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*/
241 struct icmphdr *icmph;
243 struct ipv6hdr *ipv6h;
255 struct dst_entry *dst;
259 * This is the control buffer. It is free to use for every
260 * layer. Please put your private variables there. If you
261 * want to keep them across layers you have to do a skb_clone()
262 * first. This is owned by whoever has the skb queued ATM.
284 void (*destructor)(struct sk_buff *skb);
285 #ifdef CONFIG_NETFILTER
286 struct nf_conntrack *nfct;
287 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
288 struct sk_buff *nfct_reasm;
290 #ifdef CONFIG_BRIDGE_NETFILTER
291 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
309 /* These elements must be at the end, see alloc_skb() for details. */
310 unsigned int truesize;
320 * Handling routines are only of interest to the kernel
322 #include <linux/slab.h>
324 #include <asm/system.h>
326 extern void kfree_skb(struct sk_buff *skb);
327 extern void __kfree_skb(struct sk_buff *skb);
328 extern struct sk_buff *__alloc_skb(unsigned int size,
329 gfp_t priority, int fclone, int node);
330 static inline struct sk_buff *alloc_skb(unsigned int size,
333 return __alloc_skb(size, priority, 0, -1);
336 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
339 return __alloc_skb(size, priority, 1, -1);
342 extern void kfree_skbmem(struct sk_buff *skb);
343 extern struct sk_buff *skb_clone(struct sk_buff *skb,
345 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
347 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
349 extern int pskb_expand_head(struct sk_buff *skb,
350 int nhead, int ntail,
352 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
353 unsigned int headroom);
354 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
355 int newheadroom, int newtailroom,
357 extern int skb_pad(struct sk_buff *skb, int pad);
358 #define dev_kfree_skb(a) kfree_skb(a)
359 extern void skb_over_panic(struct sk_buff *skb, int len,
361 extern void skb_under_panic(struct sk_buff *skb, int len,
363 extern void skb_truesize_bug(struct sk_buff *skb);
365 static inline void skb_truesize_check(struct sk_buff *skb)
367 if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len))
368 skb_truesize_bug(skb);
371 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
372 int getfrag(void *from, char *to, int offset,
373 int len,int odd, struct sk_buff *skb),
374 void *from, int length);
381 __u32 stepped_offset;
382 struct sk_buff *root_skb;
383 struct sk_buff *cur_skb;
387 extern void skb_prepare_seq_read(struct sk_buff *skb,
388 unsigned int from, unsigned int to,
389 struct skb_seq_state *st);
390 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
391 struct skb_seq_state *st);
392 extern void skb_abort_seq_read(struct skb_seq_state *st);
394 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
395 unsigned int to, struct ts_config *config,
396 struct ts_state *state);
399 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
402 * skb_queue_empty - check if a queue is empty
405 * Returns true if the queue is empty, false otherwise.
407 static inline int skb_queue_empty(const struct sk_buff_head *list)
409 return list->next == (struct sk_buff *)list;
413 * skb_get - reference buffer
414 * @skb: buffer to reference
416 * Makes another reference to a socket buffer and returns a pointer
419 static inline struct sk_buff *skb_get(struct sk_buff *skb)
421 atomic_inc(&skb->users);
426 * If users == 1, we are the only owner and are can avoid redundant
431 * skb_cloned - is the buffer a clone
432 * @skb: buffer to check
434 * Returns true if the buffer was generated with skb_clone() and is
435 * one of multiple shared copies of the buffer. Cloned buffers are
436 * shared data so must not be written to under normal circumstances.
438 static inline int skb_cloned(const struct sk_buff *skb)
440 return skb->cloned &&
441 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
445 * skb_header_cloned - is the header a clone
446 * @skb: buffer to check
448 * Returns true if modifying the header part of the buffer requires
449 * the data to be copied.
451 static inline int skb_header_cloned(const struct sk_buff *skb)
458 dataref = atomic_read(&skb_shinfo(skb)->dataref);
459 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
464 * skb_header_release - release reference to header
465 * @skb: buffer to operate on
467 * Drop a reference to the header part of the buffer. This is done
468 * by acquiring a payload reference. You must not read from the header
469 * part of skb->data after this.
