2 * Definitions for the 'struct sk_buff' memory handlers.
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/kernel.h>
18 #include <linux/compiler.h>
19 #include <linux/time.h>
20 #include <linux/cache.h>
22 #include <asm/atomic.h>
23 #include <asm/types.h>
24 #include <linux/spinlock.h>
26 #include <linux/highmem.h>
27 #include <linux/poll.h>
28 #include <linux/net.h>
29 #include <linux/textsearch.h>
30 #include <net/checksum.h>
31 #include <linux/dmaengine.h>
33 #define HAVE_ALLOC_SKB /* For the drivers to know */
34 #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
36 #define CHECKSUM_NONE 0
38 #define CHECKSUM_UNNECESSARY 2
40 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
41 ~(SMP_CACHE_BYTES - 1))
42 #define SKB_MAX_ORDER(X, ORDER) (((PAGE_SIZE << (ORDER)) - (X) - \
43 sizeof(struct skb_shared_info)) & \
44 ~(SMP_CACHE_BYTES - 1))
45 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
46 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
48 /* A. Checksumming of received packets by device.
50 * NONE: device failed to checksum this packet.
51 * skb->csum is undefined.
53 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
54 * skb->csum is undefined.
55 * It is bad option, but, unfortunately, many of vendors do this.
56 * Apparently with secret goal to sell you new device, when you
57 * will add new protocol to your host. F.e. IPv6. 8)
59 * HW: the most generic way. Device supplied checksum of _all_
60 * the packet as seen by netif_rx in skb->csum.
61 * NOTE: Even if device supports only some protocols, but
62 * is able to produce some skb->csum, it MUST use HW,
65 * B. Checksumming on output.
67 * NONE: skb is checksummed by protocol or csum is not required.
69 * HW: device is required to csum packet as seen by hard_start_xmit
70 * from skb->h.raw to the end and to record the checksum
71 * at skb->h.raw+skb->csum.
73 * Device must show its capabilities in dev->features, set
74 * at device setup time.
75 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
77 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
78 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
79 * TCP/UDP over IPv4. Sigh. Vendors like this
80 * way by an unknown reason. Though, see comment above
81 * about CHECKSUM_UNNECESSARY. 8)
83 * Any questions? No questions, good. --ANK
88 #ifdef CONFIG_NETFILTER
91 void (*destroy)(struct nf_conntrack *);
94 #ifdef CONFIG_BRIDGE_NETFILTER
95 struct nf_bridge_info {
97 struct net_device *physindev;
98 struct net_device *physoutdev;
99 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
100 struct net_device *netoutdev;
103 unsigned long data[32 / sizeof(unsigned long)];
109 struct sk_buff_head {
110 /* These two members must be first. */
111 struct sk_buff *next;
112 struct sk_buff *prev;
120 /* To allow 64K frame to be packed as single skb without frag_list */
121 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
123 typedef struct skb_frag_struct skb_frag_t;
125 struct skb_frag_struct {
131 /* This data is invariant across clones and lives at
132 * the end of the header data, ie. at skb->end.
134 struct skb_shared_info {
136 unsigned short nr_frags;
137 unsigned short gso_size;
138 /* Warning: this field is not always filled in (UFO)! */
139 unsigned short gso_segs;
140 unsigned short gso_type;
141 unsigned int ip6_frag_id;
142 struct sk_buff *frag_list;
143 skb_frag_t frags[MAX_SKB_FRAGS];
146 /* We divide dataref into two halves. The higher 16 bits hold references
147 * to the payload part of skb->data. The lower 16 bits hold references to
148 * the entire skb->data. It is up to the users of the skb to agree on
149 * where the payload starts.
151 * All users must obey the rule that the skb->data reference count must be
152 * greater than or equal to the payload reference count.
154 * Holding a reference to the payload part means that the user does not
155 * care about modifications to the header part of skb->data.
