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/kernel.h>
18 #include <linux/compiler.h>
19 #include <linux/time.h>
20 #include <linux/cache.h>
22 #include <asm/atomic.h>
23 #include <asm/types.h>
24 #include <linux/spinlock.h>
26 #include <linux/highmem.h>
27 #include <linux/poll.h>
28 #include <linux/net.h>
29 #include <linux/textsearch.h>
30 #include <net/checksum.h>
31 #include <linux/dmaengine.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 gso_size;
138 /* Warning: this field is not always filled in (UFO)! */
139 unsigned short gso_segs;
140 unsigned short gso_type;
141 unsigned int ip6_frag_id;
142 struct sk_buff *frag_list;
143 skb_frag_t frags[MAX_SKB_FRAGS];
146 /* We divide dataref into two halves. The higher 16 bits hold references
147 * to the payload part of skb->data. The lower 16 bits hold references to
148 * the entire skb->data. It is up to the users of the skb to agree on
149 * where the payload starts.
151 * All users must obey the rule that the skb->data reference count must be
152 * greater than or equal to the payload reference count.
154 * Holding a reference to the payload part means that the user does not
155 * care about modifications to the header part of skb->data.
157 #define SKB_DATAREF_SHIFT 16
158 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
167 SKB_FCLONE_UNAVAILABLE,
173 SKB_GSO_TCPV4 = 1 << 0,
174 SKB_GSO_UDP = 1 << 1,
176 /* This indicates the skb is from an untrusted source. */
177 SKB_GSO_DODGY = 1 << 2,
179 /* This indicates the tcp segment has CWR set. */
180 SKB_GSO_TCP_ECN = 1 << 3,
182 SKB_GSO_TCPV6 = 1 << 4,
186 * struct sk_buff - socket buffer
187 * @next: Next buffer in list
188 * @prev: Previous buffer in list
189 * @sk: Socket we are owned by
190 * @tstamp: Time we arrived
191 * @dev: Device we arrived on/are leaving by
192 * @input_dev: Device we arrived on
193 * @h: Transport layer header
194 * @nh: Network layer header
195 * @mac: Link layer header
196 * @dst: destination entry
197 * @sp: the security path, used for xfrm
198 * @cb: Control buffer. Free for use by every layer. Put private vars here
199 * @len: Length of actual data
200 * @data_len: Data length
201 * @mac_len: Length of link layer header
203 * @local_df: allow local fragmentation
204 * @cloned: Head may be cloned (check refcnt to be sure)
205 * @nohdr: Payload reference only, must not modify header
206 * @proto_data_valid: Protocol data validated since arriving at localhost
207 * @proto_csum_blank: Protocol csum must be added before leaving localhost
208 * @pkt_type: Packet class
209 * @fclone: skbuff clone status
210 * @ip_summed: Driver fed us an IP checksum
211 * @priority: Packet queueing priority
212 * @users: User count - see {datagram,tcp}.c
213 * @protocol: Packet protocol from driver
214 * @truesize: Buffer size
215 * @head: Head of buffer
216 * @data: Data head pointer
217 * @tail: Tail pointer
219 * @destructor: Destruct function
220 * @nfmark: Can be used for communication between hooks
221 * @nfct: Associated connection, if any
222 * @ipvs_property: skbuff is owned by ipvs
223 * @nfctinfo: Relationship of this skb to the connection
224 * @nfct_reasm: netfilter conntrack re-assembly pointer
225 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
226 * @tc_index: Traffic control index
227 * @tc_verd: traffic control verdict
228 * @dma_cookie: a cookie to one of several possible DMA operations
229 * done by skb DMA functions
230 * @secmark: security marking
234 /* These two members must be first. */
235 struct sk_buff *next;
236 struct sk_buff *prev;
239 struct skb_timeval tstamp;
240 struct net_device *dev;
241 struct net_device *input_dev;
246 struct icmphdr *icmph;
247 struct igmphdr *igmph;
249 struct ipv6hdr *ipv6h;
255 struct ipv6hdr *ipv6h;
264 struct dst_entry *dst;
268 * This is the control buffer. It is free to use for every
269 * layer. Please put your private variables there. If you
270 * want to keep them across layers you have to do a skb_clone()
271 * first. This is owned by whoever has the skb queued ATM.
