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;
258 struct dst_entry *dst;
262 * This is the control buffer. It is free to use for every
263 * layer. Please put your private variables there. If you
264 * want to keep them across layers you have to do a skb_clone()
265 * first. This is owned by whoever has the skb queued ATM.
287 void (*destructor)(struct sk_buff *skb);
288 #ifdef CONFIG_NETFILTER
289 struct nf_conntrack *nfct;
290 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
291 struct sk_buff *nfct_reasm;
293 #ifdef CONFIG_BRIDGE_NETFILTER
294 struct nf_bridge_info *nf_bridge;
296 #endif /* CONFIG_NETFILTER */
297 #ifdef CONFIG_NET_SCHED
298 __u16 tc_index; /* traffic control index */
299 #ifdef CONFIG_NET_CLS_ACT
300 __u16 tc_verd; /* traffic control verdict */
303 #ifdef CONFIG_NET_DMA
304 dma_cookie_t dma_cookie;
306 #ifdef CONFIG_NETWORK_SECMARK
312 /* These elements must be at the end, see alloc_skb() for details. */
313 unsigned int truesize;
323 * Handling routines are only of interest to the kernel
325 #include <linux/slab.h>
327 #include <asm/system.h>
329 extern void kfree_skb(struct sk_buff *skb);
330 extern void __kfree_skb(struct sk_buff *skb);
331 extern struct sk_buff *__alloc_skb(unsigned int size,
332 gfp_t priority, int fclone, int node);
333 static inline struct sk_buff *alloc_skb(unsigned int size,
336 return __alloc_skb(size, priority, 0, -1);
339 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
342 return __alloc_skb(size, priority, 1, -1);
345 extern void kfree_skbmem(struct sk_buff *skb);
346 extern struct sk_buff *skb_clone(struct sk_buff *skb,
348 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
350 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
352 extern int pskb_expand_head(struct sk_buff *skb,
353 int nhead, int ntail,
355 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
356 unsigned int headroom);
357 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
358 int newheadroom, int newtailroom,
360 extern int skb_pad(struct sk_buff *skb, int pad);
361 #define dev_kfree_skb(a) kfree_skb(a)
362 extern void skb_over_panic(struct sk_buff *skb, int len,
364 extern void skb_under_panic(struct sk_buff *skb, int len,
366 extern void skb_truesize_bug(struct sk_buff *skb);
368 static inline void skb_truesize_check(struct sk_buff *skb)
370 if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len))
371 skb_truesize_bug(skb);
374 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
375 int getfrag(void *from, char *to, int offset,
376 int len,int odd, struct sk_buff *skb),
377 void *from, int length);
384 __u32 stepped_offset;
385 struct sk_buff *root_skb;
386 struct sk_buff *cur_skb;
390 extern void skb_prepare_seq_read(struct sk_buff *skb,
391 unsigned int from, unsigned int to,
392 struct skb_seq_state *st);
393 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
394 struct skb_seq_state *st);
395 extern void skb_abort_seq_read(struct skb_seq_state *st);
397 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
398 unsigned int to, struct ts_config *config,
399 struct ts_state *state);
402 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
405 * skb_queue_empty - check if a queue is empty
408 * Returns true if the queue is empty, false otherwise.
410 static inline int skb_queue_empty(const struct sk_buff_head *list)
412 return list->next == (struct sk_buff *)list;
416 * skb_get - reference buffer
417 * @skb: buffer to reference
419 * Makes another reference to a socket buffer and returns a pointer
422 static inline struct sk_buff *skb_get(struct sk_buff *skb)
424 atomic_inc(&skb->users);
429 * If users == 1, we are the only owner and are can avoid redundant
434 * skb_cloned - is the buffer a clone
435 * @skb: buffer to check
437 * Returns true if the buffer was generated with skb_clone() and is
438 * one of multiple shared copies of the buffer. Cloned buffers are
439 * shared data so must not be written to under normal circumstances.
441 static inline int skb_cloned(const struct sk_buff *skb)
443 return skb->cloned &&
444 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
448 * skb_header_cloned - is the header a clone
449 * @skb: buffer to check
451 * Returns true if modifying the header part of the buffer requires
452 * the data to be copied.
454 static inline int skb_header_cloned(const struct sk_buff *skb)
461 dataref = atomic_read(&skb_shinfo(skb)->dataref);
462 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
467 * skb_header_release - release reference to header
468 * @skb: buffer to operate on
470 * Drop a reference to the header part of the buffer. This is done
471 * by acquiring a payload reference. You must not read from the header
472 * part of skb->data after this.
