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) */
36 #define CHECKSUM_NONE 0
38 #define CHECKSUM_UNNECESSARY 2
40 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
41 ~(SMP_CACHE_BYTES - 1))
42 #define SKB_MAX_ORDER(X, ORDER) (((PAGE_SIZE << (ORDER)) - (X) - \
43 sizeof(struct skb_shared_info)) & \
44 ~(SMP_CACHE_BYTES - 1))
45 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
46 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
48 /* A. Checksumming of received packets by device.
50 * NONE: device failed to checksum this packet.
51 * skb->csum is undefined.
53 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
54 * skb->csum is undefined.
55 * It is bad option, but, unfortunately, many of vendors do this.
56 * Apparently with secret goal to sell you new device, when you
57 * will add new protocol to your host. F.e. IPv6. 8)
59 * HW: the most generic way. Device supplied checksum of _all_
60 * the packet as seen by netif_rx in skb->csum.
61 * NOTE: Even if device supports only some protocols, but
62 * is able to produce some skb->csum, it MUST use HW,
65 * B. Checksumming on output.
67 * NONE: skb is checksummed by protocol or csum is not required.
69 * HW: device is required to csum packet as seen by hard_start_xmit
70 * from skb->h.raw to the end and to record the checksum
71 * at skb->h.raw+skb->csum.
73 * Device must show its capabilities in dev->features, set
74 * at device setup time.
75 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
77 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
78 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
79 * TCP/UDP over IPv4. Sigh. Vendors like this
80 * way by an unknown reason. Though, see comment above
81 * about CHECKSUM_UNNECESSARY. 8)
83 * Any questions? No questions, good. --ANK
88 #ifdef CONFIG_NETFILTER
91 void (*destroy)(struct nf_conntrack *);
94 #ifdef CONFIG_BRIDGE_NETFILTER
95 struct nf_bridge_info {
97 struct net_device *physindev;
98 struct net_device *physoutdev;
99 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
100 struct net_device *netoutdev;
103 unsigned long data[32 / sizeof(unsigned long)];
109 struct sk_buff_head {
110 /* These two members must be first. */
111 struct sk_buff *next;
112 struct sk_buff *prev;
120 /* To allow 64K frame to be packed as single skb without frag_list */
121 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
123 typedef struct skb_frag_struct skb_frag_t;
125 struct skb_frag_struct {
131 /* This data is invariant across clones and lives at
132 * the end of the header data, ie. at skb->end.
134 struct skb_shared_info {
136 unsigned short nr_frags;
137 unsigned short tso_size;
138 unsigned short tso_segs;
139 unsigned short ufo_size;
140 unsigned int ip6_frag_id;
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 * struct sk_buff - socket buffer
173 * @next: Next buffer in list
174 * @prev: Previous buffer in list
175 * @sk: Socket we are owned by
176 * @tstamp: Time we arrived
177 * @dev: Device we arrived on/are leaving by
178 * @input_dev: Device we arrived on
179 * @h: Transport layer header
180 * @nh: Network layer header
181 * @mac: Link layer header
182 * @dst: destination entry
183 * @sp: the security path, used for xfrm
184 * @cb: Control buffer. Free for use by every layer. Put private vars here
185 * @len: Length of actual data
186 * @data_len: Data length
187 * @mac_len: Length of link layer header
189 * @local_df: allow local fragmentation
190 * @cloned: Head may be cloned (check refcnt to be sure)
191 * @nohdr: Payload reference only, must not modify header
192 * @pkt_type: Packet class
193 * @fclone: skbuff clone status
194 * @ip_summed: Driver fed us an IP checksum
195 * @priority: Packet queueing priority
196 * @users: User count - see {datagram,tcp}.c
197 * @protocol: Packet protocol from driver
198 * @truesize: Buffer size
199 * @head: Head of buffer
200 * @data: Data head pointer
201 * @tail: Tail pointer
203 * @destructor: Destruct function
204 * @nfmark: Can be used for communication between hooks
205 * @nfct: Associated connection, if any
206 * @ipvs_property: skbuff is owned by ipvs
207 * @nfctinfo: Relationship of this skb to the connection
208 * @nfct_reasm: netfilter conntrack re-assembly pointer
209 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
210 * @tc_index: Traffic control index
211 * @tc_verd: traffic control verdict
215 /* These two members must be first. */
216 struct sk_buff *next;
217 struct sk_buff *prev;
220 struct skb_timeval tstamp;
221 struct net_device *dev;
222 struct net_device *input_dev;
227 struct icmphdr *icmph;
228 struct igmphdr *igmph;
230 struct ipv6hdr *ipv6h;
236 struct ipv6hdr *ipv6h;
245 struct dst_entry *dst;
249 * This is the control buffer. It is free to use for every
250 * layer. Please put your private variables there. If you
251 * want to keep them across layers you have to do a skb_clone()
252 * first. This is owned by whoever has the skb queued ATM.
