2 * Definitions for the 'struct sk_buff' memory handlers.
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/kernel.h>
18 #include <linux/compiler.h>
19 #include <linux/time.h>
20 #include <linux/cache.h>
22 #include <asm/atomic.h>
23 #include <asm/types.h>
24 #include <linux/spinlock.h>
25 #include <linux/net.h>
26 #include <linux/textsearch.h>
27 #include <net/checksum.h>
28 #include <linux/rcupdate.h>
29 #include <linux/dmaengine.h>
30 #include <linux/hrtimer.h>
32 #define HAVE_ALLOC_SKB /* For the drivers to know */
33 #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
35 #define CHECKSUM_NONE 0
36 #define CHECKSUM_PARTIAL 1
37 #define CHECKSUM_UNNECESSARY 2
38 #define CHECKSUM_COMPLETE 3
40 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
41 ~(SMP_CACHE_BYTES - 1))
42 #define SKB_WITH_OVERHEAD(X) \
43 (((X) - sizeof(struct skb_shared_info)) & \
44 ~(SMP_CACHE_BYTES - 1))
45 #define SKB_MAX_ORDER(X, ORDER) \
46 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
47 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
48 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
50 /* A. Checksumming of received packets by device.
52 * NONE: device failed to checksum this packet.
53 * skb->csum is undefined.
55 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
56 * skb->csum is undefined.
57 * It is bad option, but, unfortunately, many of vendors do this.
58 * Apparently with secret goal to sell you new device, when you
59 * will add new protocol to your host. F.e. IPv6. 8)
61 * COMPLETE: the most generic way. Device supplied checksum of _all_
62 * the packet as seen by netif_rx in skb->csum.
63 * NOTE: Even if device supports only some protocols, but
64 * is able to produce some skb->csum, it MUST use COMPLETE,
67 * B. Checksumming on output.
69 * NONE: skb is checksummed by protocol or csum is not required.
71 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
72 * from skb->transport_header to the end and to record the checksum
73 * at skb->transport_header + skb->csum.
75 * Device must show its capabilities in dev->features, set
76 * at device setup time.
77 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
79 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
80 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
81 * TCP/UDP over IPv4. Sigh. Vendors like this
82 * way by an unknown reason. Though, see comment above
83 * about CHECKSUM_UNNECESSARY. 8)
85 * Any questions? No questions, good. --ANK
90 #ifdef CONFIG_NETFILTER
93 void (*destroy)(struct nf_conntrack *);
96 #ifdef CONFIG_BRIDGE_NETFILTER
97 struct nf_bridge_info {
99 struct net_device *physindev;
100 struct net_device *physoutdev;
101 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
102 struct net_device *netoutdev;
105 unsigned long data[32 / sizeof(unsigned long)];
111 struct sk_buff_head {
112 /* These two members must be first. */
113 struct sk_buff *next;
114 struct sk_buff *prev;
122 /* To allow 64K frame to be packed as single skb without frag_list */
123 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
125 typedef struct skb_frag_struct skb_frag_t;
127 struct skb_frag_struct {
133 /* This data is invariant across clones and lives at
134 * the end of the header data, ie. at skb->end.
136 struct skb_shared_info {
138 unsigned short nr_frags;
139 unsigned short gso_size;
140 /* Warning: this field is not always filled in (UFO)! */
141 unsigned short gso_segs;
142 unsigned short gso_type;
144 struct sk_buff *frag_list;
145 skb_frag_t frags[MAX_SKB_FRAGS];
148 /* We divide dataref into two halves. The higher 16 bits hold references
149 * to the payload part of skb->data. The lower 16 bits hold references to
150 * the entire skb->data. It is up to the users of the skb to agree on
151 * where the payload starts.
153 * All users must obey the rule that the skb->data reference count must be
154 * greater than or equal to the payload reference count.
156 * Holding a reference to the payload part means that the user does not
157 * care about modifications to the header part of skb->data.
