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/config.h>
18 #include <linux/kernel.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/cache.h>
23 #include <asm/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
27 #include <linux/highmem.h>
28 #include <linux/poll.h>
29 #include <linux/net.h>
30 #include <linux/textsearch.h>
31 #include <net/checksum.h>
32 #include <linux/dmaengine.h>
34 #define HAVE_ALLOC_SKB /* For the drivers to know */
35 #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
37 #define CHECKSUM_NONE 0
39 #define CHECKSUM_UNNECESSARY 2
41 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
42 ~(SMP_CACHE_BYTES - 1))
43 #define SKB_MAX_ORDER(X, ORDER) (((PAGE_SIZE << (ORDER)) - (X) - \
44 sizeof(struct skb_shared_info)) & \
45 ~(SMP_CACHE_BYTES - 1))
46 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
47 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
49 /* A. Checksumming of received packets by device.
51 * NONE: device failed to checksum this packet.
52 * skb->csum is undefined.
54 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
55 * skb->csum is undefined.
56 * It is bad option, but, unfortunately, many of vendors do this.
57 * Apparently with secret goal to sell you new device, when you
58 * will add new protocol to your host. F.e. IPv6. 8)
60 * HW: the most generic way. Device supplied checksum of _all_
61 * the packet as seen by netif_rx in skb->csum.
62 * NOTE: Even if device supports only some protocols, but
63 * is able to produce some skb->csum, it MUST use HW,
66 * B. Checksumming on output.
68 * NONE: skb is checksummed by protocol or csum is not required.
70 * HW: device is required to csum packet as seen by hard_start_xmit
71 * from skb->h.raw to the end and to record the checksum
72 * at skb->h.raw+skb->csum.
74 * Device must show its capabilities in dev->features, set
75 * at device setup time.
76 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
78 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
79 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
80 * TCP/UDP over IPv4. Sigh. Vendors like this
81 * way by an unknown reason. Though, see comment above
82 * about CHECKSUM_UNNECESSARY. 8)
84 * Any questions? No questions, good. --ANK
89 #ifdef CONFIG_NETFILTER
92 void (*destroy)(struct nf_conntrack *);
95 #ifdef CONFIG_BRIDGE_NETFILTER
96 struct nf_bridge_info {
98 struct net_device *physindev;
99 struct net_device *physoutdev;
100 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
101 struct net_device *netoutdev;
104 unsigned long data[32 / sizeof(unsigned long)];
110 struct sk_buff_head {
111 /* These two members must be first. */
112 struct sk_buff *next;
113 struct sk_buff *prev;
121 /* To allow 64K frame to be packed as single skb without frag_list */
122 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
124 typedef struct skb_frag_struct skb_frag_t;
126 struct skb_frag_struct {
132 /* This data is invariant across clones and lives at
133 * the end of the header data, ie. at skb->end.
135 struct skb_shared_info {
137 unsigned short nr_frags;
138 unsigned short tso_size;
139 unsigned short tso_segs;
140 unsigned short ufo_size;
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 * struct sk_buff - socket buffer
174 * @next: Next buffer in list
175 * @prev: Previous buffer in list
176 * @sk: Socket we are owned by
177 * @tstamp: Time we arrived
178 * @dev: Device we arrived on/are leaving by
179 * @input_dev: Device we arrived on
180 * @h: Transport layer header
181 * @nh: Network layer header
182 * @mac: Link layer header
183 * @dst: destination entry
184 * @sp: the security path, used for xfrm
185 * @cb: Control buffer. Free for use by every layer. Put private vars here
186 * @len: Length of actual data
187 * @data_len: Data length
188 * @mac_len: Length of link layer header
190 * @local_df: allow local fragmentation
191 * @cloned: Head may be cloned (check refcnt to be sure)
192 * @nohdr: Payload reference only, must not modify header
193 * @pkt_type: Packet class
194 * @fclone: skbuff clone status
195 * @ip_summed: Driver fed us an IP checksum
196 * @priority: Packet queueing priority
197 * @users: User count - see {datagram,tcp}.c
198 * @protocol: Packet protocol from driver
199 * @truesize: Buffer size
200 * @head: Head of buffer
201 * @data: Data head pointer
202 * @tail: Tail pointer
204 * @destructor: Destruct function
205 * @nfmark: Can be used for communication between hooks
206 * @nfct: Associated connection, if any
207 * @ipvs_property: skbuff is owned by ipvs
208 * @nfctinfo: Relationship of this skb to the connection
209 * @nfct_reasm: netfilter conntrack re-assembly pointer
210 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
211 * @tc_index: Traffic control index
212 * @tc_verd: traffic control verdict
216 /* These two members must be first. */
217 struct sk_buff *next;
218 struct sk_buff *prev;
221 struct skb_timeval tstamp;
222 struct net_device *dev;
223 struct net_device *input_dev;
228 struct icmphdr *icmph;
229 struct igmphdr *igmph;
231 struct ipv6hdr *ipv6h;
237 struct ipv6hdr *ipv6h;
246 struct dst_entry *dst;
250 * This is the control buffer. It is free to use for every
251 * layer. Please put your private variables there. If you
252 * want to keep them across layers you have to do a skb_clone()
253 * first. This is owned by whoever has the skb queued ATM.