157 #define SKB_DATAREF_SHIFT 16
158 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
167 SKB_FCLONE_UNAVAILABLE,
173 SKB_GSO_TCPV4 = 1 << 0,
174 SKB_GSO_UDPV4 = 1 << 1,
176 /* This indicates the skb is from an untrusted source. */
177 SKB_GSO_DODGY = 1 << 2,
179 /* This indicates the tcp segment has CWR set. */
180 SKB_GSO_TCPV4_ECN = 1 << 3,
184 * struct sk_buff - socket buffer
185 * @next: Next buffer in list
186 * @prev: Previous buffer in list
187 * @sk: Socket we are owned by
188 * @tstamp: Time we arrived
189 * @dev: Device we arrived on/are leaving by
190 * @input_dev: Device we arrived on
191 * @h: Transport layer header
192 * @nh: Network layer header
193 * @mac: Link layer header
194 * @dst: destination entry
195 * @sp: the security path, used for xfrm
196 * @cb: Control buffer. Free for use by every layer. Put private vars here
197 * @len: Length of actual data
198 * @data_len: Data length
199 * @mac_len: Length of link layer header
201 * @local_df: allow local fragmentation
202 * @cloned: Head may be cloned (check refcnt to be sure)
203 * @nohdr: Payload reference only, must not modify header
204 * @pkt_type: Packet class
205 * @fclone: skbuff clone status
206 * @ip_summed: Driver fed us an IP checksum
207 * @priority: Packet queueing priority
208 * @users: User count - see {datagram,tcp}.c
209 * @protocol: Packet protocol from driver
210 * @truesize: Buffer size
211 * @head: Head of buffer
212 * @data: Data head pointer
213 * @tail: Tail pointer
215 * @destructor: Destruct function
216 * @nfmark: Can be used for communication between hooks
217 * @nfct: Associated connection, if any
218 * @ipvs_property: skbuff is owned by ipvs
219 * @nfctinfo: Relationship of this skb to the connection
220 * @nfct_reasm: netfilter conntrack re-assembly pointer
221 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
222 * @tc_index: Traffic control index
223 * @tc_verd: traffic control verdict
224 * @dma_cookie: a cookie to one of several possible DMA operations
225 * done by skb DMA functions
226 * @secmark: security marking
230 /* These two members must be first. */
231 struct sk_buff *next;
232 struct sk_buff *prev;
235 struct skb_timeval tstamp;
236 struct net_device *dev;
237 struct net_device *input_dev;
242 struct icmphdr *icmph;
243 struct igmphdr *igmph;
245 struct ipv6hdr *ipv6h;
251 struct ipv6hdr *ipv6h;
260 struct dst_entry *dst;
264 * This is the control buffer. It is free to use for every
265 * layer. Please put your private variables there. If you
266 * want to keep them across layers you have to do a skb_clone()
267 * first. This is owned by whoever has the skb queued ATM.