159 #define SKB_DATAREF_SHIFT 16
160 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
164 SKB_FCLONE_UNAVAILABLE,
170 SKB_GSO_TCPV4 = 1 << 0,
171 SKB_GSO_UDP = 1 << 1,
173 /* This indicates the skb is from an untrusted source. */
174 SKB_GSO_DODGY = 1 << 2,
176 /* This indicates the tcp segment has CWR set. */
177 SKB_GSO_TCP_ECN = 1 << 3,
179 SKB_GSO_TCPV6 = 1 << 4,
182 #if BITS_PER_LONG > 32
183 #define NET_SKBUFF_DATA_USES_OFFSET 1
186 #ifdef NET_SKBUFF_DATA_USES_OFFSET
187 typedef unsigned int sk_buff_data_t;
189 typedef unsigned char *sk_buff_data_t;
193 * struct sk_buff - socket buffer
194 * @next: Next buffer in list
195 * @prev: Previous buffer in list
196 * @sk: Socket we are owned by
197 * @tstamp: Time we arrived
198 * @dev: Device we arrived on/are leaving by
199 * @iif: ifindex of device we arrived on
200 * @h: Transport layer header
201 * @network_header: Network layer header
202 * @mac_header: Link layer header
203 * @dst: destination entry
204 * @sp: the security path, used for xfrm
205 * @cb: Control buffer. Free for use by every layer. Put private vars here
206 * @len: Length of actual data
207 * @data_len: Data length
208 * @mac_len: Length of link layer header
210 * @local_df: allow local fragmentation
211 * @cloned: Head may be cloned (check refcnt to be sure)
212 * @nohdr: Payload reference only, must not modify header
213 * @pkt_type: Packet class
214 * @fclone: skbuff clone status
215 * @ip_summed: Driver fed us an IP checksum
216 * @priority: Packet queueing priority
217 * @users: User count - see {datagram,tcp}.c
218 * @protocol: Packet protocol from driver
219 * @truesize: Buffer size
220 * @head: Head of buffer
221 * @data: Data head pointer
222 * @tail: Tail pointer
224 * @destructor: Destruct function
225 * @mark: Generic packet mark
226 * @nfct: Associated connection, if any
227 * @ipvs_property: skbuff is owned by ipvs
228 * @nfctinfo: Relationship of this skb to the connection
229 * @nfct_reasm: netfilter conntrack re-assembly pointer
230 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
231 * @tc_index: Traffic control index
232 * @tc_verd: traffic control verdict
233 * @dma_cookie: a cookie to one of several possible DMA operations
234 * done by skb DMA functions
235 * @secmark: security marking
239 /* These two members must be first. */
240 struct sk_buff *next;
241 struct sk_buff *prev;
245 struct net_device *dev;
247 /* 4 byte hole on 64 bit*/
249 sk_buff_data_t transport_header;
250 sk_buff_data_t network_header;
251 sk_buff_data_t mac_header;
252 struct dst_entry *dst;
256 * This is the control buffer. It is free to use for every
257 * layer. Please put your private variables there. If you
258 * want to keep them across layers you have to do a skb_clone()
259 * first. This is owned by whoever has the skb queued ATM.
281 void (*destructor)(struct sk_buff *skb);
282 #ifdef CONFIG_NETFILTER
283 struct nf_conntrack *nfct;
284 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
285 struct sk_buff *nfct_reasm;
287 #ifdef CONFIG_BRIDGE_NETFILTER
288 struct nf_bridge_info *nf_bridge;
290 #endif /* CONFIG_NETFILTER */
291 #ifdef CONFIG_NET_SCHED
292 __u16 tc_index; /* traffic control index */
293 #ifdef CONFIG_NET_CLS_ACT
294 __u16 tc_verd; /* traffic control verdict */
297 #ifdef CONFIG_NET_DMA
298 dma_cookie_t dma_cookie;
300 #ifdef CONFIG_NETWORK_SECMARK
306 /* These elements must be at the end, see alloc_skb() for details. */
307 unsigned int truesize;
317 * Handling routines are only of interest to the kernel
319 #include <linux/slab.h>
321 #include <asm/system.h>
323 extern void kfree_skb(struct sk_buff *skb);
324 extern void __kfree_skb(struct sk_buff *skb);
325 extern struct sk_buff *__alloc_skb(unsigned int size,
326 gfp_t priority, int fclone, int node);
327 static inline struct sk_buff *alloc_skb(unsigned int size,
330 return __alloc_skb(size, priority, 0, -1);
333 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
336 return __alloc_skb(size, priority, 1, -1);
339 extern void kfree_skbmem(struct sk_buff *skb);
340 extern struct sk_buff *skb_clone(struct sk_buff *skb,
342 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
344 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
346 extern int pskb_expand_head(struct sk_buff *skb,
347 int nhead, int ntail,
349 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
350 unsigned int headroom);
351 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
352 int newheadroom, int newtailroom,
354 extern int skb_pad(struct sk_buff *skb, int pad);
355 #define dev_kfree_skb(a) kfree_skb(a)
356 extern void skb_over_panic(struct sk_buff *skb, int len,
358 extern void skb_under_panic(struct sk_buff *skb, int len,
360 extern void skb_truesize_bug(struct sk_buff *skb);
362 static inline void skb_truesize_check(struct sk_buff *skb)
364 if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len))
365 skb_truesize_bug(skb);
368 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
369 int getfrag(void *from, char *to, int offset,
370 int len,int odd, struct sk_buff *skb),
371 void *from, int length);
378 __u32 stepped_offset;
379 struct sk_buff *root_skb;
380 struct sk_buff *cur_skb;
384 extern void skb_prepare_seq_read(struct sk_buff *skb,
385 unsigned int from, unsigned int to,
386 struct skb_seq_state *st);
387 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
388 struct skb_seq_state *st);
389 extern void skb_abort_seq_read(struct skb_seq_state *st);
391 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
392 unsigned int to, struct ts_config *config,
393 struct ts_state *state);
396 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
399 * skb_queue_empty - check if a queue is empty
402 * Returns true if the queue is empty, false otherwise.