296 void (*destructor)(struct sk_buff *skb);
297 #ifdef CONFIG_NETFILTER
298 struct nf_conntrack *nfct;
299 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
300 struct sk_buff *nfct_reasm;
302 #ifdef CONFIG_BRIDGE_NETFILTER
303 struct nf_bridge_info *nf_bridge;
306 #endif /* CONFIG_NETFILTER */
307 #ifdef CONFIG_NET_SCHED
308 __u16 tc_index; /* traffic control index */
309 #ifdef CONFIG_NET_CLS_ACT
310 __u16 tc_verd; /* traffic control verdict */
313 #ifdef CONFIG_NET_DMA
314 dma_cookie_t dma_cookie;
316 #ifdef CONFIG_NETWORK_SECMARK
321 /* These elements must be at the end, see alloc_skb() for details. */
322 unsigned int truesize;
332 * Handling routines are only of interest to the kernel
334 #include <linux/slab.h>
336 #include <asm/system.h>
338 extern void kfree_skb(struct sk_buff *skb);
339 extern void __kfree_skb(struct sk_buff *skb);
340 extern struct sk_buff *__alloc_skb(unsigned int size,
341 gfp_t priority, int fclone);
342 static inline struct sk_buff *alloc_skb(unsigned int size,
345 return __alloc_skb(size, priority, 0);
348 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
351 return __alloc_skb(size, priority, 1);
354 extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
358 extern void kfree_skbmem(struct sk_buff *skb);
359 extern struct sk_buff *skb_clone(struct sk_buff *skb,
361 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
363 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
365 extern int pskb_expand_head(struct sk_buff *skb,
366 int nhead, int ntail,
368 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
369 unsigned int headroom);
370 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
371 int newheadroom, int newtailroom,
373 extern int skb_pad(struct sk_buff *skb, int pad);
374 #define dev_kfree_skb(a) kfree_skb(a)
375 extern void skb_over_panic(struct sk_buff *skb, int len,
377 extern void skb_under_panic(struct sk_buff *skb, int len,
379 extern void skb_truesize_bug(struct sk_buff *skb);
381 static inline void skb_truesize_check(struct sk_buff *skb)
383 if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len))
384 skb_truesize_bug(skb);
387 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
388 int getfrag(void *from, char *to, int offset,
389 int len,int odd, struct sk_buff *skb),
390 void *from, int length);
397 __u32 stepped_offset;
398 struct sk_buff *root_skb;
399 struct sk_buff *cur_skb;
403 extern void skb_prepare_seq_read(struct sk_buff *skb,
404 unsigned int from, unsigned int to,
405 struct skb_seq_state *st);
406 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
407 struct skb_seq_state *st);
408 extern void skb_abort_seq_read(struct skb_seq_state *st);
410 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
411 unsigned int to, struct ts_config *config,
412 struct ts_state *state);
415 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
418 * skb_queue_empty - check if a queue is empty
421 * Returns true if the queue is empty, false otherwise.
423 static inline int skb_queue_empty(const struct sk_buff_head *list)
425 return list->next == (struct sk_buff *)list;
429 * skb_get - reference buffer
430 * @skb: buffer to reference
432 * Makes another reference to a socket buffer and returns a pointer
435 static inline struct sk_buff *skb_get(struct sk_buff *skb)
437 atomic_inc(&skb->users);
442 * If users == 1, we are the only owner and are can avoid redundant
447 * skb_cloned - is the buffer a clone
448 * @skb: buffer to check
450 * Returns true if the buffer was generated with skb_clone() and is
451 * one of multiple shared copies of the buffer. Cloned buffers are
452 * shared data so must not be written to under normal circumstances.
454 static inline int skb_cloned(const struct sk_buff *skb)
456 return skb->cloned &&
457 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
461 * skb_header_cloned - is the header a clone
462 * @skb: buffer to check
464 * Returns true if modifying the header part of the buffer requires
465 * the data to be copied.
467 static inline int skb_header_cloned(const struct sk_buff *skb)
474 dataref = atomic_read(&skb_shinfo(skb)->dataref);
475 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
480 * skb_header_release - release reference to header
481 * @skb: buffer to operate on
483 * Drop a reference to the header part of the buffer. This is done
484 * by acquiring a payload reference. You must not read from the header
485 * part of skb->data after this.
