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_UDPV4 = 1 << 1,
176 /* This indicates the skb is from an untrusted source. */
177 SKB_GSO_DODGY = 1 << 2,
181 * struct sk_buff - socket buffer
182 * @next: Next buffer in list
183 * @prev: Previous buffer in list
184 * @sk: Socket we are owned by
185 * @tstamp: Time we arrived
186 * @dev: Device we arrived on/are leaving by
187 * @input_dev: Device we arrived on
188 * @h: Transport layer header
189 * @nh: Network layer header
190 * @mac: Link layer header
191 * @dst: destination entry
192 * @sp: the security path, used for xfrm
193 * @cb: Control buffer. Free for use by every layer. Put private vars here
194 * @len: Length of actual data
195 * @data_len: Data length
196 * @mac_len: Length of link layer header
198 * @local_df: allow local fragmentation
199 * @cloned: Head may be cloned (check refcnt to be sure)
200 * @nohdr: Payload reference only, must not modify header
201 * @pkt_type: Packet class
202 * @fclone: skbuff clone status
203 * @ip_summed: Driver fed us an IP checksum
204 * @priority: Packet queueing priority
205 * @users: User count - see {datagram,tcp}.c
206 * @protocol: Packet protocol from driver
207 * @truesize: Buffer size
208 * @head: Head of buffer
209 * @data: Data head pointer
210 * @tail: Tail pointer
212 * @destructor: Destruct function
213 * @nfmark: Can be used for communication between hooks
214 * @nfct: Associated connection, if any
215 * @ipvs_property: skbuff is owned by ipvs
216 * @nfctinfo: Relationship of this skb to the connection
217 * @nfct_reasm: netfilter conntrack re-assembly pointer
218 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
219 * @tc_index: Traffic control index
220 * @tc_verd: traffic control verdict
221 * @dma_cookie: a cookie to one of several possible DMA operations
222 * done by skb DMA functions
223 * @secmark: security marking
227 /* These two members must be first. */
228 struct sk_buff *next;
229 struct sk_buff *prev;
232 struct skb_timeval tstamp;
233 struct net_device *dev;
234 struct net_device *input_dev;
239 struct icmphdr *icmph;
240 struct igmphdr *igmph;
242 struct ipv6hdr *ipv6h;
248 struct ipv6hdr *ipv6h;
257 struct dst_entry *dst;
261 * This is the control buffer. It is free to use for every
262 * layer. Please put your private variables there. If you
263 * want to keep them across layers you have to do a skb_clone()
264 * first. This is owned by whoever has the skb queued ATM.
283 void (*destructor)(struct sk_buff *skb);
284 #ifdef CONFIG_NETFILTER
285 struct nf_conntrack *nfct;
286 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
287 struct sk_buff *nfct_reasm;
289 #ifdef CONFIG_BRIDGE_NETFILTER
290 struct nf_bridge_info *nf_bridge;
293 #endif /* CONFIG_NETFILTER */
294 #ifdef CONFIG_NET_SCHED
295 __u16 tc_index; /* traffic control index */
296 #ifdef CONFIG_NET_CLS_ACT
297 __u16 tc_verd; /* traffic control verdict */
300 #ifdef CONFIG_NET_DMA
301 dma_cookie_t dma_cookie;
303 #ifdef CONFIG_NETWORK_SECMARK
308 /* These elements must be at the end, see alloc_skb() for details. */
309 unsigned int truesize;
319 * Handling routines are only of interest to the kernel
321 #include <linux/slab.h>
323 #include <asm/system.h>
325 extern void kfree_skb(struct sk_buff *skb);
326 extern void __kfree_skb(struct sk_buff *skb);
327 extern struct sk_buff *__alloc_skb(unsigned int size,
328 gfp_t priority, int fclone);
329 static inline struct sk_buff *alloc_skb(unsigned int size,
332 return __alloc_skb(size, priority, 0);
335 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
338 return __alloc_skb(size, priority, 1);
341 extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
344 extern void kfree_skbmem(struct sk_buff *skb);
345 extern struct sk_buff *skb_clone(struct sk_buff *skb,
347 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
349 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
351 extern int pskb_expand_head(struct sk_buff *skb,
352 int nhead, int ntail,
354 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
355 unsigned int headroom);