471 static inline void skb_header_release(struct sk_buff *skb)
475 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
479 * skb_shared - is the buffer shared
480 * @skb: buffer to check
482 * Returns true if more than one person has a reference to this
485 static inline int skb_shared(const struct sk_buff *skb)
487 return atomic_read(&skb->users) != 1;
491 * skb_share_check - check if buffer is shared and if so clone it
492 * @skb: buffer to check
493 * @pri: priority for memory allocation
495 * If the buffer is shared the buffer is cloned and the old copy
496 * drops a reference. A new clone with a single reference is returned.
497 * If the buffer is not shared the original buffer is returned. When
498 * being called from interrupt status or with spinlocks held pri must
501 * NULL is returned on a memory allocation failure.
503 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
506 might_sleep_if(pri & __GFP_WAIT);
507 if (skb_shared(skb)) {
508 struct sk_buff *nskb = skb_clone(skb, pri);
516 * Copy shared buffers into a new sk_buff. We effectively do COW on
517 * packets to handle cases where we have a local reader and forward
518 * and a couple of other messy ones. The normal one is tcpdumping
519 * a packet thats being forwarded.
523 * skb_unshare - make a copy of a shared buffer
524 * @skb: buffer to check
525 * @pri: priority for memory allocation
527 * If the socket buffer is a clone then this function creates a new
528 * copy of the data, drops a reference count on the old copy and returns
529 * the new copy with the reference count at 1. If the buffer is not a clone
530 * the original buffer is returned. When called with a spinlock held or
531 * from interrupt state @pri must be %GFP_ATOMIC
533 * %NULL is returned on a memory allocation failure.
535 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
538 might_sleep_if(pri & __GFP_WAIT);
539 if (skb_cloned(skb)) {
540 struct sk_buff *nskb = skb_copy(skb, pri);
541 kfree_skb(skb); /* Free our shared copy */
549 * @list_: list to peek at
551 * Peek an &sk_buff. Unlike most other operations you _MUST_
552 * be careful with this one. A peek leaves the buffer on the
553 * list and someone else may run off with it. You must hold
554 * the appropriate locks or have a private queue to do this.
556 * Returns %NULL for an empty list or a pointer to the head element.
557 * The reference count is not incremented and the reference is therefore
558 * volatile. Use with caution.
560 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
562 struct sk_buff *list = ((struct sk_buff *)list_)->next;
563 if (list == (struct sk_buff *)list_)
570 * @list_: list to peek at
572 * Peek an &sk_buff. Unlike most other operations you _MUST_
573 * be careful with this one. A peek leaves the buffer on the
574 * list and someone else may run off with it. You must hold
575 * the appropriate locks or have a private queue to do this.
577 * Returns %NULL for an empty list or a pointer to the tail element.
578 * The reference count is not incremented and the reference is therefore
579 * volatile. Use with caution.
581 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
583 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
584 if (list == (struct sk_buff *)list_)
590 * skb_queue_len - get queue length
591 * @list_: list to measure
593 * Return the length of an &sk_buff queue.
595 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
601 * This function creates a split out lock class for each invocation;
602 * this is needed for now since a whole lot of users of the skb-queue
603 * infrastructure in drivers have different locking usage (in hardirq)
604 * than the networking core (in softirq only). In the long run either the
605 * network layer or drivers should need annotation to consolidate the
606 * main types of usage into 3 classes.
608 static inline void skb_queue_head_init(struct sk_buff_head *list)
610 spin_lock_init(&list->lock);
611 list->prev = list->next = (struct sk_buff *)list;
615 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
616 struct lock_class_key *class)
618 skb_queue_head_init(list);
619 lockdep_set_class(&list->lock, class);
623 * Insert an sk_buff at the start of a list.
625 * The "__skb_xxxx()" functions are the non-atomic ones that
626 * can only be called with interrupts disabled.
630 * __skb_queue_after - queue a buffer at the list head
632 * @prev: place after this buffer
633 * @newsk: buffer to queue
635 * Queue a buffer int the middle of a list. This function takes no locks
636 * and you must therefore hold required locks before calling it.
638 * A buffer cannot be placed on two lists at the same time.
640 static inline void __skb_queue_after(struct sk_buff_head *list,
641 struct sk_buff *prev,
642 struct sk_buff *newsk)
644 struct sk_buff *next;
650 next->prev = prev->next = newsk;
654 * __skb_queue_head - queue a buffer at the list head
656 * @newsk: buffer to queue
658 * Queue a buffer at the start of a list. This function takes no locks
659 * and you must therefore hold required locks before calling it.
