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 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
93 void (*destroy)(struct nf_conntrack *);
97 #ifdef CONFIG_BRIDGE_NETFILTER
98 struct nf_bridge_info {
100 struct net_device *physindev;
101 struct net_device *physoutdev;
102 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
103 struct net_device *netoutdev;
106 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 gso_size;
139 /* Warning: this field is not always filled in (UFO)! */
140 unsigned short gso_segs;
141 unsigned short gso_type;
143 struct sk_buff *frag_list;
144 skb_frag_t frags[MAX_SKB_FRAGS];
147 /* We divide dataref into two halves. The higher 16 bits hold references
148 * to the payload part of skb->data. The lower 16 bits hold references to
149 * the entire skb->data. It is up to the users of the skb to agree on
150 * where the payload starts.
152 * All users must obey the rule that the skb->data reference count must be
153 * greater than or equal to the payload reference count.
155 * Holding a reference to the payload part means that the user does not
156 * care about modifications to the header part of skb->data.
158 #define SKB_DATAREF_SHIFT 16
159 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
163 SKB_FCLONE_UNAVAILABLE,
169 SKB_GSO_TCPV4 = 1 << 0,
170 SKB_GSO_UDP = 1 << 1,
172 /* This indicates the skb is from an untrusted source. */
173 SKB_GSO_DODGY = 1 << 2,
175 /* This indicates the tcp segment has CWR set. */
176 SKB_GSO_TCP_ECN = 1 << 3,
178 SKB_GSO_TCPV6 = 1 << 4,
181 #if BITS_PER_LONG > 32
182 #define NET_SKBUFF_DATA_USES_OFFSET 1
185 #ifdef NET_SKBUFF_DATA_USES_OFFSET
186 typedef unsigned int sk_buff_data_t;
188 typedef unsigned char *sk_buff_data_t;
192 * struct sk_buff - socket buffer
193 * @next: Next buffer in list
194 * @prev: Previous buffer in list
195 * @sk: Socket we are owned by
196 * @tstamp: Time we arrived
197 * @dev: Device we arrived on/are leaving by
198 * @iif: ifindex of device we arrived on
199 * @h: Transport layer header
200 * @network_header: Network layer header
201 * @mac_header: Link layer header
202 * @dst: destination entry
203 * @sp: the security path, used for xfrm
204 * @cb: Control buffer. Free for use by every layer. Put private vars here
205 * @len: Length of actual data
206 * @data_len: Data length
207 * @mac_len: Length of link layer header
209 * @local_df: allow local fragmentation
210 * @cloned: Head may be cloned (check refcnt to be sure)
211 * @nohdr: Payload reference only, must not modify header
212 * @pkt_type: Packet class
213 * @fclone: skbuff clone status
214 * @ip_summed: Driver fed us an IP checksum
215 * @priority: Packet queueing priority
216 * @users: User count - see {datagram,tcp}.c
217 * @protocol: Packet protocol from driver
218 * @truesize: Buffer size
219 * @head: Head of buffer
220 * @data: Data head pointer
221 * @tail: Tail pointer
223 * @destructor: Destruct function
224 * @mark: Generic packet mark
225 * @nfct: Associated connection, if any
226 * @ipvs_property: skbuff is owned by ipvs
227 * @nfctinfo: Relationship of this skb to the connection
228 * @nfct_reasm: netfilter conntrack re-assembly pointer
229 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
230 * @tc_index: Traffic control index
231 * @tc_verd: traffic control verdict
232 * @dma_cookie: a cookie to one of several possible DMA operations
233 * done by skb DMA functions
234 * @secmark: security marking
238 /* These two members must be first. */
239 struct sk_buff *next;
240 struct sk_buff *prev;
244 struct net_device *dev;
246 /* 4 byte hole on 64 bit*/
248 struct dst_entry *dst;
252 * This is the control buffer. It is free to use for every
253 * layer. Please put your private variables there. If you
254 * want to keep them across layers you have to do a skb_clone()
255 * first. This is owned by whoever has the skb queued ATM.
