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 struct sk_buff *frag_list;
141 skb_frag_t frags[MAX_SKB_FRAGS];
144 /* We divide dataref into two halves. The higher 16 bits hold references
145 * to the payload part of skb->data. The lower 16 bits hold references to
146 * the entire skb->data. It is up to the users of the skb to agree on
147 * where the payload starts.
149 * All users must obey the rule that the skb->data reference count must be
150 * greater than or equal to the payload reference count.
152 * Holding a reference to the payload part means that the user does not
153 * care about modifications to the header part of skb->data.
155 #define SKB_DATAREF_SHIFT 16
156 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
158 extern struct timeval skb_tv_base;
167 SKB_FCLONE_UNAVAILABLE,
173 * struct sk_buff - socket buffer
174 * @next: Next buffer in list
175 * @prev: Previous buffer in list
176 * @list: List we are on
177 * @sk: Socket we are owned by
178 * @tstamp: Time we arrived stored as offset to skb_tv_base
179 * @dev: Device we arrived on/are leaving by
180 * @input_dev: Device we arrived on
181 * @h: Transport layer header
182 * @nh: Network layer header
183 * @mac: Link layer header
184 * @dst: destination entry
185 * @sp: the security path, used for xfrm
186 * @cb: Control buffer. Free for use by every layer. Put private vars here
187 * @len: Length of actual data
188 * @data_len: Data length
189 * @mac_len: Length of link layer header
191 * @local_df: allow local fragmentation
192 * @cloned: Head may be cloned (check refcnt to be sure)
193 * @nohdr: Payload reference only, must not modify header
194 * @pkt_type: Packet class
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 * @nfctinfo: Relationship of this skb to the connection
208 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
209 * @tc_index: Traffic control index
210 * @tc_verd: traffic control verdict
214 /* These two members must be first. */
215 struct sk_buff *next;
216 struct sk_buff *prev;
219 struct skb_timeval tstamp;
220 struct net_device *dev;
221 struct net_device *input_dev;
226 struct icmphdr *icmph;
227 struct igmphdr *igmph;
229 struct ipv6hdr *ipv6h;
235 struct ipv6hdr *ipv6h;
244 struct dst_entry *dst;
248 * This is the control buffer. It is free to use for every
249 * layer. Please put your private variables there. If you
250 * want to keep them across layers you have to do a skb_clone()
251 * first. This is owned by whoever has the skb queued ATM.
269 void (*destructor)(struct sk_buff *skb);
270 #ifdef CONFIG_NETFILTER
272 struct nf_conntrack *nfct;
273 #if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
274 __u8 ipvs_property:1;
276 #ifdef CONFIG_BRIDGE_NETFILTER
277 struct nf_bridge_info *nf_bridge;
279 #endif /* CONFIG_NETFILTER */
280 #ifdef CONFIG_NET_SCHED
281 __u16 tc_index; /* traffic control index */
282 #ifdef CONFIG_NET_CLS_ACT
283 __u16 tc_verd; /* traffic control verdict */
288 /* These elements must be at the end, see alloc_skb() for details. */
289 unsigned int truesize;
299 * Handling routines are only of interest to the kernel
301 #include <linux/slab.h>
303 #include <asm/system.h>
305 extern void __kfree_skb(struct sk_buff *skb);
306 extern struct sk_buff *__alloc_skb(unsigned int size,
307 unsigned int __nocast priority, int fclone);
308 static inline struct sk_buff *alloc_skb(unsigned int size,
309 unsigned int __nocast priority)
311 return __alloc_skb(size, priority, 0);
314 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
315 unsigned int __nocast priority)
317 return __alloc_skb(size, priority, 1);
320 extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
322 unsigned int __nocast priority);
323 extern void kfree_skbmem(struct sk_buff *skb);
324 extern struct sk_buff *skb_clone(struct sk_buff *skb,
325 unsigned int __nocast priority);
326 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
327 unsigned int __nocast priority);
328 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
329 unsigned int __nocast gfp_mask);
330 extern int pskb_expand_head(struct sk_buff *skb,
331 int nhead, int ntail,
332 unsigned int __nocast gfp_mask);
333 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
334 unsigned int headroom);
335 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
336 int newheadroom, int newtailroom,
337 unsigned int __nocast priority);
338 extern struct sk_buff * skb_pad(struct sk_buff *skb, int pad);
339 #define dev_kfree_skb(a) kfree_skb(a)
340 extern void skb_over_panic(struct sk_buff *skb, int len,
342 extern void skb_under_panic(struct sk_buff *skb, int len,
350 __u32 stepped_offset;
351 struct sk_buff *root_skb;
352 struct sk_buff *cur_skb;
356 extern void skb_prepare_seq_read(struct sk_buff *skb,
357 unsigned int from, unsigned int to,
358 struct skb_seq_state *st);
359 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
360 struct skb_seq_state *st);
361 extern void skb_abort_seq_read(struct skb_seq_state *st);
363 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
364 unsigned int to, struct ts_config *config,
365 struct ts_state *state);
368 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
371 * skb_queue_empty - check if a queue is empty
374 * Returns true if the queue is empty, false otherwise.
