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 */
288 #if defined(CONFIG_VNET) || defined(CONFIG_VNET_MODULE)
289 xid_t xid; /* VServer context ID */
292 /* These elements must be at the end, see alloc_skb() for details. */
293 unsigned int truesize;
303 * Handling routines are only of interest to the kernel
305 #include <linux/slab.h>
307 #include <asm/system.h>
309 extern void kfree_skb(struct sk_buff *skb);
310 extern void __kfree_skb(struct sk_buff *skb);
311 extern struct sk_buff *__alloc_skb(unsigned int size,
312 gfp_t priority, int fclone);
313 static inline struct sk_buff *alloc_skb(unsigned int size,
316 return __alloc_skb(size, priority, 0);
319 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
322 return __alloc_skb(size, priority, 1);
325 extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
328 extern void kfree_skbmem(struct sk_buff *skb);
329 extern struct sk_buff *skb_clone(struct sk_buff *skb,
331 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
333 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
335 extern int pskb_expand_head(struct sk_buff *skb,
336 int nhead, int ntail,
338 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
339 unsigned int headroom);
340 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
341 int newheadroom, int newtailroom,
343 extern struct sk_buff * skb_pad(struct sk_buff *skb, int pad);
344 #define dev_kfree_skb(a) kfree_skb(a)
345 extern void skb_over_panic(struct sk_buff *skb, int len,
347 extern void skb_under_panic(struct sk_buff *skb, int len,
349 extern void skb_truesize_bug(struct sk_buff *skb);
351 static inline void skb_truesize_check(struct sk_buff *skb)
353 if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len))
354 skb_truesize_bug(skb);
357 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
358 int getfrag(void *from, char *to, int offset,
359 int len,int odd, struct sk_buff *skb),
360 void *from, int length);
367 __u32 stepped_offset;
368 struct sk_buff *root_skb;
369 struct sk_buff *cur_skb;
373 extern void skb_prepare_seq_read(struct sk_buff *skb,
374 unsigned int from, unsigned int to,
375 struct skb_seq_state *st);
376 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
377 struct skb_seq_state *st);
378 extern void skb_abort_seq_read(struct skb_seq_state *st);
380 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
381 unsigned int to, struct ts_config *config,
382 struct ts_state *state);
385 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
388 * skb_queue_empty - check if a queue is empty
391 * Returns true if the queue is empty, false otherwise.
393 static inline int skb_queue_empty(const struct sk_buff_head *list)
395 return list->next == (struct sk_buff *)list;
399 * skb_get - reference buffer
400 * @skb: buffer to reference
402 * Makes another reference to a socket buffer and returns a pointer
405 static inline struct sk_buff *skb_get(struct sk_buff *skb)
407 atomic_inc(&skb->users);
412 * If users == 1, we are the only owner and are can avoid redundant
417 * skb_cloned - is the buffer a clone
418 * @skb: buffer to check
420 * Returns true if the buffer was generated with skb_clone() and is
421 * one of multiple shared copies of the buffer. Cloned buffers are
422 * shared data so must not be written to under normal circumstances.
424 static inline int skb_cloned(const struct sk_buff *skb)
426 return skb->cloned &&
427 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
431 * skb_header_cloned - is the header a clone
432 * @skb: buffer to check
434 * Returns true if modifying the header part of the buffer requires
435 * the data to be copied.
437 static inline int skb_header_cloned(const struct sk_buff *skb)
444 dataref = atomic_read(&skb_shinfo(skb)->dataref);
445 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
450 * skb_header_release - release reference to header
451 * @skb: buffer to operate on
453 * Drop a reference to the header part of the buffer. This is done
454 * by acquiring a payload reference. You must not read from the header
455 * part of skb->data after this.
457 static inline void skb_header_release(struct sk_buff *skb)
461 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
465 * skb_shared - is the buffer shared
466 * @skb: buffer to check
468 * Returns true if more than one person has a reference to this
471 static inline int skb_shared(const struct sk_buff *skb)
473 return atomic_read(&skb->users) != 1;
477 * skb_share_check - check if buffer is shared and if so clone it
478 * @skb: buffer to check
479 * @pri: priority for memory allocation
481 * If the buffer is shared the buffer is cloned and the old copy
482 * drops a reference. A new clone with a single reference is returned.
