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>
25 #include <linux/net.h>
26 #include <linux/textsearch.h>
27 #include <net/checksum.h>
28 #include <linux/rcupdate.h>
29 #include <linux/dmaengine.h>
30 #include <linux/hrtimer.h>
32 #define HAVE_ALLOC_SKB /* For the drivers to know */
33 #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
35 #define CHECKSUM_NONE 0
36 #define CHECKSUM_PARTIAL 1
37 #define CHECKSUM_UNNECESSARY 2
38 #define CHECKSUM_COMPLETE 3
40 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
41 ~(SMP_CACHE_BYTES - 1))
42 #define SKB_WITH_OVERHEAD(X) \
43 (((X) - sizeof(struct skb_shared_info)) & \
44 ~(SMP_CACHE_BYTES - 1))
45 #define SKB_MAX_ORDER(X, ORDER) \
46 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
47 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
48 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
50 /* A. Checksumming of received packets by device.
52 * NONE: device failed to checksum this packet.
53 * skb->csum is undefined.
55 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
56 * skb->csum is undefined.
57 * It is bad option, but, unfortunately, many of vendors do this.
58 * Apparently with secret goal to sell you new device, when you
59 * will add new protocol to your host. F.e. IPv6. 8)
61 * COMPLETE: the most generic way. Device supplied checksum of _all_
62 * the packet as seen by netif_rx in skb->csum.
63 * NOTE: Even if device supports only some protocols, but
64 * is able to produce some skb->csum, it MUST use COMPLETE,
67 * B. Checksumming on output.
69 * NONE: skb is checksummed by protocol or csum is not required.
71 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
72 * from skb->h.raw to the end and to record the checksum
73 * at skb->h.raw+skb->csum.
75 * Device must show its capabilities in dev->features, set
76 * at device setup time.
77 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
79 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
80 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
81 * TCP/UDP over IPv4. Sigh. Vendors like this
82 * way by an unknown reason. Though, see comment above
83 * about CHECKSUM_UNNECESSARY. 8)
85 * Any questions? No questions, good. --ANK
90 #ifdef CONFIG_NETFILTER
93 void (*destroy)(struct nf_conntrack *);
96 #ifdef CONFIG_BRIDGE_NETFILTER
97 struct nf_bridge_info {
99 struct net_device *physindev;
100 struct net_device *physoutdev;
101 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
102 struct net_device *netoutdev;
105 unsigned long data[32 / sizeof(unsigned long)];
111 struct sk_buff_head {
112 /* These two members must be first. */
113 struct sk_buff *next;
114 struct sk_buff *prev;
122 /* To allow 64K frame to be packed as single skb without frag_list */
123 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
125 typedef struct skb_frag_struct skb_frag_t;
127 struct skb_frag_struct {
133 /* This data is invariant across clones and lives at
134 * the end of the header data, ie. at skb->end.
136 struct skb_shared_info {
138 unsigned short nr_frags;
139 unsigned short gso_size;
140 /* Warning: this field is not always filled in (UFO)! */
141 unsigned short gso_segs;
142 unsigned short gso_type;
144 struct sk_buff *frag_list;
145 skb_frag_t frags[MAX_SKB_FRAGS];
148 /* We divide dataref into two halves. The higher 16 bits hold references
149 * to the payload part of skb->data. The lower 16 bits hold references to
150 * the entire skb->data. It is up to the users of the skb to agree on
151 * where the payload starts.
153 * All users must obey the rule that the skb->data reference count must be
154 * greater than or equal to the payload reference count.
156 * Holding a reference to the payload part means that the user does not
157 * care about modifications to the header part of skb->data.