376 static inline int skb_queue_empty(const struct sk_buff_head *list)
378 return list->next == (struct sk_buff *)list;
382 * skb_get - reference buffer
383 * @skb: buffer to reference
385 * Makes another reference to a socket buffer and returns a pointer
388 static inline struct sk_buff *skb_get(struct sk_buff *skb)
390 atomic_inc(&skb->users);
395 * If users == 1, we are the only owner and are can avoid redundant
400 * kfree_skb - free an sk_buff
401 * @skb: buffer to free
403 * Drop a reference to the buffer and free it if the usage count has
406 static inline void kfree_skb(struct sk_buff *skb)
408 if (likely(atomic_read(&skb->users) == 1))
410 else if (likely(!atomic_dec_and_test(&skb->users)))
416 * skb_cloned - is the buffer a clone
417 * @skb: buffer to check
419 * Returns true if the buffer was generated with skb_clone() and is
420 * one of multiple shared copies of the buffer. Cloned buffers are
421 * shared data so must not be written to under normal circumstances.
423 static inline int skb_cloned(const struct sk_buff *skb)
425 return skb->cloned &&
426 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
430 * skb_header_cloned - is the header a clone
431 * @skb: buffer to check
433 * Returns true if modifying the header part of the buffer requires
434 * the data to be copied.
436 static inline int skb_header_cloned(const struct sk_buff *skb)
443 dataref = atomic_read(&skb_shinfo(skb)->dataref);
444 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
449 * skb_header_release - release reference to header
450 * @skb: buffer to operate on
452 * Drop a reference to the header part of the buffer. This is done
453 * by acquiring a payload reference. You must not read from the header
454 * part of skb->data after this.
456 static inline void skb_header_release(struct sk_buff *skb)
460 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
464 * skb_shared - is the buffer shared
465 * @skb: buffer to check
467 * Returns true if more than one person has a reference to this
470 static inline int skb_shared(const struct sk_buff *skb)
472 return atomic_read(&skb->users) != 1;
476 * skb_share_check - check if buffer is shared and if so clone it
477 * @skb: buffer to check
478 * @pri: priority for memory allocation
480 * If the buffer is shared the buffer is cloned and the old copy
481 * drops a reference. A new clone with a single reference is returned.
482 * If the buffer is not shared the original buffer is returned. When
483 * being called from interrupt status or with spinlocks held pri must
486 * NULL is returned on a memory allocation failure.
488 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
489 unsigned int __nocast pri)
491 might_sleep_if(pri & __GFP_WAIT);
492 if (skb_shared(skb)) {
493 struct sk_buff *nskb = skb_clone(skb, pri);
501 * Copy shared buffers into a new sk_buff. We effectively do COW on
502 * packets to handle cases where we have a local reader and forward
503 * and a couple of other messy ones. The normal one is tcpdumping
504 * a packet thats being forwarded.
508 * skb_unshare - make a copy of a shared buffer
509 * @skb: buffer to check
510 * @pri: priority for memory allocation
512 * If the socket buffer is a clone then this function creates a new
513 * copy of the data, drops a reference count on the old copy and returns
514 * the new copy with the reference count at 1. If the buffer is not a clone
515 * the original buffer is returned. When called with a spinlock held or
516 * from interrupt state @pri must be %GFP_ATOMIC
518 * %NULL is returned on a memory allocation failure.