272 void (*destructor)(struct sk_buff *skb);
273 #ifdef CONFIG_NETFILTER
274 struct nf_conntrack *nfct;
275 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
276 struct sk_buff *nfct_reasm;
278 #ifdef CONFIG_BRIDGE_NETFILTER
279 struct nf_bridge_info *nf_bridge;
282 #endif /* CONFIG_NETFILTER */
283 #ifdef CONFIG_NET_SCHED
284 __u16 tc_index; /* traffic control index */
285 #ifdef CONFIG_NET_CLS_ACT
286 __u16 tc_verd; /* traffic control verdict */
289 #ifdef CONFIG_NET_DMA
290 dma_cookie_t dma_cookie;
294 /* These elements must be at the end, see alloc_skb() for details. */
295 unsigned int truesize;
305 * Handling routines are only of interest to the kernel
307 #include <linux/slab.h>
309 #include <asm/system.h>
311 extern void kfree_skb(struct sk_buff *skb);
312 extern void __kfree_skb(struct sk_buff *skb);
313 extern struct sk_buff *__alloc_skb(unsigned int size,
314 gfp_t priority, int fclone);
315 static inline struct sk_buff *alloc_skb(unsigned int size,
318 return __alloc_skb(size, priority, 0);
321 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
324 return __alloc_skb(size, priority, 1);
327 extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
330 extern void kfree_skbmem(struct sk_buff *skb);
331 extern struct sk_buff *skb_clone(struct sk_buff *skb,
333 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
335 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
337 extern int pskb_expand_head(struct sk_buff *skb,
338 int nhead, int ntail,
340 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
341 unsigned int headroom);
342 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
343 int newheadroom, int newtailroom,
345 extern struct sk_buff * skb_pad(struct sk_buff *skb, int pad);
346 #define dev_kfree_skb(a) kfree_skb(a)
347 extern void skb_over_panic(struct sk_buff *skb, int len,
349 extern void skb_under_panic(struct sk_buff *skb, int len,
351 extern void skb_truesize_bug(struct sk_buff *skb);
353 static inline void skb_truesize_check(struct sk_buff *skb)
355 if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len))
356 skb_truesize_bug(skb);
359 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
360 int getfrag(void *from, char *to, int offset,
361 int len,int odd, struct sk_buff *skb),
362 void *from, int length);
369 __u32 stepped_offset;
370 struct sk_buff *root_skb;
371 struct sk_buff *cur_skb;
375 extern void skb_prepare_seq_read(struct sk_buff *skb,
376 unsigned int from, unsigned int to,
377 struct skb_seq_state *st);
378 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
379 struct skb_seq_state *st);
380 extern void skb_abort_seq_read(struct skb_seq_state *st);
382 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
383 unsigned int to, struct ts_config *config,
384 struct ts_state *state);
387 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
390 * skb_queue_empty - check if a queue is empty
393 * Returns true if the queue is empty, false otherwise.