271 void (*destructor)(struct sk_buff *skb);
272 #ifdef CONFIG_NETFILTER
273 struct nf_conntrack *nfct;
274 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
275 struct sk_buff *nfct_reasm;
277 #ifdef CONFIG_BRIDGE_NETFILTER
278 struct nf_bridge_info *nf_bridge;
281 #endif /* CONFIG_NETFILTER */
282 #ifdef CONFIG_NET_SCHED
283 __u16 tc_index; /* traffic control index */
284 #ifdef CONFIG_NET_CLS_ACT
285 __u16 tc_verd; /* traffic control verdict */
290 /* These elements must be at the end, see alloc_skb() for details. */
291 unsigned int truesize;
301 * Handling routines are only of interest to the kernel
303 #include <linux/slab.h>
305 #include <asm/system.h>
307 extern void kfree_skb(struct sk_buff *skb);
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,
347 extern void skb_truesize_bug(struct sk_buff *skb);
349 static inline void skb_truesize_check(struct sk_buff *skb)
351 if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len))
352 skb_truesize_bug(skb);
355 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
356 int getfrag(void *from, char *to, int offset,
357 int len,int odd, struct sk_buff *skb),
358 void *from, int length);
365 __u32 stepped_offset;
366 struct sk_buff *root_skb;
367 struct sk_buff *cur_skb;
371 extern void skb_prepare_seq_read(struct sk_buff *skb,
372 unsigned int from, unsigned int to,
373 struct skb_seq_state *st);
374 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
375 struct skb_seq_state *st);
376 extern void skb_abort_seq_read(struct skb_seq_state *st);
378 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
379 unsigned int to, struct ts_config *config,
380 struct ts_state *state);
383 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
386 * skb_queue_empty - check if a queue is empty
389 * Returns true if the queue is empty, false otherwise.
391 static inline int skb_queue_empty(const struct sk_buff_head *list)
393 return list->next == (struct sk_buff *)list;
397 * skb_get - reference buffer
398 * @skb: buffer to reference
400 * Makes another reference to a socket buffer and returns a pointer
403 static inline struct sk_buff *skb_get(struct sk_buff *skb)
405 atomic_inc(&skb->users);
410 * If users == 1, we are the only owner and are can avoid redundant
415 * skb_cloned - is the buffer a clone
416 * @skb: buffer to check
418 * Returns true if the buffer was generated with skb_clone() and is
419 * one of multiple shared copies of the buffer. Cloned buffers are
420 * shared data so must not be written to under normal circumstances.
422 static inline int skb_cloned(const struct sk_buff *skb)
424 return skb->cloned &&
425 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
429 * skb_header_cloned - is the header a clone
430 * @skb: buffer to check
432 * Returns true if modifying the header part of the buffer requires
433 * the data to be copied.
435 static inline int skb_header_cloned(const struct sk_buff *skb)
442 dataref = atomic_read(&skb_shinfo(skb)->dataref);
443 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
448 * skb_header_release - release reference to header
449 * @skb: buffer to operate on
451 * Drop a reference to the header part of the buffer. This is done
452 * by acquiring a payload reference. You must not read from the header
453 * part of skb->data after this.
