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_UDP = 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_TCP_ECN = 1 << 3,
182 SKB_GSO_TCPV6 = 1 << 4,
186 * struct sk_buff - socket buffer
187 * @next: Next buffer in list
188 * @prev: Previous buffer in list
189 * @sk: Socket we are owned by
190 * @tstamp: Time we arrived
191 * @dev: Device we arrived on/are leaving by
192 * @input_dev: Device we arrived on
193 * @h: Transport layer header
194 * @nh: Network layer header
195 * @mac: Link layer header
196 * @dst: destination entry
197 * @sp: the security path, used for xfrm
198 * @cb: Control buffer. Free for use by every layer. Put private vars here
199 * @len: Length of actual data
200 * @data_len: Data length
201 * @mac_len: Length of link layer header
203 * @local_df: allow local fragmentation
204 * @cloned: Head may be cloned (check refcnt to be sure)
205 * @nohdr: Payload reference only, must not modify header
206 * @proto_data_valid: Protocol data validated since arriving at localhost
207 * @proto_csum_blank: Protocol csum must be added before leaving localhost
208 * @pkt_type: Packet class
209 * @fclone: skbuff clone status
210 * @ip_summed: Driver fed us an IP checksum
211 * @priority: Packet queueing priority
212 * @users: User count - see {datagram,tcp}.c
213 * @protocol: Packet protocol from driver
214 * @truesize: Buffer size
215 * @head: Head of buffer
216 * @data: Data head pointer
217 * @tail: Tail pointer
219 * @destructor: Destruct function
220 * @nfmark: Can be used for communication between hooks
221 * @nfct: Associated connection, if any
222 * @ipvs_property: skbuff is owned by ipvs
223 * @nfctinfo: Relationship of this skb to the connection
224 * @nfct_reasm: netfilter conntrack re-assembly pointer
225 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
226 * @tc_index: Traffic control index
227 * @tc_verd: traffic control verdict
228 * @dma_cookie: a cookie to one of several possible DMA operations
229 * done by skb DMA functions
230 * @secmark: security marking
234 /* These two members must be first. */
235 struct sk_buff *next;
236 struct sk_buff *prev;
239 struct skb_timeval tstamp;
240 struct net_device *dev;
241 struct net_device *input_dev;
246 struct icmphdr *icmph;
247 struct igmphdr *igmph;
249 struct ipv6hdr *ipv6h;
255 struct ipv6hdr *ipv6h;
264 struct dst_entry *dst;
268 * This is the control buffer. It is free to use for every
269 * layer. Please put your private variables there. If you
270 * want to keep them across layers you have to do a skb_clone()
271 * first. This is owned by whoever has the skb queued ATM.
296 void (*destructor)(struct sk_buff *skb);
297 #ifdef CONFIG_NETFILTER
298 struct nf_conntrack *nfct;
299 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
300 struct sk_buff *nfct_reasm;
302 #ifdef CONFIG_BRIDGE_NETFILTER
303 struct nf_bridge_info *nf_bridge;
306 #endif /* CONFIG_NETFILTER */
307 #ifdef CONFIG_NET_SCHED
308 __u16 tc_index; /* traffic control index */
309 #ifdef CONFIG_NET_CLS_ACT
310 __u16 tc_verd; /* traffic control verdict */
313 #if defined(CONFIG_VNET) || defined(CONFIG_VNET_MODULE)
314 xid_t xid; /* VServer context ID */
316 #ifdef CONFIG_NET_DMA
317 dma_cookie_t dma_cookie;
319 #ifdef CONFIG_NETWORK_SECMARK
323 /* These elements must be at the end, see alloc_skb() for details. */
324 unsigned int truesize;
334 * Handling routines are only of interest to the kernel
336 #include <linux/slab.h>
338 #include <asm/system.h>
340 extern void kfree_skb(struct sk_buff *skb);
341 extern void __kfree_skb(struct sk_buff *skb);
342 extern struct sk_buff *__alloc_skb(unsigned int size,
343 gfp_t priority, int fclone);
344 static inline struct sk_buff *alloc_skb(unsigned int size,
347 return __alloc_skb(size, priority, 0);
350 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
353 return __alloc_skb(size, priority, 1);
356 extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
360 extern void kfree_skbmem(struct sk_buff *skb);
361 extern struct sk_buff *skb_clone(struct sk_buff *skb,
363 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
365 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
367 extern int pskb_expand_head(struct sk_buff *skb,