395 static inline int skb_queue_empty(const struct sk_buff_head *list)
397 return list->next == (struct sk_buff *)list;
401 * skb_get - reference buffer
402 * @skb: buffer to reference
404 * Makes another reference to a socket buffer and returns a pointer
407 static inline struct sk_buff *skb_get(struct sk_buff *skb)
409 atomic_inc(&skb->users);
414 * If users == 1, we are the only owner and are can avoid redundant
419 * skb_cloned - is the buffer a clone
420 * @skb: buffer to check
422 * Returns true if the buffer was generated with skb_clone() and is
423 * one of multiple shared copies of the buffer. Cloned buffers are
424 * shared data so must not be written to under normal circumstances.
426 static inline int skb_cloned(const struct sk_buff *skb)
428 return skb->cloned &&
429 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
433 * skb_header_cloned - is the header a clone
434 * @skb: buffer to check
436 * Returns true if modifying the header part of the buffer requires
437 * the data to be copied.
439 static inline int skb_header_cloned(const struct sk_buff *skb)
446 dataref = atomic_read(&skb_shinfo(skb)->dataref);
447 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
452 * skb_header_release - release reference to header
453 * @skb: buffer to operate on
455 * Drop a reference to the header part of the buffer. This is done
456 * by acquiring a payload reference. You must not read from the header
457 * part of skb->data after this.
459 static inline void skb_header_release(struct sk_buff *skb)
463 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
467 * skb_shared - is the buffer shared
468 * @skb: buffer to check
470 * Returns true if more than one person has a reference to this
473 static inline int skb_shared(const struct sk_buff *skb)
475 return atomic_read(&skb->users) != 1;
479 * skb_share_check - check if buffer is shared and if so clone it
480 * @skb: buffer to check
481 * @pri: priority for memory allocation
483 * If the buffer is shared the buffer is cloned and the old copy
484 * drops a reference. A new clone with a single reference is returned.
485 * If the buffer is not shared the original buffer is returned. When
486 * being called from interrupt status or with spinlocks held pri must
489 * NULL is returned on a memory allocation failure.
491 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
494 might_sleep_if(pri & __GFP_WAIT);
495 if (skb_shared(skb)) {
496 struct sk_buff *nskb = skb_clone(skb, pri);
504 * Copy shared buffers into a new sk_buff. We effectively do COW on
505 * packets to handle cases where we have a local reader and forward
506 * and a couple of other messy ones. The normal one is tcpdumping
507 * a packet thats being forwarded.
511 * skb_unshare - make a copy of a shared buffer
512 * @skb: buffer to check
513 * @pri: priority for memory allocation
515 * If the socket buffer is a clone then this function creates a new
516 * copy of the data, drops a reference count on the old copy and returns
517 * the new copy with the reference count at 1. If the buffer is not a clone
518 * the original buffer is returned. When called with a spinlock held or
519 * from interrupt state @pri must be %GFP_ATOMIC
521 * %NULL is returned on a memory allocation failure.
523 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
526 might_sleep_if(pri & __GFP_WAIT);
527 if (skb_cloned(skb)) {
528 struct sk_buff *nskb = skb_copy(skb, pri);
529 kfree_skb(skb); /* Free our shared copy */
537 * @list_: list to peek at
539 * Peek an &sk_buff. Unlike most other operations you _MUST_
540 * be careful with this one. A peek leaves the buffer on the
541 * list and someone else may run off with it. You must hold
542 * the appropriate locks or have a private queue to do this.
544 * Returns %NULL for an empty list or a pointer to the head element.
545 * The reference count is not incremented and the reference is therefore
546 * volatile. Use with caution.
548 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
550 struct sk_buff *list = ((struct sk_buff *)list_)->next;
551 if (list == (struct sk_buff *)list_)
558 * @list_: list to peek at
560 * Peek an &sk_buff. Unlike most other operations you _MUST_
561 * be careful with this one. A peek leaves the buffer on the
562 * list and someone else may run off with it. You must hold
563 * the appropriate locks or have a private queue to do this.
565 * Returns %NULL for an empty list or a pointer to the tail element.
566 * The reference count is not incremented and the reference is therefore
567 * volatile. Use with caution.
569 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
571 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
572 if (list == (struct sk_buff *)list_)
578 * skb_queue_len - get queue length
579 * @list_: list to measure
581 * Return the length of an &sk_buff queue.
583 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
588 static inline void skb_queue_head_init(struct sk_buff_head *list)
590 spin_lock_init(&list->lock);
591 list->prev = list->next = (struct sk_buff *)list;
596 * Insert an sk_buff at the start of a list.
