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,
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 * skb_cloned - is the buffer a clone
409 * @skb: buffer to check
411 * Returns true if the buffer was generated with skb_clone() and is
412 * one of multiple shared copies of the buffer. Cloned buffers are
413 * shared data so must not be written to under normal circumstances.
415 static inline int skb_cloned(const struct sk_buff *skb)
417 return skb->cloned &&
418 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
422 * skb_header_cloned - is the header a clone
423 * @skb: buffer to check
425 * Returns true if modifying the header part of the buffer requires
426 * the data to be copied.
428 static inline int skb_header_cloned(const struct sk_buff *skb)
435 dataref = atomic_read(&skb_shinfo(skb)->dataref);
436 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
441 * skb_header_release - release reference to header
442 * @skb: buffer to operate on
444 * Drop a reference to the header part of the buffer. This is done
445 * by acquiring a payload reference. You must not read from the header
446 * part of skb->data after this.
448 static inline void skb_header_release(struct sk_buff *skb)
452 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
456 * skb_shared - is the buffer shared
457 * @skb: buffer to check
459 * Returns true if more than one person has a reference to this
462 static inline int skb_shared(const struct sk_buff *skb)
464 return atomic_read(&skb->users) != 1;
468 * skb_share_check - check if buffer is shared and if so clone it
469 * @skb: buffer to check
470 * @pri: priority for memory allocation
472 * If the buffer is shared the buffer is cloned and the old copy
473 * drops a reference. A new clone with a single reference is returned.
474 * If the buffer is not shared the original buffer is returned. When
475 * being called from interrupt status or with spinlocks held pri must
478 * NULL is returned on a memory allocation failure.
480 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
483 might_sleep_if(pri & __GFP_WAIT);
484 if (skb_shared(skb)) {
485 struct sk_buff *nskb = skb_clone(skb, pri);
493 * Copy shared buffers into a new sk_buff. We effectively do COW on
494 * packets to handle cases where we have a local reader and forward
495 * and a couple of other messy ones. The normal one is tcpdumping
496 * a packet thats being forwarded.
500 * skb_unshare - make a copy of a shared buffer
501 * @skb: buffer to check
502 * @pri: priority for memory allocation
504 * If the socket buffer is a clone then this function creates a new
505 * copy of the data, drops a reference count on the old copy and returns
506 * the new copy with the reference count at 1. If the buffer is not a clone
507 * the original buffer is returned. When called with a spinlock held or
508 * from interrupt state @pri must be %GFP_ATOMIC
510 * %NULL is returned on a memory allocation failure.
512 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
515 might_sleep_if(pri & __GFP_WAIT);
516 if (skb_cloned(skb)) {
517 struct sk_buff *nskb = skb_copy(skb, pri);
518 kfree_skb(skb); /* Free our shared copy */
526 * @list_: list to peek at
528 * Peek an &sk_buff. Unlike most other operations you _MUST_
529 * be careful with this one. A peek leaves the buffer on the
530 * list and someone else may run off with it. You must hold
531 * the appropriate locks or have a private queue to do this.
533 * Returns %NULL for an empty list or a pointer to the head element.
534 * The reference count is not incremented and the reference is therefore
535 * volatile. Use with caution.
537 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
539 struct sk_buff *list = ((struct sk_buff *)list_)->next;
540 if (list == (struct sk_buff *)list_)
547 * @list_: list to peek at
549 * Peek an &sk_buff. Unlike most other operations you _MUST_
550 * be careful with this one. A peek leaves the buffer on the
551 * list and someone else may run off with it. You must hold
552 * the appropriate locks or have a private queue to do this.
554 * Returns %NULL for an empty list or a pointer to the tail element.
555 * The reference count is not incremented and the reference is therefore
556 * volatile. Use with caution.
558 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
560 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
561 if (list == (struct sk_buff *)list_)
567 * skb_queue_len - get queue length
568 * @list_: list to measure
570 * Return the length of an &sk_buff queue.
572 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
577 static inline void skb_queue_head_init(struct sk_buff_head *list)
579 spin_lock_init(&list->lock);
580 list->prev = list->next = (struct sk_buff *)list;
585 * Insert an sk_buff at the start of a list.