368 int nhead, int ntail,
370 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
371 unsigned int headroom);
372 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
373 int newheadroom, int newtailroom,
375 extern int skb_pad(struct sk_buff *skb, int pad);
376 #define dev_kfree_skb(a) kfree_skb(a)
377 extern void skb_over_panic(struct sk_buff *skb, int len,
379 extern void skb_under_panic(struct sk_buff *skb, int len,
381 extern void skb_truesize_bug(struct sk_buff *skb);
383 static inline void skb_truesize_check(struct sk_buff *skb)
385 if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len))
386 skb_truesize_bug(skb);
389 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
390 int getfrag(void *from, char *to, int offset,
391 int len,int odd, struct sk_buff *skb),
392 void *from, int length);
399 __u32 stepped_offset;
400 struct sk_buff *root_skb;
401 struct sk_buff *cur_skb;
405 extern void skb_prepare_seq_read(struct sk_buff *skb,
406 unsigned int from, unsigned int to,
407 struct skb_seq_state *st);
408 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
409 struct skb_seq_state *st);
410 extern void skb_abort_seq_read(struct skb_seq_state *st);
412 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
413 unsigned int to, struct ts_config *config,
414 struct ts_state *state);
417 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
420 * skb_queue_empty - check if a queue is empty
423 * Returns true if the queue is empty, false otherwise.
425 static inline int skb_queue_empty(const struct sk_buff_head *list)
427 return list->next == (struct sk_buff *)list;
431 * skb_get - reference buffer
432 * @skb: buffer to reference
434 * Makes another reference to a socket buffer and returns a pointer
437 static inline struct sk_buff *skb_get(struct sk_buff *skb)
439 atomic_inc(&skb->users);
444 * If users == 1, we are the only owner and are can avoid redundant
449 * skb_cloned - is the buffer a clone
450 * @skb: buffer to check
452 * Returns true if the buffer was generated with skb_clone() and is
453 * one of multiple shared copies of the buffer. Cloned buffers are
454 * shared data so must not be written to under normal circumstances.
456 static inline int skb_cloned(const struct sk_buff *skb)
458 return skb->cloned &&
459 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
463 * skb_header_cloned - is the header a clone
464 * @skb: buffer to check
466 * Returns true if modifying the header part of the buffer requires
467 * the data to be copied.
469 static inline int skb_header_cloned(const struct sk_buff *skb)
476 dataref = atomic_read(&skb_shinfo(skb)->dataref);
477 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
482 * skb_header_release - release reference to header
483 * @skb: buffer to operate on
485 * Drop a reference to the header part of the buffer. This is done
486 * by acquiring a payload reference. You must not read from the header
487 * part of skb->data after this.
489 static inline void skb_header_release(struct sk_buff *skb)
493 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
497 * skb_shared - is the buffer shared
498 * @skb: buffer to check
500 * Returns true if more than one person has a reference to this
503 static inline int skb_shared(const struct sk_buff *skb)
505 return atomic_read(&skb->users) != 1;
509 * skb_share_check - check if buffer is shared and if so clone it
510 * @skb: buffer to check
511 * @pri: priority for memory allocation
513 * If the buffer is shared the buffer is cloned and the old copy
514 * drops a reference. A new clone with a single reference is returned.
515 * If the buffer is not shared the original buffer is returned. When
516 * being called from interrupt status or with spinlocks held pri must
519 * NULL is returned on a memory allocation failure.
521 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
524 might_sleep_if(pri & __GFP_WAIT);
525 if (skb_shared(skb)) {
526 struct sk_buff *nskb = skb_clone(skb, pri);
534 * Copy shared buffers into a new sk_buff. We effectively do COW on
535 * packets to handle cases where we have a local reader and forward
536 * and a couple of other messy ones. The normal one is tcpdumping
537 * a packet thats being forwarded.