474 static inline void skb_header_release(struct sk_buff *skb)
478 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
482 * skb_shared - is the buffer shared
483 * @skb: buffer to check
485 * Returns true if more than one person has a reference to this
488 static inline int skb_shared(const struct sk_buff *skb)
490 return atomic_read(&skb->users) != 1;
494 * skb_share_check - check if buffer is shared and if so clone it
495 * @skb: buffer to check
496 * @pri: priority for memory allocation
498 * If the buffer is shared the buffer is cloned and the old copy
499 * drops a reference. A new clone with a single reference is returned.
500 * If the buffer is not shared the original buffer is returned. When
501 * being called from interrupt status or with spinlocks held pri must
504 * NULL is returned on a memory allocation failure.
506 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
509 might_sleep_if(pri & __GFP_WAIT);
510 if (skb_shared(skb)) {
511 struct sk_buff *nskb = skb_clone(skb, pri);
519 * Copy shared buffers into a new sk_buff. We effectively do COW on
520 * packets to handle cases where we have a local reader and forward
521 * and a couple of other messy ones. The normal one is tcpdumping
522 * a packet thats being forwarded.
526 * skb_unshare - make a copy of a shared buffer
527 * @skb: buffer to check
528 * @pri: priority for memory allocation
530 * If the socket buffer is a clone then this function creates a new
531 * copy of the data, drops a reference count on the old copy and returns
532 * the new copy with the reference count at 1. If the buffer is not a clone
533 * the original buffer is returned. When called with a spinlock held or
534 * from interrupt state @pri must be %GFP_ATOMIC
536 * %NULL is returned on a memory allocation failure.
538 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
541 might_sleep_if(pri & __GFP_WAIT);
542 if (skb_cloned(skb)) {
543 struct sk_buff *nskb = skb_copy(skb, pri);
544 kfree_skb(skb); /* Free our shared copy */
552 * @list_: list to peek at
554 * Peek an &sk_buff. Unlike most other operations you _MUST_
555 * be careful with this one. A peek leaves the buffer on the
556 * list and someone else may run off with it. You must hold
557 * the appropriate locks or have a private queue to do this.
559 * Returns %NULL for an empty list or a pointer to the head element.
560 * The reference count is not incremented and the reference is therefore
561 * volatile. Use with caution.
563 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
565 struct sk_buff *list = ((struct sk_buff *)list_)->next;
566 if (list == (struct sk_buff *)list_)
573 * @list_: list to peek at
575 * Peek an &sk_buff. Unlike most other operations you _MUST_
576 * be careful with this one. A peek leaves the buffer on the
577 * list and someone else may run off with it. You must hold
578 * the appropriate locks or have a private queue to do this.
580 * Returns %NULL for an empty list or a pointer to the tail element.
581 * The reference count is not incremented and the reference is therefore
582 * volatile. Use with caution.
584 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
586 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
587 if (list == (struct sk_buff *)list_)
593 * skb_queue_len - get queue length
594 * @list_: list to measure
596 * Return the length of an &sk_buff queue.
598 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
604 * This function creates a split out lock class for each invocation;
605 * this is needed for now since a whole lot of users of the skb-queue
606 * infrastructure in drivers have different locking usage (in hardirq)
607 * than the networking core (in softirq only). In the long run either the
608 * network layer or drivers should need annotation to consolidate the
609 * main types of usage into 3 classes.
611 static inline void skb_queue_head_init(struct sk_buff_head *list)
613 spin_lock_init(&list->lock);
614 list->prev = list->next = (struct sk_buff *)list;
618 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
619 struct lock_class_key *class)
621 skb_queue_head_init(list);
622 lockdep_set_class(&list->lock, class);
626 * Insert an sk_buff at the start of a list.
628 * The "__skb_xxxx()" functions are the non-atomic ones that
629 * can only be called with interrupts disabled.
633 * __skb_queue_after - queue a buffer at the list head
635 * @prev: place after this buffer
636 * @newsk: buffer to queue
638 * Queue a buffer int the middle of a list. This function takes no locks
639 * and you must therefore hold required locks before calling it.
641 * A buffer cannot be placed on two lists at the same time.