520 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
521 unsigned int __nocast pri)
523 might_sleep_if(pri & __GFP_WAIT);
524 if (skb_cloned(skb)) {
525 struct sk_buff *nskb = skb_copy(skb, pri);
526 kfree_skb(skb); /* Free our shared copy */
534 * @list_: list to peek at
536 * Peek an &sk_buff. Unlike most other operations you _MUST_
537 * be careful with this one. A peek leaves the buffer on the
538 * list and someone else may run off with it. You must hold
539 * the appropriate locks or have a private queue to do this.
541 * Returns %NULL for an empty list or a pointer to the head element.
542 * The reference count is not incremented and the reference is therefore
543 * volatile. Use with caution.
545 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
547 struct sk_buff *list = ((struct sk_buff *)list_)->next;
548 if (list == (struct sk_buff *)list_)
555 * @list_: list to peek at
557 * Peek an &sk_buff. Unlike most other operations you _MUST_
558 * be careful with this one. A peek leaves the buffer on the
559 * list and someone else may run off with it. You must hold
560 * the appropriate locks or have a private queue to do this.
562 * Returns %NULL for an empty list or a pointer to the tail element.
563 * The reference count is not incremented and the reference is therefore
564 * volatile. Use with caution.
566 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
568 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
569 if (list == (struct sk_buff *)list_)
575 * skb_queue_len - get queue length
576 * @list_: list to measure
578 * Return the length of an &sk_buff queue.
580 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
585 static inline void skb_queue_head_init(struct sk_buff_head *list)
587 spin_lock_init(&list->lock);
588 list->prev = list->next = (struct sk_buff *)list;
593 * Insert an sk_buff at the start of a list.
595 * The "__skb_xxxx()" functions are the non-atomic ones that
596 * can only be called with interrupts disabled.
600 * __skb_queue_head - queue a buffer at the list head
602 * @newsk: buffer to queue
604 * Queue a buffer at the start 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 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
610 static inline void __skb_queue_head(struct sk_buff_head *list,
611 struct sk_buff *newsk)
613 struct sk_buff *prev, *next;
616 prev = (struct sk_buff *)list;
620 next->prev = prev->next = newsk;
624 * __skb_queue_tail - queue a buffer at the list tail
626 * @newsk: buffer to queue
628 * Queue a buffer at the end of a list. This function takes no locks
629 * and you must therefore hold required locks before calling it.
631 * A buffer cannot be placed on two lists at the same time.
633 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
634 static inline void __skb_queue_tail(struct sk_buff_head *list,
635 struct sk_buff *newsk)
637 struct sk_buff *prev, *next;
640 next = (struct sk_buff *)list;
644 next->prev = prev->next = newsk;
649 * __skb_dequeue - remove from the head of the queue
650 * @list: list to dequeue from
652 * Remove the head of the list. This function does not take any locks
653 * so must be used with appropriate locks held only. The head item is
654 * returned or %NULL if the list is empty.
656 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
657 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
659 struct sk_buff *next, *prev, *result;
661 prev = (struct sk_buff *) list;
670 result->next = result->prev = NULL;
677 * Insert a packet on a list.
679 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
680 static inline void __skb_insert(struct sk_buff *newsk,
681 struct sk_buff *prev, struct sk_buff *next,
682 struct sk_buff_head *list)
686 next->prev = prev->next = newsk;
691 * Place a packet after a given packet in a list.
693 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
694 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
696 __skb_insert(newsk, old, old->next, list);
700 * remove sk_buff from list. _Must_ be called atomically, and with
703 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
704 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
706 struct sk_buff *next, *prev;
711 skb->next = skb->prev = NULL;
717 /* XXX: more streamlined implementation */
720 * __skb_dequeue_tail - remove from the tail of the queue
721 * @list: list to dequeue from
723 * Remove the tail of the list. This function does not take any locks
724 * so must be used with appropriate locks held only. The tail item is
725 * returned or %NULL if the list is empty.