455 static inline void skb_header_release(struct sk_buff *skb)
459 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
463 * skb_shared - is the buffer shared
464 * @skb: buffer to check
466 * Returns true if more than one person has a reference to this
469 static inline int skb_shared(const struct sk_buff *skb)
471 return atomic_read(&skb->users) != 1;
475 * skb_share_check - check if buffer is shared and if so clone it
476 * @skb: buffer to check
477 * @pri: priority for memory allocation
479 * If the buffer is shared the buffer is cloned and the old copy
480 * drops a reference. A new clone with a single reference is returned.
481 * If the buffer is not shared the original buffer is returned. When
482 * being called from interrupt status or with spinlocks held pri must
485 * NULL is returned on a memory allocation failure.
487 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
490 might_sleep_if(pri & __GFP_WAIT);
491 if (skb_shared(skb)) {
492 struct sk_buff *nskb = skb_clone(skb, pri);
500 * Copy shared buffers into a new sk_buff. We effectively do COW on
501 * packets to handle cases where we have a local reader and forward
502 * and a couple of other messy ones. The normal one is tcpdumping
503 * a packet thats being forwarded.
507 * skb_unshare - make a copy of a shared buffer
508 * @skb: buffer to check
509 * @pri: priority for memory allocation
511 * If the socket buffer is a clone then this function creates a new
512 * copy of the data, drops a reference count on the old copy and returns
513 * the new copy with the reference count at 1. If the buffer is not a clone
514 * the original buffer is returned. When called with a spinlock held or
515 * from interrupt state @pri must be %GFP_ATOMIC
517 * %NULL is returned on a memory allocation failure.
519 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
522 might_sleep_if(pri & __GFP_WAIT);
523 if (skb_cloned(skb)) {
524 struct sk_buff *nskb = skb_copy(skb, pri);
525 kfree_skb(skb); /* Free our shared copy */
533 * @list_: list to peek at
535 * Peek an &sk_buff. Unlike most other operations you _MUST_
536 * be careful with this one. A peek leaves the buffer on the
537 * list and someone else may run off with it. You must hold
538 * the appropriate locks or have a private queue to do this.
540 * Returns %NULL for an empty list or a pointer to the head element.
541 * The reference count is not incremented and the reference is therefore
542 * volatile. Use with caution.
544 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
546 struct sk_buff *list = ((struct sk_buff *)list_)->next;
547 if (list == (struct sk_buff *)list_)
554 * @list_: list to peek at
556 * Peek an &sk_buff. Unlike most other operations you _MUST_
557 * be careful with this one. A peek leaves the buffer on the
558 * list and someone else may run off with it. You must hold
559 * the appropriate locks or have a private queue to do this.
561 * Returns %NULL for an empty list or a pointer to the tail element.
562 * The reference count is not incremented and the reference is therefore
563 * volatile. Use with caution.
565 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
567 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
568 if (list == (struct sk_buff *)list_)
574 * skb_queue_len - get queue length
575 * @list_: list to measure
577 * Return the length of an &sk_buff queue.
579 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
584 static inline void skb_queue_head_init(struct sk_buff_head *list)
586 spin_lock_init(&list->lock);
587 list->prev = list->next = (struct sk_buff *)list;
592 * Insert an sk_buff at the start of a list.
594 * The "__skb_xxxx()" functions are the non-atomic ones that
595 * can only be called with interrupts disabled.
599 * __skb_queue_after - queue a buffer at the list head
601 * @prev: place after this buffer
602 * @newsk: buffer to queue
604 * Queue a buffer int the middle of a list. This function takes no locks
605 * and you must therefore hold required locks before calling it.
607 * A buffer cannot be placed on two lists at the same time.
609 static inline void __skb_queue_after(struct sk_buff_head *list,
610 struct sk_buff *prev,
611 struct sk_buff *newsk)
613 struct sk_buff *next;
619 next->prev = prev->next = newsk;
623 * __skb_queue_head - queue a buffer at the list head
625 * @newsk: buffer to queue
627 * Queue a buffer at the start of a list. This function takes no locks
628 * and you must therefore hold required locks before calling it.