483 * If the buffer is not shared the original buffer is returned. When
484 * being called from interrupt status or with spinlocks held pri must
487 * NULL is returned on a memory allocation failure.
489 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
492 might_sleep_if(pri & __GFP_WAIT);
493 if (skb_shared(skb)) {
494 struct sk_buff *nskb = skb_clone(skb, pri);
502 * Copy shared buffers into a new sk_buff. We effectively do COW on
503 * packets to handle cases where we have a local reader and forward
504 * and a couple of other messy ones. The normal one is tcpdumping
505 * a packet thats being forwarded.
509 * skb_unshare - make a copy of a shared buffer
510 * @skb: buffer to check
511 * @pri: priority for memory allocation
513 * If the socket buffer is a clone then this function creates a new
514 * copy of the data, drops a reference count on the old copy and returns
515 * the new copy with the reference count at 1. If the buffer is not a clone
516 * the original buffer is returned. When called with a spinlock held or
517 * from interrupt state @pri must be %GFP_ATOMIC
519 * %NULL is returned on a memory allocation failure.
521 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
524 might_sleep_if(pri & __GFP_WAIT);
525 if (skb_cloned(skb)) {
526 struct sk_buff *nskb = skb_copy(skb, pri);
527 kfree_skb(skb); /* Free our shared copy */
535 * @list_: list to peek at
537 * Peek an &sk_buff. Unlike most other operations you _MUST_
538 * be careful with this one. A peek leaves the buffer on the
539 * list and someone else may run off with it. You must hold
540 * the appropriate locks or have a private queue to do this.
542 * Returns %NULL for an empty list or a pointer to the head element.
543 * The reference count is not incremented and the reference is therefore
544 * volatile. Use with caution.
546 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
548 struct sk_buff *list = ((struct sk_buff *)list_)->next;
549 if (list == (struct sk_buff *)list_)
556 * @list_: list to peek at
558 * Peek an &sk_buff. Unlike most other operations you _MUST_
559 * be careful with this one. A peek leaves the buffer on the
560 * list and someone else may run off with it. You must hold
561 * the appropriate locks or have a private queue to do this.
563 * Returns %NULL for an empty list or a pointer to the tail element.
564 * The reference count is not incremented and the reference is therefore
565 * volatile. Use with caution.
567 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
569 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
570 if (list == (struct sk_buff *)list_)
576 * skb_queue_len - get queue length
577 * @list_: list to measure
579 * Return the length of an &sk_buff queue.
581 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
586 static inline void skb_queue_head_init(struct sk_buff_head *list)
588 spin_lock_init(&list->lock);
589 list->prev = list->next = (struct sk_buff *)list;
594 * Insert an sk_buff at the start of a list.
596 * The "__skb_xxxx()" functions are the non-atomic ones that
597 * can only be called with interrupts disabled.
601 * __skb_queue_after - queue a buffer at the list head
603 * @prev: place after this buffer
604 * @newsk: buffer to queue
606 * Queue a buffer int the middle of a list. This function takes no locks
607 * and you must therefore hold required locks before calling it.
609 * A buffer cannot be placed on two lists at the same time.
611 static inline void __skb_queue_after(struct sk_buff_head *list,
612 struct sk_buff *prev,
613 struct sk_buff *newsk)
615 struct sk_buff *next;
621 next->prev = prev->next = newsk;
625 * __skb_queue_head - queue a buffer at the list head
627 * @newsk: buffer to queue
629 * Queue a buffer at the start of a list. This function takes no locks
630 * and you must therefore hold required locks before calling it.
632 * A buffer cannot be placed on two lists at the same time.