643 static inline void __skb_queue_after(struct sk_buff_head *list,
644 struct sk_buff *prev,
645 struct sk_buff *newsk)
647 struct sk_buff *next;
653 next->prev = prev->next = newsk;
657 * __skb_queue_head - queue a buffer at the list head
659 * @newsk: buffer to queue
661 * Queue a buffer at the start of a list. This function takes no locks
662 * and you must therefore hold required locks before calling it.
664 * A buffer cannot be placed on two lists at the same time.
666 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
667 static inline void __skb_queue_head(struct sk_buff_head *list,
668 struct sk_buff *newsk)
670 __skb_queue_after(list, (struct sk_buff *)list, newsk);
674 * __skb_queue_tail - queue a buffer at the list tail
676 * @newsk: buffer to queue
678 * Queue a buffer at the end of a list. This function takes no locks
679 * and you must therefore hold required locks before calling it.
681 * A buffer cannot be placed on two lists at the same time.
683 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
684 static inline void __skb_queue_tail(struct sk_buff_head *list,
685 struct sk_buff *newsk)
687 struct sk_buff *prev, *next;
690 next = (struct sk_buff *)list;
694 next->prev = prev->next = newsk;
699 * __skb_dequeue - remove from the head of the queue
700 * @list: list to dequeue from
702 * Remove the head of the list. This function does not take any locks
703 * so must be used with appropriate locks held only. The head item is
704 * returned or %NULL if the list is empty.
706 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
707 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
709 struct sk_buff *next, *prev, *result;
711 prev = (struct sk_buff *) list;
720 result->next = result->prev = NULL;
727 * Insert a packet on a list.
729 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
730 static inline void __skb_insert(struct sk_buff *newsk,
731 struct sk_buff *prev, struct sk_buff *next,
732 struct sk_buff_head *list)
736 next->prev = prev->next = newsk;
741 * Place a packet after a given packet in a list.
743 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
744 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
746 __skb_insert(newsk, old, old->next, list);
750 * remove sk_buff from list. _Must_ be called atomically, and with
753 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
754 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
756 struct sk_buff *next, *prev;
761 skb->next = skb->prev = NULL;
767 /* XXX: more streamlined implementation */
770 * __skb_dequeue_tail - remove from the tail of the queue
771 * @list: list to dequeue from
773 * Remove the tail of the list. This function does not take any locks
774 * so must be used with appropriate locks held only. The tail item is
775 * returned or %NULL if the list is empty.
777 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
778 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
780 struct sk_buff *skb = skb_peek_tail(list);
782 __skb_unlink(skb, list);
787 static inline int skb_is_nonlinear(const struct sk_buff *skb)
789 return skb->data_len;
792 static inline unsigned int skb_headlen(const struct sk_buff *skb)
794 return skb->len - skb->data_len;
797 static inline int skb_pagelen(const struct sk_buff *skb)
801 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
802 len += skb_shinfo(skb)->frags[i].size;
803 return len + skb_headlen(skb);
806 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
807 struct page *page, int off, int size)
809 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
812 frag->page_offset = off;
814 skb_shinfo(skb)->nr_frags = i + 1;
817 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
818 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
819 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
822 * Add data to an sk_buff
824 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
826 unsigned char *tmp = skb->tail;
827 SKB_LINEAR_ASSERT(skb);
834 * skb_put - add data to a buffer
835 * @skb: buffer to use
836 * @len: amount of data to add
838 * This function extends the used data area of the buffer. If this would
839 * exceed the total buffer size the kernel will panic. A pointer to the
840 * first byte of the extra data is returned.
842 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
844 unsigned char *tmp = skb->tail;
845 SKB_LINEAR_ASSERT(skb);
848 if (unlikely(skb->tail>skb->end))
849 skb_over_panic(skb, len, current_text_addr());
853 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
861 * skb_push - add data to the start of a buffer
862 * @skb: buffer to use
863 * @len: amount of data to add
865 * This function extends the used data area of the buffer at the buffer
866 * start. If this would exceed the total buffer headroom the kernel will
867 * panic. A pointer to the first byte of the extra data is returned.