404 static inline int skb_queue_empty(const struct sk_buff_head *list)
406 return list->next == (struct sk_buff *)list;
410 * skb_get - reference buffer
411 * @skb: buffer to reference
413 * Makes another reference to a socket buffer and returns a pointer
416 static inline struct sk_buff *skb_get(struct sk_buff *skb)
418 atomic_inc(&skb->users);
423 * If users == 1, we are the only owner and are can avoid redundant
428 * skb_cloned - is the buffer a clone
429 * @skb: buffer to check
431 * Returns true if the buffer was generated with skb_clone() and is
432 * one of multiple shared copies of the buffer. Cloned buffers are
433 * shared data so must not be written to under normal circumstances.
435 static inline int skb_cloned(const struct sk_buff *skb)
437 return skb->cloned &&
438 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
442 * skb_header_cloned - is the header a clone
443 * @skb: buffer to check
445 * Returns true if modifying the header part of the buffer requires
446 * the data to be copied.
448 static inline int skb_header_cloned(const struct sk_buff *skb)
455 dataref = atomic_read(&skb_shinfo(skb)->dataref);
456 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
461 * skb_header_release - release reference to header
462 * @skb: buffer to operate on
464 * Drop a reference to the header part of the buffer. This is done
465 * by acquiring a payload reference. You must not read from the header
466 * part of skb->data after this.
468 static inline void skb_header_release(struct sk_buff *skb)
472 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
476 * skb_shared - is the buffer shared
477 * @skb: buffer to check
479 * Returns true if more than one person has a reference to this
482 static inline int skb_shared(const struct sk_buff *skb)
484 return atomic_read(&skb->users) != 1;
488 * skb_share_check - check if buffer is shared and if so clone it
489 * @skb: buffer to check
490 * @pri: priority for memory allocation
492 * If the buffer is shared the buffer is cloned and the old copy
493 * drops a reference. A new clone with a single reference is returned.
494 * If the buffer is not shared the original buffer is returned. When
495 * being called from interrupt status or with spinlocks held pri must
498 * NULL is returned on a memory allocation failure.
500 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
503 might_sleep_if(pri & __GFP_WAIT);
504 if (skb_shared(skb)) {
505 struct sk_buff *nskb = skb_clone(skb, pri);
513 * Copy shared buffers into a new sk_buff. We effectively do COW on
514 * packets to handle cases where we have a local reader and forward
515 * and a couple of other messy ones. The normal one is tcpdumping
516 * a packet thats being forwarded.
520 * skb_unshare - make a copy of a shared buffer
521 * @skb: buffer to check
522 * @pri: priority for memory allocation
524 * If the socket buffer is a clone then this function creates a new
525 * copy of the data, drops a reference count on the old copy and returns
526 * the new copy with the reference count at 1. If the buffer is not a clone
527 * the original buffer is returned. When called with a spinlock held or
528 * from interrupt state @pri must be %GFP_ATOMIC
530 * %NULL is returned on a memory allocation failure.