630 * A buffer cannot be placed on two lists at the same time.
632 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
633 static inline void __skb_queue_head(struct sk_buff_head *list,
634 struct sk_buff *newsk)
636 __skb_queue_after(list, (struct sk_buff *)list, newsk);
640 * __skb_queue_tail - queue a buffer at the list tail
642 * @newsk: buffer to queue
644 * Queue a buffer at the end of a list. This function takes no locks
645 * and you must therefore hold required locks before calling it.
647 * A buffer cannot be placed on two lists at the same time.
649 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
650 static inline void __skb_queue_tail(struct sk_buff_head *list,
651 struct sk_buff *newsk)
653 struct sk_buff *prev, *next;
656 next = (struct sk_buff *)list;
660 next->prev = prev->next = newsk;
665 * __skb_dequeue - remove from the head of the queue
666 * @list: list to dequeue from
668 * Remove the head of the list. This function does not take any locks
669 * so must be used with appropriate locks held only. The head item is
670 * returned or %NULL if the list is empty.
672 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
673 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
675 struct sk_buff *next, *prev, *result;
677 prev = (struct sk_buff *) list;
686 result->next = result->prev = NULL;
693 * Insert a packet on a list.
695 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
696 static inline void __skb_insert(struct sk_buff *newsk,
697 struct sk_buff *prev, struct sk_buff *next,
698 struct sk_buff_head *list)
702 next->prev = prev->next = newsk;
707 * Place a packet after a given packet in a list.
709 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
710 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
712 __skb_insert(newsk, old, old->next, list);
716 * remove sk_buff from list. _Must_ be called atomically, and with
719 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
720 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
722 struct sk_buff *next, *prev;
727 skb->next = skb->prev = NULL;
733 /* XXX: more streamlined implementation */
736 * __skb_dequeue_tail - remove from the tail of the queue
737 * @list: list to dequeue from
739 * Remove the tail of the list. This function does not take any locks
740 * so must be used with appropriate locks held only. The tail item is
741 * returned or %NULL if the list is empty.
743 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
744 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
746 struct sk_buff *skb = skb_peek_tail(list);
748 __skb_unlink(skb, list);
753 static inline int skb_is_nonlinear(const struct sk_buff *skb)
755 return skb->data_len;
758 static inline unsigned int skb_headlen(const struct sk_buff *skb)
760 return skb->len - skb->data_len;
763 static inline int skb_pagelen(const struct sk_buff *skb)
767 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
768 len += skb_shinfo(skb)->frags[i].size;
769 return len + skb_headlen(skb);
772 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
773 struct page *page, int off, int size)
775 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
778 frag->page_offset = off;
780 skb_shinfo(skb)->nr_frags = i + 1;
783 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
784 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
785 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
788 * Add data to an sk_buff
790 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
792 unsigned char *tmp = skb->tail;
793 SKB_LINEAR_ASSERT(skb);
800 * skb_put - add data to a buffer
801 * @skb: buffer to use
802 * @len: amount of data to add
804 * This function extends the used data area of the buffer. If this would
805 * exceed the total buffer size the kernel will panic. A pointer to the
806 * first byte of the extra data is returned.
808 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
810 unsigned char *tmp = skb->tail;
811 SKB_LINEAR_ASSERT(skb);
814 if (unlikely(skb->tail>skb->end))
815 skb_over_panic(skb, len, current_text_addr());
819 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
827 * skb_push - add data to the start of a buffer
828 * @skb: buffer to use
829 * @len: amount of data to add
831 * This function extends the used data area of the buffer at the buffer
832 * start. If this would exceed the total buffer headroom the kernel will
833 * panic. A pointer to the first byte of the extra data is returned.