869 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
873 if (unlikely(skb->data<skb->head))
874 skb_under_panic(skb, len, current_text_addr());
878 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
881 BUG_ON(skb->len < skb->data_len);
882 return skb->data += len;
886 * skb_pull - remove data from the start of a buffer
887 * @skb: buffer to use
888 * @len: amount of data to remove
890 * This function removes data from the start of a buffer, returning
891 * the memory to the headroom. A pointer to the next data in the buffer
892 * is returned. Once the data has been pulled future pushes will overwrite
895 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
897 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
900 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
902 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
904 if (len > skb_headlen(skb) &&
905 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
908 return skb->data += len;
911 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
913 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
916 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
918 if (likely(len <= skb_headlen(skb)))
920 if (unlikely(len > skb->len))
922 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
926 * skb_headroom - bytes at buffer head
927 * @skb: buffer to check
929 * Return the number of bytes of free space at the head of an &sk_buff.
931 static inline int skb_headroom(const struct sk_buff *skb)
933 return skb->data - skb->head;
937 * skb_tailroom - bytes at buffer end
938 * @skb: buffer to check
940 * Return the number of bytes of free space at the tail of an sk_buff
942 static inline int skb_tailroom(const struct sk_buff *skb)
944 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
948 * skb_reserve - adjust headroom
949 * @skb: buffer to alter
950 * @len: bytes to move
952 * Increase the headroom of an empty &sk_buff by reducing the tail
953 * room. This is only allowed for an empty buffer.
955 static inline void skb_reserve(struct sk_buff *skb, int len)
961 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
966 static inline void skb_reset_network_header(struct sk_buff *skb)
968 skb->nh.raw = skb->data;
971 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
973 skb->nh.raw = skb->data + offset;
976 static inline int skb_network_offset(const struct sk_buff *skb)
978 return skb->nh.raw - skb->data;
981 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
986 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
988 return skb->mac.raw != NULL;
991 static inline void skb_reset_mac_header(struct sk_buff *skb)
993 skb->mac.raw = skb->data;
996 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
998 skb->mac.raw = skb->data + offset;
1002 * CPUs often take a performance hit when accessing unaligned memory
1003 * locations. The actual performance hit varies, it can be small if the
1004 * hardware handles it or large if we have to take an exception and fix it
1007 * Since an ethernet header is 14 bytes network drivers often end up with
1008 * the IP header at an unaligned offset. The IP header can be aligned by
1009 * shifting the start of the packet by 2 bytes. Drivers should do this
1012 * skb_reserve(NET_IP_ALIGN);
1014 * The downside to this alignment of the IP header is that the DMA is now
1015 * unaligned. On some architectures the cost of an unaligned DMA is high
1016 * and this cost outweighs the gains made by aligning the IP header.
1018 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1021 #ifndef NET_IP_ALIGN
1022 #define NET_IP_ALIGN 2
1026 * The networking layer reserves some headroom in skb data (via
1027 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1028 * the header has to grow. In the default case, if the header has to grow
1029 * 16 bytes or less we avoid the reallocation.
1031 * Unfortunately this headroom changes the DMA alignment of the resulting
1032 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1033 * on some architectures. An architecture can override this value,
1034 * perhaps setting it to a cacheline in size (since that will maintain
1035 * cacheline alignment of the DMA). It must be a power of 2.
1037 * Various parts of the networking layer expect at least 16 bytes of
1038 * headroom, you should not reduce this.
1041 #define NET_SKB_PAD 16
1044 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1046 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1048 if (unlikely(skb->data_len)) {
1053 skb->tail = skb->data + len;
1057 * skb_trim - remove end from a buffer
1058 * @skb: buffer to alter
1061 * Cut the length of a buffer down by removing data from the tail. If
1062 * the buffer is already under the length specified it is not modified.
1063 * The skb must be linear.
1065 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
1068 __skb_trim(skb, len);
1072 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1075 return ___pskb_trim(skb, len);
1076 __skb_trim(skb, len);
1080 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1082 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1086 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1087 * @skb: buffer to alter
1090 * This is identical to pskb_trim except that the caller knows that
1091 * the skb is not cloned so we should never get an error due to out-
1094 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1096 int err = pskb_trim(skb, len);
1101 * skb_orphan - orphan a buffer
1102 * @skb: buffer to orphan
1104 * If a buffer currently has an owner then we call the owner's
1105 * destructor function and make the @skb unowned. The buffer continues
1106 * to exist but is no longer charged to its former owner.
1108 static inline void skb_orphan(struct sk_buff *skb)
1110 if (skb->destructor)
1111 skb->destructor(skb);
1112 skb->destructor = NULL;
1117 * __skb_queue_purge - empty a list
1118 * @list: list to empty
1120 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1121 * the list and one reference dropped. This function does not take the
1122 * list lock and the caller must hold the relevant locks to use it.