159 #define SKB_DATAREF_SHIFT 16
160 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
164 SKB_FCLONE_UNAVAILABLE,
170 SKB_GSO_TCPV4 = 1 << 0,
171 SKB_GSO_UDP = 1 << 1,
173 /* This indicates the skb is from an untrusted source. */
174 SKB_GSO_DODGY = 1 << 2,
176 /* This indicates the tcp segment has CWR set. */
177 SKB_GSO_TCP_ECN = 1 << 3,
179 SKB_GSO_TCPV6 = 1 << 4,
183 * struct sk_buff - socket buffer
184 * @next: Next buffer in list
185 * @prev: Previous buffer in list
186 * @sk: Socket we are owned by
187 * @tstamp: Time we arrived
188 * @dev: Device we arrived on/are leaving by
189 * @iif: ifindex of device we arrived on
190 * @h: Transport layer header
191 * @nh: Network layer header
192 * @mac: Link layer header
193 * @dst: destination entry
194 * @sp: the security path, used for xfrm
195 * @cb: Control buffer. Free for use by every layer. Put private vars here
196 * @len: Length of actual data
197 * @data_len: Data length
198 * @mac_len: Length of link layer header
200 * @local_df: allow local fragmentation
201 * @cloned: Head may be cloned (check refcnt to be sure)
202 * @nohdr: Payload reference only, must not modify header
203 * @pkt_type: Packet class
204 * @fclone: skbuff clone status
205 * @ip_summed: Driver fed us an IP checksum
206 * @priority: Packet queueing priority
207 * @users: User count - see {datagram,tcp}.c
208 * @protocol: Packet protocol from driver
209 * @truesize: Buffer size
210 * @head: Head of buffer
211 * @data: Data head pointer
212 * @tail: Tail pointer
214 * @destructor: Destruct function
215 * @mark: Generic packet mark
216 * @nfct: Associated connection, if any
217 * @ipvs_property: skbuff is owned by ipvs
218 * @nfctinfo: Relationship of this skb to the connection
219 * @nfct_reasm: netfilter conntrack re-assembly pointer
220 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
221 * @tc_index: Traffic control index
222 * @tc_verd: traffic control verdict
223 * @dma_cookie: a cookie to one of several possible DMA operations
224 * done by skb DMA functions
225 * @secmark: security marking
229 /* These two members must be first. */
230 struct sk_buff *next;
231 struct sk_buff *prev;
235 struct net_device *dev;
237 /* 4 byte hole on 64 bit*/
242 struct icmphdr *icmph;
243 struct igmphdr *igmph;
245 struct ipv6hdr *ipv6h;
250 struct ipv6hdr *ipv6h;
259 struct dst_entry *dst;
263 * This is the control buffer. It is free to use for every
264 * layer. Please put your private variables there. If you
265 * want to keep them across layers you have to do a skb_clone()
266 * first. This is owned by whoever has the skb queued ATM.
288 void (*destructor)(struct sk_buff *skb);
289 #ifdef CONFIG_NETFILTER
290 struct nf_conntrack *nfct;
291 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
292 struct sk_buff *nfct_reasm;
294 #ifdef CONFIG_BRIDGE_NETFILTER
295 struct nf_bridge_info *nf_bridge;
297 #endif /* CONFIG_NETFILTER */
298 #ifdef CONFIG_NET_SCHED
299 __u16 tc_index; /* traffic control index */
300 #ifdef CONFIG_NET_CLS_ACT
301 __u16 tc_verd; /* traffic control verdict */
304 #ifdef CONFIG_NET_DMA
305 dma_cookie_t dma_cookie;
307 #ifdef CONFIG_NETWORK_SECMARK
313 /* These elements must be at the end, see alloc_skb() for details. */
314 unsigned int truesize;
324 * Handling routines are only of interest to the kernel
326 #include <linux/slab.h>
328 #include <asm/system.h>
330 extern void kfree_skb(struct sk_buff *skb);
331 extern void __kfree_skb(struct sk_buff *skb);
332 extern struct sk_buff *__alloc_skb(unsigned int size,
333 gfp_t priority, int fclone, int node);
334 static inline struct sk_buff *alloc_skb(unsigned int size,
337 return __alloc_skb(size, priority, 0, -1);
340 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
343 return __alloc_skb(size, priority, 1, -1);
346 extern void kfree_skbmem(struct sk_buff *skb);
347 extern struct sk_buff *skb_clone(struct sk_buff *skb,
349 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
351 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
353 extern int pskb_expand_head(struct sk_buff *skb,
354 int nhead, int ntail,
356 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