286 void (*destructor)(struct sk_buff *skb);
287 #ifdef CONFIG_NETFILTER
288 struct nf_conntrack *nfct;
289 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
290 struct sk_buff *nfct_reasm;
292 #ifdef CONFIG_BRIDGE_NETFILTER
293 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
311 /* These elements must be at the end, see alloc_skb() for details. */
312 unsigned int truesize;
322 * Handling routines are only of interest to the kernel
324 #include <linux/slab.h>
326 #include <asm/system.h>
328 extern void kfree_skb(struct sk_buff *skb);
329 extern void __kfree_skb(struct sk_buff *skb);
330 extern struct sk_buff *__alloc_skb(unsigned int size,
331 gfp_t priority, int fclone);
332 static inline struct sk_buff *alloc_skb(unsigned int size,
335 return __alloc_skb(size, priority, 0);
338 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
341 return __alloc_skb(size, priority, 1);
344 extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
347 extern void kfree_skbmem(struct sk_buff *skb);
348 extern struct sk_buff *skb_clone(struct sk_buff *skb,
350 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
352 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
354 extern int pskb_expand_head(struct sk_buff *skb,
355 int nhead, int ntail,
357 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
358 unsigned int headroom);
359 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
360 int newheadroom, int newtailroom,
362 extern int skb_pad(struct sk_buff *skb, int pad);
363 #define dev_kfree_skb(a) kfree_skb(a)
364 extern void skb_over_panic(struct sk_buff *skb, int len,
366 extern void skb_under_panic(struct sk_buff *skb, int len,
368 extern void skb_truesize_bug(struct sk_buff *skb);
370 static inline void skb_truesize_check(struct sk_buff *skb)
372 if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len))
373 skb_truesize_bug(skb);
376 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
377 int getfrag(void *from, char *to, int offset,
378 int len,int odd, struct sk_buff *skb),
379 void *from, int length);
386 __u32 stepped_offset;
387 struct sk_buff *root_skb;
388 struct sk_buff *cur_skb;
392 extern void skb_prepare_seq_read(struct sk_buff *skb,
393 unsigned int from, unsigned int to,
394 struct skb_seq_state *st);
395 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
396 struct skb_seq_state *st);
397 extern void skb_abort_seq_read(struct skb_seq_state *st);
399 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
400 unsigned int to, struct ts_config *config,
401 struct ts_state *state);
404 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
407 * skb_queue_empty - check if a queue is empty
410 * Returns true if the queue is empty, false otherwise.
412 static inline int skb_queue_empty(const struct sk_buff_head *list)
414 return list->next == (struct sk_buff *)list;
418 * skb_get - reference buffer
419 * @skb: buffer to reference
421 * Makes another reference to a socket buffer and returns a pointer
424 static inline struct sk_buff *skb_get(struct sk_buff *skb)
426 atomic_inc(&skb->users);
431 * If users == 1, we are the only owner and are can avoid redundant
436 * skb_cloned - is the buffer a clone
437 * @skb: buffer to check
439 * Returns true if the buffer was generated with skb_clone() and is
440 * one of multiple shared copies of the buffer. Cloned buffers are
441 * shared data so must not be written to under normal circumstances.
443 static inline int skb_cloned(const struct sk_buff *skb)
445 return skb->cloned &&
446 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
450 * skb_header_cloned - is the header a clone
451 * @skb: buffer to check
453 * Returns true if modifying the header part of the buffer requires
454 * the data to be copied.
456 static inline int skb_header_cloned(const struct sk_buff *skb)
463 dataref = atomic_read(&skb_shinfo(skb)->dataref);
464 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
469 * skb_header_release - release reference to header
470 * @skb: buffer to operate on
472 * Drop a reference to the header part of the buffer. This is done
473 * by acquiring a payload reference. You must not read from the header
474 * part of skb->data after this.
476 static inline void skb_header_release(struct sk_buff *skb)
480 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
484 * skb_shared - is the buffer shared
485 * @skb: buffer to check
487 * Returns true if more than one person has a reference to this
490 static inline int skb_shared(const struct sk_buff *skb)
492 return atomic_read(&skb->users) != 1;
496 * skb_share_check - check if buffer is shared and if so clone it
497 * @skb: buffer to check
498 * @pri: priority for memory allocation
500 * If the buffer is shared the buffer is cloned and the old copy
501 * drops a reference. A new clone with a single reference is returned.
502 * If the buffer is not shared the original buffer is returned. When
503 * being called from interrupt status or with spinlocks held pri must
506 * NULL is returned on a memory allocation failure.
508 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
511 might_sleep_if(pri & __GFP_WAIT);
512 if (skb_shared(skb)) {
513 struct sk_buff *nskb = skb_clone(skb, pri);
521 * Copy shared buffers into a new sk_buff. We effectively do COW on
522 * packets to handle cases where we have a local reader and forward
523 * and a couple of other messy ones. The normal one is tcpdumping
524 * a packet thats being forwarded.