1124 extern void skb_queue_purge(struct sk_buff_head *list);
1125 static inline void __skb_queue_purge(struct sk_buff_head *list)
1127 struct sk_buff *skb;
1128 while ((skb = __skb_dequeue(list)) != NULL)
1133 * __dev_alloc_skb - allocate an skbuff for receiving
1134 * @length: length to allocate
1135 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1137 * Allocate a new &sk_buff and assign it a usage count of one. The
1138 * buffer has unspecified headroom built in. Users should allocate
1139 * the headroom they think they need without accounting for the
1140 * built in space. The built in space is used for optimisations.
1142 * %NULL is returned if there is no free memory.
1144 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1147 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1149 skb_reserve(skb, NET_SKB_PAD);
1154 * dev_alloc_skb - allocate an skbuff for receiving
1155 * @length: length to allocate
1157 * Allocate a new &sk_buff and assign it a usage count of one. The
1158 * buffer has unspecified headroom built in. Users should allocate
1159 * the headroom they think they need without accounting for the
1160 * built in space. The built in space is used for optimisations.
1162 * %NULL is returned if there is no free memory. Although this function
1163 * allocates memory it can be called from an interrupt.
1165 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1167 return __dev_alloc_skb(length, GFP_ATOMIC);
1170 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1171 unsigned int length, gfp_t gfp_mask);
1174 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1175 * @dev: network device to receive on
1176 * @length: length to allocate
1178 * Allocate a new &sk_buff and assign it a usage count of one. The
1179 * buffer has unspecified headroom built in. Users should allocate
1180 * the headroom they think they need without accounting for the
1181 * built in space. The built in space is used for optimisations.
1183 * %NULL is returned if there is no free memory. Although this function
1184 * allocates memory it can be called from an interrupt.
1186 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1187 unsigned int length)
1189 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1193 * skb_cow - copy header of skb when it is required
1194 * @skb: buffer to cow
1195 * @headroom: needed headroom
1197 * If the skb passed lacks sufficient headroom or its data part
1198 * is shared, data is reallocated. If reallocation fails, an error
1199 * is returned and original skb is not changed.
1201 * The result is skb with writable area skb->head...skb->tail
1202 * and at least @headroom of space at head.
1204 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1206 int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) -
1212 if (delta || skb_cloned(skb))
1213 return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) &
1214 ~(NET_SKB_PAD-1), 0, GFP_ATOMIC);
1219 * skb_padto - pad an skbuff up to a minimal size
1220 * @skb: buffer to pad
1221 * @len: minimal length
1223 * Pads up a buffer to ensure the trailing bytes exist and are
1224 * blanked. If the buffer already contains sufficient data it
1225 * is untouched. Otherwise it is extended. Returns zero on
1226 * success. The skb is freed on error.
1229 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1231 unsigned int size = skb->len;
1232 if (likely(size >= len))
1234 return skb_pad(skb, len-size);
1237 static inline int skb_add_data(struct sk_buff *skb,
1238 char __user *from, int copy)
1240 const int off = skb->len;
1242 if (skb->ip_summed == CHECKSUM_NONE) {
1244 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1247 skb->csum = csum_block_add(skb->csum, csum, off);
1250 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1253 __skb_trim(skb, off);
1257 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1258 struct page *page, int off)
1261 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1263 return page == frag->page &&
1264 off == frag->page_offset + frag->size;
1269 static inline int __skb_linearize(struct sk_buff *skb)
1271 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1275 * skb_linearize - convert paged skb to linear one
1276 * @skb: buffer to linarize
1278 * If there is no free memory -ENOMEM is returned, otherwise zero
1279 * is returned and the old skb data released.
1281 static inline int skb_linearize(struct sk_buff *skb)
1283 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1287 * skb_linearize_cow - make sure skb is linear and writable
1288 * @skb: buffer to process
1290 * If there is no free memory -ENOMEM is returned, otherwise zero
1291 * is returned and the old skb data released.
1293 static inline int skb_linearize_cow(struct sk_buff *skb)
1295 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1296 __skb_linearize(skb) : 0;
1300 * skb_postpull_rcsum - update checksum for received skb after pull
1301 * @skb: buffer to update
1302 * @start: start of data before pull
1303 * @len: length of data pulled
1305 * After doing a pull on a received packet, you need to call this to
1306 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1307 * CHECKSUM_NONE so that it can be recomputed from scratch.