357 unsigned int headroom);
358 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
359 int newheadroom, int newtailroom,
361 extern int skb_pad(struct sk_buff *skb, int pad);
362 #define dev_kfree_skb(a) kfree_skb(a)
363 extern void skb_over_panic(struct sk_buff *skb, int len,
365 extern void skb_under_panic(struct sk_buff *skb, int len,
367 extern void skb_truesize_bug(struct sk_buff *skb);
369 static inline void skb_truesize_check(struct sk_buff *skb)
371 if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len))
372 skb_truesize_bug(skb);
375 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
376 int getfrag(void *from, char *to, int offset,
377 int len,int odd, struct sk_buff *skb),
378 void *from, int length);
385 __u32 stepped_offset;
386 struct sk_buff *root_skb;
387 struct sk_buff *cur_skb;
391 extern void skb_prepare_seq_read(struct sk_buff *skb,
392 unsigned int from, unsigned int to,
393 struct skb_seq_state *st);
394 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
395 struct skb_seq_state *st);
396 extern void skb_abort_seq_read(struct skb_seq_state *st);
398 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
399 unsigned int to, struct ts_config *config,
400 struct ts_state *state);
403 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
406 * skb_queue_empty - check if a queue is empty
409 * Returns true if the queue is empty, false otherwise.
411 static inline int skb_queue_empty(const struct sk_buff_head *list)
413 return list->next == (struct sk_buff *)list;
417 * skb_get - reference buffer
418 * @skb: buffer to reference
420 * Makes another reference to a socket buffer and returns a pointer
423 static inline struct sk_buff *skb_get(struct sk_buff *skb)
425 atomic_inc(&skb->users);
430 * If users == 1, we are the only owner and are can avoid redundant
435 * skb_cloned - is the buffer a clone
436 * @skb: buffer to check
438 * Returns true if the buffer was generated with skb_clone() and is
439 * one of multiple shared copies of the buffer. Cloned buffers are
440 * shared data so must not be written to under normal circumstances.
442 static inline int skb_cloned(const struct sk_buff *skb)
444 return skb->cloned &&
445 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
449 * skb_header_cloned - is the header a clone
450 * @skb: buffer to check
452 * Returns true if modifying the header part of the buffer requires
453 * the data to be copied.
455 static inline int skb_header_cloned(const struct sk_buff *skb)
462 dataref = atomic_read(&skb_shinfo(skb)->dataref);
463 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
468 * skb_header_release - release reference to header
469 * @skb: buffer to operate on
471 * Drop a reference to the header part of the buffer. This is done
472 * by acquiring a payload reference. You must not read from the header
473 * part of skb->data after this.
475 static inline void skb_header_release(struct sk_buff *skb)
479 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
483 * skb_shared - is the buffer shared
484 * @skb: buffer to check
486 * Returns true if more than one person has a reference to this
489 static inline int skb_shared(const struct sk_buff *skb)
491 return atomic_read(&skb->users) != 1;
495 * skb_share_check - check if buffer is shared and if so clone it
496 * @skb: buffer to check
497 * @pri: priority for memory allocation
499 * If the buffer is shared the buffer is cloned and the old copy
500 * drops a reference. A new clone with a single reference is returned.
501 * If the buffer is not shared the original buffer is returned. When
502 * being called from interrupt status or with spinlocks held pri must
505 * NULL is returned on a memory allocation failure.
507 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
510 might_sleep_if(pri & __GFP_WAIT);
511 if (skb_shared(skb)) {
512 struct sk_buff *nskb = skb_clone(skb, pri);
520 * Copy shared buffers into a new sk_buff. We effectively do COW on
521 * packets to handle cases where we have a local reader and forward
522 * and a couple of other messy ones. The normal one is tcpdumping
523 * a packet thats being forwarded.