598 * The "__skb_xxxx()" functions are the non-atomic ones that
599 * can only be called with interrupts disabled.
603 * __skb_queue_after - queue a buffer at the list head
605 * @prev: place after this buffer
606 * @newsk: buffer to queue
608 * Queue a buffer int the middle of a list. This function takes no locks
609 * and you must therefore hold required locks before calling it.
611 * A buffer cannot be placed on two lists at the same time.
613 static inline void __skb_queue_after(struct sk_buff_head *list,
614 struct sk_buff *prev,
615 struct sk_buff *newsk)
617 struct sk_buff *next;
623 next->prev = prev->next = newsk;
627 * __skb_queue_head - queue a buffer at the list head
629 * @newsk: buffer to queue
631 * Queue a buffer at the start of a list. This function takes no locks
632 * and you must therefore hold required locks before calling it.
634 * A buffer cannot be placed on two lists at the same time.
636 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
637 static inline void __skb_queue_head(struct sk_buff_head *list,
638 struct sk_buff *newsk)
640 __skb_queue_after(list, (struct sk_buff *)list, newsk);
644 * __skb_queue_tail - queue a buffer at the list tail
646 * @newsk: buffer to queue
648 * Queue a buffer at the end 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_tail(struct sk_buff_head *list, struct sk_buff *newsk);
654 static inline void __skb_queue_tail(struct sk_buff_head *list,
655 struct sk_buff *newsk)
657 struct sk_buff *prev, *next;
660 next = (struct sk_buff *)list;
664 next->prev = prev->next = newsk;
669 * __skb_dequeue - remove from the head of the queue
670 * @list: list to dequeue from
672 * Remove the head of the list. This function does not take any locks
673 * so must be used with appropriate locks held only. The head item is
674 * returned or %NULL if the list is empty.
676 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
677 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
679 struct sk_buff *next, *prev, *result;
681 prev = (struct sk_buff *) list;
690 result->next = result->prev = NULL;
697 * Insert a packet on a list.
699 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
700 static inline void __skb_insert(struct sk_buff *newsk,
701 struct sk_buff *prev, struct sk_buff *next,
702 struct sk_buff_head *list)
706 next->prev = prev->next = newsk;
711 * Place a packet after a given packet in a list.
713 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
714 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
716 __skb_insert(newsk, old, old->next, list);
720 * remove sk_buff from list. _Must_ be called atomically, and with
723 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
724 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
726 struct sk_buff *next, *prev;
731 skb->next = skb->prev = NULL;
737 /* XXX: more streamlined implementation */
740 * __skb_dequeue_tail - remove from the tail of the queue
741 * @list: list to dequeue from
743 * Remove the tail of the list. This function does not take any locks
744 * so must be used with appropriate locks held only. The tail item is
745 * returned or %NULL if the list is empty.
747 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
748 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
750 struct sk_buff *skb = skb_peek_tail(list);
752 __skb_unlink(skb, list);
757 static inline int skb_is_nonlinear(const struct sk_buff *skb)
759 return skb->data_len;
762 static inline unsigned int skb_headlen(const struct sk_buff *skb)
764 return skb->len - skb->data_len;
767 static inline int skb_pagelen(const struct sk_buff *skb)
771 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
772 len += skb_shinfo(skb)->frags[i].size;
773 return len + skb_headlen(skb);
776 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
777 struct page *page, int off, int size)
779 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
782 frag->page_offset = off;
784 skb_shinfo(skb)->nr_frags = i + 1;
787 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
788 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
789 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
792 * Add data to an sk_buff
794 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
796 unsigned char *tmp = skb->tail;
797 SKB_LINEAR_ASSERT(skb);
804 * skb_put - add data to a buffer
805 * @skb: buffer to use
806 * @len: amount of data to add
808 * This function extends the used data area of the buffer. If this would
809 * exceed the total buffer size the kernel will panic. A pointer to the
810 * first byte of the extra data is returned.
812 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
814 unsigned char *tmp = skb->tail;
815 SKB_LINEAR_ASSERT(skb);
818 if (unlikely(skb->tail>skb->end))
819 skb_over_panic(skb, len, current_text_addr());
823 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
831 * skb_push - add data to the start of a buffer
832 * @skb: buffer to use
833 * @len: amount of data to add
835 * This function extends the used data area of the buffer at the buffer
836 * start. If this would exceed the total buffer headroom the kernel will
837 * panic. A pointer to the first byte of the extra data is returned.