727 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
728 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
730 struct sk_buff *skb = skb_peek_tail(list);
732 __skb_unlink(skb, list);
737 static inline int skb_is_nonlinear(const struct sk_buff *skb)
739 return skb->data_len;
742 static inline unsigned int skb_headlen(const struct sk_buff *skb)
744 return skb->len - skb->data_len;
747 static inline int skb_pagelen(const struct sk_buff *skb)
751 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
752 len += skb_shinfo(skb)->frags[i].size;
753 return len + skb_headlen(skb);
756 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
757 struct page *page, int off, int size)
759 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
762 frag->page_offset = off;
764 skb_shinfo(skb)->nr_frags = i + 1;
767 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
768 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
769 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
772 * Add data to an sk_buff
774 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
776 unsigned char *tmp = skb->tail;
777 SKB_LINEAR_ASSERT(skb);
784 * skb_put - add data to a buffer
785 * @skb: buffer to use
786 * @len: amount of data to add
788 * This function extends the used data area of the buffer. If this would
789 * exceed the total buffer size the kernel will panic. A pointer to the
790 * first byte of the extra data is returned.
792 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
794 unsigned char *tmp = skb->tail;
795 SKB_LINEAR_ASSERT(skb);
798 if (unlikely(skb->tail>skb->end))
799 skb_over_panic(skb, len, current_text_addr());
803 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
811 * skb_push - add data to the start of a buffer
812 * @skb: buffer to use
813 * @len: amount of data to add
815 * This function extends the used data area of the buffer at the buffer
816 * start. If this would exceed the total buffer headroom the kernel will
817 * panic. A pointer to the first byte of the extra data is returned.
819 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
823 if (unlikely(skb->data<skb->head))
824 skb_under_panic(skb, len, current_text_addr());
828 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
831 BUG_ON(skb->len < skb->data_len);
832 return skb->data += len;
836 * skb_pull - remove data from the start of a buffer
837 * @skb: buffer to use
838 * @len: amount of data to remove
840 * This function removes data from the start of a buffer, returning
841 * the memory to the headroom. A pointer to the next data in the buffer
842 * is returned. Once the data has been pulled future pushes will overwrite
845 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
847 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
850 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
852 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
854 if (len > skb_headlen(skb) &&
855 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
858 return skb->data += len;
861 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
863 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
866 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
868 if (likely(len <= skb_headlen(skb)))
870 if (unlikely(len > skb->len))
872 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
876 * skb_headroom - bytes at buffer head
877 * @skb: buffer to check
879 * Return the number of bytes of free space at the head of an &sk_buff.
881 static inline int skb_headroom(const struct sk_buff *skb)
883 return skb->data - skb->head;
887 * skb_tailroom - bytes at buffer end
888 * @skb: buffer to check
890 * Return the number of bytes of free space at the tail of an sk_buff
892 static inline int skb_tailroom(const struct sk_buff *skb)
894 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
898 * skb_reserve - adjust headroom
899 * @skb: buffer to alter
900 * @len: bytes to move
902 * Increase the headroom of an empty &sk_buff by reducing the tail
903 * room. This is only allowed for an empty buffer.
905 static inline void skb_reserve(struct sk_buff *skb, unsigned int len)
912 * CPUs often take a performance hit when accessing unaligned memory
913 * locations. The actual performance hit varies, it can be small if the
914 * hardware handles it or large if we have to take an exception and fix it
917 * Since an ethernet header is 14 bytes network drivers often end up with
918 * the IP header at an unaligned offset. The IP header can be aligned by
919 * shifting the start of the packet by 2 bytes. Drivers should do this
922 * skb_reserve(NET_IP_ALIGN);
924 * The downside to this alignment of the IP header is that the DMA is now
925 * unaligned. On some architectures the cost of an unaligned DMA is high
926 * and this cost outweighs the gains made by aligning the IP header.
928 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
932 #define NET_IP_ALIGN 2
935 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
937 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
939 if (!skb->data_len) {
941 skb->tail = skb->data + len;
943 ___pskb_trim(skb, len, 0);
947 * skb_trim - remove end from a buffer
948 * @skb: buffer to alter
951 * Cut the length of a buffer down by removing data from the tail. If
952 * the buffer is already under the length specified it is not modified.
954 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
957 __skb_trim(skb, len);
961 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
963 if (!skb->data_len) {
965 skb->tail = skb->data+len;
968 return ___pskb_trim(skb, len, 1);
971 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
973 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
977 * skb_orphan - orphan a buffer
978 * @skb: buffer to orphan
980 * If a buffer currently has an owner then we call the owner's
981 * destructor function and make the @skb unowned. The buffer continues
982 * to exist but is no longer charged to its former owner.