527 * skb_unshare - make a copy of a shared buffer
528 * @skb: buffer to check
529 * @pri: priority for memory allocation
531 * If the socket buffer is a clone then this function creates a new
532 * copy of the data, drops a reference count on the old copy and returns
533 * the new copy with the reference count at 1. If the buffer is not a clone
534 * the original buffer is returned. When called with a spinlock held or
535 * from interrupt state @pri must be %GFP_ATOMIC
537 * %NULL is returned on a memory allocation failure.
539 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
542 might_sleep_if(pri & __GFP_WAIT);
543 if (skb_cloned(skb)) {
544 struct sk_buff *nskb = skb_copy(skb, pri);
545 kfree_skb(skb); /* Free our shared copy */
553 * @list_: list to peek at
555 * Peek an &sk_buff. Unlike most other operations you _MUST_
556 * be careful with this one. A peek leaves the buffer on the
557 * list and someone else may run off with it. You must hold
558 * the appropriate locks or have a private queue to do this.
560 * Returns %NULL for an empty list or a pointer to the head element.
561 * The reference count is not incremented and the reference is therefore
562 * volatile. Use with caution.
564 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
566 struct sk_buff *list = ((struct sk_buff *)list_)->next;
567 if (list == (struct sk_buff *)list_)
574 * @list_: list to peek at
576 * Peek an &sk_buff. Unlike most other operations you _MUST_
577 * be careful with this one. A peek leaves the buffer on the
578 * list and someone else may run off with it. You must hold
579 * the appropriate locks or have a private queue to do this.
581 * Returns %NULL for an empty list or a pointer to the tail element.
582 * The reference count is not incremented and the reference is therefore
583 * volatile. Use with caution.
585 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
587 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
588 if (list == (struct sk_buff *)list_)
594 * skb_queue_len - get queue length
595 * @list_: list to measure
597 * Return the length of an &sk_buff queue.
599 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
605 * This function creates a split out lock class for each invocation;
606 * this is needed for now since a whole lot of users of the skb-queue
607 * infrastructure in drivers have different locking usage (in hardirq)
608 * than the networking core (in softirq only). In the long run either the
609 * network layer or drivers should need annotation to consolidate the
610 * main types of usage into 3 classes.
612 static inline void skb_queue_head_init(struct sk_buff_head *list)
614 spin_lock_init(&list->lock);
615 list->prev = list->next = (struct sk_buff *)list;
619 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
620 struct lock_class_key *class)
622 skb_queue_head_init(list);
623 lockdep_set_class(&list->lock, class);
627 * Insert an sk_buff at the start of a list.
629 * The "__skb_xxxx()" functions are the non-atomic ones that
630 * can only be called with interrupts disabled.
634 * __skb_queue_after - queue a buffer at the list head
636 * @prev: place after this buffer
637 * @newsk: buffer to queue
639 * Queue a buffer int the middle of a list. This function takes no locks
640 * and you must therefore hold required locks before calling it.
642 * A buffer cannot be placed on two lists at the same time.
644 static inline void __skb_queue_after(struct sk_buff_head *list,
645 struct sk_buff *prev,
646 struct sk_buff *newsk)
648 struct sk_buff *next;
654 next->prev = prev->next = newsk;
658 * __skb_queue_head - queue a buffer at the list head
660 * @newsk: buffer to queue
662 * Queue a buffer at the start 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_head(struct sk_buff_head *list, struct sk_buff *newsk);
668 static inline void __skb_queue_head(struct sk_buff_head *list,
669 struct sk_buff *newsk)
671 __skb_queue_after(list, (struct sk_buff *)list, newsk);
675 * __skb_queue_tail - queue a buffer at the list tail
677 * @newsk: buffer to queue
679 * Queue a buffer at the end of a list. This function takes no locks
680 * and you must therefore hold required locks before calling it.
682 * A buffer cannot be placed on two lists at the same time.
684 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
685 static inline void __skb_queue_tail(struct sk_buff_head *list,
686 struct sk_buff *newsk)
688 struct sk_buff *prev, *next;
691 next = (struct sk_buff *)list;
695 next->prev = prev->next = newsk;
700 * __skb_dequeue - remove from the head of the queue
701 * @list: list to dequeue from
703 * Remove the head of the list. This function does not take any locks
704 * so must be used with appropriate locks held only. The head item is
705 * returned or %NULL if the list is empty.