661 * A buffer cannot be placed on two lists at the same time.
663 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
664 static inline void __skb_queue_head(struct sk_buff_head *list,
665 struct sk_buff *newsk)
667 __skb_queue_after(list, (struct sk_buff *)list, newsk);
671 * __skb_queue_tail - queue a buffer at the list tail
673 * @newsk: buffer to queue
675 * Queue a buffer at the end of a list. This function takes no locks
676 * and you must therefore hold required locks before calling it.
678 * A buffer cannot be placed on two lists at the same time.
680 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
681 static inline void __skb_queue_tail(struct sk_buff_head *list,
682 struct sk_buff *newsk)
684 struct sk_buff *prev, *next;
687 next = (struct sk_buff *)list;
691 next->prev = prev->next = newsk;
696 * __skb_dequeue - remove from the head of the queue
697 * @list: list to dequeue from
699 * Remove the head of the list. This function does not take any locks
700 * so must be used with appropriate locks held only. The head item is
701 * returned or %NULL if the list is empty.
703 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
704 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
706 struct sk_buff *next, *prev, *result;
708 prev = (struct sk_buff *) list;
717 result->next = result->prev = NULL;
724 * Insert a packet on a list.
726 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
727 static inline void __skb_insert(struct sk_buff *newsk,
728 struct sk_buff *prev, struct sk_buff *next,
729 struct sk_buff_head *list)
733 next->prev = prev->next = newsk;
738 * Place a packet after a given packet in a list.
740 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
741 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
743 __skb_insert(newsk, old, old->next, list);
747 * remove sk_buff from list. _Must_ be called atomically, and with
750 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
751 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
753 struct sk_buff *next, *prev;
758 skb->next = skb->prev = NULL;
764 /* XXX: more streamlined implementation */
767 * __skb_dequeue_tail - remove from the tail of the queue
768 * @list: list to dequeue from
770 * Remove the tail of the list. This function does not take any locks
771 * so must be used with appropriate locks held only. The tail item is
772 * returned or %NULL if the list is empty.
774 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
775 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
777 struct sk_buff *skb = skb_peek_tail(list);
779 __skb_unlink(skb, list);
784 static inline int skb_is_nonlinear(const struct sk_buff *skb)
786 return skb->data_len;
789 static inline unsigned int skb_headlen(const struct sk_buff *skb)
791 return skb->len - skb->data_len;
794 static inline int skb_pagelen(const struct sk_buff *skb)
798 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
799 len += skb_shinfo(skb)->frags[i].size;
800 return len + skb_headlen(skb);
803 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
804 struct page *page, int off, int size)
806 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
809 frag->page_offset = off;
811 skb_shinfo(skb)->nr_frags = i + 1;
814 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
815 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
816 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
819 * Add data to an sk_buff
821 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
823 unsigned char *tmp = skb->tail;
824 SKB_LINEAR_ASSERT(skb);
831 * skb_put - add data to a buffer
832 * @skb: buffer to use
833 * @len: amount of data to add
835 * This function extends the used data area of the buffer. If this would
836 * exceed the total buffer size the kernel will panic. A pointer to the
837 * first byte of the extra data is returned.
839 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
841 unsigned char *tmp = skb->tail;
842 SKB_LINEAR_ASSERT(skb);
845 if (unlikely(skb->tail>skb->end))
846 skb_over_panic(skb, len, current_text_addr());
850 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
858 * skb_push - add data to the start of a buffer
859 * @skb: buffer to use
860 * @len: amount of data to add
862 * This function extends the used data area of the buffer at the buffer
863 * start. If this would exceed the total buffer headroom the kernel will
864 * panic. A pointer to the first byte of the extra data is returned.
866 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
870 if (unlikely(skb->data<skb->head))
871 skb_under_panic(skb, len, current_text_addr());
875 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
878 BUG_ON(skb->len < skb->data_len);
879 return skb->data += len;
883 * skb_pull - remove data from the start of a buffer
884 * @skb: buffer to use
885 * @len: amount of data to remove
887 * This function removes data from the start of a buffer, returning
888 * the memory to the headroom. A pointer to the next data in the buffer
889 * is returned. Once the data has been pulled future pushes will overwrite
892 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
894 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
897 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
899 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
901 if (len > skb_headlen(skb) &&
902 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
905 return skb->data += len;
908 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
910 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
913 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
915 if (likely(len <= skb_headlen(skb)))
917 if (unlikely(len > skb->len))
919 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
923 * skb_headroom - bytes at buffer head
924 * @skb: buffer to check
926 * Return the number of bytes of free space at the head of an &sk_buff.