356 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
357 int newheadroom, int newtailroom,
359 extern int skb_pad(struct sk_buff *skb, int pad);
360 #define dev_kfree_skb(a) kfree_skb(a)
361 extern void skb_over_panic(struct sk_buff *skb, int len,
363 extern void skb_under_panic(struct sk_buff *skb, int len,
365 extern void skb_truesize_bug(struct sk_buff *skb);
367 static inline void skb_truesize_check(struct sk_buff *skb)
369 if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len))
370 skb_truesize_bug(skb);
373 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
374 int getfrag(void *from, char *to, int offset,
375 int len,int odd, struct sk_buff *skb),
376 void *from, int length);
383 __u32 stepped_offset;
384 struct sk_buff *root_skb;
385 struct sk_buff *cur_skb;
389 extern void skb_prepare_seq_read(struct sk_buff *skb,
390 unsigned int from, unsigned int to,
391 struct skb_seq_state *st);
392 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
393 struct skb_seq_state *st);
394 extern void skb_abort_seq_read(struct skb_seq_state *st);
396 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
397 unsigned int to, struct ts_config *config,
398 struct ts_state *state);
401 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
404 * skb_queue_empty - check if a queue is empty
407 * Returns true if the queue is empty, false otherwise.
409 static inline int skb_queue_empty(const struct sk_buff_head *list)
411 return list->next == (struct sk_buff *)list;
415 * skb_get - reference buffer
416 * @skb: buffer to reference
418 * Makes another reference to a socket buffer and returns a pointer
421 static inline struct sk_buff *skb_get(struct sk_buff *skb)
423 atomic_inc(&skb->users);
428 * If users == 1, we are the only owner and are can avoid redundant
433 * skb_cloned - is the buffer a clone
434 * @skb: buffer to check
436 * Returns true if the buffer was generated with skb_clone() and is
437 * one of multiple shared copies of the buffer. Cloned buffers are
438 * shared data so must not be written to under normal circumstances.
440 static inline int skb_cloned(const struct sk_buff *skb)
442 return skb->cloned &&
443 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
447 * skb_header_cloned - is the header a clone
448 * @skb: buffer to check
450 * Returns true if modifying the header part of the buffer requires
451 * the data to be copied.
453 static inline int skb_header_cloned(const struct sk_buff *skb)
460 dataref = atomic_read(&skb_shinfo(skb)->dataref);
461 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
466 * skb_header_release - release reference to header
467 * @skb: buffer to operate on
469 * Drop a reference to the header part of the buffer. This is done
470 * by acquiring a payload reference. You must not read from the header
471 * part of skb->data after this.
473 static inline void skb_header_release(struct sk_buff *skb)
477 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
481 * skb_shared - is the buffer shared
482 * @skb: buffer to check
484 * Returns true if more than one person has a reference to this
487 static inline int skb_shared(const struct sk_buff *skb)
489 return atomic_read(&skb->users) != 1;
493 * skb_share_check - check if buffer is shared and if so clone it
494 * @skb: buffer to check
495 * @pri: priority for memory allocation
497 * If the buffer is shared the buffer is cloned and the old copy
498 * drops a reference. A new clone with a single reference is returned.
499 * If the buffer is not shared the original buffer is returned. When
500 * being called from interrupt status or with spinlocks held pri must
503 * NULL is returned on a memory allocation failure.
505 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
508 might_sleep_if(pri & __GFP_WAIT);
509 if (skb_shared(skb)) {
510 struct sk_buff *nskb = skb_clone(skb, pri);
518 * Copy shared buffers into a new sk_buff. We effectively do COW on
519 * packets to handle cases where we have a local reader and forward
520 * and a couple of other messy ones. The normal one is tcpdumping
521 * a packet thats being forwarded.