541 * skb_unshare - make a copy of a shared buffer
542 * @skb: buffer to check
543 * @pri: priority for memory allocation
545 * If the socket buffer is a clone then this function creates a new
546 * copy of the data, drops a reference count on the old copy and returns
547 * the new copy with the reference count at 1. If the buffer is not a clone
548 * the original buffer is returned. When called with a spinlock held or
549 * from interrupt state @pri must be %GFP_ATOMIC
551 * %NULL is returned on a memory allocation failure.
553 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
556 might_sleep_if(pri & __GFP_WAIT);
557 if (skb_cloned(skb)) {
558 struct sk_buff *nskb = skb_copy(skb, pri);
559 kfree_skb(skb); /* Free our shared copy */
567 * @list_: list to peek at
569 * Peek an &sk_buff. Unlike most other operations you _MUST_
570 * be careful with this one. A peek leaves the buffer on the
571 * list and someone else may run off with it. You must hold
572 * the appropriate locks or have a private queue to do this.
574 * Returns %NULL for an empty list or a pointer to the head element.
575 * The reference count is not incremented and the reference is therefore
576 * volatile. Use with caution.
578 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
580 struct sk_buff *list = ((struct sk_buff *)list_)->next;
581 if (list == (struct sk_buff *)list_)
588 * @list_: list to peek at
590 * Peek an &sk_buff. Unlike most other operations you _MUST_
591 * be careful with this one. A peek leaves the buffer on the
592 * list and someone else may run off with it. You must hold
593 * the appropriate locks or have a private queue to do this.
595 * Returns %NULL for an empty list or a pointer to the tail element.
596 * The reference count is not incremented and the reference is therefore
597 * volatile. Use with caution.
599 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
601 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
602 if (list == (struct sk_buff *)list_)
608 * skb_queue_len - get queue length
609 * @list_: list to measure
611 * Return the length of an &sk_buff queue.
613 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
619 * This function creates a split out lock class for each invocation;
620 * this is needed for now since a whole lot of users of the skb-queue
621 * infrastructure in drivers have different locking usage (in hardirq)
622 * than the networking core (in softirq only). In the long run either the
623 * network layer or drivers should need annotation to consolidate the
624 * main types of usage into 3 classes.
626 static inline void skb_queue_head_init(struct sk_buff_head *list)
628 spin_lock_init(&list->lock);
629 list->prev = list->next = (struct sk_buff *)list;
634 * Insert an sk_buff at the start of a list.
636 * The "__skb_xxxx()" functions are the non-atomic ones that
637 * can only be called with interrupts disabled.
641 * __skb_queue_after - queue a buffer at the list head
643 * @prev: place after this buffer
644 * @newsk: buffer to queue
646 * Queue a buffer int the middle of a list. This function takes no locks
647 * and you must therefore hold required locks before calling it.
649 * A buffer cannot be placed on two lists at the same time.
651 static inline void __skb_queue_after(struct sk_buff_head *list,
652 struct sk_buff *prev,
653 struct sk_buff *newsk)
655 struct sk_buff *next;
661 next->prev = prev->next = newsk;
665 * __skb_queue_head - queue a buffer at the list head
667 * @newsk: buffer to queue
669 * Queue a buffer at the start of a list. This function takes no locks
670 * and you must therefore hold required locks before calling it.
672 * A buffer cannot be placed on two lists at the same time.
674 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
675 static inline void __skb_queue_head(struct sk_buff_head *list,
676 struct sk_buff *newsk)
678 __skb_queue_after(list, (struct sk_buff *)list, newsk);
682 * __skb_queue_tail - queue a buffer at the list tail
684 * @newsk: buffer to queue
686 * Queue a buffer at the end of a list. This function takes no locks
687 * and you must therefore hold required locks before calling it.
689 * A buffer cannot be placed on two lists at the same time.
691 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
692 static inline void __skb_queue_tail(struct sk_buff_head *list,
693 struct sk_buff *newsk)
695 struct sk_buff *prev, *next;
698 next = (struct sk_buff *)list;
702 next->prev = prev->next = newsk;
707 * __skb_dequeue - remove from the head of the queue
708 * @list: list to dequeue from
710 * Remove the head of the list. This function does not take any locks
711 * so must be used with appropriate locks held only. The head item is
712 * returned or %NULL if the list is empty.