532 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
535 might_sleep_if(pri & __GFP_WAIT);
536 if (skb_cloned(skb)) {
537 struct sk_buff *nskb = skb_copy(skb, pri);
538 kfree_skb(skb); /* Free our shared copy */
546 * @list_: list to peek at
548 * Peek an &sk_buff. Unlike most other operations you _MUST_
549 * be careful with this one. A peek leaves the buffer on the
550 * list and someone else may run off with it. You must hold
551 * the appropriate locks or have a private queue to do this.
553 * Returns %NULL for an empty list or a pointer to the head element.
554 * The reference count is not incremented and the reference is therefore
555 * volatile. Use with caution.
557 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
559 struct sk_buff *list = ((struct sk_buff *)list_)->next;
560 if (list == (struct sk_buff *)list_)
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 tail 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_tail(struct sk_buff_head *list_)
580 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
581 if (list == (struct sk_buff *)list_)
587 * skb_queue_len - get queue length
588 * @list_: list to measure
590 * Return the length of an &sk_buff queue.
592 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
598 * This function creates a split out lock class for each invocation;
599 * this is needed for now since a whole lot of users of the skb-queue
600 * infrastructure in drivers have different locking usage (in hardirq)
601 * than the networking core (in softirq only). In the long run either the
602 * network layer or drivers should need annotation to consolidate the
603 * main types of usage into 3 classes.
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;
612 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
613 struct lock_class_key *class)
615 skb_queue_head_init(list);
616 lockdep_set_class(&list->lock, class);
620 * Insert an sk_buff at the start of a list.
622 * The "__skb_xxxx()" functions are the non-atomic ones that
623 * can only be called with interrupts disabled.
627 * __skb_queue_after - queue a buffer at the list head
629 * @prev: place after this buffer
630 * @newsk: buffer to queue
632 * Queue a buffer int the middle of a list. This function takes no locks
633 * and you must therefore hold required locks before calling it.
635 * A buffer cannot be placed on two lists at the same time.
637 static inline void __skb_queue_after(struct sk_buff_head *list,
638 struct sk_buff *prev,
639 struct sk_buff *newsk)
641 struct sk_buff *next;
647 next->prev = prev->next = newsk;
651 * __skb_queue_head - queue a buffer at the list head
653 * @newsk: buffer to queue
655 * Queue a buffer at the start of a list. This function takes no locks
656 * and you must therefore hold required locks before calling it.
658 * A buffer cannot be placed on two lists at the same time.
660 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
661 static inline void __skb_queue_head(struct sk_buff_head *list,
662 struct sk_buff *newsk)
664 __skb_queue_after(list, (struct sk_buff *)list, newsk);
668 * __skb_queue_tail - queue a buffer at the list tail
670 * @newsk: buffer to queue
672 * Queue a buffer at the end of a list. This function takes no locks
673 * and you must therefore hold required locks before calling it.
675 * A buffer cannot be placed on two lists at the same time.
677 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
678 static inline void __skb_queue_tail(struct sk_buff_head *list,
679 struct sk_buff *newsk)
681 struct sk_buff *prev, *next;
684 next = (struct sk_buff *)list;
688 next->prev = prev->next = newsk;
693 * __skb_dequeue - remove from the head of the queue
694 * @list: list to dequeue from
696 * Remove the head of the list. This function does not take any locks
697 * so must be used with appropriate locks held only. The head item is
698 * returned or %NULL if the list is empty.
700 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
701 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
703 struct sk_buff *next, *prev, *result;
705 prev = (struct sk_buff *) list;
714 result->next = result->prev = NULL;
721 * Insert a packet on a list.
723 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
724 static inline void __skb_insert(struct sk_buff *newsk,
725 struct sk_buff *prev, struct sk_buff *next,
726 struct sk_buff_head *list)
730 next->prev = prev->next = newsk;
735 * Place a packet after a given packet in a list.
737 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
738 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
740 __skb_insert(newsk, old, old->next, list);
744 * remove sk_buff from list. _Must_ be called atomically, and with
747 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
748 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
750 struct sk_buff *next, *prev;
755 skb->next = skb->prev = NULL;
761 /* XXX: more streamlined implementation */
764 * __skb_dequeue_tail - remove from the tail of the queue
765 * @list: list to dequeue from
767 * Remove the tail of the list. This function does not take any locks
768 * so must be used with appropriate locks held only. The tail item is
769 * returned or %NULL if the list is empty.