525 * skb_unshare - make a copy of a shared buffer
526 * @skb: buffer to check
527 * @pri: priority for memory allocation
529 * If the socket buffer is a clone then this function creates a new
530 * copy of the data, drops a reference count on the old copy and returns
531 * the new copy with the reference count at 1. If the buffer is not a clone
532 * the original buffer is returned. When called with a spinlock held or
533 * from interrupt state @pri must be %GFP_ATOMIC
535 * %NULL is returned on a memory allocation failure.
537 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
540 might_sleep_if(pri & __GFP_WAIT);
541 if (skb_cloned(skb)) {
542 struct sk_buff *nskb = skb_copy(skb, pri);
543 kfree_skb(skb); /* Free our shared copy */
551 * @list_: list to peek at
553 * Peek an &sk_buff. Unlike most other operations you _MUST_
554 * be careful with this one. A peek leaves the buffer on the
555 * list and someone else may run off with it. You must hold
556 * the appropriate locks or have a private queue to do this.
558 * Returns %NULL for an empty list or a pointer to the head element.
559 * The reference count is not incremented and the reference is therefore
560 * volatile. Use with caution.
562 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
564 struct sk_buff *list = ((struct sk_buff *)list_)->next;
565 if (list == (struct sk_buff *)list_)
572 * @list_: list to peek at
574 * Peek an &sk_buff. Unlike most other operations you _MUST_
575 * be careful with this one. A peek leaves the buffer on the
576 * list and someone else may run off with it. You must hold
577 * the appropriate locks or have a private queue to do this.
579 * Returns %NULL for an empty list or a pointer to the tail element.
580 * The reference count is not incremented and the reference is therefore
581 * volatile. Use with caution.
583 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
585 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
586 if (list == (struct sk_buff *)list_)
592 * skb_queue_len - get queue length
593 * @list_: list to measure
595 * Return the length of an &sk_buff queue.
597 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
602 static inline void skb_queue_head_init(struct sk_buff_head *list)
604 spin_lock_init(&list->lock);
605 list->prev = list->next = (struct sk_buff *)list;
610 * Insert an sk_buff at the start of a list.
612 * The "__skb_xxxx()" functions are the non-atomic ones that
613 * can only be called with interrupts disabled.
617 * __skb_queue_after - queue a buffer at the list head
619 * @prev: place after this buffer
620 * @newsk: buffer to queue
622 * Queue a buffer int the middle of a list. This function takes no locks
623 * and you must therefore hold required locks before calling it.
625 * A buffer cannot be placed on two lists at the same time.
627 static inline void __skb_queue_after(struct sk_buff_head *list,
628 struct sk_buff *prev,
629 struct sk_buff *newsk)
631 struct sk_buff *next;
637 next->prev = prev->next = newsk;
641 * __skb_queue_head - queue a buffer at the list head
643 * @newsk: buffer to queue
645 * Queue a buffer at the start of a list. This function takes no locks
646 * and you must therefore hold required locks before calling it.
648 * A buffer cannot be placed on two lists at the same time.
650 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
651 static inline void __skb_queue_head(struct sk_buff_head *list,
652 struct sk_buff *newsk)
654 __skb_queue_after(list, (struct sk_buff *)list, newsk);
658 * __skb_queue_tail - queue a buffer at the list tail
660 * @newsk: buffer to queue
662 * Queue a buffer at the end of a list. This function takes no locks
663 * and you must therefore hold required locks before calling it.
665 * A buffer cannot be placed on two lists at the same time.
667 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
668 static inline void __skb_queue_tail(struct sk_buff_head *list,
669 struct sk_buff *newsk)
671 struct sk_buff *prev, *next;
674 next = (struct sk_buff *)list;
678 next->prev = prev->next = newsk;
683 * __skb_dequeue - remove from the head of the queue
684 * @list: list to dequeue from
686 * Remove the head of the list. This function does not take any locks
687 * so must be used with appropriate locks held only. The head item is
688 * returned or %NULL if the list is empty.