528 * skb_unshare - make a copy of a shared buffer
529 * @skb: buffer to check
530 * @pri: priority for memory allocation
532 * If the socket buffer is a clone then this function creates a new
533 * copy of the data, drops a reference count on the old copy and returns
534 * the new copy with the reference count at 1. If the buffer is not a clone
535 * the original buffer is returned. When called with a spinlock held or
536 * from interrupt state @pri must be %GFP_ATOMIC
538 * %NULL is returned on a memory allocation failure.
540 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
543 might_sleep_if(pri & __GFP_WAIT);
544 if (skb_cloned(skb)) {
545 struct sk_buff *nskb = skb_copy(skb, pri);
546 kfree_skb(skb); /* Free our shared copy */
554 * @list_: list to peek at
556 * Peek an &sk_buff. Unlike most other operations you _MUST_
557 * be careful with this one. A peek leaves the buffer on the
558 * list and someone else may run off with it. You must hold
559 * the appropriate locks or have a private queue to do this.
561 * Returns %NULL for an empty list or a pointer to the head element.
562 * The reference count is not incremented and the reference is therefore
563 * volatile. Use with caution.
565 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
567 struct sk_buff *list = ((struct sk_buff *)list_)->next;
568 if (list == (struct sk_buff *)list_)
575 * @list_: list to peek at
577 * Peek an &sk_buff. Unlike most other operations you _MUST_
578 * be careful with this one. A peek leaves the buffer on the
579 * list and someone else may run off with it. You must hold
580 * the appropriate locks or have a private queue to do this.
582 * Returns %NULL for an empty list or a pointer to the tail element.
583 * The reference count is not incremented and the reference is therefore
584 * volatile. Use with caution.
586 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
588 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
589 if (list == (struct sk_buff *)list_)
595 * skb_queue_len - get queue length
596 * @list_: list to measure
598 * Return the length of an &sk_buff queue.
600 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
605 static inline void skb_queue_head_init(struct sk_buff_head *list)
607 spin_lock_init(&list->lock);
608 list->prev = list->next = (struct sk_buff *)list;
613 * Insert an sk_buff at the start of a list.
615 * The "__skb_xxxx()" functions are the non-atomic ones that
616 * can only be called with interrupts disabled.
620 * __skb_queue_after - queue a buffer at the list head
622 * @prev: place after this buffer
623 * @newsk: buffer to queue
625 * Queue a buffer int the middle of a list. This function takes no locks
626 * and you must therefore hold required locks before calling it.
628 * A buffer cannot be placed on two lists at the same time.
630 static inline void __skb_queue_after(struct sk_buff_head *list,
631 struct sk_buff *prev,
632 struct sk_buff *newsk)
634 struct sk_buff *next;
640 next->prev = prev->next = newsk;
644 * __skb_queue_head - queue a buffer at the list head
646 * @newsk: buffer to queue
648 * Queue a buffer at the start of a list. This function takes no locks
649 * and you must therefore hold required locks before calling it.
651 * A buffer cannot be placed on two lists at the same time.
653 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
654 static inline void __skb_queue_head(struct sk_buff_head *list,
655 struct sk_buff *newsk)
657 __skb_queue_after(list, (struct sk_buff *)list, newsk);
661 * __skb_queue_tail - queue a buffer at the list tail
663 * @newsk: buffer to queue
665 * Queue a buffer at the end of a list. This function takes no locks
666 * and you must therefore hold required locks before calling it.
668 * A buffer cannot be placed on two lists at the same time.
670 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
671 static inline void __skb_queue_tail(struct sk_buff_head *list,
672 struct sk_buff *newsk)
674 struct sk_buff *prev, *next;
677 next = (struct sk_buff *)list;
681 next->prev = prev->next = newsk;
686 * __skb_dequeue - remove from the head of the queue
687 * @list: list to dequeue from
689 * Remove the head of the list. This function does not take any locks
690 * so must be used with appropriate locks held only. The head item is
691 * returned or %NULL if the list is empty.
