or %NULL if the list is empty.
681 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
682 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
684 struct sk_buff *next, *prev, *result;
686 prev = (struct sk_buff *) list;
695 result->next = result->prev = NULL;
702 * Insert a packet on a list.
704 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
705 static inline void __skb_insert(struct sk_buff *newsk,
706 struct sk_buff *prev, struct sk_buff *next,
707 struct sk_buff_head *list)
711 next->prev = prev->next = newsk;
716 * Place a packet after a given packet in a list.
718 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
719 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
721 __skb_insert(newsk, old, old->next, list);
725 * remove sk_buff from list. _Must_ be called atomically, and with
728 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
729 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
731 struct sk_buff *next, *prev;
736 skb->next = skb->prev = NULL;
742 /* XXX: more streamlined implementation */
745 * __skb_dequeue_tail - remove from the tail of the queue
746 * @list: list to dequeue from
748 * Remove the tail of the list. This function does not take any locks
749 * so must be used with appropriate locks held only. The tail item is
750 * returned or %NULL if the list is empty.
752 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
753 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
755 struct sk_buff *skb = skb_peek_tail(list);
757 __skb_unlink(skb, list);
762 static inline int skb_is_nonlinear(const struct sk_buff *skb)
764 return skb->data_len;
767 static inline unsigned int skb_headlen(const struct sk_buff *skb)
769 return skb->len - skb->data_len;
772 static inline int skb_pagelen(const struct sk_buff *skb)
776 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
777 len += skb_shinfo(skb)->frags[i].size;
778 return len + skb_headlen(skb);
781 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
782 struct page *page, int off, int size)
784 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
787 frag->page_offset = off;
789 skb_shinfo(skb)->nr_frags = i + 1;
792 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
793 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
794 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
797 * Add data to an sk_buff
799 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
801 unsigned char *tmp = skb->tail;
802 SKB_LINEAR_ASSERT(skb);
809 * skb_put - add data to a buffer
810 * @skb: buffer to use
811 * @len: amount of data to add
813 * This function extends the used data area of the buffer. If this would
814 * exceed the total buffer size the kernel will panic. A pointer to the
815 * first byte of the extra data is returned.
817 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
819 unsigned char *tmp = skb->tail;
820 SKB_LINEAR_ASSERT(skb);
823 if (unlikely(skb->tail>skb->end))
824 skb_over_panic(skb, len, current_text_addr());
828 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
836 * skb_push - add data to the start of a buffer
837 * @skb: buffer to use
838 * @len: amount of data to add
840 * This function extends the used data area of the buffer at the buffer
841 * start. If this would exceed the total buffer headroom the kernel will
842 * panic. A pointer to the first byte of the extra data is returned.
844 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
848 if (unlikely(skb->data<skb->head))
849 skb_under_panic(skb, len, current_text_addr());
853 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
856 BUG_ON(skb->len < skb->data_len);
857 return skb->data += len;
861 * skb_pull - remove data from the start of a buffer
862 * @skb: buffer to use
863 * @len: amount of data to remove
865 * This function removes data from the start of a buffer, returning
866 * the memory to the headroom. A pointer to the next data in the buffer
867 * is returned. Once the data has been pulled future pushes will overwrite
870 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
872 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
875 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
877 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
879 if (len > skb_headlen(skb) &&
880 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
883 return skb->data += len;
886 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
888 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
891 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
893 if (likely(len <= skb_headlen(skb)))
895 if (unlikely(len > skb->len))
897 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
901 * skb_headroom - bytes at buffer head
902 * @skb: buffer to check
904 * Return the number of bytes of free space at the head of an &sk_buff.
906 static inline int skb_headroom(const struct sk_buff *skb)
908 return skb->data - skb->head;
912 * skb_tailroom - bytes at buffer end
913 * @skb: buffer to check
915 * Return the number of bytes of free space at the tail of an sk_buff
917 static inline int skb_tailroom(const struct sk_buff *skb)
919 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
923 * skb_reserve - adjust headroom
924 * @skb: buffer to alter
925 * @len: bytes to move
927 * Increase the headroom of an empty &sk_buff by reducing the tail
928 * room. This is only allowed for an empty buffer.
930 static inline void skb_reserve(struct sk_buff *skb, unsigned int len)
937 * CPUs often take a performance hit when accessing unaligned memory
938 * locations. The actual performance hit varies, it can be small if the
939 * hardware handles it or large if we have to take an exception and fix it
942 * Since an ethernet header is 14 bytes network drivers often end up with
943 * the IP header at an unaligned offset. The IP header can be aligned by
944 * shifting the start of the packet by 2 bytes. Drivers should do this
947 * skb_reserve(NET_IP_ALIGN);
949 * The downside to this alignment of the IP header is that the DMA is now
950 * unaligned. On some architectures the cost of an unaligned DMA is high
951 * and this cost outweighs the gains made by aligning the IP header.