634 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
635 static inline void __skb_queue_head(struct sk_buff_head *list,
636 struct sk_buff *newsk)
638 __skb_queue_after(list, (struct sk_buff *)list, newsk);
642 * __skb_queue_tail - queue a buffer at the list tail
644 * @newsk: buffer to queue
646 * Queue a buffer at the end of a list. This function takes no locks
647 * and you must therefore hold required locks before calling it.
649 * A buffer cannot be placed on two lists at the same time.
651 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
652 static inline void __skb_queue_tail(struct sk_buff_head *list,
653 struct sk_buff *newsk)
655 struct sk_buff *prev, *next;
658 next = (struct sk_buff *)list;
662 next->prev = prev->next = newsk;
667 * __skb_dequeue - remove from the head of the queue
668 * @list: list to dequeue from
670 * Remove the head of the list. This function does not take any locks
671 * so must be used with appropriate locks held only. The head item is
672 * returned or %NULL if the list is empty.
674 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
675 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
677 struct sk_buff *next, *prev, *result;
679 prev = (struct sk_buff *) list;
688 result->next = result->prev = NULL;
695 * Insert a packet on a list.
697 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
698 static inline void __skb_insert(struct sk_buff *newsk,
699 struct sk_buff *prev, struct sk_buff *next,
700 struct sk_buff_head *list)
704 next->prev = prev->next = newsk;
709 * Place a packet after a given packet in a list.
711 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
712 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
714 __skb_insert(newsk, old, old->next, list);
718 * remove sk_buff from list. _Must_ be called atomically, and with
721 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
722 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
724 struct sk_buff *next, *prev;
729 skb->next = skb->prev = NULL;
735 /* XXX: more streamlined implementation */
738 * __skb_dequeue_tail - remove from the tail of the queue
739 * @list: list to dequeue from
741 * Remove the tail of the list. This function does not take any locks
742 * so must be used with appropriate locks held only. The tail item is
743 * returned or %NULL if the list is empty.
745 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
746 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
748 struct sk_buff *skb = skb_peek_tail(list);
750 __skb_unlink(skb, list);
755 static inline int skb_is_nonlinear(const struct sk_buff *skb)
757 return skb->data_len;
760 static inline unsigned int skb_headlen(const struct sk_buff *skb)
762 return skb->len - skb->data_len;
765 static inline int skb_pagelen(const struct sk_buff *skb)
769 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
770 len += skb_shinfo(skb)->frags[i].size;
771 return len + skb_headlen(skb);
774 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
775 struct page *page, int off, int size)
777 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
780 frag->page_offset = off;
782 skb_shinfo(skb)->nr_frags = i + 1;
785 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
786 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
787 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
790 * Add data to an sk_buff
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);
802 * skb_put - add data to a buffer
803 * @skb: buffer to use
804 * @len: amount of data to add
806 * This function extends the used data area of the buffer. If this would
807 * exceed the total buffer size the kernel will panic. A pointer to the
808 * first byte of the extra data is returned.
810 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
812 unsigned char *tmp = skb->tail;
813 SKB_LINEAR_ASSERT(skb);
816 if (unlikely(skb->tail>skb->end))
817 skb_over_panic(skb, len, current_text_addr());
821 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
829 * skb_push - add data to the start of a buffer
830 * @skb: buffer to use
831 * @len: amount of data to add
833 * This function extends the used data area of the buffer at the buffer
834 * start. If this would exceed the total buffer headroom the kernel will
835 * panic. A pointer to the first byte of the extra data is returned.
837 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
841 if (unlikely(skb->data<skb->head))
842 skb_under_panic(skb, len, current_text_addr());
846 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
849 BUG_ON(skb->len < skb->data_len);
850 return skb->data += len;
854 * skb_pull - remove data from the start of a buffer
855 * @skb: buffer to use
856 * @len: amount of data to remove
858 * This function removes data from the start of a buffer, returning
859 * the memory to the headroom. A pointer to the next data in the buffer
860 * is returned. Once the data has been pulled future pushes will overwrite
863 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
865 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
868 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
870 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
872 if (len > skb_headlen(skb) &&
873 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
876 return skb->data += len;
879 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
881 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
884 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
886 if (likely(len <= skb_headlen(skb)))
888 if (unlikely(len > skb->len))
890 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
894 * skb_headroom - bytes at buffer head
895 * @skb: buffer to check
897 * Return the number of bytes of free space at the head of an &sk_buff.