487 static inline void skb_header_release(struct sk_buff *skb)
491 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
495 * skb_shared - is the buffer shared
496 * @skb: buffer to check
498 * Returns true if more than one person has a reference to this
501 static inline int skb_shared(const struct sk_buff *skb)
503 return atomic_read(&skb->users) != 1;
507 * skb_share_check - check if buffer is shared and if so clone it
508 * @skb: buffer to check
509 * @pri: priority for memory allocation
511 * If the buffer is shared the buffer is cloned and the old copy
512 * drops a reference. A new clone with a single reference is returned.
513 * If the buffer is not shared the original buffer is returned. When
514 * being called from interrupt status or with spinlocks held pri must
517 * NULL is returned on a memory allocation failure.
519 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
522 might_sleep_if(pri & __GFP_WAIT);
523 if (skb_shared(skb)) {
524 struct sk_buff *nskb = skb_clone(skb, pri);
532 * Copy shared buffers into a new sk_buff. We effectively do COW on
533 * packets to handle cases where we have a local reader and forward
534 * and a couple of other messy ones. The normal one is tcpdumping
535 * a packet thats being forwarded.
539 * skb_unshare - make a copy of a shared buffer
540 * @skb: buffer to check
541 * @pri: priority for memory allocation
543 * If the socket buffer is a clone then this function creates a new
544 * copy of the data, drops a reference count on the old copy and returns
545 * the new copy with the reference count at 1. If the buffer is not a clone
546 * the original buffer is returned. When called with a spinlock held or
547 * from interrupt state @pri must be %GFP_ATOMIC
549 * %NULL is returned on a memory allocation failure.
551 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
554 might_sleep_if(pri & __GFP_WAIT);
555 if (skb_cloned(skb)) {
556 struct sk_buff *nskb = skb_copy(skb, pri);
557 kfree_skb(skb); /* Free our shared copy */
565 * @list_: list to peek at
567 * Peek an &sk_buff. Unlike most other operations you _MUST_
568 * be careful with this one. A peek leaves the buffer on the
569 * list and someone else may run off with it. You must hold
570 * the appropriate locks or have a private queue to do this.
572 * Returns %NULL for an empty list or a pointer to the head element.
573 * The reference count is not incremented and the reference is therefore
574 * volatile. Use with caution.
576 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
578 struct sk_buff *list = ((struct sk_buff *)list_)->next;
579 if (list == (struct sk_buff *)list_)
586 * @list_: list to peek at
588 * Peek an &sk_buff. Unlike most other operations you _MUST_
589 * be careful with this one. A peek leaves the buffer on the
590 * list and someone else may run off with it. You must hold
591 * the appropriate locks or have a private queue to do this.
593 * Returns %NULL for an empty list or a pointer to the tail element.
594 * The reference count is not incremented and the reference is therefore
595 * volatile. Use with caution.
597 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
599 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
600 if (list == (struct sk_buff *)list_)
606 * skb_queue_len - get queue length
607 * @list_: list to measure
609 * Return the length of an &sk_buff queue.
611 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
617 * This function creates a split out lock class for each invocation;
618 * this is needed for now since a whole lot of users of the skb-queue
619 * infrastructure in drivers have different locking usage (in hardirq)
620 * than the networking core (in softirq only). In the long run either the
621 * network layer or drivers should need annotation to consolidate the
622 * main types of usage into 3 classes.
624 static inline void skb_queue_head_init(struct sk_buff_head *list)
626 spin_lock_init(&list->lock);
627 list->prev = list->next = (struct sk_buff *)list;
632 * Insert an sk_buff at the start of a list.
634 * The "__skb_xxxx()" functions are the non-atomic ones that
635 * can only be called with interrupts disabled.
639 * __skb_queue_after - queue a buffer at the list head
641 * @prev: place after this buffer
642 * @newsk: buffer to queue
644 * Queue a buffer int the middle of a list. This function takes no locks
645 * and you must therefore hold required locks before calling it.
647 * A buffer cannot be placed on two lists at the same time.
649 static inline void __skb_queue_after(struct sk_buff_head *list,
650 struct sk_buff *prev,
651 struct sk_buff *newsk)
653 struct sk_buff *next;
659 next->prev = prev->next = newsk;
663 * __skb_queue_head - queue a buffer at the list head
665 * @newsk: buffer to queue
667 * Queue a buffer at the start of a list. This function takes no locks
668 * and you must therefore hold required locks before calling it.
670 * A buffer cannot be placed on two lists at the same time.