771 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
772 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
774 struct sk_buff *skb = skb_peek_tail(list);
776 __skb_unlink(skb, list);
781 static inline int skb_is_nonlinear(const struct sk_buff *skb)
783 return skb->data_len;
786 static inline unsigned int skb_headlen(const struct sk_buff *skb)
788 return skb->len - skb->data_len;
791 static inline int skb_pagelen(const struct sk_buff *skb)
795 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
796 len += skb_shinfo(skb)->frags[i].size;
797 return len + skb_headlen(skb);
800 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
801 struct page *page, int off, int size)
803 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
806 frag->page_offset = off;
808 skb_shinfo(skb)->nr_frags = i + 1;
811 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
812 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
813 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
816 * Add data to an sk_buff
818 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
820 unsigned char *tmp = skb->tail;
821 SKB_LINEAR_ASSERT(skb);
828 * skb_put - add data to a buffer
829 * @skb: buffer to use
830 * @len: amount of data to add
832 * This function extends the used data area of the buffer. If this would
833 * exceed the total buffer size the kernel will panic. A pointer to the
834 * first byte of the extra data is returned.
836 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
838 unsigned char *tmp = skb->tail;
839 SKB_LINEAR_ASSERT(skb);
842 if (unlikely(skb->tail>skb->end))
843 skb_over_panic(skb, len, current_text_addr());
847 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
855 * skb_push - add data to the start of a buffer
856 * @skb: buffer to use
857 * @len: amount of data to add
859 * This function extends the used data area of the buffer at the buffer
860 * start. If this would exceed the total buffer headroom the kernel will
861 * panic. A pointer to the first byte of the extra data is returned.
863 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
867 if (unlikely(skb->data<skb->head))
868 skb_under_panic(skb, len, current_text_addr());
872 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
875 BUG_ON(skb->len < skb->data_len);
876 return skb->data += len;
880 * skb_pull - remove data from the start of a buffer
881 * @skb: buffer to use
882 * @len: amount of data to remove
884 * This function removes data from the start of a buffer, returning
885 * the memory to the headroom. A pointer to the next data in the buffer
886 * is returned. Once the data has been pulled future pushes will overwrite
889 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
891 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
894 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
896 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
898 if (len > skb_headlen(skb) &&
899 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
902 return skb->data += len;
905 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
907 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
910 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
912 if (likely(len <= skb_headlen(skb)))
914 if (unlikely(len > skb->len))
916 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
920 * skb_headroom - bytes at buffer head
921 * @skb: buffer to check
923 * Return the number of bytes of free space at the head of an &sk_buff.
925 static inline int skb_headroom(const struct sk_buff *skb)
927 return skb->data - skb->head;
931 * skb_tailroom - bytes at buffer end
932 * @skb: buffer to check
934 * Return the number of bytes of free space at the tail of an sk_buff
936 static inline int skb_tailroom(const struct sk_buff *skb)
938 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
942 * skb_reserve - adjust headroom
943 * @skb: buffer to alter
944 * @len: bytes to move
946 * Increase the headroom of an empty &sk_buff by reducing the tail
947 * room. This is only allowed for an empty buffer.
949 static inline void skb_reserve(struct sk_buff *skb, int len)
955 #ifdef NET_SKBUFF_DATA_USES_OFFSET
956 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
958 return skb->head + skb->transport_header;
961 static inline void skb_reset_transport_header(struct sk_buff *skb)
963 skb->transport_header = skb->data - skb->head;
966 static inline void skb_set_transport_header(struct sk_buff *skb,
969 skb_reset_transport_header(skb);
970 skb->transport_header += offset;
973 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
975 return skb->head + skb->network_header;
978 static inline void skb_reset_network_header(struct sk_buff *skb)
980 skb->network_header = skb->data - skb->head;
983 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
985 skb_reset_network_header(skb);
986 skb->network_header += offset;
989 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
991 return skb->head + skb->mac_header;
994 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
996 return skb->mac_header != ~0U;
999 static inline void skb_reset_mac_header(struct sk_buff *skb)
1001 skb->mac_header = skb->data - skb->head;
1004 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1006 skb_reset_mac_header(skb);
1007 skb->mac_header += offset;
1010 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1012 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1014 return skb->transport_header;