899 static inline int skb_headroom(const struct sk_buff *skb)
901 return skb->data - skb->head;
905 * skb_tailroom - bytes at buffer end
906 * @skb: buffer to check
908 * Return the number of bytes of free space at the tail of an sk_buff
910 static inline int skb_tailroom(const struct sk_buff *skb)
912 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
916 * skb_reserve - adjust headroom
917 * @skb: buffer to alter
918 * @len: bytes to move
920 * Increase the headroom of an empty &sk_buff by reducing the tail
921 * room. This is only allowed for an empty buffer.
923 static inline void skb_reserve(struct sk_buff *skb, int len)
930 * CPUs often take a performance hit when accessing unaligned memory
931 * locations. The actual performance hit varies, it can be small if the
932 * hardware handles it or large if we have to take an exception and fix it
935 * Since an ethernet header is 14 bytes network drivers often end up with
936 * the IP header at an unaligned offset. The IP header can be aligned by
937 * shifting the start of the packet by 2 bytes. Drivers should do this
940 * skb_reserve(NET_IP_ALIGN);
942 * The downside to this alignment of the IP header is that the DMA is now
943 * unaligned. On some architectures the cost of an unaligned DMA is high
944 * and this cost outweighs the gains made by aligning the IP header.
946 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
950 #define NET_IP_ALIGN 2
954 * The networking layer reserves some headroom in skb data (via
955 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
956 * the header has to grow. In the default case, if the header has to grow
957 * 16 bytes or less we avoid the reallocation.
959 * Unfortunately this headroom changes the DMA alignment of the resulting
960 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
961 * on some architectures. An architecture can override this value,
962 * perhaps setting it to a cacheline in size (since that will maintain
963 * cacheline alignment of the DMA). It must be a power of 2.
965 * Various parts of the networking layer expect at least 16 bytes of
966 * headroom, you should not reduce this.
969 #define NET_SKB_PAD 16
972 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
974 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
976 if (!skb->data_len) {
978 skb->tail = skb->data + len;
980 ___pskb_trim(skb, len, 0);
984 * skb_trim - remove end from a buffer
985 * @skb: buffer to alter
988 * Cut the length of a buffer down by removing data from the tail. If
989 * the buffer is already under the length specified it is not modified.
991 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
994 __skb_trim(skb, len);
998 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1000 if (!skb->data_len) {
1002 skb->tail = skb->data+len;
1005 return ___pskb_trim(skb, len, 1);
1008 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1010 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1014 * skb_orphan - orphan a buffer
1015 * @skb: buffer to orphan
1017 * If a buffer currently has an owner then we call the owner's
1018 * destructor function and make the @skb unowned. The buffer continues
1019 * to exist but is no longer charged to its former owner.
1021 static inline void skb_orphan(struct sk_buff *skb)
1023 if (skb->destructor)
1024 skb->destructor(skb);
1025 skb->destructor = NULL;
1030 * __skb_queue_purge - empty a list
1031 * @list: list to empty
1033 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1034 * the list and one reference dropped. This function does not take the
1035 * list lock and the caller must hold the relevant locks to use it.
1037 extern void skb_queue_purge(struct sk_buff_head *list);
1038 static inline void __skb_queue_purge(struct sk_buff_head *list)
1040 struct sk_buff *skb;
1041 while ((skb = __skb_dequeue(list)) != NULL)
1045 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
1047 * __dev_alloc_skb - allocate an skbuff for sending
1048 * @length: length to allocate
1049 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1051 * Allocate a new &sk_buff and assign it a usage count of one. The
1052 * buffer has unspecified headroom built in. Users should allocate
1053 * the headroom they think they need without accounting for the
1054 * built in space. The built in space is used for optimisations.