672 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
673 static inline void __skb_queue_head(struct sk_buff_head *list,
674 struct sk_buff *newsk)
676 __skb_queue_after(list, (struct sk_buff *)list, newsk);
680 * __skb_queue_tail - queue a buffer at the list tail
682 * @newsk: buffer to queue
684 * Queue a buffer at the end of a list. This function takes no locks
685 * and you must therefore hold required locks before calling it.
687 * A buffer cannot be placed on two lists at the same time.
689 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
690 static inline void __skb_queue_tail(struct sk_buff_head *list,
691 struct sk_buff *newsk)
693 struct sk_buff *prev, *next;
696 next = (struct sk_buff *)list;
700 next->prev = prev->next = newsk;
705 * __skb_dequeue - remove from the head of the queue
706 * @list: list to dequeue from
708 * Remove the head of the list. This function does not take any locks
709 * so must be used with appropriate locks held only. The head item is
710 * returned or %NULL if the list is empty.
712 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
713 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
715 struct sk_buff *next, *prev, *result;
717 prev = (struct sk_buff *) list;
726 result->next = result->prev = NULL;
733 * Insert a packet on a list.
735 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
736 static inline void __skb_insert(struct sk_buff *newsk,
737 struct sk_buff *prev, struct sk_buff *next,
738 struct sk_buff_head *list)
742 next->prev = prev->next = newsk;
747 * Place a packet after a given packet in a list.
749 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
750 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
752 __skb_insert(newsk, old, old->next, list);
756 * remove sk_buff from list. _Must_ be called atomically, and with
759 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
760 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
762 struct sk_buff *next, *prev;
767 skb->next = skb->prev = NULL;
773 /* XXX: more streamlined implementation */
776 * __skb_dequeue_tail - remove from the tail of the queue
777 * @list: list to dequeue from
779 * Remove the tail of the list. This function does not take any locks
780 * so must be used with appropriate locks held only. The tail item is
781 * returned or %NULL if the list is empty.
783 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
784 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
786 struct sk_buff *skb = skb_peek_tail(list);
788 __skb_unlink(skb, list);
793 static inline int skb_is_nonlinear(const struct sk_buff *skb)
795 return skb->data_len;
798 static inline unsigned int skb_headlen(const struct sk_buff *skb)
800 return skb->len - skb->data_len;
803 static inline int skb_pagelen(const struct sk_buff *skb)
807 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
808 len += skb_shinfo(skb)->frags[i].size;
809 return len + skb_headlen(skb);
812 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
813 struct page *page, int off, int size)
815 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
818 frag->page_offset = off;
820 skb_shinfo(skb)->nr_frags = i + 1;
823 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
824 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
825 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
828 * Add data to an sk_buff
830 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
832 unsigned char *tmp = skb->tail;
833 SKB_LINEAR_ASSERT(skb);
840 * skb_put - add data to a buffer
841 * @skb: buffer to use
842 * @len: amount of data to add
844 * This function extends the used data area of the buffer. If this would
845 * exceed the total buffer size the kernel will panic. A pointer to the
846 * first byte of the extra data is returned.
848 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
850 unsigned char *tmp = skb->tail;
851 SKB_LINEAR_ASSERT(skb);
854 if (unlikely(skb->tail>skb->end))
855 skb_over_panic(skb, len, current_text_addr());
859 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
867 * skb_push - add data to the start of a buffer
868 * @skb: buffer to use
869 * @len: amount of data to add
871 * This function extends the used data area of the buffer at the buffer
872 * start. If this would exceed the total buffer headroom the kernel will
873 * panic. A pointer to the first byte of the extra data is returned.
875 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
879 if (unlikely(skb->data<skb->head))
880 skb_under_panic(skb, len, current_text_addr());
884 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
887 BUG_ON(skb->len < skb->data_len);
888 return skb->data += len;
892 * skb_pull - remove data from the start of a buffer
893 * @skb: buffer to use
894 * @len: amount of data to remove
896 * This function removes data from the start of a buffer, returning
897 * the memory to the headroom. A pointer to the next data in the buffer
898 * is returned. Once the data has been pulled future pushes will overwrite
901 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
903 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
906 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
908 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
910 if (len > skb_headlen(skb) &&
911 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
914 return skb->data += len;
917 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
919 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
922 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
924 if (likely(len <= skb_headlen(skb)))
926 if (unlikely(len > skb->len))
928 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
932 * skb_headroom - bytes at buffer head
933 * @skb: buffer to check
935 * Return the number of bytes of free space at the head of an &sk_buff.