1017 static inline void skb_reset_transport_header(struct sk_buff *skb)
1019 skb->transport_header = skb->data;
1022 static inline void skb_set_transport_header(struct sk_buff *skb,
1025 skb->transport_header = skb->data + offset;
1028 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1030 return skb->network_header;
1033 static inline void skb_reset_network_header(struct sk_buff *skb)
1035 skb->network_header = skb->data;
1038 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1040 skb->network_header = skb->data + offset;
1043 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1045 return skb->mac_header;
1048 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1050 return skb->mac_header != NULL;
1053 static inline void skb_reset_mac_header(struct sk_buff *skb)
1055 skb->mac_header = skb->data;
1058 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1060 skb->mac_header = skb->data + offset;
1062 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1064 static inline int skb_transport_offset(const struct sk_buff *skb)
1066 return skb_transport_header(skb) - skb->data;
1069 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1071 return skb->transport_header - skb->network_header;
1074 static inline int skb_network_offset(const struct sk_buff *skb)
1076 return skb_network_header(skb) - skb->data;
1080 * CPUs often take a performance hit when accessing unaligned memory
1081 * locations. The actual performance hit varies, it can be small if the
1082 * hardware handles it or large if we have to take an exception and fix it
1085 * Since an ethernet header is 14 bytes network drivers often end up with
1086 * the IP header at an unaligned offset. The IP header can be aligned by
1087 * shifting the start of the packet by 2 bytes. Drivers should do this
1090 * skb_reserve(NET_IP_ALIGN);
1092 * The downside to this alignment of the IP header is that the DMA is now
1093 * unaligned. On some architectures the cost of an unaligned DMA is high
1094 * and this cost outweighs the gains made by aligning the IP header.
1096 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1099 #ifndef NET_IP_ALIGN
1100 #define NET_IP_ALIGN 2
1104 * The networking layer reserves some headroom in skb data (via
1105 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1106 * the header has to grow. In the default case, if the header has to grow
1107 * 16 bytes or less we avoid the reallocation.
1109 * Unfortunately this headroom changes the DMA alignment of the resulting
1110 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1111 * on some architectures. An architecture can override this value,
1112 * perhaps setting it to a cacheline in size (since that will maintain
1113 * cacheline alignment of the DMA). It must be a power of 2.
1115 * Various parts of the networking layer expect at least 16 bytes of
1116 * headroom, you should not reduce this.
1119 #define NET_SKB_PAD 16
1122 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1124 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1126 if (unlikely(skb->data_len)) {
1131 skb->tail = skb->data + len;
1135 * skb_trim - remove end from a buffer
1136 * @skb: buffer to alter
1139 * Cut the length of a buffer down by removing data from the tail. If
1140 * the buffer is already under the length specified it is not modified.
1141 * The skb must be linear.
1143 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
1146 __skb_trim(skb, len);
1150 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1153 return ___pskb_trim(skb, len);
1154 __skb_trim(skb, len);
1158 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1160 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1164 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1165 * @skb: buffer to alter
1168 * This is identical to pskb_trim except that the caller knows that
1169 * the skb is not cloned so we should never get an error due to out-
1172 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1174 int err = pskb_trim(skb, len);
1179 * skb_orphan - orphan a buffer
1180 * @skb: buffer to orphan
1182 * If a buffer currently has an owner then we call the owner's
1183 * destructor function and make the @skb unowned. The buffer continues
1184 * to exist but is no longer charged to its former owner.
1186 static inline void skb_orphan(struct sk_buff *skb)
1188 if (skb->destructor)
1189 skb->destructor(skb);
1190 skb->destructor = NULL;
1195 * __skb_queue_purge - empty a list
1196 * @list: list to empty
1198 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1199 * the list and one reference dropped. This function does not take the
1200 * list lock and the caller must hold the relevant locks to use it.
1202 extern void skb_queue_purge(struct sk_buff_head *list);
1203 static inline void __skb_queue_purge(struct sk_buff_head *list)
1205 struct sk_buff *skb;
1206 while ((skb = __skb_dequeue(list)) != NULL)
1211 * __dev_alloc_skb - allocate an skbuff for receiving
1212 * @length: length to allocate
1213 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1215 * Allocate a new &sk_buff and assign it a usage count of one. The
1216 * buffer has unspecified headroom built in. Users should allocate
1217 * the headroom they think they need without accounting for the
1218 * built in space. The built in space is used for optimisations.
1220 * %NULL is returned if there is no free memory.