690 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
691 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
693 struct sk_buff *next, *prev, *result;
695 prev = (struct sk_buff *) list;
704 result->next = result->prev = NULL;
711 * Insert a packet on a list.
713 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
714 static inline void __skb_insert(struct sk_buff *newsk,
715 struct sk_buff *prev, struct sk_buff *next,
716 struct sk_buff_head *list)
720 next->prev = prev->next = newsk;
725 * Place a packet after a given packet in a list.
727 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
728 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
730 __skb_insert(newsk, old, old->next, list);
734 * remove sk_buff from list. _Must_ be called atomically, and with
737 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
738 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
740 struct sk_buff *next, *prev;
745 skb->next = skb->prev = NULL;
751 /* XXX: more streamlined implementation */
754 * __skb_dequeue_tail - remove from the tail of the queue
755 * @list: list to dequeue from
757 * Remove the tail of the list. This function does not take any locks
758 * so must be used with appropriate locks held only. The tail item is
759 * returned or %NULL if the list is empty.
761 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
762 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
764 struct sk_buff *skb = skb_peek_tail(list);
766 __skb_unlink(skb, list);
771 static inline int skb_is_nonlinear(const struct sk_buff *skb)
773 return skb->data_len;
776 static inline unsigned int skb_headlen(const struct sk_buff *skb)
778 return skb->len - skb->data_len;
781 static inline int skb_pagelen(const struct sk_buff *skb)
785 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
786 len += skb_shinfo(skb)->frags[i].size;
787 return len + skb_headlen(skb);
790 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
791 struct page *page, int off, int size)
793 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
796 frag->page_offset = off;
798 skb_shinfo(skb)->nr_frags = i + 1;
801 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
802 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
803 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
806 * Add data to an sk_buff
808 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
810 unsigned char *tmp = skb->tail;
811 SKB_LINEAR_ASSERT(skb);
818 * skb_put - add data to a buffer
819 * @skb: buffer to use
820 * @len: amount of data to add
822 * This function extends the used data area of the buffer. If this would
823 * exceed the total buffer size the kernel will panic. A pointer to the
824 * first byte of the extra data is returned.
826 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
828 unsigned char *tmp = skb->tail;
829 SKB_LINEAR_ASSERT(skb);
832 if (unlikely(skb->tail>skb->end))
833 skb_over_panic(skb, len, current_text_addr());
837 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
845 * skb_push - add data to the start of a buffer
846 * @skb: buffer to use
847 * @len: amount of data to add
849 * This function extends the used data area of the buffer at the buffer
850 * start. If this would exceed the total buffer headroom the kernel will
851 * panic. A pointer to the first byte of the extra data is returned.
853 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
857 if (unlikely(skb->data<skb->head))
858 skb_under_panic(skb, len, current_text_addr());
862 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
865 BUG_ON(skb->len < skb->data_len);
866 return skb->data += len;
870 * skb_pull - remove data from the start of a buffer
871 * @skb: buffer to use
872 * @len: amount of data to remove
874 * This function removes data from the start of a buffer, returning
875 * the memory to the headroom. A pointer to the next data in the buffer
876 * is returned. Once the data has been pulled future pushes will overwrite
879 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
881 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
884 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
886 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
888 if (len > skb_headlen(skb) &&
889 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
892 return skb->data += len;
895 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
897 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
900 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
902 if (likely(len <= skb_headlen(skb)))
904 if (unlikely(len > skb->len))
906 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
910 * skb_headroom - bytes at buffer head
911 * @skb: buffer to check
913 * Return the number of bytes of free space at the head of an &sk_buff.