707 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
708 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
710 struct sk_buff *next, *prev, *result;
712 prev = (struct sk_buff *) list;
721 result->next = result->prev = NULL;
728 * Insert a packet on a list.
730 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
731 static inline void __skb_insert(struct sk_buff *newsk,
732 struct sk_buff *prev, struct sk_buff *next,
733 struct sk_buff_head *list)
737 next->prev = prev->next = newsk;
742 * Place a packet after a given packet in a list.
744 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
745 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
747 __skb_insert(newsk, old, old->next, list);
751 * remove sk_buff from list. _Must_ be called atomically, and with
754 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
755 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
757 struct sk_buff *next, *prev;
762 skb->next = skb->prev = NULL;
768 /* XXX: more streamlined implementation */
771 * __skb_dequeue_tail - remove from the tail of the queue
772 * @list: list to dequeue from
774 * Remove the tail of the list. This function does not take any locks
775 * so must be used with appropriate locks held only. The tail item is
776 * returned or %NULL if the list is empty.
778 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
779 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
781 struct sk_buff *skb = skb_peek_tail(list);
783 __skb_unlink(skb, list);
788 static inline int skb_is_nonlinear(const struct sk_buff *skb)
790 return skb->data_len;
793 static inline unsigned int skb_headlen(const struct sk_buff *skb)
795 return skb->len - skb->data_len;
798 static inline int skb_pagelen(const struct sk_buff *skb)
802 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
803 len += skb_shinfo(skb)->frags[i].size;
804 return len + skb_headlen(skb);
807 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
808 struct page *page, int off, int size)
810 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
813 frag->page_offset = off;
815 skb_shinfo(skb)->nr_frags = i + 1;
818 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
819 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
820 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
823 * Add data to an sk_buff
825 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
827 unsigned char *tmp = skb->tail;
828 SKB_LINEAR_ASSERT(skb);
835 * skb_put - add data to a buffer
836 * @skb: buffer to use
837 * @len: amount of data to add
839 * This function extends the used data area of the buffer. If this would
840 * exceed the total buffer size the kernel will panic. A pointer to the
841 * first byte of the extra data is returned.
843 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
845 unsigned char *tmp = skb->tail;
846 SKB_LINEAR_ASSERT(skb);
849 if (unlikely(skb->tail>skb->end))
850 skb_over_panic(skb, len, current_text_addr());
854 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
862 * skb_push - add data to the start of a buffer
863 * @skb: buffer to use
864 * @len: amount of data to add
866 * This function extends the used data area of the buffer at the buffer
867 * start. If this would exceed the total buffer headroom the kernel will
868 * panic. A pointer to the first byte of the extra data is returned.
870 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
874 if (unlikely(skb->data<skb->head))
875 skb_under_panic(skb, len, current_text_addr());
879 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
882 BUG_ON(skb->len < skb->data_len);
883 return skb->data += len;
887 * skb_pull - remove data from the start of a buffer
888 * @skb: buffer to use
889 * @len: amount of data to remove
891 * This function removes data from the start of a buffer, returning
892 * the memory to the headroom. A pointer to the next data in the buffer
893 * is returned. Once the data has been pulled future pushes will overwrite
896 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
898 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
901 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
903 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
905 if (len > skb_headlen(skb) &&
906 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
909 return skb->data += len;
912 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
914 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
917 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
919 if (likely(len <= skb_headlen(skb)))
921 if (unlikely(len > skb->len))
923 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
927 * skb_headroom - bytes at buffer head
928 * @skb: buffer to check
930 * Return the number of bytes of free space at the head of an &sk_buff.
932 static inline int skb_headroom(const struct sk_buff *skb)
934 return skb->data - skb->head;
938 * skb_tailroom - bytes at buffer end
939 * @skb: buffer to check
941 * Return the number of bytes of free space at the tail of an sk_buff
943 static inline int skb_tailroom(const struct sk_buff *skb)
945 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
949 * skb_reserve - adjust headroom
950 * @skb: buffer to alter
951 * @len: bytes to move
953 * Increase the headroom of an empty &sk_buff by reducing the tail
954 * room. This is only allowed for an empty buffer.