839 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
843 if (unlikely(skb->data<skb->head))
844 skb_under_panic(skb, len, current_text_addr());
848 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
851 BUG_ON(skb->len < skb->data_len);
852 return skb->data += len;
856 * skb_pull - remove data from the start of a buffer
857 * @skb: buffer to use
858 * @len: amount of data to remove
860 * This function removes data from the start of a buffer, returning
861 * the memory to the headroom. A pointer to the next data in the buffer
862 * is returned. Once the data has been pulled future pushes will overwrite
865 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
867 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
870 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
872 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
874 if (len > skb_headlen(skb) &&
875 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
878 return skb->data += len;
881 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
883 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
886 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
888 if (likely(len <= skb_headlen(skb)))
890 if (unlikely(len > skb->len))
892 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
896 * skb_headroom - bytes at buffer head
897 * @skb: buffer to check
899 * Return the number of bytes of free space at the head of an &sk_buff.
901 static inline int skb_headroom(const struct sk_buff *skb)
903 return skb->data - skb->head;
907 * skb_tailroom - bytes at buffer end
908 * @skb: buffer to check
910 * Return the number of bytes of free space at the tail of an sk_buff
912 static inline int skb_tailroom(const struct sk_buff *skb)
914 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
918 * skb_reserve - adjust headroom
919 * @skb: buffer to alter
920 * @len: bytes to move
922 * Increase the headroom of an empty &sk_buff by reducing the tail
923 * room. This is only allowed for an empty buffer.
925 static inline void skb_reserve(struct sk_buff *skb, int len)
932 * CPUs often take a performance hit when accessing unaligned memory
933 * locations. The actual performance hit varies, it can be small if the
934 * hardware handles it or large if we have to take an exception and fix it
937 * Since an ethernet header is 14 bytes network drivers often end up with
938 * the IP header at an unaligned offset. The IP header can be aligned by
939 * shifting the start of the packet by 2 bytes. Drivers should do this
942 * skb_reserve(NET_IP_ALIGN);
944 * The downside to this alignment of the IP header is that the DMA is now
945 * unaligned. On some architectures the cost of an unaligned DMA is high
946 * and this cost outweighs the gains made by aligning the IP header.
948 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
952 #define NET_IP_ALIGN 2
956 * The networking layer reserves some headroom in skb data (via
957 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
958 * the header has to grow. In the default case, if the header has to grow
959 * 16 bytes or less we avoid the reallocation.
961 * Unfortunately this headroom changes the DMA alignment of the resulting
962 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
963 * on some architectures. An architecture can override this value,
964 * perhaps setting it to a cacheline in size (since that will maintain
965 * cacheline alignment of the DMA). It must be a power of 2.
967 * Various parts of the networking layer expect at least 16 bytes of
968 * headroom, you should not reduce this.
971 #define NET_SKB_PAD 16
974 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
976 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
978 if (!skb->data_len) {
980 skb->tail = skb->data + len;
982 ___pskb_trim(skb, len, 0);
986 * skb_trim - remove end from a buffer
987 * @skb: buffer to alter
990 * Cut the length of a buffer down by removing data from the tail. If
991 * the buffer is already under the length specified it is not modified.
993 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
996 __skb_trim(skb, len);
1000 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1002 if (!skb->data_len) {
1004 skb->tail = skb->data+len;
1007 return ___pskb_trim(skb, len, 1);
1010 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1012 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1016 * skb_orphan - orphan a buffer
1017 * @skb: buffer to orphan
1019 * If a buffer currently has an owner then we call the owner's
1020 * destructor function and make the @skb unowned. The buffer continues
1021 * to exist but is no longer charged to its former owner.
1023 static inline void skb_orphan(struct sk_buff *skb)
1025 if (skb->destructor)
1026 skb->destructor(skb);
1027 skb->destructor = NULL;
1032 * __skb_queue_purge - empty a list
1033 * @list: list to empty
1035 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1036 * the list and one reference dropped. This function does not take the
1037 * list lock and the caller must hold the relevant locks to use it.