915 static inline int skb_headroom(const struct sk_buff *skb)
917 return skb->data - skb->head;
921 * skb_tailroom - bytes at buffer end
922 * @skb: buffer to check
924 * Return the number of bytes of free space at the tail of an sk_buff
926 static inline int skb_tailroom(const struct sk_buff *skb)
928 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
932 * skb_reserve - adjust headroom
933 * @skb: buffer to alter
934 * @len: bytes to move
936 * Increase the headroom of an empty &sk_buff by reducing the tail
937 * room. This is only allowed for an empty buffer.
939 static inline void skb_reserve(struct sk_buff *skb, int len)
946 * CPUs often take a performance hit when accessing unaligned memory
947 * locations. The actual performance hit varies, it can be small if the
948 * hardware handles it or large if we have to take an exception and fix it
951 * Since an ethernet header is 14 bytes network drivers often end up with
952 * the IP header at an unaligned offset. The IP header can be aligned by
953 * shifting the start of the packet by 2 bytes. Drivers should do this
956 * skb_reserve(NET_IP_ALIGN);
958 * The downside to this alignment of the IP header is that the DMA is now
959 * unaligned. On some architectures the cost of an unaligned DMA is high
960 * and this cost outweighs the gains made by aligning the IP header.
962 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
966 #define NET_IP_ALIGN 2
970 * The networking layer reserves some headroom in skb data (via
971 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
972 * the header has to grow. In the default case, if the header has to grow
973 * 16 bytes or less we avoid the reallocation.
975 * Unfortunately this headroom changes the DMA alignment of the resulting
976 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
977 * on some architectures. An architecture can override this value,
978 * perhaps setting it to a cacheline in size (since that will maintain
979 * cacheline alignment of the DMA). It must be a power of 2.
981 * Various parts of the networking layer expect at least 16 bytes of
982 * headroom, you should not reduce this.
985 #define NET_SKB_PAD 16
988 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
990 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
992 if (unlikely(skb->data_len)) {
997 skb->tail = skb->data + len;
1001 * skb_trim - remove end from a buffer
1002 * @skb: buffer to alter
1005 * Cut the length of a buffer down by removing data from the tail. If
1006 * the buffer is already under the length specified it is not modified.
1007 * The skb must be linear.
1009 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
1012 __skb_trim(skb, len);
1016 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1019 return ___pskb_trim(skb, len);
1020 __skb_trim(skb, len);
1024 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1026 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1030 * skb_orphan - orphan a buffer
1031 * @skb: buffer to orphan
1033 * If a buffer currently has an owner then we call the owner's
1034 * destructor function and make the @skb unowned. The buffer continues
1035 * to exist but is no longer charged to its former owner.
1037 static inline void skb_orphan(struct sk_buff *skb)
1039 if (skb->destructor)
1040 skb->destructor(skb);
1041 skb->destructor = NULL;
1046 * __skb_queue_purge - empty a list
1047 * @list: list to empty
1049 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1050 * the list and one reference dropped. This function does not take the
1051 * list lock and the caller must hold the relevant locks to use it.
1053 extern void skb_queue_purge(struct sk_buff_head *list);
1054 static inline void __skb_queue_purge(struct sk_buff_head *list)
1056 struct sk_buff *skb;
1057 while ((skb = __skb_dequeue(list)) != NULL)
1061 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
1063 * __dev_alloc_skb - allocate an skbuff for sending
1064 * @length: length to allocate
1065 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1067 * Allocate a new &sk_buff and assign it a usage count of one. The
1068 * buffer has unspecified headroom built in. Users should allocate
1069 * the headroom they think they need without accounting for the
1070 * built in space. The built in space is used for optimisations.
1072 * %NULL is returned in there is no free memory.
1074 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1077 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1079 skb_reserve(skb, NET_SKB_PAD);
1083 extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
1087 * dev_alloc_skb - allocate an skbuff for sending
1088 * @length: length to allocate
1090 * Allocate a new &sk_buff and assign it a usage count of one. The
1091 * buffer has unspecified headroom built in. Users should allocate
1092 * the headroom they think they need without accounting for the
1093 * built in space. The built in space is used for optimisations.