1310 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1311 const void *start, unsigned int len)
1313 if (skb->ip_summed == CHECKSUM_COMPLETE)
1314 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1317 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1320 * pskb_trim_rcsum - trim received skb and update checksum
1321 * @skb: buffer to trim
1324 * This is exactly the same as pskb_trim except that it ensures the
1325 * checksum of received packets are still valid after the operation.
1328 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1330 if (likely(len >= skb->len))
1332 if (skb->ip_summed == CHECKSUM_COMPLETE)
1333 skb->ip_summed = CHECKSUM_NONE;
1334 return __pskb_trim(skb, len);
1337 #define skb_queue_walk(queue, skb) \
1338 for (skb = (queue)->next; \
1339 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1342 #define skb_queue_reverse_walk(queue, skb) \
1343 for (skb = (queue)->prev; \
1344 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1348 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1349 int noblock, int *err);
1350 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1351 struct poll_table_struct *wait);
1352 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1353 int offset, struct iovec *to,
1355 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1358 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1359 extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1360 unsigned int flags);
1361 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1362 int len, __wsum csum);
1363 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1365 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1366 void *from, int len);
1367 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1368 int offset, u8 *to, int len,
1370 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1371 extern void skb_split(struct sk_buff *skb,
1372 struct sk_buff *skb1, const u32 len);
1374 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1376 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1377 int len, void *buffer)
1379 int hlen = skb_headlen(skb);
1381 if (hlen - offset >= len)
1382 return skb->data + offset;
1384 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1390 extern void skb_init(void);
1391 extern void skb_add_mtu(int mtu);
1394 * skb_get_timestamp - get timestamp from a skb
1395 * @skb: skb to get stamp from
1396 * @stamp: pointer to struct timeval to store stamp in
1398 * Timestamps are stored in the skb as offsets to a base timestamp.
1399 * This function converts the offset back to a struct timeval and stores
1402 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1404 *stamp = ktime_to_timeval(skb->tstamp);
1407 static inline void __net_timestamp(struct sk_buff *skb)
1409 skb->tstamp = ktime_get_real();
1413 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
1414 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
1417 * skb_checksum_complete - Calculate checksum of an entire packet
1418 * @skb: packet to process
1420 * This function calculates the checksum over the entire packet plus
1421 * the value of skb->csum. The latter can be used to supply the
1422 * checksum of a pseudo header as used by TCP/UDP. It returns the
1425 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1426 * this function can be used to verify that checksum on received
1427 * packets. In that case the function should return zero if the
1428 * checksum is correct. In particular, this function will return zero
1429 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1430 * hardware has already verified the correctness of the checksum.
1432 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1434 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1435 __skb_checksum_complete(skb);
1438 #ifdef CONFIG_NETFILTER
1439 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1441 if (nfct && atomic_dec_and_test(&nfct->use))
1442 nfct->destroy(nfct);
1444 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1447 atomic_inc(&nfct->use);
1449 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1450 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1453 atomic_inc(&skb->users);
1455 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1461 #ifdef CONFIG_BRIDGE_NETFILTER
1462 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1464 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1467 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1470 atomic_inc(&nf_bridge->use);
1472 #endif /* CONFIG_BRIDGE_NETFILTER */
1473 static inline void nf_reset(struct sk_buff *skb)
1475 nf_conntrack_put(skb->nfct);
1477 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1478 nf_conntrack_put_reasm(skb->nfct_reasm);
1479 skb->nfct_reasm = NULL;
1481 #ifdef CONFIG_BRIDGE_NETFILTER
1482 nf_bridge_put(skb->nf_bridge);
1483 skb->nf_bridge = NULL;
1487 #else /* CONFIG_NETFILTER */
1488 static inline void nf_reset(struct sk_buff *skb) {}
1489 #endif /* CONFIG_NETFILTER */
1491 #ifdef CONFIG_NETWORK_SECMARK
1492 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1494 to->secmark = from->secmark;
1497 static inline void skb_init_secmark(struct sk_buff *skb)
1502 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1505 static inline void skb_init_secmark(struct sk_buff *skb)
1509 static inline int skb_is_gso(const struct sk_buff *skb)
1511 return skb_shinfo(skb)->gso_size;
1514 #endif /* __KERNEL__ */
1515 #endif /* _LINUX_SKBUFF_H */