984 static inline void skb_orphan(struct sk_buff *skb)
987 skb->destructor(skb);
988 skb->destructor = NULL;
993 * __skb_queue_purge - empty a list
994 * @list: list to empty
996 * Delete all buffers on an &sk_buff list. Each buffer is removed from
997 * the list and one reference dropped. This function does not take the
998 * list lock and the caller must hold the relevant locks to use it.
1000 extern void skb_queue_purge(struct sk_buff_head *list);
1001 static inline void __skb_queue_purge(struct sk_buff_head *list)
1003 struct sk_buff *skb;
1004 while ((skb = __skb_dequeue(list)) != NULL)
1008 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
1010 * __dev_alloc_skb - allocate an skbuff for sending
1011 * @length: length to allocate
1012 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1014 * Allocate a new &sk_buff and assign it a usage count of one. The
1015 * buffer has unspecified headroom built in. Users should allocate
1016 * the headroom they think they need without accounting for the
1017 * built in space. The built in space is used for optimisations.
1019 * %NULL is returned in there is no free memory.
1021 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1022 unsigned int __nocast gfp_mask)
1024 struct sk_buff *skb = alloc_skb(length + 16, gfp_mask);
1026 skb_reserve(skb, 16);
1030 extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
1034 * dev_alloc_skb - allocate an skbuff for sending
1035 * @length: length to allocate
1037 * Allocate a new &sk_buff and assign it a usage count of one. The
1038 * buffer has unspecified headroom built in. Users should allocate
1039 * the headroom they think they need without accounting for the
1040 * built in space. The built in space is used for optimisations.
1042 * %NULL is returned in there is no free memory. Although this function
1043 * allocates memory it can be called from an interrupt.
1045 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1047 return __dev_alloc_skb(length, GFP_ATOMIC);
1051 * skb_cow - copy header of skb when it is required
1052 * @skb: buffer to cow
1053 * @headroom: needed headroom
1055 * If the skb passed lacks sufficient headroom or its data part
1056 * is shared, data is reallocated. If reallocation fails, an error
1057 * is returned and original skb is not changed.
1059 * The result is skb with writable area skb->head...skb->tail
1060 * and at least @headroom of space at head.
1062 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1064 int delta = (headroom > 16 ? headroom : 16) - skb_headroom(skb);
1069 if (delta || skb_cloned(skb))
1070 return pskb_expand_head(skb, (delta + 15) & ~15, 0, GFP_ATOMIC);
1075 * skb_padto - pad an skbuff up to a minimal size
1076 * @skb: buffer to pad
1077 * @len: minimal length
1079 * Pads up a buffer to ensure the trailing bytes exist and are
1080 * blanked. If the buffer already contains sufficient data it
1081 * is untouched. Returns the buffer, which may be a replacement
1082 * for the original, or NULL for out of memory - in which case
1083 * the original buffer is still freed.
1086 static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
1088 unsigned int size = skb->len;
1089 if (likely(size >= len))
1091 return skb_pad(skb, len-size);
1094 static inline int skb_add_data(struct sk_buff *skb,
1095 char __user *from, int copy)
1097 const int off = skb->len;
1099 if (skb->ip_summed == CHECKSUM_NONE) {
1101 unsigned int csum = csum_and_copy_from_user(from,
1105 skb->csum = csum_block_add(skb->csum, csum, off);
1108 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1111 __skb_trim(skb, off);
1115 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1116 struct page *page, int off)
1119 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1121 return page == frag->page &&
1122 off == frag->page_offset + frag->size;
1128 * skb_linearize - convert paged skb to linear one
1129 * @skb: buffer to linarize
1130 * @gfp: allocation mode
1132 * If there is no free memory -ENOMEM is returned, otherwise zero
1133 * is returned and the old skb data released.
1135 extern int __skb_linearize(struct sk_buff *skb, unsigned int __nocast gfp);
1136 static inline int skb_linearize(struct sk_buff *skb, unsigned int __nocast gfp)
1138 return __skb_linearize(skb, gfp);
1142 * skb_postpull_rcsum - update checksum for received skb after pull
1143 * @skb: buffer to update
1144 * @start: start of data before pull
1145 * @len: length of data pulled
1147 * After doing a pull on a received packet, you need to call this to
1148 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1149 * so that it can be recomputed from scratch.