1222 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1225 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1227 skb_reserve(skb, NET_SKB_PAD);
1232 * dev_alloc_skb - allocate an skbuff for receiving
1233 * @length: length to allocate
1235 * Allocate a new &sk_buff and assign it a usage count of one. The
1236 * buffer has unspecified headroom built in. Users should allocate
1237 * the headroom they think they need without accounting for the
1238 * built in space. The built in space is used for optimisations.
1240 * %NULL is returned if there is no free memory. Although this function
1241 * allocates memory it can be called from an interrupt.
1243 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1245 return __dev_alloc_skb(length, GFP_ATOMIC);
1248 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1249 unsigned int length, gfp_t gfp_mask);
1252 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1253 * @dev: network device to receive on
1254 * @length: length to allocate
1256 * Allocate a new &sk_buff and assign it a usage count of one. The
1257 * buffer has unspecified headroom built in. Users should allocate
1258 * the headroom they think they need without accounting for the
1259 * built in space. The built in space is used for optimisations.
1261 * %NULL is returned if there is no free memory. Although this function
1262 * allocates memory it can be called from an interrupt.
1264 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1265 unsigned int length)
1267 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1271 * skb_cow - copy header of skb when it is required
1272 * @skb: buffer to cow
1273 * @headroom: needed headroom
1275 * If the skb passed lacks sufficient headroom or its data part
1276 * is shared, data is reallocated. If reallocation fails, an error
1277 * is returned and original skb is not changed.
1279 * The result is skb with writable area skb->head...skb->tail
1280 * and at least @headroom of space at head.
1282 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1284 int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) -
1290 if (delta || skb_cloned(skb))
1291 return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) &
1292 ~(NET_SKB_PAD-1), 0, GFP_ATOMIC);
1297 * skb_padto - pad an skbuff up to a minimal size
1298 * @skb: buffer to pad
1299 * @len: minimal length
1301 * Pads up a buffer to ensure the trailing bytes exist and are
1302 * blanked. If the buffer already contains sufficient data it
1303 * is untouched. Otherwise it is extended. Returns zero on
1304 * success. The skb is freed on error.
1307 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1309 unsigned int size = skb->len;
1310 if (likely(size >= len))
1312 return skb_pad(skb, len-size);
1315 static inline int skb_add_data(struct sk_buff *skb,
1316 char __user *from, int copy)
1318 const int off = skb->len;
1320 if (skb->ip_summed == CHECKSUM_NONE) {
1322 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1325 skb->csum = csum_block_add(skb->csum, csum, off);
1328 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1331 __skb_trim(skb, off);
1335 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1336 struct page *page, int off)
1339 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1341 return page == frag->page &&
1342 off == frag->page_offset + frag->size;
1347 static inline int __skb_linearize(struct sk_buff *skb)
1349 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1353 * skb_linearize - convert paged skb to linear one
1354 * @skb: buffer to linarize
1356 * If there is no free memory -ENOMEM is returned, otherwise zero
1357 * is returned and the old skb data released.
1359 static inline int skb_linearize(struct sk_buff *skb)
1361 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1365 * skb_linearize_cow - make sure skb is linear and writable
1366 * @skb: buffer to process
1368 * If there is no free memory -ENOMEM is returned, otherwise zero
1369 * is returned and the old skb data released.
1371 static inline int skb_linearize_cow(struct sk_buff *skb)
1373 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1374 __skb_linearize(skb) : 0;
1378 * skb_postpull_rcsum - update checksum for received skb after pull
1379 * @skb: buffer to update
1380 * @start: start of data before pull
1381 * @len: length of data pulled
1383 * After doing a pull on a received packet, you need to call this to
1384 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1385 * CHECKSUM_NONE so that it can be recomputed from scratch.
1388 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1389 const void *start, unsigned int len)
1391 if (skb->ip_summed == CHECKSUM_COMPLETE)
1392 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1395 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1398 * pskb_trim_rcsum - trim received skb and update checksum
1399 * @skb: buffer to trim
1402 * This is exactly the same as pskb_trim except that it ensures the
1403 * checksum of received packets are still valid after the operation.