937 static inline int skb_headroom(const struct sk_buff *skb)
939 return skb->data - skb->head;
943 * skb_tailroom - bytes at buffer end
944 * @skb: buffer to check
946 * Return the number of bytes of free space at the tail of an sk_buff
948 static inline int skb_tailroom(const struct sk_buff *skb)
950 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
954 * skb_reserve - adjust headroom
955 * @skb: buffer to alter
956 * @len: bytes to move
958 * Increase the headroom of an empty &sk_buff by reducing the tail
959 * room. This is only allowed for an empty buffer.
961 static inline void skb_reserve(struct sk_buff *skb, int len)
968 * CPUs often take a performance hit when accessing unaligned memory
969 * locations. The actual performance hit varies, it can be small if the
970 * hardware handles it or large if we have to take an exception and fix it
973 * Since an ethernet header is 14 bytes network drivers often end up with
974 * the IP header at an unaligned offset. The IP header can be aligned by
975 * shifting the start of the packet by 2 bytes. Drivers should do this
978 * skb_reserve(NET_IP_ALIGN);
980 * The downside to this alignment of the IP header is that the DMA is now
981 * unaligned. On some architectures the cost of an unaligned DMA is high
982 * and this cost outweighs the gains made by aligning the IP header.
984 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
988 #define NET_IP_ALIGN 2
992 * The networking layer reserves some headroom in skb data (via
993 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
994 * the header has to grow. In the default case, if the header has to grow
995 * 16 bytes or less we avoid the reallocation.
997 * Unfortunately this headroom changes the DMA alignment of the resulting
998 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
999 * on some architectures. An architecture can override this value,
1000 * perhaps setting it to a cacheline in size (since that will maintain
1001 * cacheline alignment of the DMA). It must be a power of 2.
1003 * Various parts of the networking layer expect at least 16 bytes of
1004 * headroom, you should not reduce this.
1007 #define NET_SKB_PAD 16
1010 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1012 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1014 if (unlikely(skb->data_len)) {
1019 skb->tail = skb->data + len;
1023 * skb_trim - remove end from a buffer
1024 * @skb: buffer to alter
1027 * Cut the length of a buffer down by removing data from the tail. If
1028 * the buffer is already under the length specified it is not modified.
1029 * The skb must be linear.
1031 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
1034 __skb_trim(skb, len);
1038 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1041 return ___pskb_trim(skb, len);
1042 __skb_trim(skb, len);
1046 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1048 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1052 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1053 * @skb: buffer to alter
1056 * This is identical to pskb_trim except that the caller knows that
1057 * the skb is not cloned so we should never get an error due to out-
1060 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1062 int err = pskb_trim(skb, len);
1067 * skb_orphan - orphan a buffer
1068 * @skb: buffer to orphan
1070 * If a buffer currently has an owner then we call the owner's
1071 * destructor function and make the @skb unowned. The buffer continues
1072 * to exist but is no longer charged to its former owner.
1074 static inline void skb_orphan(struct sk_buff *skb)
1076 if (skb->destructor)
1077 skb->destructor(skb);
1078 skb->destructor = NULL;
1083 * __skb_queue_purge - empty a list
1084 * @list: list to empty
1086 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1087 * the list and one reference dropped. This function does not take the
1088 * list lock and the caller must hold the relevant locks to use it.
1090 extern void skb_queue_purge(struct sk_buff_head *list);
1091 static inline void __skb_queue_purge(struct sk_buff_head *list)
1093 struct sk_buff *skb;
1094 while ((skb = __skb_dequeue(list)) != NULL)
1098 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
1100 * __dev_alloc_skb - allocate an skbuff for receiving
1101 * @length: length to allocate
1102 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1104 * Allocate a new &sk_buff and assign it a usage count of one. The
1105 * buffer has unspecified headroom built in. Users should allocate
1106 * the headroom they think they need without accounting for the
1107 * built in space. The built in space is used for optimisations.
1109 * %NULL is returned if there is no free memory.
1111 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1114 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1116 skb_reserve(skb, NET_SKB_PAD);
1120 extern struct sk_buff *__dev_alloc_skb(unsigned int length, gfp_t gfp_mask);
1124 * dev_alloc_skb - allocate an skbuff for receiving
1125 * @length: length to allocate
1127 * Allocate a new &sk_buff and assign it a usage count of one. The
1128 * buffer has unspecified headroom built in. Users should allocate
1129 * the headroom they think they need without accounting for the
1130 * built in space. The built in space is used for optimisations.