928 static inline int skb_headroom(const struct sk_buff *skb)
930 return skb->data - skb->head;
934 * skb_tailroom - bytes at buffer end
935 * @skb: buffer to check
937 * Return the number of bytes of free space at the tail of an sk_buff
939 static inline int skb_tailroom(const struct sk_buff *skb)
941 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
945 * skb_reserve - adjust headroom
946 * @skb: buffer to alter
947 * @len: bytes to move
949 * Increase the headroom of an empty &sk_buff by reducing the tail
950 * room. This is only allowed for an empty buffer.
952 static inline void skb_reserve(struct sk_buff *skb, int len)
958 static inline void skb_reset_transport_header(struct sk_buff *skb)
960 skb->h.raw = skb->data;
963 static inline void skb_set_transport_header(struct sk_buff *skb,
966 skb->h.raw = skb->data + offset;
969 static inline int skb_transport_offset(const struct sk_buff *skb)
971 return skb->h.raw - skb->data;
974 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
979 static inline void skb_reset_network_header(struct sk_buff *skb)
981 skb->nh.raw = skb->data;
984 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
986 skb->nh.raw = skb->data + offset;
989 static inline int skb_network_offset(const struct sk_buff *skb)
991 return skb->nh.raw - skb->data;
994 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
999 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1001 return skb->mac.raw != NULL;
1004 static inline void skb_reset_mac_header(struct sk_buff *skb)
1006 skb->mac.raw = skb->data;
1009 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1011 skb->mac.raw = skb->data + offset;
1015 * CPUs often take a performance hit when accessing unaligned memory
1016 * locations. The actual performance hit varies, it can be small if the
1017 * hardware handles it or large if we have to take an exception and fix it
1020 * Since an ethernet header is 14 bytes network drivers often end up with
1021 * the IP header at an unaligned offset. The IP header can be aligned by
1022 * shifting the start of the packet by 2 bytes. Drivers should do this
1025 * skb_reserve(NET_IP_ALIGN);
1027 * The downside to this alignment of the IP header is that the DMA is now
1028 * unaligned. On some architectures the cost of an unaligned DMA is high
1029 * and this cost outweighs the gains made by aligning the IP header.
1031 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1034 #ifndef NET_IP_ALIGN
1035 #define NET_IP_ALIGN 2
1039 * The networking layer reserves some headroom in skb data (via
1040 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1041 * the header has to grow. In the default case, if the header has to grow
1042 * 16 bytes or less we avoid the reallocation.
1044 * Unfortunately this headroom changes the DMA alignment of the resulting
1045 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1046 * on some architectures. An architecture can override this value,
1047 * perhaps setting it to a cacheline in size (since that will maintain
1048 * cacheline alignment of the DMA). It must be a power of 2.
1050 * Various parts of the networking layer expect at least 16 bytes of
1051 * headroom, you should not reduce this.
1054 #define NET_SKB_PAD 16
1057 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1059 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1061 if (unlikely(skb->data_len)) {
1066 skb->tail = skb->data + len;
1070 * skb_trim - remove end from a buffer
1071 * @skb: buffer to alter
1074 * Cut the length of a buffer down by removing data from the tail. If
1075 * the buffer is already under the length specified it is not modified.
1076 * The skb must be linear.
1078 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
1081 __skb_trim(skb, len);
1085 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1088 return ___pskb_trim(skb, len);
1089 __skb_trim(skb, len);
1093 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1095 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1099 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1100 * @skb: buffer to alter
1103 * This is identical to pskb_trim except that the caller knows that
1104 * the skb is not cloned so we should never get an error due to out-
1107 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1109 int err = pskb_trim(skb, len);
1114 * skb_orphan - orphan a buffer
1115 * @skb: buffer to orphan
1117 * If a buffer currently has an owner then we call the owner's
1118 * destructor function and make the @skb unowned. The buffer continues
1119 * to exist but is no longer charged to its former owner.
1121 static inline void skb_orphan(struct sk_buff *skb)
1123 if (skb->destructor)
1124 skb->destructor(skb);
1125 skb->destructor = NULL;
1130 * __skb_queue_purge - empty a list
1131 * @list: list to empty
1133 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1134 * the list and one reference dropped. This function does not take the
1135 * list lock and the caller must hold the relevant locks to use it.