835 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
839 if (unlikely(skb->data<skb->head))
840 skb_under_panic(skb, len, current_text_addr());
844 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
847 BUG_ON(skb->len < skb->data_len);
848 return skb->data += len;
852 * skb_pull - remove data from the start of a buffer
853 * @skb: buffer to use
854 * @len: amount of data to remove
856 * This function removes data from the start of a buffer, returning
857 * the memory to the headroom. A pointer to the next data in the buffer
858 * is returned. Once the data has been pulled future pushes will overwrite
861 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
863 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
866 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
868 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
870 if (len > skb_headlen(skb) &&
871 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
874 return skb->data += len;
877 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
879 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
882 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
884 if (likely(len <= skb_headlen(skb)))
886 if (unlikely(len > skb->len))
888 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
892 * skb_headroom - bytes at buffer head
893 * @skb: buffer to check
895 * Return the number of bytes of free space at the head of an &sk_buff.
897 static inline int skb_headroom(const struct sk_buff *skb)
899 return skb->data - skb->head;
903 * skb_tailroom - bytes at buffer end
904 * @skb: buffer to check
906 * Return the number of bytes of free space at the tail of an sk_buff
908 static inline int skb_tailroom(const struct sk_buff *skb)
910 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
914 * skb_reserve - adjust headroom
915 * @skb: buffer to alter
916 * @len: bytes to move
918 * Increase the headroom of an empty &sk_buff by reducing the tail
919 * room. This is only allowed for an empty buffer.
921 static inline void skb_reserve(struct sk_buff *skb, int len)
928 * CPUs often take a performance hit when accessing unaligned memory
929 * locations. The actual performance hit varies, it can be small if the
930 * hardware handles it or large if we have to take an exception and fix it
933 * Since an ethernet header is 14 bytes network drivers often end up with
934 * the IP header at an unaligned offset. The IP header can be aligned by
935 * shifting the start of the packet by 2 bytes. Drivers should do this
938 * skb_reserve(NET_IP_ALIGN);
940 * The downside to this alignment of the IP header is that the DMA is now
941 * unaligned. On some architectures the cost of an unaligned DMA is high
942 * and this cost outweighs the gains made by aligning the IP header.
944 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
948 #define NET_IP_ALIGN 2
952 * The networking layer reserves some headroom in skb data (via
953 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
954 * the header has to grow. In the default case, if the header has to grow
955 * 16 bytes or less we avoid the reallocation.
957 * Unfortunately this headroom changes the DMA alignment of the resulting
958 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
959 * on some architectures. An architecture can override this value,
960 * perhaps setting it to a cacheline in size (since that will maintain
961 * cacheline alignment of the DMA). It must be a power of 2.
963 * Various parts of the networking layer expect at least 16 bytes of
964 * headroom, you should not reduce this.
967 #define NET_SKB_PAD 16
970 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
972 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
974 if (!skb->data_len) {
976 skb->tail = skb->data + len;
978 ___pskb_trim(skb, len, 0);
982 * skb_trim - remove end from a buffer
983 * @skb: buffer to alter
986 * Cut the length of a buffer down by removing data from the tail. If
987 * the buffer is already under the length specified it is not modified.
989 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
992 __skb_trim(skb, len);
996 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
998 if (!skb->data_len) {
1000 skb->tail = skb->data+len;
1003 return ___pskb_trim(skb, len, 1);
1006 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1008 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1012 * skb_orphan - orphan a buffer
1013 * @skb: buffer to orphan
1015 * If a buffer currently has an owner then we call the owner's
1016 * destructor function and make the @skb unowned. The buffer continues
1017 * to exist but is no longer charged to its former owner.
1019 static inline void skb_orphan(struct sk_buff *skb)
1021 if (skb->destructor)
1022 skb->destructor(skb);
1023 skb->destructor = NULL;
1028 * __skb_queue_purge - empty a list
1029 * @list: list to empty
1031 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1032 * the list and one reference dropped. This function does not take the
1033 * list lock and the caller must hold the relevant locks to use it.
1035 extern void skb_queue_purge(struct sk_buff_head *list);
1036 static inline void __skb_queue_purge(struct sk_buff_head *list)
1038 struct sk_buff *skb;
1039 while ((skb = __skb_dequeue(list)) != NULL)
1043 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
1045 * __dev_alloc_skb - allocate an skbuff for sending
1046 * @length: length to allocate
1047 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1049 * Allocate a new &sk_buff and assign it a usage count of one. The
1050 * buffer has unspecified headroom built in. Users should allocate
1051 * the headroom they think they need without accounting for the
1052 * built in space. The built in space is used for optimisations.