1039 extern void skb_queue_purge(struct sk_buff_head *list);
1040 static inline void __skb_queue_purge(struct sk_buff_head *list)
1042 struct sk_buff *skb;
1043 while ((skb = __skb_dequeue(list)) != NULL)
1047 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
1049 * __dev_alloc_skb - allocate an skbuff for sending
1050 * @length: length to allocate
1051 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1053 * Allocate a new &sk_buff and assign it a usage count of one. The
1054 * buffer has unspecified headroom built in. Users should allocate
1055 * the headroom they think they need without accounting for the
1056 * built in space. The built in space is used for optimisations.
1058 * %NULL is returned in there is no free memory.
1060 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1063 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1065 skb_reserve(skb, NET_SKB_PAD);
1069 extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
1073 * dev_alloc_skb - allocate an skbuff for sending
1074 * @length: length to allocate
1076 * Allocate a new &sk_buff and assign it a usage count of one. The
1077 * buffer has unspecified headroom built in. Users should allocate
1078 * the headroom they think they need without accounting for the
1079 * built in space. The built in space is used for optimisations.
1081 * %NULL is returned in there is no free memory. Although this function
1082 * allocates memory it can be called from an interrupt.
1084 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1086 return __dev_alloc_skb(length, GFP_ATOMIC);
1090 * skb_cow - copy header of skb when it is required
1091 * @skb: buffer to cow
1092 * @headroom: needed headroom
1094 * If the skb passed lacks sufficient headroom or its data part
1095 * is shared, data is reallocated. If reallocation fails, an error
1096 * is returned and original skb is not changed.
1098 * The result is skb with writable area skb->head...skb->tail
1099 * and at least @headroom of space at head.
1101 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1103 int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) -
1109 if (delta || skb_cloned(skb))
1110 return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) &
1111 ~(NET_SKB_PAD-1), 0, GFP_ATOMIC);
1116 * skb_padto - pad an skbuff up to a minimal size
1117 * @skb: buffer to pad
1118 * @len: minimal length
1120 * Pads up a buffer to ensure the trailing bytes exist and are
1121 * blanked. If the buffer already contains sufficient data it
1122 * is untouched. Returns the buffer, which may be a replacement
1123 * for the original, or NULL for out of memory - in which case
1124 * the original buffer is still freed.
1127 static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
1129 unsigned int size = skb->len;
1130 if (likely(size >= len))
1132 return skb_pad(skb, len-size);
1135 static inline int skb_add_data(struct sk_buff *skb,
1136 char __user *from, int copy)
1138 const int off = skb->len;
1140 if (skb->ip_summed == CHECKSUM_NONE) {
1142 unsigned int csum = csum_and_copy_from_user(from,
1146 skb->csum = csum_block_add(skb->csum, csum, off);
1149 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1152 __skb_trim(skb, off);
1156 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1157 struct page *page, int off)
1160 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1162 return page == frag->page &&
1163 off == frag->page_offset + frag->size;
1169 * skb_linearize - convert paged skb to linear one
1170 * @skb: buffer to linarize
1171 * @gfp: allocation mode
1173 * If there is no free memory -ENOMEM is returned, otherwise zero
1174 * is returned and the old skb data released.
1176 extern int __skb_linearize(struct sk_buff *skb, gfp_t gfp);
1177 static inline int skb_linearize(struct sk_buff *skb, gfp_t gfp)
1179 return __skb_linearize(skb, gfp);
1183 * skb_postpull_rcsum - update checksum for received skb after pull
1184 * @skb: buffer to update
1185 * @start: start of data before pull
1186 * @len: length of data pulled
1188 * After doing a pull on a received packet, you need to call this to
1189 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1190 * so that it can be recomputed from scratch.
1193 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1194 const void *start, unsigned int len)
1196 if (skb->ip_summed == CHECKSUM_HW)
1197 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1200 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1203 * pskb_trim_rcsum - trim received skb and update checksum
1204 * @skb: buffer to trim
1207 * This is exactly the same as pskb_trim except that it ensures the
1208 * checksum of received packets are still valid after the operation.