953 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
957 #define NET_IP_ALIGN 2
960 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
962 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
964 if (!skb->data_len) {
966 skb->tail = skb->data + len;
968 ___pskb_trim(skb, len, 0);
972 * skb_trim - remove end from a buffer
973 * @skb: buffer to alter
976 * Cut the length of a buffer down by removing data from the tail. If
977 * the buffer is already under the length specified it is not modified.
979 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
982 __skb_trim(skb, len);
986 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
988 if (!skb->data_len) {
990 skb->tail = skb->data+len;
993 return ___pskb_trim(skb, len, 1);
996 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
998 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1002 * skb_orphan - orphan a buffer
1003 * @skb: buffer to orphan
1005 * If a buffer currently has an owner then we call the owner's
1006 * destructor function and make the @skb unowned. The buffer continues
1007 * to exist but is no longer charged to its former owner.
1009 static inline void skb_orphan(struct sk_buff *skb)
1011 if (skb->destructor)
1012 skb->destructor(skb);
1013 skb->destructor = NULL;
1018 * __skb_queue_purge - empty a list
1019 * @list: list to empty
1021 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1022 * the list and one reference dropped. This function does not take the
1023 * list lock and the caller must hold the relevant locks to use it.
1025 extern void skb_queue_purge(struct sk_buff_head *list);
1026 static inline void __skb_queue_purge(struct sk_buff_head *list)
1028 struct sk_buff *skb;
1029 while ((skb = __skb_dequeue(list)) != NULL)
1033 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
1035 * __dev_alloc_skb - allocate an skbuff for sending
1036 * @length: length to allocate
1037 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1039 * Allocate a new &sk_buff and assign it a usage count of one. The
1040 * buffer has unspecified headroom built in. Users should allocate
1041 * the headroom they think they need without accounting for the
1042 * built in space. The built in space is used for optimisations.
1044 * %NULL is returned in there is no free memory.
1046 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1049 struct sk_buff *skb = alloc_skb(length + 16, gfp_mask);
1051 skb_reserve(skb, 16);
1055 extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
1059 * dev_alloc_skb - allocate an skbuff for sending
1060 * @length: length to allocate
1062 * Allocate a new &sk_buff and assign it a usage count of one. The
1063 * buffer has unspecified headroom built in. Users should allocate
1064 * the headroom they think they need without accounting for the
1065 * built in space. The built in space is used for optimisations.
1067 * %NULL is returned in there is no free memory. Although this function
1068 * allocates memory it can be called from an interrupt.
1070 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1072 return __dev_alloc_skb(length, GFP_ATOMIC);
1076 * skb_cow - copy header of skb when it is required
1077 * @skb: buffer to cow
1078 * @headroom: needed headroom
1080 * If the skb passed lacks sufficient headroom or its data part
1081 * is shared, data is reallocated. If reallocation fails, an error
1082 * is returned and original skb is not changed.
1084 * The result is skb with writable area skb->head...skb->tail
1085 * and at least @headroom of space at head.
1087 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1089 int delta = (headroom > 16 ? headroom : 16) - skb_headroom(skb);
1094 if (delta || skb_cloned(skb))
1095 return pskb_expand_head(skb, (delta + 15) & ~15, 0, GFP_ATOMIC);
1100 * skb_padto - pad an skbuff up to a minimal size
1101 * @skb: buffer to pad
1102 * @len: minimal length
1104 * Pads up a buffer to ensure the trailing bytes exist and are
1105 * blanked. If the buffer already contains sufficient data it
1106 * is untouched. Returns the buffer, which may be a replacement
1107 * for the original, or NULL for out of memory - in which case
1108 * the original buffer is still freed.
1111 static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
1113 unsigned int size = skb->len;
1114 if (likely(size >= len))
1116 return skb_pad(skb, len-size);
1119 static inline int skb_add_data(struct sk_buff *skb,
1120 char __user *from, int copy)
1122 const int off = skb->len;
1124 if (skb->ip_summed == CHECKSUM_NONE) {
1126 unsigned int csum = csum_and_copy_from_user(from,
1130 skb->csum = csum_block_add(skb->csum, csum, off);
1133 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1136 __skb_trim(skb, off);
1140 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1141 struct page *page, int off)
1144 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1146 return page == frag->page &&
1147 off == frag->page_offset + frag->size;
1153 * skb_linearize - convert paged skb to linear one
1154 * @skb: buffer to linarize
1155 * @gfp: allocation mode
1157 * If there is no free memory -ENOMEM is returned, otherwise zero
1158 * is returned and the old skb data released.