587 * The "__skb_xxxx()" functions are the non-atomic ones that
588 * can only be called with interrupts disabled.
592 * __skb_queue_after - queue a buffer at the list head
594 * @prev: place after this buffer
595 * @newsk: buffer to queue
597 * Queue a buffer int the middle of a list. This function takes no locks
598 * and you must therefore hold required locks before calling it.
600 * A buffer cannot be placed on two lists at the same time.
602 static inline void __skb_queue_after(struct sk_buff_head *list,
603 struct sk_buff *prev,
604 struct sk_buff *newsk)
606 struct sk_buff *next;
612 next->prev = prev->next = newsk;
616 * __skb_queue_head - queue a buffer at the list head
618 * @newsk: buffer to queue
620 * Queue a buffer at the start of a list. This function takes no locks
621 * and you must therefore hold required locks before calling it.
623 * A buffer cannot be placed on two lists at the same time.
625 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
626 static inline void __skb_queue_head(struct sk_buff_head *list,
627 struct sk_buff *newsk)
629 __skb_queue_after(list, (struct sk_buff *)list, newsk);
633 * __skb_queue_tail - queue a buffer at the list tail
635 * @newsk: buffer to queue
637 * Queue a buffer at the end of a list. This function takes no locks
638 * and you must therefore hold required locks before calling it.
640 * A buffer cannot be placed on two lists at the same time.
642 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
643 static inline void __skb_queue_tail(struct sk_buff_head *list,
644 struct sk_buff *newsk)
646 struct sk_buff *prev, *next;
649 next = (struct sk_buff *)list;
653 next->prev = prev->next = newsk;
658 * __skb_dequeue - remove from the head of the queue
659 * @list: list to dequeue from
661 * Remove the head of the list. This function does not take any locks
662 * so must be used with appropriate locks held only. The head item is
663 * returned or %NULL if the list is empty.
665 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
666 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
668 struct sk_buff *next, *prev, *result;
670 prev = (struct sk_buff *) list;
679 result->next = result->prev = NULL;
686 * Insert a packet on a list.
688 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
689 static inline void __skb_insert(struct sk_buff *newsk,
690 struct sk_buff *prev, struct sk_buff *next,
691 struct sk_buff_head *list)
695 next->prev = prev->next = newsk;
700 * Place a packet after a given packet in a list.
702 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
703 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
705 __skb_insert(newsk, old, old->next, list);
709 * remove sk_buff from list. _Must_ be called atomically, and with
712 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
713 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
715 struct sk_buff *next, *prev;
720 skb->next = skb->prev = NULL;
726 /* XXX: more streamlined implementation */
729 * __skb_dequeue_tail - remove from the tail of the queue
730 * @list: list to dequeue from
732 * Remove the tail of the list. This function does not take any locks
733 * so must be used with appropriate locks held only. The tail item is
734 * returned or %NULL if the list is empty.
736 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
737 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
739 struct sk_buff *skb = skb_peek_tail(list);
741 __skb_unlink(skb, list);
746 static inline int skb_is_nonlinear(const struct sk_buff *skb)
748 return skb->data_len;
751 static inline unsigned int skb_headlen(const struct sk_buff *skb)
753 return skb->len - skb->data_len;
756 static inline int skb_pagelen(const struct sk_buff *skb)
760 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
761 len += skb_shinfo(skb)->frags[i].size;
762 return len + skb_headlen(skb);
765 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
766 struct page *page, int off, int size)
768 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
771 frag->page_offset = off;
773 skb_shinfo(skb)->nr_frags = i + 1;
776 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
777 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
778 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
781 * Add data to an sk_buff
783 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
785 unsigned char *tmp = skb->tail;
786 SKB_LINEAR_ASSERT(skb);
793 * skb_put - add data to a buffer
794 * @skb: buffer to use
795 * @len: amount of data to add
797 * This function extends the used data area of the buffer. If this would
798 * exceed the total buffer size the kernel will panic. A pointer to the
799 * first byte of the extra data is returned.