1406 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1408 if (likely(len >= skb->len))
1410 if (skb->ip_summed == CHECKSUM_COMPLETE)
1411 skb->ip_summed = CHECKSUM_NONE;
1412 return __pskb_trim(skb, len);
1415 #define skb_queue_walk(queue, skb) \
1416 for (skb = (queue)->next; \
1417 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1420 #define skb_queue_reverse_walk(queue, skb) \
1421 for (skb = (queue)->prev; \
1422 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1426 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1427 int noblock, int *err);
1428 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1429 struct poll_table_struct *wait);
1430 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1431 int offset, struct iovec *to,
1433 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1436 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1437 extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1438 unsigned int flags);
1439 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1440 int len, __wsum csum);
1441 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1443 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1444 void *from, int len);
1445 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1446 int offset, u8 *to, int len,
1448 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1449 extern void skb_split(struct sk_buff *skb,
1450 struct sk_buff *skb1, const u32 len);
1452 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1454 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1455 int len, void *buffer)
1457 int hlen = skb_headlen(skb);
1459 if (hlen - offset >= len)
1460 return skb->data + offset;
1462 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1468 extern void skb_init(void);
1469 extern void skb_add_mtu(int mtu);
1472 * skb_get_timestamp - get timestamp from a skb
1473 * @skb: skb to get stamp from
1474 * @stamp: pointer to struct timeval to store stamp in
1476 * Timestamps are stored in the skb as offsets to a base timestamp.
1477 * This function converts the offset back to a struct timeval and stores
1480 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1482 *stamp = ktime_to_timeval(skb->tstamp);
1485 static inline void __net_timestamp(struct sk_buff *skb)
1487 skb->tstamp = ktime_get_real();
1491 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
1492 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
1495 * skb_checksum_complete - Calculate checksum of an entire packet
1496 * @skb: packet to process
1498 * This function calculates the checksum over the entire packet plus
1499 * the value of skb->csum. The latter can be used to supply the
1500 * checksum of a pseudo header as used by TCP/UDP. It returns the
1503 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1504 * this function can be used to verify that checksum on received
1505 * packets. In that case the function should return zero if the
1506 * checksum is correct. In particular, this function will return zero
1507 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1508 * hardware has already verified the correctness of the checksum.
1510 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1512 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1513 __skb_checksum_complete(skb);
1516 #ifdef CONFIG_NETFILTER
1517 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1519 if (nfct && atomic_dec_and_test(&nfct->use))
1520 nfct->destroy(nfct);
1522 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1525 atomic_inc(&nfct->use);
1527 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1528 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1531 atomic_inc(&skb->users);
1533 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1539 #ifdef CONFIG_BRIDGE_NETFILTER
1540 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1542 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1545 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1548 atomic_inc(&nf_bridge->use);
1550 #endif /* CONFIG_BRIDGE_NETFILTER */
1551 static inline void nf_reset(struct sk_buff *skb)
1553 nf_conntrack_put(skb->nfct);
1555 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1556 nf_conntrack_put_reasm(skb->nfct_reasm);
1557 skb->nfct_reasm = NULL;
1559 #ifdef CONFIG_BRIDGE_NETFILTER
1560 nf_bridge_put(skb->nf_bridge);
1561 skb->nf_bridge = NULL;
1565 /* Note: This doesn't put any conntrack and bridge info in dst. */
1566 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1568 dst->nfct = src->nfct;
1569 nf_conntrack_get(src->nfct);
1570 dst->nfctinfo = src->nfctinfo;
1571 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1572 dst->nfct_reasm = src->nfct_reasm;
1573 nf_conntrack_get_reasm(src->nfct_reasm);
1575 #ifdef CONFIG_BRIDGE_NETFILTER
1576 dst->nf_bridge = src->nf_bridge;
1577 nf_bridge_get(src->nf_bridge);
1581 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1583 nf_conntrack_put(dst->nfct);
1584 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1585 nf_conntrack_put_reasm(dst->nfct_reasm);
1587 #ifdef CONFIG_BRIDGE_NETFILTER
1588 nf_bridge_put(dst->nf_bridge);
1590 __nf_copy(dst, src);
1593 #else /* CONFIG_NETFILTER */
1594 static inline void nf_reset(struct sk_buff *skb) {}
1595 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src) {}
1596 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) {}
1597 #endif /* CONFIG_NETFILTER */
1599 #ifdef CONFIG_NETWORK_SECMARK
1600 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1602 to->secmark = from->secmark;
1605 static inline void skb_init_secmark(struct sk_buff *skb)
1610 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1613 static inline void skb_init_secmark(struct sk_buff *skb)
1617 static inline int skb_is_gso(const struct sk_buff *skb)
1619 return skb_shinfo(skb)->gso_size;
1622 #endif /* __KERNEL__ */
1623 #endif /* _LINUX_SKBUFF_H */