693 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
694 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
696 struct sk_buff *next, *prev, *result;
698 prev = (struct sk_buff *) list;
707 result->next = result->prev = NULL;
714 * Insert a packet on a list.
716 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
717 static inline void __skb_insert(struct sk_buff *newsk,
718 struct sk_buff *prev, struct sk_buff *next,
719 struct sk_buff_head *list)
723 next->prev = prev->next = newsk;
728 * Place a packet after a given packet in a list.
730 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
731 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
733 __skb_insert(newsk, old, old->next, list);
737 * remove sk_buff from list. _Must_ be called atomically, and with
740 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
741 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
743 struct sk_buff *next, *prev;
748 skb->next = skb->prev = NULL;
754 /* XXX: more streamlined implementation */
757 * __skb_dequeue_tail - remove from the tail of the queue
758 * @list: list to dequeue from
760 * Remove the tail of the list. This function does not take any locks
761 * so must be used with appropriate locks held only. The tail item is
762 * returned or %NULL if the list is empty.
764 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
765 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
767 struct sk_buff *skb = skb_peek_tail(list);
769 __skb_unlink(skb, list);
774 static inline int skb_is_nonlinear(const struct sk_buff *skb)
776 return skb->data_len;
779 static inline unsigned int skb_headlen(const struct sk_buff *skb)
781 return skb->len - skb->data_len;
784 static inline int skb_pagelen(const struct sk_buff *skb)
788 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
789 len += skb_shinfo(skb)->frags[i].size;
790 return len + skb_headlen(skb);
793 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
794 struct page *page, int off, int size)
796 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
799 frag->page_offset = off;
801 skb_shinfo(skb)->nr_frags = i + 1;
804 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
805 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
806 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
809 * Add data to an sk_buff
811 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
813 unsigned char *tmp = skb->tail;
814 SKB_LINEAR_ASSERT(skb);
821 * skb_put - add data to a buffer
822 * @skb: buffer to use
823 * @len: amount of data to add
825 * This function extends the used data area of the buffer. If this would
826 * exceed the total buffer size the kernel will panic. A pointer to the
827 * first byte of the extra data is returned.
829 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
831 unsigned char *tmp = skb->tail;
832 SKB_LINEAR_ASSERT(skb);
835 if (unlikely(skb->tail>skb->end))
836 skb_over_panic(skb, len, current_text_addr());
840 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
848 * skb_push - add data to the start of a buffer
849 * @skb: buffer to use
850 * @len: amount of data to add
852 * This function extends the used data area of the buffer at the buffer
853 * start. If this would exceed the total buffer headroom the kernel will
854 * panic. A pointer to the first byte of the extra data is returned.
856 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
860 if (unlikely(skb->data<skb->head))
861 skb_under_panic(skb, len, current_text_addr());
865 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
868 BUG_ON(skb->len < skb->data_len);
869 return skb->data += len;
873 * skb_pull - remove data from the start of a buffer
874 * @skb: buffer to use
875 * @len: amount of data to remove
877 * This function removes data from the start of a buffer, returning
878 * the memory to the headroom. A pointer to the next data in the buffer
879 * is returned. Once the data has been pulled future pushes will overwrite
882 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
884 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
887 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
889 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
891 if (len > skb_headlen(skb) &&
892 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
895 return skb->data += len;
898 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
900 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
903 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
905 if (likely(len <= skb_headlen(skb)))
907 if (unlikely(len > skb->len))
909 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
913 * skb_headroom - bytes at buffer head
914 * @skb: buffer to check
916 * Return the number of bytes of free space at the head of an &sk_buff.