714 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
715 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
717 struct sk_buff *next, *prev, *result;
719 prev = (struct sk_buff *) list;
728 result->next = result->prev = NULL;
735 * Insert a packet on a list.
737 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
738 static inline void __skb_insert(struct sk_buff *newsk,
739 struct sk_buff *prev, struct sk_buff *next,
740 struct sk_buff_head *list)
744 next->prev = prev->next = newsk;
749 * Place a packet after a given packet in a list.
751 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
752 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
754 __skb_insert(newsk, old, old->next, list);
758 * remove sk_buff from list. _Must_ be called atomically, and with
761 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
762 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
764 struct sk_buff *next, *prev;
769 skb->next = skb->prev = NULL;
775 /* XXX: more streamlined implementation */
778 * __skb_dequeue_tail - remove from the tail of the queue
779 * @list: list to dequeue from
781 * Remove the tail of the list. This function does not take any locks
782 * so must be used with appropriate locks held only. The tail item is
783 * returned or %NULL if the list is empty.
785 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
786 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
788 struct sk_buff *skb = skb_peek_tail(list);
790 __skb_unlink(skb, list);
795 static inline int skb_is_nonlinear(const struct sk_buff *skb)
797 return skb->data_len;
800 static inline unsigned int skb_headlen(const struct sk_buff *skb)
802 return skb->len - skb->data_len;
805 static inline int skb_pagelen(const struct sk_buff *skb)
809 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
810 len += skb_shinfo(skb)->frags[i].size;
811 return len + skb_headlen(skb);
814 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
815 struct page *page, int off, int size)
817 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
820 frag->page_offset = off;
822 skb_shinfo(skb)->nr_frags = i + 1;
825 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
826 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
827 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
830 * Add data to an sk_buff
832 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
834 unsigned char *tmp = skb->tail;
835 SKB_LINEAR_ASSERT(skb);
842 * skb_put - add data to a buffer
843 * @skb: buffer to use
844 * @len: amount of data to add
846 * This function extends the used data area of the buffer. If this would
847 * exceed the total buffer size the kernel will panic. A pointer to the
848 * first byte of the extra data is returned.
850 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
852 unsigned char *tmp = skb->tail;
853 SKB_LINEAR_ASSERT(skb);
856 if (unlikely(skb->tail>skb->end))
857 skb_over_panic(skb, len, current_text_addr());
861 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
869 * skb_push - add data to the start of a buffer
870 * @skb: buffer to use
871 * @len: amount of data to add
873 * This function extends the used data area of the buffer at the buffer
874 * start. If this would exceed the total buffer headroom the kernel will
875 * panic. A pointer to the first byte of the extra data is returned.
877 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
881 if (unlikely(skb->data<skb->head))
882 skb_under_panic(skb, len, current_text_addr());
886 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
889 BUG_ON(skb->len < skb->data_len);
890 return skb->data += len;
894 * skb_pull - remove data from the start of a buffer
895 * @skb: buffer to use
896 * @len: amount of data to remove
898 * This function removes data from the start of a buffer, returning
899 * the memory to the headroom. A pointer to the next data in the buffer
900 * is returned. Once the data has been pulled future pushes will overwrite
903 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
905 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
908 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
910 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
912 if (len > skb_headlen(skb) &&
913 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
916 return skb->data += len;
919 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
921 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
924 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
926 if (likely(len <= skb_headlen(skb)))
928 if (unlikely(len > skb->len))
930 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
934 * skb_headroom - bytes at buffer head
935 * @skb: buffer to check
937 * Return the number of bytes of free space at the head of an &sk_buff.
939 static inline int skb_headroom(const struct sk_buff *skb)
941 return skb->data - skb->head;
945 * skb_tailroom - bytes at buffer end
946 * @skb: buffer to check
948 * Return the number of bytes of free space at the tail of an sk_buff
950 static inline int skb_tailroom(const struct sk_buff *skb)
952 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
956 * skb_reserve - adjust headroom
957 * @skb: buffer to alter
958 * @len: bytes to move
960 * Increase the headroom of an empty &sk_buff by reducing the tail
961 * room. This is only allowed for an empty buffer.