1132 * %NULL is returned if there is no free memory. Although this function
1133 * allocates memory it can be called from an interrupt.
1135 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1137 return __dev_alloc_skb(length, GFP_ATOMIC);
1140 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1141 unsigned int length, gfp_t gfp_mask);
1144 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1145 * @dev: network device to receive on
1146 * @length: length to allocate
1148 * Allocate a new &sk_buff and assign it a usage count of one. The
1149 * buffer has unspecified headroom built in. Users should allocate
1150 * the headroom they think they need without accounting for the
1151 * built in space. The built in space is used for optimisations.
1153 * %NULL is returned if there is no free memory. Although this function
1154 * allocates memory it can be called from an interrupt.
1156 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1157 unsigned int length)
1159 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1163 * skb_cow - copy header of skb when it is required
1164 * @skb: buffer to cow
1165 * @headroom: needed headroom
1167 * If the skb passed lacks sufficient headroom or its data part
1168 * is shared, data is reallocated. If reallocation fails, an error
1169 * is returned and original skb is not changed.
1171 * The result is skb with writable area skb->head...skb->tail
1172 * and at least @headroom of space at head.
1174 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1176 int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) -
1182 if (delta || skb_cloned(skb))
1183 return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) &
1184 ~(NET_SKB_PAD-1), 0, GFP_ATOMIC);
1189 * skb_padto - pad an skbuff up to a minimal size
1190 * @skb: buffer to pad
1191 * @len: minimal length
1193 * Pads up a buffer to ensure the trailing bytes exist and are
1194 * blanked. If the buffer already contains sufficient data it
1195 * is untouched. Otherwise it is extended. Returns zero on
1196 * success. The skb is freed on error.
1199 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1201 unsigned int size = skb->len;
1202 if (likely(size >= len))
1204 return skb_pad(skb, len-size);
1207 static inline int skb_add_data(struct sk_buff *skb,
1208 char __user *from, int copy)
1210 const int off = skb->len;
1212 if (skb->ip_summed == CHECKSUM_NONE) {
1214 unsigned int csum = csum_and_copy_from_user(from,
1218 skb->csum = csum_block_add(skb->csum, csum, off);
1221 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1224 __skb_trim(skb, off);
1228 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1229 struct page *page, int off)
1232 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1234 return page == frag->page &&
1235 off == frag->page_offset + frag->size;
1240 static inline int __skb_linearize(struct sk_buff *skb)
1242 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1246 * skb_linearize - convert paged skb to linear one
1247 * @skb: buffer to linarize
1249 * If there is no free memory -ENOMEM is returned, otherwise zero
1250 * is returned and the old skb data released.
1252 static inline int skb_linearize(struct sk_buff *skb)
1254 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1258 * skb_linearize_cow - make sure skb is linear and writable
1259 * @skb: buffer to process
1261 * If there is no free memory -ENOMEM is returned, otherwise zero
1262 * is returned and the old skb data released.
1264 static inline int skb_linearize_cow(struct sk_buff *skb)
1266 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1267 __skb_linearize(skb) : 0;
1271 * skb_postpull_rcsum - update checksum for received skb after pull
1272 * @skb: buffer to update
1273 * @start: start of data before pull
1274 * @len: length of data pulled
1276 * After doing a pull on a received packet, you need to call this to
1277 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1278 * so that it can be recomputed from scratch.
1281 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1282 const void *start, unsigned int len)
1284 if (skb->ip_summed == CHECKSUM_HW)
1285 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1288 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1291 * pskb_trim_rcsum - trim received skb and update checksum
1292 * @skb: buffer to trim
1295 * This is exactly the same as pskb_trim except that it ensures the
1296 * checksum of received packets are still valid after the operation.