918 static inline int skb_headroom(const struct sk_buff *skb)
920 return skb->data - skb->head;
924 * skb_tailroom - bytes at buffer end
925 * @skb: buffer to check
927 * Return the number of bytes of free space at the tail of an sk_buff
929 static inline int skb_tailroom(const struct sk_buff *skb)
931 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
935 * skb_reserve - adjust headroom
936 * @skb: buffer to alter
937 * @len: bytes to move
939 * Increase the headroom of an empty &sk_buff by reducing the tail
940 * room. This is only allowed for an empty buffer.
942 static inline void skb_reserve(struct sk_buff *skb, int len)
949 * CPUs often take a performance hit when accessing unaligned memory
950 * locations. The actual performance hit varies, it can be small if the
951 * hardware handles it or large if we have to take an exception and fix it
954 * Since an ethernet header is 14 bytes network drivers often end up with
955 * the IP header at an unaligned offset. The IP header can be aligned by
956 * shifting the start of the packet by 2 bytes. Drivers should do this
959 * skb_reserve(NET_IP_ALIGN);
961 * The downside to this alignment of the IP header is that the DMA is now
962 * unaligned. On some architectures the cost of an unaligned DMA is high
963 * and this cost outweighs the gains made by aligning the IP header.
965 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
969 #define NET_IP_ALIGN 2
973 * The networking layer reserves some headroom in skb data (via
974 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
975 * the header has to grow. In the default case, if the header has to grow
976 * 16 bytes or less we avoid the reallocation.
978 * Unfortunately this headroom changes the DMA alignment of the resulting
979 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
980 * on some architectures. An architecture can override this value,
981 * perhaps setting it to a cacheline in size (since that will maintain
982 * cacheline alignment of the DMA). It must be a power of 2.
984 * Various parts of the networking layer expect at least 16 bytes of
985 * headroom, you should not reduce this.
988 #define NET_SKB_PAD 16
991 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
993 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
995 if (unlikely(skb->data_len)) {
1000 skb->tail = skb->data + len;
1004 * skb_trim - remove end from a buffer
1005 * @skb: buffer to alter
1008 * Cut the length of a buffer down by removing data from the tail. If
1009 * the buffer is already under the length specified it is not modified.
1010 * The skb must be linear.
1012 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
1015 __skb_trim(skb, len);
1019 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1022 return ___pskb_trim(skb, len);
1023 __skb_trim(skb, len);
1027 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1029 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1033 * skb_orphan - orphan a buffer
1034 * @skb: buffer to orphan
1036 * If a buffer currently has an owner then we call the owner's
1037 * destructor function and make the @skb unowned. The buffer continues
1038 * to exist but is no longer charged to its former owner.
1040 static inline void skb_orphan(struct sk_buff *skb)
1042 if (skb->destructor)
1043 skb->destructor(skb);
1044 skb->destructor = NULL;
1049 * __skb_queue_purge - empty a list
1050 * @list: list to empty
1052 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1053 * the list and one reference dropped. This function does not take the
1054 * list lock and the caller must hold the relevant locks to use it.
1056 extern void skb_queue_purge(struct sk_buff_head *list);
1057 static inline void __skb_queue_purge(struct sk_buff_head *list)
1059 struct sk_buff *skb;
1060 while ((skb = __skb_dequeue(list)) != NULL)
1064 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
1066 * __dev_alloc_skb - allocate an skbuff for sending
1067 * @length: length to allocate
1068 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1070 * Allocate a new &sk_buff and assign it a usage count of one. The
1071 * buffer has unspecified headroom built in. Users should allocate
1072 * the headroom they think they need without accounting for the
1073 * built in space. The built in space is used for optimisations.
1075 * %NULL is returned in there is no free memory.
1077 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1080 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1082 skb_reserve(skb, NET_SKB_PAD);
1086 extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
1090 * dev_alloc_skb - allocate an skbuff for sending
1091 * @length: length to allocate
1093 * Allocate a new &sk_buff and assign it a usage count of one. The
1094 * buffer has unspecified headroom built in. Users should allocate
1095 * the headroom they think they need without accounting for the
1096 * built in space. The built in space is used for optimisations.