1137 extern void skb_queue_purge(struct sk_buff_head *list);
1138 static inline void __skb_queue_purge(struct sk_buff_head *list)
1140 struct sk_buff *skb;
1141 while ((skb = __skb_dequeue(list)) != NULL)
1146 * __dev_alloc_skb - allocate an skbuff for receiving
1147 * @length: length to allocate
1148 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1150 * Allocate a new &sk_buff and assign it a usage count of one. The
1151 * buffer has unspecified headroom built in. Users should allocate
1152 * the headroom they think they need without accounting for the
1153 * built in space. The built in space is used for optimisations.
1155 * %NULL is returned if there is no free memory.
1157 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1160 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1162 skb_reserve(skb, NET_SKB_PAD);
1167 * dev_alloc_skb - allocate an skbuff for receiving
1168 * @length: length to allocate
1170 * Allocate a new &sk_buff and assign it a usage count of one. The
1171 * buffer has unspecified headroom built in. Users should allocate
1172 * the headroom they think they need without accounting for the
1173 * built in space. The built in space is used for optimisations.
1175 * %NULL is returned if there is no free memory. Although this function
1176 * allocates memory it can be called from an interrupt.
1178 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1180 return __dev_alloc_skb(length, GFP_ATOMIC);
1183 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1184 unsigned int length, gfp_t gfp_mask);
1187 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1188 * @dev: network device to receive on
1189 * @length: length to allocate
1191 * Allocate a new &sk_buff and assign it a usage count of one. The
1192 * buffer has unspecified headroom built in. Users should allocate
1193 * the headroom they think they need without accounting for the
1194 * built in space. The built in space is used for optimisations.
1196 * %NULL is returned if there is no free memory. Although this function
1197 * allocates memory it can be called from an interrupt.
1199 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1200 unsigned int length)
1202 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1206 * skb_cow - copy header of skb when it is required
1207 * @skb: buffer to cow
1208 * @headroom: needed headroom
1210 * If the skb passed lacks sufficient headroom or its data part
1211 * is shared, data is reallocated. If reallocation fails, an error
1212 * is returned and original skb is not changed.
1214 * The result is skb with writable area skb->head...skb->tail
1215 * and at least @headroom of space at head.
1217 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1219 int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) -
1225 if (delta || skb_cloned(skb))
1226 return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) &
1227 ~(NET_SKB_PAD-1), 0, GFP_ATOMIC);
1232 * skb_padto - pad an skbuff up to a minimal size
1233 * @skb: buffer to pad
1234 * @len: minimal length
1236 * Pads up a buffer to ensure the trailing bytes exist and are
1237 * blanked. If the buffer already contains sufficient data it
1238 * is untouched. Otherwise it is extended. Returns zero on
1239 * success. The skb is freed on error.
1242 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1244 unsigned int size = skb->len;
1245 if (likely(size >= len))
1247 return skb_pad(skb, len-size);
1250 static inline int skb_add_data(struct sk_buff *skb,
1251 char __user *from, int copy)
1253 const int off = skb->len;
1255 if (skb->ip_summed == CHECKSUM_NONE) {
1257 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1260 skb->csum = csum_block_add(skb->csum, csum, off);
1263 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1266 __skb_trim(skb, off);
1270 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1271 struct page *page, int off)
1274 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1276 return page == frag->page &&
1277 off == frag->page_offset + frag->size;
1282 static inline int __skb_linearize(struct sk_buff *skb)
1284 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1288 * skb_linearize - convert paged skb to linear one
1289 * @skb: buffer to linarize
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(struct sk_buff *skb)
1296 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1300 * skb_linearize_cow - make sure skb is linear and writable
1301 * @skb: buffer to process
1303 * If there is no free memory -ENOMEM is returned, otherwise zero
1304 * is returned and the old skb data released.
1306 static inline int skb_linearize_cow(struct sk_buff *skb)
1308 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1309 __skb_linearize(skb) : 0;
1313 * skb_postpull_rcsum - update checksum for received skb after pull
1314 * @skb: buffer to update
1315 * @start: start of data before pull
1316 * @len: length of data pulled
1318 * After doing a pull on a received packet, you need to call this to
1319 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1320 * CHECKSUM_NONE so that it can be recomputed from scratch.