963 static inline void skb_reserve(struct sk_buff *skb, int len)
970 * CPUs often take a performance hit when accessing unaligned memory
971 * locations. The actual performance hit varies, it can be small if the
972 * hardware handles it or large if we have to take an exception and fix it
975 * Since an ethernet header is 14 bytes network drivers often end up with
976 * the IP header at an unaligned offset. The IP header can be aligned by
977 * shifting the start of the packet by 2 bytes. Drivers should do this
980 * skb_reserve(NET_IP_ALIGN);
982 * The downside to this alignment of the IP header is that the DMA is now
983 * unaligned. On some architectures the cost of an unaligned DMA is high
984 * and this cost outweighs the gains made by aligning the IP header.
986 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
990 #define NET_IP_ALIGN 2
994 * The networking layer reserves some headroom in skb data (via
995 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
996 * the header has to grow. In the default case, if the header has to grow
997 * 16 bytes or less we avoid the reallocation.
999 * Unfortunately this headroom changes the DMA alignment of the resulting
1000 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1001 * on some architectures. An architecture can override this value,
1002 * perhaps setting it to a cacheline in size (since that will maintain
1003 * cacheline alignment of the DMA). It must be a power of 2.
1005 * Various parts of the networking layer expect at least 16 bytes of
1006 * headroom, you should not reduce this.
1009 #define NET_SKB_PAD 16
1012 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1014 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1016 if (unlikely(skb->data_len)) {
1021 skb->tail = skb->data + len;
1025 * skb_trim - remove end from a buffer
1026 * @skb: buffer to alter
1029 * Cut the length of a buffer down by removing data from the tail. If
1030 * the buffer is already under the length specified it is not modified.
1031 * The skb must be linear.
1033 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
1036 __skb_trim(skb, len);
1040 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1043 return ___pskb_trim(skb, len);
1044 __skb_trim(skb, len);
1048 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1050 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1054 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1055 * @skb: buffer to alter
1058 * This is identical to pskb_trim except that the caller knows that
1059 * the skb is not cloned so we should never get an error due to out-
1062 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1064 int err = pskb_trim(skb, len);
1069 * skb_orphan - orphan a buffer
1070 * @skb: buffer to orphan
1072 * If a buffer currently has an owner then we call the owner's
1073 * destructor function and make the @skb unowned. The buffer continues
1074 * to exist but is no longer charged to its former owner.
1076 static inline void skb_orphan(struct sk_buff *skb)
1078 if (skb->destructor)
1079 skb->destructor(skb);
1080 skb->destructor = NULL;
1085 * __skb_queue_purge - empty a list
1086 * @list: list to empty
1088 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1089 * the list and one reference dropped. This function does not take the
1090 * list lock and the caller must hold the relevant locks to use it.
1092 extern void skb_queue_purge(struct sk_buff_head *list);
1093 static inline void __skb_queue_purge(struct sk_buff_head *list)
1095 struct sk_buff *skb;
1096 while ((skb = __skb_dequeue(list)) != NULL)
1100 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
1102 * __dev_alloc_skb - allocate an skbuff for receiving
1103 * @length: length to allocate
1104 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1106 * Allocate a new &sk_buff and assign it a usage count of one. The
1107 * buffer has unspecified headroom built in. Users should allocate
1108 * the headroom they think they need without accounting for the
1109 * built in space. The built in space is used for optimisations.
1111 * %NULL is returned if there is no free memory.
1113 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1116 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1118 skb_reserve(skb, NET_SKB_PAD);
1122 extern struct sk_buff *__dev_alloc_skb(unsigned int length, gfp_t gfp_mask);
1126 * dev_alloc_skb - allocate an skbuff for receiving
1127 * @length: length to allocate
1129 * Allocate a new &sk_buff and assign it a usage count of one. The
1130 * buffer has unspecified headroom built in. Users should allocate
1131 * the headroom they think they need without accounting for the
1132 * built in space. The built in space is used for optimisations.