1056 * %NULL is returned in there is no free memory.
1058 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1061 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1063 skb_reserve(skb, NET_SKB_PAD);
1067 extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
1071 * dev_alloc_skb - allocate an skbuff for sending
1072 * @length: length to allocate
1074 * Allocate a new &sk_buff and assign it a usage count of one. The
1075 * buffer has unspecified headroom built in. Users should allocate
1076 * the headroom they think they need without accounting for the
1077 * built in space. The built in space is used for optimisations.
1079 * %NULL is returned in there is no free memory. Although this function
1080 * allocates memory it can be called from an interrupt.
1082 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1084 return __dev_alloc_skb(length, GFP_ATOMIC);
1088 * skb_cow - copy header of skb when it is required
1089 * @skb: buffer to cow
1090 * @headroom: needed headroom
1092 * If the skb passed lacks sufficient headroom or its data part
1093 * is shared, data is reallocated. If reallocation fails, an error
1094 * is returned and original skb is not changed.
1096 * The result is skb with writable area skb->head...skb->tail
1097 * and at least @headroom of space at head.
1099 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1101 int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) -
1107 if (delta || skb_cloned(skb))
1108 return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) &
1109 ~(NET_SKB_PAD-1), 0, GFP_ATOMIC);
1114 * skb_padto - pad an skbuff up to a minimal size
1115 * @skb: buffer to pad
1116 * @len: minimal length
1118 * Pads up a buffer to ensure the trailing bytes exist and are
1119 * blanked. If the buffer already contains sufficient data it
1120 * is untouched. Returns the buffer, which may be a replacement
1121 * for the original, or NULL for out of memory - in which case
1122 * the original buffer is still freed.
1125 static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
1127 unsigned int size = skb->len;
1128 if (likely(size >= len))
1130 return skb_pad(skb, len-size);
1133 static inline int skb_add_data(struct sk_buff *skb,
1134 char __user *from, int copy)
1136 const int off = skb->len;
1138 if (skb->ip_summed == CHECKSUM_NONE) {
1140 unsigned int csum = csum_and_copy_from_user(from,
1144 skb->csum = csum_block_add(skb->csum, csum, off);
1147 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1150 __skb_trim(skb, off);
1154 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1155 struct page *page, int off)
1158 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1160 return page == frag->page &&
1161 off == frag->page_offset + frag->size;
1167 * skb_linearize - convert paged skb to linear one
1168 * @skb: buffer to linarize
1169 * @gfp: allocation mode
1171 * If there is no free memory -ENOMEM is returned, otherwise zero
1172 * is returned and the old skb data released.
1174 extern int __skb_linearize(struct sk_buff *skb, gfp_t gfp);
1175 static inline int skb_linearize(struct sk_buff *skb, gfp_t gfp)
1177 return __skb_linearize(skb, gfp);
1181 * skb_postpull_rcsum - update checksum for received skb after pull
1182 * @skb: buffer to update
1183 * @start: start of data before pull
1184 * @len: length of data pulled
1186 * After doing a pull on a received packet, you need to call this to
1187 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1188 * so that it can be recomputed from scratch.
1191 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1192 const void *start, unsigned int len)
1194 if (skb->ip_summed == CHECKSUM_HW)
1195 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1198 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1201 * pskb_trim_rcsum - trim received skb and update checksum
1202 * @skb: buffer to trim
1205 * This is exactly the same as pskb_trim except that it ensures the
1206 * checksum of received packets are still valid after the operation.