1211 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1213 if (likely(len >= skb->len))
1215 if (skb->ip_summed == CHECKSUM_HW)
1216 skb->ip_summed = CHECKSUM_NONE;
1217 return __pskb_trim(skb, len);
1220 static inline void *kmap_skb_frag(const skb_frag_t *frag)
1222 #ifdef CONFIG_HIGHMEM
1227 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1230 static inline void kunmap_skb_frag(void *vaddr)
1232 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1233 #ifdef CONFIG_HIGHMEM
1238 #define skb_queue_walk(queue, skb) \
1239 for (skb = (queue)->next; \
1240 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1243 #define skb_queue_reverse_walk(queue, skb) \
1244 for (skb = (queue)->prev; \
1245 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1249 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1250 int noblock, int *err);
1251 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1252 struct poll_table_struct *wait);
1253 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1254 int offset, struct iovec *to,
1256 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1259 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1260 extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1261 unsigned int flags);
1262 extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1263 int len, unsigned int csum);
1264 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1266 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1267 void *from, int len);
1268 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1269 int offset, u8 *to, int len,
1271 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1272 extern void skb_split(struct sk_buff *skb,
1273 struct sk_buff *skb1, const u32 len);
1275 extern void skb_release_data(struct sk_buff *skb);
1277 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1278 int len, void *buffer)
1280 int hlen = skb_headlen(skb);
1282 if (hlen - offset >= len)
1283 return skb->data + offset;
1285 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1291 extern void skb_init(void);
1292 extern void skb_add_mtu(int mtu);
1295 * skb_get_timestamp - get timestamp from a skb
1296 * @skb: skb to get stamp from
1297 * @stamp: pointer to struct timeval to store stamp in
1299 * Timestamps are stored in the skb as offsets to a base timestamp.
1300 * This function converts the offset back to a struct timeval and stores
1303 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1305 stamp->tv_sec = skb->tstamp.off_sec;
1306 stamp->tv_usec = skb->tstamp.off_usec;
1310 * skb_set_timestamp - set timestamp of a skb
1311 * @skb: skb to set stamp of
1312 * @stamp: pointer to struct timeval to get stamp from
1314 * Timestamps are stored in the skb as offsets to a base timestamp.
1315 * This function converts a struct timeval to an offset and stores
1318 static inline void skb_set_timestamp(struct sk_buff *skb, const struct timeval *stamp)
1320 skb->tstamp.off_sec = stamp->tv_sec;
1321 skb->tstamp.off_usec = stamp->tv_usec;
1324 extern void __net_timestamp(struct sk_buff *skb);
1326 extern unsigned int __skb_checksum_complete(struct sk_buff *skb);
1329 * skb_checksum_complete - Calculate checksum of an entire packet
1330 * @skb: packet to process
1332 * This function calculates the checksum over the entire packet plus
1333 * the value of skb->csum. The latter can be used to supply the
1334 * checksum of a pseudo header as used by TCP/UDP. It returns the
1337 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1338 * this function can be used to verify that checksum on received
1339 * packets. In that case the function should return zero if the
1340 * checksum is correct. In particular, this function will return zero
1341 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1342 * hardware has already verified the correctness of the checksum.
1344 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1346 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1347 __skb_checksum_complete(skb);
1350 #ifdef CONFIG_NETFILTER
1351 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1353 if (nfct && atomic_dec_and_test(&nfct->use))
1354 nfct->destroy(nfct);
1356 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1359 atomic_inc(&nfct->use);
1361 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1362 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1365 atomic_inc(&skb->users);
1367 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1373 #ifdef CONFIG_BRIDGE_NETFILTER
1374 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1376 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1379 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1382 atomic_inc(&nf_bridge->use);
1384 #endif /* CONFIG_BRIDGE_NETFILTER */
1385 static inline void nf_reset(struct sk_buff *skb)
1387 nf_conntrack_put(skb->nfct);
1389 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1390 nf_conntrack_put_reasm(skb->nfct_reasm);
1391 skb->nfct_reasm = NULL;
1393 #ifdef CONFIG_BRIDGE_NETFILTER
1394 nf_bridge_put(skb->nf_bridge);
1395 skb->nf_bridge = NULL;
1399 #else /* CONFIG_NETFILTER */
1400 static inline void nf_reset(struct sk_buff *skb) {}
1401 #endif /* CONFIG_NETFILTER */
1403 #endif /* __KERNEL__ */
1404 #endif /* _LINUX_SKBUFF_H */