1098 * %NULL is returned in there is no free memory. Although this function
1099 * allocates memory it can be called from an interrupt.
1101 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1103 return __dev_alloc_skb(length, GFP_ATOMIC);
1107 * skb_cow - copy header of skb when it is required
1108 * @skb: buffer to cow
1109 * @headroom: needed headroom
1111 * If the skb passed lacks sufficient headroom or its data part
1112 * is shared, data is reallocated. If reallocation fails, an error
1113 * is returned and original skb is not changed.
1115 * The result is skb with writable area skb->head...skb->tail
1116 * and at least @headroom of space at head.
1118 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1120 int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) -
1126 if (delta || skb_cloned(skb))
1127 return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) &
1128 ~(NET_SKB_PAD-1), 0, GFP_ATOMIC);
1133 * skb_padto - pad an skbuff up to a minimal size
1134 * @skb: buffer to pad
1135 * @len: minimal length
1137 * Pads up a buffer to ensure the trailing bytes exist and are
1138 * blanked. If the buffer already contains sufficient data it
1139 * is untouched. Otherwise it is extended. Returns zero on
1140 * success. The skb is freed on error.
1143 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1145 unsigned int size = skb->len;
1146 if (likely(size >= len))
1148 return skb_pad(skb, len-size);
1151 static inline int skb_add_data(struct sk_buff *skb,
1152 char __user *from, int copy)
1154 const int off = skb->len;
1156 if (skb->ip_summed == CHECKSUM_NONE) {
1158 unsigned int csum = csum_and_copy_from_user(from,
1162 skb->csum = csum_block_add(skb->csum, csum, off);
1165 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1168 __skb_trim(skb, off);
1172 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1173 struct page *page, int off)
1176 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1178 return page == frag->page &&
1179 off == frag->page_offset + frag->size;
1184 static inline int __skb_linearize(struct sk_buff *skb)
1186 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1190 * skb_linearize - convert paged skb to linear one
1191 * @skb: buffer to linarize
1193 * If there is no free memory -ENOMEM is returned, otherwise zero
1194 * is returned and the old skb data released.
1196 static inline int skb_linearize(struct sk_buff *skb)
1198 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1202 * skb_linearize_cow - make sure skb is linear and writable
1203 * @skb: buffer to process
1205 * If there is no free memory -ENOMEM is returned, otherwise zero
1206 * is returned and the old skb data released.
1208 static inline int skb_linearize_cow(struct sk_buff *skb)
1210 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1211 __skb_linearize(skb) : 0;
1215 * skb_postpull_rcsum - update checksum for received skb after pull
1216 * @skb: buffer to update
1217 * @start: start of data before pull
1218 * @len: length of data pulled
1220 * After doing a pull on a received packet, you need to call this to
1221 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1222 * so that it can be recomputed from scratch.
1225 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1226 const void *start, unsigned int len)
1228 if (skb->ip_summed == CHECKSUM_HW)
1229 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1232 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1235 * pskb_trim_rcsum - trim received skb and update checksum
1236 * @skb: buffer to trim
1239 * This is exactly the same as pskb_trim except that it ensures the
1240 * checksum of received packets are still valid after the operation.