1054 * %NULL is returned in there is no free memory.
1056 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1059 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1061 skb_reserve(skb, NET_SKB_PAD);
1065 extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
1069 * dev_alloc_skb - allocate an skbuff for sending
1070 * @length: length to allocate
1072 * Allocate a new &sk_buff and assign it a usage count of one. The
1073 * buffer has unspecified headroom built in. Users should allocate
1074 * the headroom they think they need without accounting for the
1075 * built in space. The built in space is used for optimisations.
1077 * %NULL is returned in there is no free memory. Although this function
1078 * allocates memory it can be called from an interrupt.
1080 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1082 return __dev_alloc_skb(length, GFP_ATOMIC);
1086 * skb_cow - copy header of skb when it is required
1087 * @skb: buffer to cow
1088 * @headroom: needed headroom
1090 * If the skb passed lacks sufficient headroom or its data part
1091 * is shared, data is reallocated. If reallocation fails, an error
1092 * is returned and original skb is not changed.
1094 * The result is skb with writable area skb->head...skb->tail
1095 * and at least @headroom of space at head.
1097 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1099 int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) -
1105 if (delta || skb_cloned(skb))
1106 return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) &
1107 ~(NET_SKB_PAD-1), 0, GFP_ATOMIC);
1112 * skb_padto - pad an skbuff up to a minimal size
1113 * @skb: buffer to pad
1114 * @len: minimal length
1116 * Pads up a buffer to ensure the trailing bytes exist and are
1117 * blanked. If the buffer already contains sufficient data it
1118 * is untouched. Returns the buffer, which may be a replacement
1119 * for the original, or NULL for out of memory - in which case
1120 * the original buffer is still freed.
1123 static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
1125 unsigned int size = skb->len;
1126 if (likely(size >= len))
1128 return skb_pad(skb, len-size);
1131 static inline int skb_add_data(struct sk_buff *skb,
1132 char __user *from, int copy)
1134 const int off = skb->len;
1136 if (skb->ip_summed == CHECKSUM_NONE) {
1138 unsigned int csum = csum_and_copy_from_user(from,
1142 skb->csum = csum_block_add(skb->csum, csum, off);
1145 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1148 __skb_trim(skb, off);
1152 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1153 struct page *page, int off)
1156 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1158 return page == frag->page &&
1159 off == frag->page_offset + frag->size;
1165 * skb_linearize - convert paged skb to linear one
1166 * @skb: buffer to linarize
1167 * @gfp: allocation mode
1169 * If there is no free memory -ENOMEM is returned, otherwise zero
1170 * is returned and the old skb data released.
1172 extern int __skb_linearize(struct sk_buff *skb, gfp_t gfp);
1173 static inline int skb_linearize(struct sk_buff *skb, gfp_t gfp)
1175 return __skb_linearize(skb, gfp);
1179 * skb_postpull_rcsum - update checksum for received skb after pull
1180 * @skb: buffer to update
1181 * @start: start of data before pull
1182 * @len: length of data pulled
1184 * After doing a pull on a received packet, you need to call this to
1185 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1186 * so that it can be recomputed from scratch.
1189 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1190 const void *start, unsigned int len)
1192 if (skb->ip_summed == CHECKSUM_HW)
1193 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1196 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1199 * pskb_trim_rcsum - trim received skb and update checksum
1200 * @skb: buffer to trim
1203 * This is exactly the same as pskb_trim except that it ensures the
1204 * checksum of received packets are still valid after the operation.