1160 extern int __skb_linearize(struct sk_buff *skb, gfp_t gfp);
1161 static inline int skb_linearize(struct sk_buff *skb, gfp_t gfp)
1163 return __skb_linearize(skb, gfp);
1167 * skb_postpull_rcsum - update checksum for received skb after pull
1168 * @skb: buffer to update
1169 * @start: start of data before pull
1170 * @len: length of data pulled
1172 * After doing a pull on a received packet, you need to call this to
1173 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1174 * so that it can be recomputed from scratch.
1177 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1178 const void *start, int len)
1180 if (skb->ip_summed == CHECKSUM_HW)
1181 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1185 * pskb_trim_rcsum - trim received skb and update checksum
1186 * @skb: buffer to trim
1189 * This is exactly the same as pskb_trim except that it ensures the
1190 * checksum of received packets are still valid after the operation.
1193 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1195 if (likely(len >= skb->len))
1197 if (skb->ip_summed == CHECKSUM_HW)
1198 skb->ip_summed = CHECKSUM_NONE;
1199 return __pskb_trim(skb, len);
1202 static inline void *kmap_skb_frag(const skb_frag_t *frag)
1204 #ifdef CONFIG_HIGHMEM
1209 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1212 static inline void kunmap_skb_frag(void *vaddr)
1214 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1215 #ifdef CONFIG_HIGHMEM
1220 #define skb_queue_walk(queue, skb) \
1221 for (skb = (queue)->next; \
1222 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1225 #define skb_queue_reverse_walk(queue, skb) \
1226 for (skb = (queue)->prev; \
1227 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1231 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1232 int noblock, int *err);
1233 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1234 struct poll_table_struct *wait);
1235 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1236 int offset, struct iovec *to,
1238 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1241 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1242 extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1243 int len, unsigned int csum);
1244 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1246 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1247 void *from, int len);
1248 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1249 int offset, u8 *to, int len,
1251 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1252 extern void skb_split(struct sk_buff *skb,
1253 struct sk_buff *skb1, const u32 len);
1255 extern void skb_release_data(struct sk_buff *skb);
1257 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1258 int len, void *buffer)
1260 int hlen = skb_headlen(skb);
1262 if (hlen - offset >= len)
1263 return skb->data + offset;
1265 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1271 extern void skb_init(void);
1272 extern void skb_add_mtu(int mtu);
1275 * skb_get_timestamp - get timestamp from a skb
1276 * @skb: skb to get stamp from
1277 * @stamp: pointer to struct timeval to store stamp in
1279 * Timestamps are stored in the skb as offsets to a base timestamp.
1280 * This function converts the offset back to a struct timeval and stores
1283 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1285 stamp->tv_sec = skb->tstamp.off_sec;
1286 stamp->tv_usec = skb->tstamp.off_usec;
1290 * skb_set_timestamp - set timestamp of a skb
1291 * @skb: skb to set stamp of
1292 * @stamp: pointer to struct timeval to get stamp from
1294 * Timestamps are stored in the skb as offsets to a base timestamp.
1295 * This function converts a struct timeval to an offset and stores
1298 static inline void skb_set_timestamp(struct sk_buff *skb, const struct timeval *stamp)
1300 skb->tstamp.off_sec = stamp->tv_sec;
1301 skb->tstamp.off_usec = stamp->tv_usec;
1304 extern void __net_timestamp(struct sk_buff *skb);
1306 extern unsigned int __skb_checksum_complete(struct sk_buff *skb);
1309 * skb_checksum_complete - Calculate checksum of an entire packet
1310 * @skb: packet to process
1312 * This function calculates the checksum over the entire packet plus
1313 * the value of skb->csum. The latter can be used to supply the
1314 * checksum of a pseudo header as used by TCP/UDP. It returns the
1317 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1318 * this function can be used to verify that checksum on received
1319 * packets. In that case the function should return zero if the
1320 * checksum is correct. In particular, this function will return zero
1321 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1322 * hardware has already verified the correctness of the checksum.
1324 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1326 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1327 __skb_checksum_complete(skb);
1330 #ifdef CONFIG_NETFILTER
1331 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1333 if (nfct && atomic_dec_and_test(&nfct->use))
1334 nfct->destroy(nfct);
1336 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1339 atomic_inc(&nfct->use);
1341 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1342 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1345 atomic_inc(&skb->users);
1347 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1353 static inline void nf_reset(struct sk_buff *skb)
1355 nf_conntrack_put(skb->nfct);
1357 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1358 nf_conntrack_put_reasm(skb->nfct_reasm);
1359 skb->nfct_reasm = NULL;
1363 #ifdef CONFIG_BRIDGE_NETFILTER
1364 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1366 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1369 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1372 atomic_inc(&nf_bridge->use);
1374 #endif /* CONFIG_BRIDGE_NETFILTER */
1375 #else /* CONFIG_NETFILTER */
1376 static inline void nf_reset(struct sk_buff *skb) {}
1377 #endif /* CONFIG_NETFILTER */
1379 #endif /* __KERNEL__ */
1380 #endif /* _LINUX_SKBUFF_H */