1209 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1211 if (likely(len >= skb->len))
1213 if (skb->ip_summed == CHECKSUM_HW)
1214 skb->ip_summed = CHECKSUM_NONE;
1215 return __pskb_trim(skb, len);
1218 static inline void *kmap_skb_frag(const skb_frag_t *frag)
1220 #ifdef CONFIG_HIGHMEM
1225 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1228 static inline void kunmap_skb_frag(void *vaddr)
1230 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1231 #ifdef CONFIG_HIGHMEM
1236 #define skb_queue_walk(queue, skb) \
1237 for (skb = (queue)->next; \
1238 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1241 #define skb_queue_reverse_walk(queue, skb) \
1242 for (skb = (queue)->prev; \
1243 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1247 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1248 int noblock, int *err);
1249 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1250 struct poll_table_struct *wait);
1251 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1252 int offset, struct iovec *to,
1254 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1257 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1258 extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1259 unsigned int flags);
1260 extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1261 int len, unsigned int csum);
1262 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1264 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1265 void *from, int len);
1266 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1267 int offset, u8 *to, int len,
1269 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1270 extern void skb_split(struct sk_buff *skb,
1271 struct sk_buff *skb1, const u32 len);
1273 extern void skb_release_data(struct sk_buff *skb);
1275 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1276 int len, void *buffer)
1278 int hlen = skb_headlen(skb);
1280 if (hlen - offset >= len)
1281 return skb->data + offset;
1283 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1289 extern void skb_init(void);
1290 extern void skb_add_mtu(int mtu);
1293 * skb_get_timestamp - get timestamp from a skb
1294 * @skb: skb to get stamp from
1295 * @stamp: pointer to struct timeval to store stamp in
1297 * Timestamps are stored in the skb as offsets to a base timestamp.
1298 * This function converts the offset back to a struct timeval and stores
1301 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1303 stamp->tv_sec = skb->tstamp.off_sec;
1304 stamp->tv_usec = skb->tstamp.off_usec;
1308 * skb_set_timestamp - set timestamp of a skb
1309 * @skb: skb to set stamp of
1310 * @stamp: pointer to struct timeval to get stamp from
1312 * Timestamps are stored in the skb as offsets to a base timestamp.
1313 * This function converts a struct timeval to an offset and stores
1316 static inline void skb_set_timestamp(struct sk_buff *skb, const struct timeval *stamp)
1318 skb->tstamp.off_sec = stamp->tv_sec;
1319 skb->tstamp.off_usec = stamp->tv_usec;
1322 extern void __net_timestamp(struct sk_buff *skb);
1324 extern unsigned int __skb_checksum_complete(struct sk_buff *skb);
1327 * skb_checksum_complete - Calculate checksum of an entire packet
1328 * @skb: packet to process
1330 * This function calculates the checksum over the entire packet plus
1331 * the value of skb->csum. The latter can be used to supply the
1332 * checksum of a pseudo header as used by TCP/UDP. It returns the
1335 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1336 * this function can be used to verify that checksum on received
1337 * packets. In that case the function should return zero if the
1338 * checksum is correct. In particular, this function will return zero
1339 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1340 * hardware has already verified the correctness of the checksum.
1342 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1344 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1345 __skb_checksum_complete(skb);
1348 struct tux_req_struct;
1350 #ifdef CONFIG_NETFILTER
1351 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1353 if (nfct && atomic_dec_and_test(&nfct->use))
1354 nfct->destroy(nfct);
1356 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1359 atomic_inc(&nfct->use);
1361 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1362 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1365 atomic_inc(&skb->users);
1367 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1373 #ifdef CONFIG_BRIDGE_NETFILTER
1374 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1376 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1379 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1382 atomic_inc(&nf_bridge->use);
1384 #endif /* CONFIG_BRIDGE_NETFILTER */
1385 static inline void nf_reset(struct sk_buff *skb)
1387 nf_conntrack_put(skb->nfct);
1389 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1390 nf_conntrack_put_reasm(skb->nfct_reasm);
1391 skb->nfct_reasm = NULL;
1393 #ifdef CONFIG_BRIDGE_NETFILTER
1394 nf_bridge_put(skb->nf_bridge);
1395 skb->nf_bridge = NULL;
1399 #else /* CONFIG_NETFILTER */
1400 static inline void nf_reset(struct sk_buff *skb) {}
1401 #endif /* CONFIG_NETFILTER */
1403 #endif /* __KERNEL__ */
1404 #endif /* _LINUX_SKBUFF_H */