277 void (*destructor)(struct sk_buff *skb);
278 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
279 struct nf_conntrack *nfct;
280 struct sk_buff *nfct_reasm;
282 #ifdef CONFIG_BRIDGE_NETFILTER
283 struct nf_bridge_info *nf_bridge;
285 #ifdef CONFIG_NET_SCHED
286 __u16 tc_index; /* traffic control index */
287 #ifdef CONFIG_NET_CLS_ACT
288 __u16 tc_verd; /* traffic control verdict */
291 #ifdef CONFIG_NET_DMA
292 dma_cookie_t dma_cookie;
294 #ifdef CONFIG_NETWORK_SECMARK
300 sk_buff_data_t transport_header;
301 sk_buff_data_t network_header;
302 sk_buff_data_t mac_header;
303 /* These elements must be at the end, see alloc_skb() for details. */
308 unsigned int truesize;
314 * Handling routines are only of interest to the kernel
316 #include <linux/slab.h>
318 #include <asm/system.h>
320 extern void kfree_skb(struct sk_buff *skb);
321 extern void __kfree_skb(struct sk_buff *skb);
322 extern struct sk_buff *__alloc_skb(unsigned int size,
323 gfp_t priority, int fclone, int node);
324 static inline struct sk_buff *alloc_skb(unsigned int size,
327 return __alloc_skb(size, priority, 0, -1);
330 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
333 return __alloc_skb(size, priority, 1, -1);
336 extern void kfree_skbmem(struct sk_buff *skb);
337 extern struct sk_buff *skb_clone(struct sk_buff *skb,
339 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
341 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
343 extern int pskb_expand_head(struct sk_buff *skb,
344 int nhead, int ntail,
346 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
347 unsigned int headroom);
348 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
349 int newheadroom, int newtailroom,
351 extern int skb_pad(struct sk_buff *skb, int pad);
352 #define dev_kfree_skb(a) kfree_skb(a)
353 extern void skb_over_panic(struct sk_buff *skb, int len,
355 extern void skb_under_panic(struct sk_buff *skb, int len,
357 extern void skb_truesize_bug(struct sk_buff *skb);
359 static inline void skb_truesize_check(struct sk_buff *skb)
361 if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len))
362 skb_truesize_bug(skb);
365 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
366 int getfrag(void *from, char *to, int offset,
367 int len,int odd, struct sk_buff *skb),
368 void *from, int length);
375 __u32 stepped_offset;
376 struct sk_buff *root_skb;
377 struct sk_buff *cur_skb;
381 extern void skb_prepare_seq_read(struct sk_buff *skb,
382 unsigned int from, unsigned int to,
383 struct skb_seq_state *st);
384 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
385 struct skb_seq_state *st);
386 extern void skb_abort_seq_read(struct skb_seq_state *st);
388 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
389 unsigned int to, struct ts_config *config,
390 struct ts_state *state);
392 #ifdef NET_SKBUFF_DATA_USES_OFFSET
393 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
395 return skb->head + skb->end;
398 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
405 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
408 * skb_queue_empty - check if a queue is empty
411 * Returns true if the queue is empty, false otherwise.
413 static inline int skb_queue_empty(const struct sk_buff_head *list)
415 return list->next == (struct sk_buff *)list;
419 * skb_get - reference buffer
420 * @skb: buffer to reference
422 * Makes another reference to a socket buffer and returns a pointer
425 static inline struct sk_buff *skb_get(struct sk_buff *skb)
427 atomic_inc(&skb->users);
432 * If users == 1, we are the only owner and are can avoid redundant
437 * skb_cloned - is the buffer a clone
438 * @skb: buffer to check
440 * Returns true if the buffer was generated with skb_clone() and is
441 * one of multiple shared copies of the buffer. Cloned buffers are
442 * shared data so must not be written to under normal circumstances.
444 static inline int skb_cloned(const struct sk_buff *skb)
446 return skb->cloned &&
447 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
451 * skb_header_cloned - is the header a clone
452 * @skb: buffer to check
454 * Returns true if modifying the header part of the buffer requires
455 * the data to be copied.