801 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
803 unsigned char *tmp = skb->tail;
804 SKB_LINEAR_ASSERT(skb);
807 if (unlikely(skb->tail>skb->end))
808 skb_over_panic(skb, len, current_text_addr());
812 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
820 * skb_push - add data to the start of a buffer
821 * @skb: buffer to use
822 * @len: amount of data to add
824 * This function extends the used data area of the buffer at the buffer
825 * start. If this would exceed the total buffer headroom the kernel will
826 * panic. A pointer to the first byte of the extra data is returned.
828 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
832 if (unlikely(skb->data<skb->head))
833 skb_under_panic(skb, len, current_text_addr());
837 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
840 BUG_ON(skb->len < skb->data_len);
841 return skb->data += len;
845 * skb_pull - remove data from the start of a buffer
846 * @skb: buffer to use
847 * @len: amount of data to remove
849 * This function removes data from the start of a buffer, returning
850 * the memory to the headroom. A pointer to the next data in the buffer
851 * is returned. Once the data has been pulled future pushes will overwrite
854 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
856 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
859 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
861 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
863 if (len > skb_headlen(skb) &&
864 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
867 return skb->data += len;
870 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
872 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
875 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
877 if (likely(len <= skb_headlen(skb)))
879 if (unlikely(len > skb->len))
881 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
885 * skb_headroom - bytes at buffer head
886 * @skb: buffer to check
888 * Return the number of bytes of free space at the head of an &sk_buff.
890 static inline int skb_headroom(const struct sk_buff *skb)
892 return skb->data - skb->head;
896 * skb_tailroom - bytes at buffer end
897 * @skb: buffer to check
899 * Return the number of bytes of free space at the tail of an sk_buff
901 static inline int skb_tailroom(const struct sk_buff *skb)
903 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
907 * skb_reserve - adjust headroom
908 * @skb: buffer to alter
909 * @len: bytes to move
911 * Increase the headroom of an empty &sk_buff by reducing the tail
912 * room. This is only allowed for an empty buffer.
914 static inline void skb_reserve(struct sk_buff *skb, int len)
921 * CPUs often take a performance hit when accessing unaligned memory
922 * locations. The actual performance hit varies, it can be small if the
923 * hardware handles it or large if we have to take an exception and fix it
926 * Since an ethernet header is 14 bytes network drivers often end up with
927 * the IP header at an unaligned offset. The IP header can be aligned by
928 * shifting the start of the packet by 2 bytes. Drivers should do this
931 * skb_reserve(NET_IP_ALIGN);
933 * The downside to this alignment of the IP header is that the DMA is now
934 * unaligned. On some architectures the cost of an unaligned DMA is high
935 * and this cost outweighs the gains made by aligning the IP header.
937 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
941 #define NET_IP_ALIGN 2
944 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
946 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
948 if (!skb->data_len) {
950 skb->tail = skb->data + len;
952 ___pskb_trim(skb, len, 0);
956 * skb_trim - remove end from a buffer
957 * @skb: buffer to alter
960 * Cut the length of a buffer down by removing data from the tail. If
961 * the buffer is already under the length specified it is not modified.
963 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
966 __skb_trim(skb, len);
970 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
972 if (!skb->data_len) {
974 skb->tail = skb->data+len;
977 return ___pskb_trim(skb, len, 1);
980 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
982 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
986 * skb_orphan - orphan a buffer
987 * @skb: buffer to orphan
989 * If a buffer currently has an owner then we call the owner's
990 * destructor function and make the @skb unowned. The buffer continues
991 * to exist but is no longer charged to its former owner.
993 static inline void skb_orphan(struct sk_buff *skb)
996 skb->destructor(skb);
997 skb->destructor = NULL;
1002 * __skb_queue_purge - empty a list
1003 * @list: list to empty
1005 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1006 * the list and one reference dropped. This function does not take the
1007 * list lock and the caller must hold the relevant locks to use it.