956 static inline void skb_reserve(struct sk_buff *skb, int len)
962 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
967 static inline void skb_reset_network_header(struct sk_buff *skb)
969 skb->nh.raw = skb->data;
972 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
974 skb->nh.raw = skb->data + offset;
977 static inline int skb_network_offset(const struct sk_buff *skb)
979 return skb->nh.raw - skb->data;
982 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
987 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
989 return skb->mac.raw != NULL;
992 static inline void skb_reset_mac_header(struct sk_buff *skb)
994 skb->mac.raw = skb->data;
997 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
999 skb->mac.raw = skb->data + offset;
1003 * CPUs often take a performance hit when accessing unaligned memory
1004 * locations. The actual performance hit varies, it can be small if the
1005 * hardware handles it or large if we have to take an exception and fix it
1008 * Since an ethernet header is 14 bytes network drivers often end up with
1009 * the IP header at an unaligned offset. The IP header can be aligned by
1010 * shifting the start of the packet by 2 bytes. Drivers should do this
1013 * skb_reserve(NET_IP_ALIGN);
1015 * The downside to this alignment of the IP header is that the DMA is now
1016 * unaligned. On some architectures the cost of an unaligned DMA is high
1017 * and this cost outweighs the gains made by aligning the IP header.
1019 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1022 #ifndef NET_IP_ALIGN
1023 #define NET_IP_ALIGN 2
1027 * The networking layer reserves some headroom in skb data (via
1028 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1029 * the header has to grow. In the default case, if the header has to grow
1030 * 16 bytes or less we avoid the reallocation.
1032 * Unfortunately this headroom changes the DMA alignment of the resulting
1033 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1034 * on some architectures. An architecture can override this value,
1035 * perhaps setting it to a cacheline in size (since that will maintain
1036 * cacheline alignment of the DMA). It must be a power of 2.
1038 * Various parts of the networking layer expect at least 16 bytes of
1039 * headroom, you should not reduce this.
1042 #define NET_SKB_PAD 16
1045 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1047 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1049 if (unlikely(skb->data_len)) {
1054 skb->tail = skb->data + len;
1058 * skb_trim - remove end from a buffer
1059 * @skb: buffer to alter
1062 * Cut the length of a buffer down by removing data from the tail. If
1063 * the buffer is already under the length specified it is not modified.
1064 * The skb must be linear.
1066 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
1069 __skb_trim(skb, len);
1073 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1076 return ___pskb_trim(skb, len);
1077 __skb_trim(skb, len);
1081 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1083 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1087 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1088 * @skb: buffer to alter
1091 * This is identical to pskb_trim except that the caller knows that
1092 * the skb is not cloned so we should never get an error due to out-
1095 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1097 int err = pskb_trim(skb, len);
1102 * skb_orphan - orphan a buffer
1103 * @skb: buffer to orphan
1105 * If a buffer currently has an owner then we call the owner's
1106 * destructor function and make the @skb unowned. The buffer continues
1107 * to exist but is no longer charged to its former owner.
1109 static inline void skb_orphan(struct sk_buff *skb)
1111 if (skb->destructor)
1112 skb->destructor(skb);
1113 skb->destructor = NULL;
1118 * __skb_queue_purge - empty a list
1119 * @list: list to empty
1121 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1122 * the list and one reference dropped. This function does not take the
1123 * list lock and the caller must hold the relevant locks to use it.
1125 extern void skb_queue_purge(struct sk_buff_head *list);
1126 static inline void __skb_queue_purge(struct sk_buff_head *list)
1128 struct sk_buff *skb;
1129 while ((skb = __skb_dequeue(list)) != NULL)
1134 * __dev_alloc_skb - allocate an skbuff for receiving
1135 * @length: length to allocate
1136 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1138 * Allocate a new &sk_buff and assign it a usage count of one. The
1139 * buffer has unspecified headroom built in. Users should allocate
1140 * the headroom they think they need without accounting for the
1141 * built in space. The built in space is used for optimisations.