457 static inline int skb_header_cloned(const struct sk_buff *skb)
464 dataref = atomic_read(&skb_shinfo(skb)->dataref);
465 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
470 * skb_header_release - release reference to header
471 * @skb: buffer to operate on
473 * Drop a reference to the header part of the buffer. This is done
474 * by acquiring a payload reference. You must not read from the header
475 * part of skb->data after this.
477 static inline void skb_header_release(struct sk_buff *skb)
481 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
485 * skb_shared - is the buffer shared
486 * @skb: buffer to check
488 * Returns true if more than one person has a reference to this
491 static inline int skb_shared(const struct sk_buff *skb)
493 return atomic_read(&skb->users) != 1;
497 * skb_share_check - check if buffer is shared and if so clone it
498 * @skb: buffer to check
499 * @pri: priority for memory allocation
501 * If the buffer is shared the buffer is cloned and the old copy
502 * drops a reference. A new clone with a single reference is returned.
503 * If the buffer is not shared the original buffer is returned. When
504 * being called from interrupt status or with spinlocks held pri must
507 * NULL is returned on a memory allocation failure.
509 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
512 might_sleep_if(pri & __GFP_WAIT);
513 if (skb_shared(skb)) {
514 struct sk_buff *nskb = skb_clone(skb, pri);
522 * Copy shared buffers into a new sk_buff. We effectively do COW on
523 * packets to handle cases where we have a local reader and forward
524 * and a couple of other messy ones. The normal one is tcpdumping
525 * a packet thats being forwarded.
529 * skb_unshare - make a copy of a shared buffer
530 * @skb: buffer to check
531 * @pri: priority for memory allocation
533 * If the socket buffer is a clone then this function creates a new
534 * copy of the data, drops a reference count on the old copy and returns
535 * the new copy with the reference count at 1. If the buffer is not a clone
536 * the original buffer is returned. When called with a spinlock held or
537 * from interrupt state @pri must be %GFP_ATOMIC
539 * %NULL is returned on a memory allocation failure.
541 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
544 might_sleep_if(pri & __GFP_WAIT);
545 if (skb_cloned(skb)) {
546 struct sk_buff *nskb = skb_copy(skb, pri);
547 kfree_skb(skb); /* Free our shared copy */
555 * @list_: list to peek at
557 * Peek an &sk_buff. Unlike most other operations you _MUST_
558 * be careful with this one. A peek leaves the buffer on the
559 * list and someone else may run off with it. You must hold
560 * the appropriate locks or have a private queue to do this.
562 * Returns %NULL for an empty list or a pointer to the head element.
563 * The reference count is not incremented and the reference is therefore
564 * volatile. Use with caution.
566 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
568 struct sk_buff *list = ((struct sk_buff *)list_)->next;
569 if (list == (struct sk_buff *)list_)
576 * @list_: list to peek at
578 * Peek an &sk_buff. Unlike most other operations you _MUST_
579 * be careful with this one. A peek leaves the buffer on the
580 * list and someone else may run off with it. You must hold
581 * the appropriate locks or have a private queue to do this.
583 * Returns %NULL for an empty list or a pointer to the tail element.
584 * The reference count is not incremented and the reference is therefore
585 * volatile. Use with caution.
587 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
589 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
590 if (list == (struct sk_buff *)list_)
596 * skb_queue_len - get queue length
597 * @list_: list to measure
599 * Return the length of an &sk_buff queue.
601 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
607 * This function creates a split out lock class for each invocation;
608 * this is needed for now since a whole lot of users of the skb-queue
609 * infrastructure in drivers have different locking usage (in hardirq)
610 * than the networking core (in softirq only). In the long run either the
611 * network layer or drivers should need annotation to consolidate the
612 * main types of usage into 3 classes.
614 static inline void skb_queue_head_init(struct sk_buff_head *list)
616 spin_lock_init(&list->lock);
617 list->prev = list->next = (struct sk_buff *)list;
621 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
622 struct lock_class_key *class)
624 skb_queue_head_init(list);
625 lockdep_set_class(&list->lock, class);
629 * Insert an sk_buff at the start of a list.