1152 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1153 const void *start, int len)
1155 if (skb->ip_summed == CHECKSUM_HW)
1156 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1160 * pskb_trim_rcsum - trim received skb and update checksum
1161 * @skb: buffer to trim
1164 * This is exactly the same as pskb_trim except that it ensures the
1165 * checksum of received packets are still valid after the operation.
1168 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1170 if (len >= skb->len)
1172 if (skb->ip_summed == CHECKSUM_HW)
1173 skb->ip_summed = CHECKSUM_NONE;
1174 return __pskb_trim(skb, len);
1177 static inline void *kmap_skb_frag(const skb_frag_t *frag)
1179 #ifdef CONFIG_HIGHMEM
1184 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1187 static inline void kunmap_skb_frag(void *vaddr)
1189 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1190 #ifdef CONFIG_HIGHMEM
1195 #define skb_queue_walk(queue, skb) \
1196 for (skb = (queue)->next; \
1197 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1201 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1202 int noblock, int *err);
1203 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1204 struct poll_table_struct *wait);
1205 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1206 int offset, struct iovec *to,
1208 extern int skb_copy_and_csum_datagram_iovec(const
1209 struct sk_buff *skb,
1212 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1213 extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1214 int len, unsigned int csum);
1215 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1217 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1218 void *from, int len);
1219 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1220 int offset, u8 *to, int len,
1222 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1223 extern void skb_split(struct sk_buff *skb,
1224 struct sk_buff *skb1, const u32 len);
1226 extern void skb_release_data(struct sk_buff *skb);
1228 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1229 int len, void *buffer)
1231 int hlen = skb_headlen(skb);
1233 if (hlen - offset >= len)
1234 return skb->data + offset;
1236 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1242 extern void skb_init(void);
1243 extern void skb_add_mtu(int mtu);
1246 * skb_get_timestamp - get timestamp from a skb
1247 * @skb: skb to get stamp from
1248 * @stamp: pointer to struct timeval to store stamp in
1250 * Timestamps are stored in the skb as offsets to a base timestamp.
1251 * This function converts the offset back to a struct timeval and stores
1254 static inline void skb_get_timestamp(struct sk_buff *skb, struct timeval *stamp)
1256 stamp->tv_sec = skb->tstamp.off_sec;
1257 stamp->tv_usec = skb->tstamp.off_usec;
1258 if (skb->tstamp.off_sec) {
1259 stamp->tv_sec += skb_tv_base.tv_sec;
1260 stamp->tv_usec += skb_tv_base.tv_usec;
1265 * skb_set_timestamp - set timestamp of a skb
1266 * @skb: skb to set stamp of
1267 * @stamp: pointer to struct timeval to get stamp from
1269 * Timestamps are stored in the skb as offsets to a base timestamp.
1270 * This function converts a struct timeval to an offset and stores
1273 static inline void skb_set_timestamp(struct sk_buff *skb, struct timeval *stamp)
1275 skb->tstamp.off_sec = stamp->tv_sec - skb_tv_base.tv_sec;
1276 skb->tstamp.off_usec = stamp->tv_usec - skb_tv_base.tv_usec;
1279 extern void __net_timestamp(struct sk_buff *skb);
1281 #ifdef CONFIG_NETFILTER
1282 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1284 if (nfct && atomic_dec_and_test(&nfct->use))
1285 nfct->destroy(nfct);
1287 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1290 atomic_inc(&nfct->use);
1292 static inline void nf_reset(struct sk_buff *skb)
1294 nf_conntrack_put(skb->nfct);
1298 #ifdef CONFIG_BRIDGE_NETFILTER
1299 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1301 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1304 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1307 atomic_inc(&nf_bridge->use);
1309 #endif /* CONFIG_BRIDGE_NETFILTER */
1310 #else /* CONFIG_NETFILTER */
1311 static inline void nf_reset(struct sk_buff *skb) {}
1312 #endif /* CONFIG_NETFILTER */
1314 #endif /* __KERNEL__ */
1315 #endif /* _LINUX_SKBUFF_H */