1143 * %NULL is returned if there is no free memory.
1145 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1148 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1150 skb_reserve(skb, NET_SKB_PAD);
1155 * dev_alloc_skb - allocate an skbuff for receiving
1156 * @length: length to allocate
1158 * Allocate a new &sk_buff and assign it a usage count of one. The
1159 * buffer has unspecified headroom built in. Users should allocate
1160 * the headroom they think they need without accounting for the
1161 * built in space. The built in space is used for optimisations.
1163 * %NULL is returned if there is no free memory. Although this function
1164 * allocates memory it can be called from an interrupt.
1166 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1168 return __dev_alloc_skb(length, GFP_ATOMIC);
1171 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1172 unsigned int length, gfp_t gfp_mask);
1175 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1176 * @dev: network device to receive on
1177 * @length: length to allocate
1179 * Allocate a new &sk_buff and assign it a usage count of one. The
1180 * buffer has unspecified headroom built in. Users should allocate
1181 * the headroom they think they need without accounting for the
1182 * built in space. The built in space is used for optimisations.
1184 * %NULL is returned if there is no free memory. Although this function
1185 * allocates memory it can be called from an interrupt.
1187 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1188 unsigned int length)
1190 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1194 * skb_cow - copy header of skb when it is required
1195 * @skb: buffer to cow
1196 * @headroom: needed headroom
1198 * If the skb passed lacks sufficient headroom or its data part
1199 * is shared, data is reallocated. If reallocation fails, an error
1200 * is returned and original skb is not changed.
1202 * The result is skb with writable area skb->head...skb->tail
1203 * and at least @headroom of space at head.
1205 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1207 int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) -
1213 if (delta || skb_cloned(skb))
1214 return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) &
1215 ~(NET_SKB_PAD-1), 0, GFP_ATOMIC);
1220 * skb_padto - pad an skbuff up to a minimal size
1221 * @skb: buffer to pad
1222 * @len: minimal length
1224 * Pads up a buffer to ensure the trailing bytes exist and are
1225 * blanked. If the buffer already contains sufficient data it
1226 * is untouched. Otherwise it is extended. Returns zero on
1227 * success. The skb is freed on error.
1230 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1232 unsigned int size = skb->len;
1233 if (likely(size >= len))
1235 return skb_pad(skb, len-size);
1238 static inline int skb_add_data(struct sk_buff *skb,
1239 char __user *from, int copy)
1241 const int off = skb->len;
1243 if (skb->ip_summed == CHECKSUM_NONE) {
1245 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1248 skb->csum = csum_block_add(skb->csum, csum, off);
1251 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1254 __skb_trim(skb, off);
1258 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1259 struct page *page, int off)
1262 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1264 return page == frag->page &&
1265 off == frag->page_offset + frag->size;
1270 static inline int __skb_linearize(struct sk_buff *skb)
1272 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1276 * skb_linearize - convert paged skb to linear one
1277 * @skb: buffer to linarize
1279 * If there is no free memory -ENOMEM is returned, otherwise zero
1280 * is returned and the old skb data released.
1282 static inline int skb_linearize(struct sk_buff *skb)
1284 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1288 * skb_linearize_cow - make sure skb is linear and writable
1289 * @skb: buffer to process
1291 * If there is no free memory -ENOMEM is returned, otherwise zero
1292 * is returned and the old skb data released.
1294 static inline int skb_linearize_cow(struct sk_buff *skb)
1296 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1297 __skb_linearize(skb) : 0;
1301 * skb_postpull_rcsum - update checksum for received skb after pull
1302 * @skb: buffer to update
1303 * @start: start of data before pull
1304 * @len: length of data pulled
1306 * After doing a pull on a received packet, you need to call this to
1307 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1308 * CHECKSUM_NONE so that it can be recomputed from scratch.
1311 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1312 const void *start, unsigned int len)
1314 if (skb->ip_summed == CHECKSUM_COMPLETE)
1315 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1318 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1321 * pskb_trim_rcsum - trim received skb and update checksum
1322 * @skb: buffer to trim
1325 * This is exactly the same as pskb_trim except that it ensures the
1326 * checksum of received packets are still valid after the operation.