1134 * %NULL is returned if there is no free memory. Although this function
1135 * allocates memory it can be called from an interrupt.
1137 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1139 return __dev_alloc_skb(length, GFP_ATOMIC);
1142 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1143 unsigned int length, gfp_t gfp_mask);
1146 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1147 * @dev: network device to receive on
1148 * @length: length to allocate
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. Although this function
1156 * allocates memory it can be called from an interrupt.
1158 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1159 unsigned int length)
1161 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1165 * skb_cow - copy header of skb when it is required
1166 * @skb: buffer to cow
1167 * @headroom: needed headroom
1169 * If the skb passed lacks sufficient headroom or its data part
1170 * is shared, data is reallocated. If reallocation fails, an error
1171 * is returned and original skb is not changed.
1173 * The result is skb with writable area skb->head...skb->tail
1174 * and at least @headroom of space at head.
1176 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1178 int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) -
1184 if (delta || skb_cloned(skb))
1185 return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) &
1186 ~(NET_SKB_PAD-1), 0, GFP_ATOMIC);
1191 * skb_padto - pad an skbuff up to a minimal size
1192 * @skb: buffer to pad
1193 * @len: minimal length
1195 * Pads up a buffer to ensure the trailing bytes exist and are
1196 * blanked. If the buffer already contains sufficient data it
1197 * is untouched. Otherwise it is extended. Returns zero on
1198 * success. The skb is freed on error.
1201 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1203 unsigned int size = skb->len;
1204 if (likely(size >= len))
1206 return skb_pad(skb, len-size);
1209 static inline int skb_add_data(struct sk_buff *skb,
1210 char __user *from, int copy)
1212 const int off = skb->len;
1214 if (skb->ip_summed == CHECKSUM_NONE) {
1216 unsigned int csum = csum_and_copy_from_user(from,
1220 skb->csum = csum_block_add(skb->csum, csum, off);
1223 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1226 __skb_trim(skb, off);
1230 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1231 struct page *page, int off)
1234 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1236 return page == frag->page &&
1237 off == frag->page_offset + frag->size;
1242 static inline int __skb_linearize(struct sk_buff *skb)
1244 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1248 * skb_linearize - convert paged skb to linear one
1249 * @skb: buffer to linarize
1251 * If there is no free memory -ENOMEM is returned, otherwise zero
1252 * is returned and the old skb data released.
1254 static inline int skb_linearize(struct sk_buff *skb)
1256 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1260 * skb_linearize_cow - make sure skb is linear and writable
1261 * @skb: buffer to process
1263 * If there is no free memory -ENOMEM is returned, otherwise zero
1264 * is returned and the old skb data released.
1266 static inline int skb_linearize_cow(struct sk_buff *skb)
1268 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1269 __skb_linearize(skb) : 0;
1273 * skb_postpull_rcsum - update checksum for received skb after pull
1274 * @skb: buffer to update
1275 * @start: start of data before pull
1276 * @len: length of data pulled
1278 * After doing a pull on a received packet, you need to call this to
1279 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1280 * so that it can be recomputed from scratch.
1283 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1284 const void *start, unsigned int len)
1286 if (skb->ip_summed == CHECKSUM_HW)
1287 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1290 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1293 * pskb_trim_rcsum - trim received skb and update checksum
1294 * @skb: buffer to trim
1297 * This is exactly the same as pskb_trim except that it ensures the
1298 * checksum of received packets are still valid after the operation.