1009 extern void skb_queue_purge(struct sk_buff_head *list);
1010 static inline void __skb_queue_purge(struct sk_buff_head *list)
1012 struct sk_buff *skb;
1013 while ((skb = __skb_dequeue(list)) != NULL)
1017 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
1019 * __dev_alloc_skb - allocate an skbuff for sending
1020 * @length: length to allocate
1021 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1023 * Allocate a new &sk_buff and assign it a usage count of one. The
1024 * buffer has unspecified headroom built in. Users should allocate
1025 * the headroom they think they need without accounting for the
1026 * built in space. The built in space is used for optimisations.
1028 * %NULL is returned in there is no free memory.
1030 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1033 struct sk_buff *skb = alloc_skb(length + 16, gfp_mask);
1035 skb_reserve(skb, 16);
1039 extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
1043 * dev_alloc_skb - allocate an skbuff for sending
1044 * @length: length to allocate
1046 * Allocate a new &sk_buff and assign it a usage count of one. The
1047 * buffer has unspecified headroom built in. Users should allocate
1048 * the headroom they think they need without accounting for the
1049 * built in space. The built in space is used for optimisations.
1051 * %NULL is returned in there is no free memory. Although this function
1052 * allocates memory it can be called from an interrupt.
1054 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1056 return __dev_alloc_skb(length, GFP_ATOMIC);
1060 * skb_cow - copy header of skb when it is required
1061 * @skb: buffer to cow
1062 * @headroom: needed headroom
1064 * If the skb passed lacks sufficient headroom or its data part
1065 * is shared, data is reallocated. If reallocation fails, an error
1066 * is returned and original skb is not changed.
1068 * The result is skb with writable area skb->head...skb->tail
1069 * and at least @headroom of space at head.
1071 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1073 int delta = (headroom > 16 ? headroom : 16) - skb_headroom(skb);
1078 if (delta || skb_cloned(skb))
1079 return pskb_expand_head(skb, (delta + 15) & ~15, 0, GFP_ATOMIC);
1084 * skb_padto - pad an skbuff up to a minimal size
1085 * @skb: buffer to pad
1086 * @len: minimal length
1088 * Pads up a buffer to ensure the trailing bytes exist and are
1089 * blanked. If the buffer already contains sufficient data it
1090 * is untouched. Returns the buffer, which may be a replacement
1091 * for the original, or NULL for out of memory - in which case
1092 * the original buffer is still freed.
1095 static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
1097 unsigned int size = skb->len;
1098 if (likely(size >= len))
1100 return skb_pad(skb, len-size);
1103 static inline int skb_add_data(struct sk_buff *skb,
1104 char __user *from, int copy)
1106 const int off = skb->len;
1108 if (skb->ip_summed == CHECKSUM_NONE) {
1110 unsigned int csum = csum_and_copy_from_user(from,
1114 skb->csum = csum_block_add(skb->csum, csum, off);
1117 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1120 __skb_trim(skb, off);
1124 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1125 struct page *page, int off)
1128 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1130 return page == frag->page &&
1131 off == frag->page_offset + frag->size;
1137 * skb_linearize - convert paged skb to linear one
1138 * @skb: buffer to linarize
1139 * @gfp: allocation mode
1141 * If there is no free memory -ENOMEM is returned, otherwise zero
1142 * is returned and the old skb data released.
1144 extern int __skb_linearize(struct sk_buff *skb, gfp_t gfp);
1145 static inline int skb_linearize(struct sk_buff *skb, gfp_t gfp)
1147 return __skb_linearize(skb, gfp);
1151 * skb_postpull_rcsum - update checksum for received skb after pull
1152 * @skb: buffer to update
1153 * @start: start of data before pull
1154 * @len: length of data pulled
1156 * After doing a pull on a received packet, you need to call this to
1157 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1158 * so that it can be recomputed from scratch.
1161 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1162 const void *start, unsigned int len)
1164 if (skb->ip_summed == CHECKSUM_HW)
1165 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1168 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1171 * pskb_trim_rcsum - trim received skb and update checksum
1172 * @skb: buffer to trim
1175 * This is exactly the same as pskb_trim except that it ensures the
1176 * checksum of received packets are still valid after the operation.