1243 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1245 if (likely(len >= skb->len))
1247 if (skb->ip_summed == CHECKSUM_HW)
1248 skb->ip_summed = CHECKSUM_NONE;
1249 return __pskb_trim(skb, len);
1252 static inline void *kmap_skb_frag(const skb_frag_t *frag)
1254 #ifdef CONFIG_HIGHMEM
1259 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1262 static inline void kunmap_skb_frag(void *vaddr)
1264 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1265 #ifdef CONFIG_HIGHMEM
1270 #define skb_queue_walk(queue, skb) \
1271 for (skb = (queue)->next; \
1272 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1275 #define skb_queue_reverse_walk(queue, skb) \
1276 for (skb = (queue)->prev; \
1277 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1281 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1282 int noblock, int *err);
1283 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1284 struct poll_table_struct *wait);
1285 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1286 int offset, struct iovec *to,
1288 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1291 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1292 extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1293 unsigned int flags);
1294 extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1295 int len, unsigned int csum);
1296 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1298 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1299 void *from, int len);
1300 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1301 int offset, u8 *to, int len,
1303 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1304 extern void skb_split(struct sk_buff *skb,
1305 struct sk_buff *skb1, const u32 len);
1307 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1309 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1310 int len, void *buffer)
1312 int hlen = skb_headlen(skb);
1314 if (hlen - offset >= len)
1315 return skb->data + offset;
1317 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1323 extern void skb_init(void);
1324 extern void skb_add_mtu(int mtu);
1327 * skb_get_timestamp - get timestamp from a skb
1328 * @skb: skb to get stamp from
1329 * @stamp: pointer to struct timeval to store stamp in
1331 * Timestamps are stored in the skb as offsets to a base timestamp.
1332 * This function converts the offset back to a struct timeval and stores
1335 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1337 stamp->tv_sec = skb->tstamp.off_sec;
1338 stamp->tv_usec = skb->tstamp.off_usec;
1342 * skb_set_timestamp - set timestamp of a skb
1343 * @skb: skb to set stamp of
1344 * @stamp: pointer to struct timeval to get stamp from
1346 * Timestamps are stored in the skb as offsets to a base timestamp.
1347 * This function converts a struct timeval to an offset and stores
1350 static inline void skb_set_timestamp(struct sk_buff *skb, const struct timeval *stamp)
1352 skb->tstamp.off_sec = stamp->tv_sec;
1353 skb->tstamp.off_usec = stamp->tv_usec;
1356 extern void __net_timestamp(struct sk_buff *skb);
1358 extern unsigned int __skb_checksum_complete(struct sk_buff *skb);
1361 * skb_checksum_complete - Calculate checksum of an entire packet
1362 * @skb: packet to process
1364 * This function calculates the checksum over the entire packet plus
1365 * the value of skb->csum. The latter can be used to supply the
1366 * checksum of a pseudo header as used by TCP/UDP. It returns the
1369 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1370 * this function can be used to verify that checksum on received
1371 * packets. In that case the function should return zero if the
1372 * checksum is correct. In particular, this function will return zero
1373 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1374 * hardware has already verified the correctness of the checksum.
1376 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1378 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1379 __skb_checksum_complete(skb);
1382 #ifdef CONFIG_NETFILTER
1383 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1385 if (nfct && atomic_dec_and_test(&nfct->use))
1386 nfct->destroy(nfct);
1388 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1391 atomic_inc(&nfct->use);
1393 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1394 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1397 atomic_inc(&skb->users);
1399 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1405 #ifdef CONFIG_BRIDGE_NETFILTER
1406 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1408 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1411 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1414 atomic_inc(&nf_bridge->use);
1416 #endif /* CONFIG_BRIDGE_NETFILTER */
1417 static inline void nf_reset(struct sk_buff *skb)
1419 nf_conntrack_put(skb->nfct);
1421 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1422 nf_conntrack_put_reasm(skb->nfct_reasm);
1423 skb->nfct_reasm = NULL;
1425 #ifdef CONFIG_BRIDGE_NETFILTER
1426 nf_bridge_put(skb->nf_bridge);
1427 skb->nf_bridge = NULL;
1431 #else /* CONFIG_NETFILTER */
1432 static inline void nf_reset(struct sk_buff *skb) {}
1433 #endif /* CONFIG_NETFILTER */
1435 #ifdef CONFIG_NETWORK_SECMARK
1436 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1438 to->secmark = from->secmark;
1441 static inline void skb_init_secmark(struct sk_buff *skb)
1446 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1449 static inline void skb_init_secmark(struct sk_buff *skb)
1453 #endif /* __KERNEL__ */
1454 #endif /* _LINUX_SKBUFF_H */