1299 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1301 if (likely(len >= skb->len))
1303 if (skb->ip_summed == CHECKSUM_HW)
1304 skb->ip_summed = CHECKSUM_NONE;
1305 return __pskb_trim(skb, len);
1308 static inline void *kmap_skb_frag(const skb_frag_t *frag)
1310 #ifdef CONFIG_HIGHMEM
1315 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1318 static inline void kunmap_skb_frag(void *vaddr)
1320 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1321 #ifdef CONFIG_HIGHMEM
1326 #define skb_queue_walk(queue, skb) \
1327 for (skb = (queue)->next; \
1328 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1331 #define skb_queue_reverse_walk(queue, skb) \
1332 for (skb = (queue)->prev; \
1333 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1337 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1338 int noblock, int *err);
1339 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1340 struct poll_table_struct *wait);
1341 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1342 int offset, struct iovec *to,
1344 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1347 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1348 extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1349 unsigned int flags);
1350 extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1351 int len, unsigned int csum);
1352 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1354 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1355 void *from, int len);
1356 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1357 int offset, u8 *to, int len,
1359 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1360 extern void skb_split(struct sk_buff *skb,
1361 struct sk_buff *skb1, const u32 len);
1363 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1365 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1366 int len, void *buffer)
1368 int hlen = skb_headlen(skb);
1370 if (hlen - offset >= len)
1371 return skb->data + offset;
1373 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1379 extern void skb_init(void);
1380 extern void skb_add_mtu(int mtu);
1383 * skb_get_timestamp - get timestamp from a skb
1384 * @skb: skb to get stamp from
1385 * @stamp: pointer to struct timeval to store stamp in
1387 * Timestamps are stored in the skb as offsets to a base timestamp.
1388 * This function converts the offset back to a struct timeval and stores
1391 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1393 stamp->tv_sec = skb->tstamp.off_sec;
1394 stamp->tv_usec = skb->tstamp.off_usec;
1398 * skb_set_timestamp - set timestamp of a skb
1399 * @skb: skb to set stamp of
1400 * @stamp: pointer to struct timeval to get stamp from
1402 * Timestamps are stored in the skb as offsets to a base timestamp.
1403 * This function converts a struct timeval to an offset and stores
1406 static inline void skb_set_timestamp(struct sk_buff *skb, const struct timeval *stamp)
1408 skb->tstamp.off_sec = stamp->tv_sec;
1409 skb->tstamp.off_usec = stamp->tv_usec;
1412 extern void __net_timestamp(struct sk_buff *skb);
1414 extern unsigned int __skb_checksum_complete(struct sk_buff *skb);
1417 * skb_checksum_complete - Calculate checksum of an entire packet
1418 * @skb: packet to process
1420 * This function calculates the checksum over the entire packet plus
1421 * the value of skb->csum. The latter can be used to supply the
1422 * checksum of a pseudo header as used by TCP/UDP. It returns the
1425 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1426 * this function can be used to verify that checksum on received
1427 * packets. In that case the function should return zero if the
1428 * checksum is correct. In particular, this function will return zero
1429 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1430 * hardware has already verified the correctness of the checksum.
1432 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1434 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1435 __skb_checksum_complete(skb);
1438 struct tux_req_struct;
1440 #ifdef CONFIG_NETFILTER
1441 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1443 if (nfct && atomic_dec_and_test(&nfct->use))
1444 nfct->destroy(nfct);
1446 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1449 atomic_inc(&nfct->use);
1451 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1452 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1455 atomic_inc(&skb->users);
1457 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1463 #ifdef CONFIG_BRIDGE_NETFILTER
1464 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1466 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1469 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1472 atomic_inc(&nf_bridge->use);
1474 #endif /* CONFIG_BRIDGE_NETFILTER */
1475 static inline void nf_reset(struct sk_buff *skb)
1477 nf_conntrack_put(skb->nfct);
1479 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1480 nf_conntrack_put_reasm(skb->nfct_reasm);
1481 skb->nfct_reasm = NULL;
1483 #ifdef CONFIG_BRIDGE_NETFILTER
1484 nf_bridge_put(skb->nf_bridge);
1485 skb->nf_bridge = NULL;
1489 #else /* CONFIG_NETFILTER */
1490 static inline void nf_reset(struct sk_buff *skb) {}
1491 #endif /* CONFIG_NETFILTER */
1493 #ifdef CONFIG_NETWORK_SECMARK
1494 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1496 to->secmark = from->secmark;
1499 static inline void skb_init_secmark(struct sk_buff *skb)
1504 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1507 static inline void skb_init_secmark(struct sk_buff *skb)
1511 static inline int skb_is_gso(const struct sk_buff *skb)
1513 return skb_shinfo(skb)->gso_size;
1516 #endif /* __KERNEL__ */
1517 #endif /* _LINUX_SKBUFF_H */