1207 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1209 if (likely(len >= skb->len))
1211 if (skb->ip_summed == CHECKSUM_HW)
1212 skb->ip_summed = CHECKSUM_NONE;
1213 return __pskb_trim(skb, len);
1216 static inline void *kmap_skb_frag(const skb_frag_t *frag)
1218 #ifdef CONFIG_HIGHMEM
1223 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1226 static inline void kunmap_skb_frag(void *vaddr)
1228 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1229 #ifdef CONFIG_HIGHMEM
1234 #define skb_queue_walk(queue, skb) \
1235 for (skb = (queue)->next; \
1236 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1239 #define skb_queue_reverse_walk(queue, skb) \
1240 for (skb = (queue)->prev; \
1241 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1245 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1246 int noblock, int *err);
1247 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1248 struct poll_table_struct *wait);
1249 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1250 int offset, struct iovec *to,
1252 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1255 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1256 extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1257 unsigned int flags);
1258 extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1259 int len, unsigned int csum);
1260 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1262 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1263 void *from, int len);
1264 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1265 int offset, u8 *to, int len,
1267 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1268 extern void skb_split(struct sk_buff *skb,
1269 struct sk_buff *skb1, const u32 len);
1271 extern void skb_release_data(struct sk_buff *skb);
1273 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1274 int len, void *buffer)
1276 int hlen = skb_headlen(skb);
1278 if (hlen - offset >= len)
1279 return skb->data + offset;
1281 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1287 extern void skb_init(void);
1288 extern void skb_add_mtu(int mtu);
1291 * skb_get_timestamp - get timestamp from a skb
1292 * @skb: skb to get stamp from
1293 * @stamp: pointer to struct timeval to store stamp in
1295 * Timestamps are stored in the skb as offsets to a base timestamp.
1296 * This function converts the offset back to a struct timeval and stores
1299 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1301 stamp->tv_sec = skb->tstamp.off_sec;
1302 stamp->tv_usec = skb->tstamp.off_usec;
1306 * skb_set_timestamp - set timestamp of a skb
1307 * @skb: skb to set stamp of
1308 * @stamp: pointer to struct timeval to get stamp from
1310 * Timestamps are stored in the skb as offsets to a base timestamp.
1311 * This function converts a struct timeval to an offset and stores
1314 static inline void skb_set_timestamp(struct sk_buff *skb, const struct timeval *stamp)
1316 skb->tstamp.off_sec = stamp->tv_sec;
1317 skb->tstamp.off_usec = stamp->tv_usec;
1320 extern void __net_timestamp(struct sk_buff *skb);
1322 extern unsigned int __skb_checksum_complete(struct sk_buff *skb);
1325 * skb_checksum_complete - Calculate checksum of an entire packet
1326 * @skb: packet to process
1328 * This function calculates the checksum over the entire packet plus
1329 * the value of skb->csum. The latter can be used to supply the
1330 * checksum of a pseudo header as used by TCP/UDP. It returns the
1333 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1334 * this function can be used to verify that checksum on received
1335 * packets. In that case the function should return zero if the
1336 * checksum is correct. In particular, this function will return zero
1337 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1338 * hardware has already verified the correctness of the checksum.
1340 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1342 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1343 __skb_checksum_complete(skb);
1346 #ifdef CONFIG_NETFILTER
1347 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1349 if (nfct && atomic_dec_and_test(&nfct->use))
1350 nfct->destroy(nfct);
1352 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1355 atomic_inc(&nfct->use);
1357 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1358 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1361 atomic_inc(&skb->users);
1363 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1369 #ifdef CONFIG_BRIDGE_NETFILTER
1370 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1372 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1375 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1378 atomic_inc(&nf_bridge->use);
1380 #endif /* CONFIG_BRIDGE_NETFILTER */
1381 static inline void nf_reset(struct sk_buff *skb)
1383 nf_conntrack_put(skb->nfct);
1385 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1386 nf_conntrack_put_reasm(skb->nfct_reasm);
1387 skb->nfct_reasm = NULL;
1389 #ifdef CONFIG_BRIDGE_NETFILTER
1390 nf_bridge_put(skb->nf_bridge);
1391 skb->nf_bridge = NULL;
1395 #else /* CONFIG_NETFILTER */
1396 static inline void nf_reset(struct sk_buff *skb) {}
1397 #endif /* CONFIG_NETFILTER */
1399 #endif /* __KERNEL__ */
1400 #endif /* _LINUX_SKBUFF_H */