1301 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1303 if (likely(len >= skb->len))
1305 if (skb->ip_summed == CHECKSUM_HW)
1306 skb->ip_summed = CHECKSUM_NONE;
1307 return __pskb_trim(skb, len);
1310 static inline void *kmap_skb_frag(const skb_frag_t *frag)
1312 #ifdef CONFIG_HIGHMEM
1317 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1320 static inline void kunmap_skb_frag(void *vaddr)
1322 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1323 #ifdef CONFIG_HIGHMEM
1328 #define skb_queue_walk(queue, skb) \
1329 for (skb = (queue)->next; \
1330 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1333 #define skb_queue_reverse_walk(queue, skb) \
1334 for (skb = (queue)->prev; \
1335 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1339 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1340 int noblock, int *err);
1341 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1342 struct poll_table_struct *wait);
1343 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1344 int offset, struct iovec *to,
1346 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1349 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1350 extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1351 unsigned int flags);
1352 extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1353 int len, unsigned int csum);
1354 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1356 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1357 void *from, int len);
1358 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1359 int offset, u8 *to, int len,
1361 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1362 extern void skb_split(struct sk_buff *skb,
1363 struct sk_buff *skb1, const u32 len);
1365 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1367 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1368 int len, void *buffer)
1370 int hlen = skb_headlen(skb);
1372 if (hlen - offset >= len)
1373 return skb->data + offset;
1375 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1381 extern void skb_init(void);
1382 extern void skb_add_mtu(int mtu);
1385 * skb_get_timestamp - get timestamp from a skb
1386 * @skb: skb to get stamp from
1387 * @stamp: pointer to struct timeval to store stamp in
1389 * Timestamps are stored in the skb as offsets to a base timestamp.
1390 * This function converts the offset back to a struct timeval and stores
1393 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1395 stamp->tv_sec = skb->tstamp.off_sec;
1396 stamp->tv_usec = skb->tstamp.off_usec;
1400 * skb_set_timestamp - set timestamp of a skb
1401 * @skb: skb to set stamp of
1402 * @stamp: pointer to struct timeval to get stamp from
1404 * Timestamps are stored in the skb as offsets to a base timestamp.
1405 * This function converts a struct timeval to an offset and stores
1408 static inline void skb_set_timestamp(struct sk_buff *skb, const struct timeval *stamp)
1410 skb->tstamp.off_sec = stamp->tv_sec;
1411 skb->tstamp.off_usec = stamp->tv_usec;
1414 extern void __net_timestamp(struct sk_buff *skb);
1416 extern unsigned int __skb_checksum_complete(struct sk_buff *skb);
1419 * skb_checksum_complete - Calculate checksum of an entire packet
1420 * @skb: packet to process
1422 * This function calculates the checksum over the entire packet plus
1423 * the value of skb->csum. The latter can be used to supply the
1424 * checksum of a pseudo header as used by TCP/UDP. It returns the
1427 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1428 * this function can be used to verify that checksum on received
1429 * packets. In that case the function should return zero if the
1430 * checksum is correct. In particular, this function will return zero
1431 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1432 * hardware has already verified the correctness of the checksum.
1434 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1436 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1437 __skb_checksum_complete(skb);
1440 struct tux_req_struct;
1442 #ifdef CONFIG_NETFILTER
1443 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1445 if (nfct && atomic_dec_and_test(&nfct->use))
1446 nfct->destroy(nfct);
1448 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1451 atomic_inc(&nfct->use);
1453 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1454 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1457 atomic_inc(&skb->users);
1459 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1465 #ifdef CONFIG_BRIDGE_NETFILTER
1466 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1468 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1471 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1474 atomic_inc(&nf_bridge->use);
1476 #endif /* CONFIG_BRIDGE_NETFILTER */
1477 static inline void nf_reset(struct sk_buff *skb)
1479 nf_conntrack_put(skb->nfct);
1481 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1482 nf_conntrack_put_reasm(skb->nfct_reasm);
1483 skb->nfct_reasm = NULL;
1485 #ifdef CONFIG_BRIDGE_NETFILTER
1486 nf_bridge_put(skb->nf_bridge);
1487 skb->nf_bridge = NULL;
1491 #else /* CONFIG_NETFILTER */
1492 static inline void nf_reset(struct sk_buff *skb) {}
1493 #endif /* CONFIG_NETFILTER */
1495 #ifdef CONFIG_NETWORK_SECMARK
1496 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1498 to->secmark = from->secmark;
1501 static inline void skb_init_secmark(struct sk_buff *skb)
1506 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1509 static inline void skb_init_secmark(struct sk_buff *skb)
1513 static inline int skb_is_gso(const struct sk_buff *skb)
1515 return skb_shinfo(skb)->gso_size;
1518 #endif /* __KERNEL__ */
1519 #endif /* _LINUX_SKBUFF_H */