1179 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1181 if (likely(len >= skb->len))
1183 if (skb->ip_summed == CHECKSUM_HW)
1184 skb->ip_summed = CHECKSUM_NONE;
1185 return __pskb_trim(skb, len);
1188 static inline void *kmap_skb_frag(const skb_frag_t *frag)
1190 #ifdef CONFIG_HIGHMEM
1195 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1198 static inline void kunmap_skb_frag(void *vaddr)
1200 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1201 #ifdef CONFIG_HIGHMEM
1206 #define skb_queue_walk(queue, skb) \
1207 for (skb = (queue)->next; \
1208 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1211 #define skb_queue_reverse_walk(queue, skb) \
1212 for (skb = (queue)->prev; \
1213 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1217 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1218 int noblock, int *err);
1219 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1220 struct poll_table_struct *wait);
1221 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1222 int offset, struct iovec *to,
1224 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1227 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1228 extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1229 unsigned int flags);
1230 extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1231 int len, unsigned int csum);
1232 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1234 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1235 void *from, int len);
1236 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1237 int offset, u8 *to, int len,
1239 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1240 extern void skb_split(struct sk_buff *skb,
1241 struct sk_buff *skb1, const u32 len);
1243 extern void skb_release_data(struct sk_buff *skb);
1245 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1246 int len, void *buffer)
1248 int hlen = skb_headlen(skb);
1250 if (hlen - offset >= len)
1251 return skb->data + offset;
1253 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1259 extern void skb_init(void);
1260 extern void skb_add_mtu(int mtu);
1263 * skb_get_timestamp - get timestamp from a skb
1264 * @skb: skb to get stamp from
1265 * @stamp: pointer to struct timeval to store stamp in
1267 * Timestamps are stored in the skb as offsets to a base timestamp.
1268 * This function converts the offset back to a struct timeval and stores
1271 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1273 stamp->tv_sec = skb->tstamp.off_sec;
1274 stamp->tv_usec = skb->tstamp.off_usec;
1278 * skb_set_timestamp - set timestamp of a skb
1279 * @skb: skb to set stamp of
1280 * @stamp: pointer to struct timeval to get stamp from
1282 * Timestamps are stored in the skb as offsets to a base timestamp.
1283 * This function converts a struct timeval to an offset and stores
1286 static inline void skb_set_timestamp(struct sk_buff *skb, const struct timeval *stamp)
1288 skb->tstamp.off_sec = stamp->tv_sec;
1289 skb->tstamp.off_usec = stamp->tv_usec;
1292 extern void __net_timestamp(struct sk_buff *skb);
1294 extern unsigned int __skb_checksum_complete(struct sk_buff *skb);
1297 * skb_checksum_complete - Calculate checksum of an entire packet
1298 * @skb: packet to process
1300 * This function calculates the checksum over the entire packet plus
1301 * the value of skb->csum. The latter can be used to supply the
1302 * checksum of a pseudo header as used by TCP/UDP. It returns the
1305 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1306 * this function can be used to verify that checksum on received
1307 * packets. In that case the function should return zero if the
1308 * checksum is correct. In particular, this function will return zero
1309 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1310 * hardware has already verified the correctness of the checksum.
1312 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1314 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1315 __skb_checksum_complete(skb);
1318 #ifdef CONFIG_NETFILTER
1319 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1321 if (nfct && atomic_dec_and_test(&nfct->use))
1322 nfct->destroy(nfct);
1324 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1327 atomic_inc(&nfct->use);
1329 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1330 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1333 atomic_inc(&skb->users);
1335 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1341 #ifdef CONFIG_BRIDGE_NETFILTER
1342 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1344 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1347 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1350 atomic_inc(&nf_bridge->use);
1352 #endif /* CONFIG_BRIDGE_NETFILTER */
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;
1361 #ifdef CONFIG_BRIDGE_NETFILTER
1362 nf_bridge_put(skb->nf_bridge);
1363 skb->nf_bridge = NULL;
1367 #else /* CONFIG_NETFILTER */
1368 static inline void nf_reset(struct sk_buff *skb) {}
1369 #endif /* CONFIG_NETFILTER */
1371 #endif /* __KERNEL__ */
1372 #endif /* _LINUX_SKBUFF_H */