1095 * %NULL is returned in there is no free memory. Although this function
1096 * allocates memory it can be called from an interrupt.
1098 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1100 return __dev_alloc_skb(length, GFP_ATOMIC);
1104 * skb_cow - copy header of skb when it is required
1105 * @skb: buffer to cow
1106 * @headroom: needed headroom
1108 * If the skb passed lacks sufficient headroom or its data part
1109 * is shared, data is reallocated. If reallocation fails, an error
1110 * is returned and original skb is not changed.
1112 * The result is skb with writable area skb->head...skb->tail
1113 * and at least @headroom of space at head.
1115 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1117 int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) -
1123 if (delta || skb_cloned(skb))
1124 return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) &
1125 ~(NET_SKB_PAD-1), 0, GFP_ATOMIC);
1130 * skb_padto - pad an skbuff up to a minimal size
1131 * @skb: buffer to pad
1132 * @len: minimal length
1134 * Pads up a buffer to ensure the trailing bytes exist and are
1135 * blanked. If the buffer already contains sufficient data it
1136 * is untouched. Otherwise it is extended. Returns zero on
1137 * success. The skb is freed on error.
1140 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1142 unsigned int size = skb->len;
1143 if (likely(size >= len))
1145 return skb_pad(skb, len-size);
1148 static inline int skb_add_data(struct sk_buff *skb,
1149 char __user *from, int copy)
1151 const int off = skb->len;
1153 if (skb->ip_summed == CHECKSUM_NONE) {
1155 unsigned int csum = csum_and_copy_from_user(from,
1159 skb->csum = csum_block_add(skb->csum, csum, off);
1162 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1165 __skb_trim(skb, off);
1169 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1170 struct page *page, int off)
1173 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1175 return page == frag->page &&
1176 off == frag->page_offset + frag->size;
1181 static inline int __skb_linearize(struct sk_buff *skb)
1183 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1187 * skb_linearize - convert paged skb to linear one
1188 * @skb: buffer to linarize
1190 * If there is no free memory -ENOMEM is returned, otherwise zero
1191 * is returned and the old skb data released.
1193 static inline int skb_linearize(struct sk_buff *skb)
1195 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1199 * skb_linearize_cow - make sure skb is linear and writable
1200 * @skb: buffer to process
1202 * If there is no free memory -ENOMEM is returned, otherwise zero
1203 * is returned and the old skb data released.
1205 static inline int skb_linearize_cow(struct sk_buff *skb)
1207 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1208 __skb_linearize(skb) : 0;
1212 * skb_postpull_rcsum - update checksum for received skb after pull
1213 * @skb: buffer to update
1214 * @start: start of data before pull
1215 * @len: length of data pulled
1217 * After doing a pull on a received packet, you need to call this to
1218 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1219 * so that it can be recomputed from scratch.
1222 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1223 const void *start, unsigned int len)
1225 if (skb->ip_summed == CHECKSUM_HW)
1226 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1229 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1232 * pskb_trim_rcsum - trim received skb and update checksum
1233 * @skb: buffer to trim
1236 * This is exactly the same as pskb_trim except that it ensures the
1237 * checksum of received packets are still valid after the operation.