1329 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1331 if (likely(len >= skb->len))
1333 if (skb->ip_summed == CHECKSUM_COMPLETE)
1334 skb->ip_summed = CHECKSUM_NONE;
1335 return __pskb_trim(skb, len);
1338 #define skb_queue_walk(queue, skb) \
1339 for (skb = (queue)->next; \
1340 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1343 #define skb_queue_reverse_walk(queue, skb) \
1344 for (skb = (queue)->prev; \
1345 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1349 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1350 int noblock, int *err);
1351 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1352 struct poll_table_struct *wait);
1353 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1354 int offset, struct iovec *to,
1356 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1359 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1360 extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1361 unsigned int flags);
1362 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1363 int len, __wsum csum);
1364 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1366 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1367 void *from, int len);
1368 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1369 int offset, u8 *to, int len,
1371 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1372 extern void skb_split(struct sk_buff *skb,
1373 struct sk_buff *skb1, const u32 len);
1375 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1377 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1378 int len, void *buffer)
1380 int hlen = skb_headlen(skb);
1382 if (hlen - offset >= len)
1383 return skb->data + offset;
1385 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1391 extern void skb_init(void);
1392 extern void skb_add_mtu(int mtu);
1395 * skb_get_timestamp - get timestamp from a skb
1396 * @skb: skb to get stamp from
1397 * @stamp: pointer to struct timeval to store stamp in
1399 * Timestamps are stored in the skb as offsets to a base timestamp.
1400 * This function converts the offset back to a struct timeval and stores
1403 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1405 *stamp = ktime_to_timeval(skb->tstamp);
1408 static inline void __net_timestamp(struct sk_buff *skb)
1410 skb->tstamp = ktime_get_real();
1414 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
1415 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
1418 * skb_checksum_complete - Calculate checksum of an entire packet
1419 * @skb: packet to process
1421 * This function calculates the checksum over the entire packet plus
1422 * the value of skb->csum. The latter can be used to supply the
1423 * checksum of a pseudo header as used by TCP/UDP. It returns the
1426 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1427 * this function can be used to verify that checksum on received
1428 * packets. In that case the function should return zero if the
1429 * checksum is correct. In particular, this function will return zero
1430 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1431 * hardware has already verified the correctness of the checksum.
1433 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1435 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1436 __skb_checksum_complete(skb);
1439 #ifdef CONFIG_NETFILTER
1440 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1442 if (nfct && atomic_dec_and_test(&nfct->use))
1443 nfct->destroy(nfct);
1445 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1448 atomic_inc(&nfct->use);
1450 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1451 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1454 atomic_inc(&skb->users);
1456 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1462 #ifdef CONFIG_BRIDGE_NETFILTER
1463 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1465 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1468 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1471 atomic_inc(&nf_bridge->use);
1473 #endif /* CONFIG_BRIDGE_NETFILTER */
1474 static inline void nf_reset(struct sk_buff *skb)
1476 nf_conntrack_put(skb->nfct);
1478 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1479 nf_conntrack_put_reasm(skb->nfct_reasm);
1480 skb->nfct_reasm = NULL;
1482 #ifdef CONFIG_BRIDGE_NETFILTER
1483 nf_bridge_put(skb->nf_bridge);
1484 skb->nf_bridge = NULL;
1488 #else /* CONFIG_NETFILTER */
1489 static inline void nf_reset(struct sk_buff *skb) {}
1490 #endif /* CONFIG_NETFILTER */
1492 #ifdef CONFIG_NETWORK_SECMARK
1493 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1495 to->secmark = from->secmark;
1498 static inline void skb_init_secmark(struct sk_buff *skb)
1503 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1506 static inline void skb_init_secmark(struct sk_buff *skb)
1510 static inline int skb_is_gso(const struct sk_buff *skb)
1512 return skb_shinfo(skb)->gso_size;
1515 #endif /* __KERNEL__ */
1516 #endif /* _LINUX_SKBUFF_H */