1323 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1324 const void *start, unsigned int len)
1326 if (skb->ip_summed == CHECKSUM_COMPLETE)
1327 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1330 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1333 * pskb_trim_rcsum - trim received skb and update checksum
1334 * @skb: buffer to trim
1337 * This is exactly the same as pskb_trim except that it ensures the
1338 * checksum of received packets are still valid after the operation.
1341 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1343 if (likely(len >= skb->len))
1345 if (skb->ip_summed == CHECKSUM_COMPLETE)
1346 skb->ip_summed = CHECKSUM_NONE;
1347 return __pskb_trim(skb, len);
1350 #define skb_queue_walk(queue, skb) \
1351 for (skb = (queue)->next; \
1352 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1355 #define skb_queue_reverse_walk(queue, skb) \
1356 for (skb = (queue)->prev; \
1357 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1361 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1362 int noblock, int *err);
1363 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1364 struct poll_table_struct *wait);
1365 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1366 int offset, struct iovec *to,
1368 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1371 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1372 extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1373 unsigned int flags);
1374 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1375 int len, __wsum csum);
1376 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1378 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1379 void *from, int len);
1380 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1381 int offset, u8 *to, int len,
1383 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1384 extern void skb_split(struct sk_buff *skb,
1385 struct sk_buff *skb1, const u32 len);
1387 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1389 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1390 int len, void *buffer)
1392 int hlen = skb_headlen(skb);
1394 if (hlen - offset >= len)
1395 return skb->data + offset;
1397 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1403 extern void skb_init(void);
1404 extern void skb_add_mtu(int mtu);
1407 * skb_get_timestamp - get timestamp from a skb
1408 * @skb: skb to get stamp from
1409 * @stamp: pointer to struct timeval to store stamp in
1411 * Timestamps are stored in the skb as offsets to a base timestamp.
1412 * This function converts the offset back to a struct timeval and stores
1415 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1417 *stamp = ktime_to_timeval(skb->tstamp);
1420 static inline void __net_timestamp(struct sk_buff *skb)
1422 skb->tstamp = ktime_get_real();
1426 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
1427 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
1430 * skb_checksum_complete - Calculate checksum of an entire packet
1431 * @skb: packet to process
1433 * This function calculates the checksum over the entire packet plus
1434 * the value of skb->csum. The latter can be used to supply the
1435 * checksum of a pseudo header as used by TCP/UDP. It returns the
1438 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1439 * this function can be used to verify that checksum on received
1440 * packets. In that case the function should return zero if the
1441 * checksum is correct. In particular, this function will return zero
1442 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1443 * hardware has already verified the correctness of the checksum.
1445 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1447 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1448 __skb_checksum_complete(skb);
1451 #ifdef CONFIG_NETFILTER
1452 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1454 if (nfct && atomic_dec_and_test(&nfct->use))
1455 nfct->destroy(nfct);
1457 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1460 atomic_inc(&nfct->use);
1462 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1463 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1466 atomic_inc(&skb->users);
1468 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1474 #ifdef CONFIG_BRIDGE_NETFILTER
1475 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1477 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1480 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1483 atomic_inc(&nf_bridge->use);
1485 #endif /* CONFIG_BRIDGE_NETFILTER */
1486 static inline void nf_reset(struct sk_buff *skb)
1488 nf_conntrack_put(skb->nfct);
1490 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1491 nf_conntrack_put_reasm(skb->nfct_reasm);
1492 skb->nfct_reasm = NULL;
1494 #ifdef CONFIG_BRIDGE_NETFILTER
1495 nf_bridge_put(skb->nf_bridge);
1496 skb->nf_bridge = NULL;
1500 #else /* CONFIG_NETFILTER */
1501 static inline void nf_reset(struct sk_buff *skb) {}
1502 #endif /* CONFIG_NETFILTER */
1504 #ifdef CONFIG_NETWORK_SECMARK
1505 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1507 to->secmark = from->secmark;
1510 static inline void skb_init_secmark(struct sk_buff *skb)
1515 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1518 static inline void skb_init_secmark(struct sk_buff *skb)
1522 static inline int skb_is_gso(const struct sk_buff *skb)
1524 return skb_shinfo(skb)->gso_size;
1527 #endif /* __KERNEL__ */
1528 #endif /* _LINUX_SKBUFF_H */