1240 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1242 if (likely(len >= skb->len))
1244 if (skb->ip_summed == CHECKSUM_HW)
1245 skb->ip_summed = CHECKSUM_NONE;
1246 return __pskb_trim(skb, len);
1249 static inline void *kmap_skb_frag(const skb_frag_t *frag)
1251 #ifdef CONFIG_HIGHMEM
1256 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1259 static inline void kunmap_skb_frag(void *vaddr)
1261 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1262 #ifdef CONFIG_HIGHMEM
1267 #define skb_queue_walk(queue, skb) \
1268 for (skb = (queue)->next; \
1269 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1272 #define skb_queue_reverse_walk(queue, skb) \
1273 for (skb = (queue)->prev; \
1274 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1278 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1279 int noblock, int *err);
1280 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1281 struct poll_table_struct *wait);
1282 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1283 int offset, struct iovec *to,
1285 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1288 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1289 extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1290 unsigned int flags);
1291 extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1292 int len, unsigned int csum);
1293 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1295 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1296 void *from, int len);
1297 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1298 int offset, u8 *to, int len,
1300 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1301 extern void skb_split(struct sk_buff *skb,
1302 struct sk_buff *skb1, const u32 len);
1304 extern void skb_release_data(struct sk_buff *skb);
1305 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1307 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1308 int len, void *buffer)
1310 int hlen = skb_headlen(skb);
1312 if (hlen - offset >= len)
1313 return skb->data + offset;
1315 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1321 extern void skb_init(void);
1322 extern void skb_add_mtu(int mtu);
1325 * skb_get_timestamp - get timestamp from a skb
1326 * @skb: skb to get stamp from
1327 * @stamp: pointer to struct timeval to store stamp in
1329 * Timestamps are stored in the skb as offsets to a base timestamp.
1330 * This function converts the offset back to a struct timeval and stores
1333 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1335 stamp->tv_sec = skb->tstamp.off_sec;
1336 stamp->tv_usec = skb->tstamp.off_usec;
1340 * skb_set_timestamp - set timestamp of a skb
1341 * @skb: skb to set stamp of
1342 * @stamp: pointer to struct timeval to get stamp from
1344 * Timestamps are stored in the skb as offsets to a base timestamp.
1345 * This function converts a struct timeval to an offset and stores
1348 static inline void skb_set_timestamp(struct sk_buff *skb, const struct timeval *stamp)
1350 skb->tstamp.off_sec = stamp->tv_sec;
1351 skb->tstamp.off_usec = stamp->tv_usec;
1354 extern void __net_timestamp(struct sk_buff *skb);
1356 extern unsigned int __skb_checksum_complete(struct sk_buff *skb);
1359 * skb_checksum_complete - Calculate checksum of an entire packet
1360 * @skb: packet to process
1362 * This function calculates the checksum over the entire packet plus
1363 * the value of skb->csum. The latter can be used to supply the
1364 * checksum of a pseudo header as used by TCP/UDP. It returns the
1367 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1368 * this function can be used to verify that checksum on received
1369 * packets. In that case the function should return zero if the
1370 * checksum is correct. In particular, this function will return zero
1371 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1372 * hardware has already verified the correctness of the checksum.
1374 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1376 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1377 __skb_checksum_complete(skb);
1380 #ifdef CONFIG_NETFILTER
1381 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1383 if (nfct && atomic_dec_and_test(&nfct->use))
1384 nfct->destroy(nfct);
1386 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1389 atomic_inc(&nfct->use);
1391 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1392 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1395 atomic_inc(&skb->users);
1397 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1403 #ifdef CONFIG_BRIDGE_NETFILTER
1404 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1406 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1409 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1412 atomic_inc(&nf_bridge->use);
1414 #endif /* CONFIG_BRIDGE_NETFILTER */
1415 static inline void nf_reset(struct sk_buff *skb)
1417 nf_conntrack_put(skb->nfct);
1419 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1420 nf_conntrack_put_reasm(skb->nfct_reasm);
1421 skb->nfct_reasm = NULL;
1423 #ifdef CONFIG_BRIDGE_NETFILTER
1424 nf_bridge_put(skb->nf_bridge);
1425 skb->nf_bridge = NULL;
1429 #else /* CONFIG_NETFILTER */
1430 static inline void nf_reset(struct sk_buff *skb) {}
1431 #endif /* CONFIG_NETFILTER */
1433 #ifdef CONFIG_NETWORK_SECMARK
1434 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1436 to->secmark = from->secmark;
1439 static inline void skb_init_secmark(struct sk_buff *skb)
1444 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1447 static inline void skb_init_secmark(struct sk_buff *skb)
1451 #endif /* __KERNEL__ */
1452 #endif /* _LINUX_SKBUFF_H */