631 * The "__skb_xxxx()" functions are the non-atomic ones that
632 * can only be called with interrupts disabled.
636 * __skb_queue_after - queue a buffer at the list head
638 * @prev: place after this buffer
639 * @newsk: buffer to queue
641 * Queue a buffer int the middle of a list. This function takes no locks
642 * and you must therefore hold required locks before calling it.
644 * A buffer cannot be placed on two lists at the same time.
646 static inline void __skb_queue_after(struct sk_buff_head *list,
647 struct sk_buff *prev,
648 struct sk_buff *newsk)
650 struct sk_buff *next;
656 next->prev = prev->next = newsk;
660 * __skb_queue_head - queue a buffer at the list head
662 * @newsk: buffer to queue
664 * Queue a buffer at the start of a list. This function takes no locks
665 * and you must therefore hold required locks before calling it.
667 * A buffer cannot be placed on two lists at the same time.
669 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
670 static inline void __skb_queue_head(struct sk_buff_head *list,
671 struct sk_buff *newsk)
673 __skb_queue_after(list, (struct sk_buff *)list, newsk);
677 * __skb_queue_tail - queue a buffer at the list tail
679 * @newsk: buffer to queue
681 * Queue a buffer at the end of a list. This function takes no locks
682 * and you must therefore hold required locks before calling it.
684 * A buffer cannot be placed on two lists at the same time.
686 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
687 static inline void __skb_queue_tail(struct sk_buff_head *list,
688 struct sk_buff *newsk)
690 struct sk_buff *prev, *next;
693 next = (struct sk_buff *)list;
697 next->prev = prev->next = newsk;
702 * __skb_dequeue - remove from the head of the queue
703 * @list: list to dequeue from
705 * Remove the head of the list. This function does not take any locks
706 * so must be used with appropriate locks held only. The head item is
707 * returned or %NULL if the list is empty.
709 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
710 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
712 struct sk_buff *next, *prev, *result;
714 prev = (struct sk_buff *) list;
723 result->next = result->prev = NULL;
730 * Insert a packet on a list.
732 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
733 static inline void __skb_insert(struct sk_buff *newsk,
734 struct sk_buff *prev, struct sk_buff *next,
735 struct sk_buff_head *list)
739 next->prev = prev->next = newsk;
744 * Place a packet after a given packet in a list.
746 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
747 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
749 __skb_insert(newsk, old, old->next, list);
753 * remove sk_buff from list. _Must_ be called atomically, and with
756 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
757 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
759 struct sk_buff *next, *prev;
764 skb->next = skb->prev = NULL;
770 /* XXX: more streamlined implementation */
773 * __skb_dequeue_tail - remove from the tail of the queue
774 * @list: list to dequeue from
776 * Remove the tail of the list. This function does not take any locks
777 * so must be used with appropriate locks held only. The tail item is
778 * returned or %NULL if the list is empty.
780 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
781 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
783 struct sk_buff *skb = skb_peek_tail(list);
785 __skb_unlink(skb, list);
790 static inline int skb_is_nonlinear(const struct sk_buff *skb)
792 return skb->data_len;
795 static inline unsigned int skb_headlen(const struct sk_buff *skb)
797 return skb->len - skb->data_len;
800 static inline int skb_pagelen(const struct sk_buff *skb)
804 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
805 len += skb_shinfo(skb)->frags[i].size;
806 return len + skb_headlen(skb);
809 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
810 struct page *page, int off, int size)
812 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
815 frag->page_offset = off;
817 skb_shinfo(skb)->nr_frags = i + 1;
820 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
821 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
822 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
824 #ifdef NET_SKBUFF_DATA_USES_OFFSET
825 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
827 return skb->head + skb->tail;
830 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
832 skb->tail = skb->data - skb->head;
835 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
837 skb_reset_tail_pointer(skb);
840 #else /* NET_SKBUFF_DATA_USES_OFFSET */
841 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
846 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
848 skb->tail = skb->data;
851 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
853 skb->tail = skb->data + offset;
856 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
859 * Add data to an sk_buff
861 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
863 unsigned char *tmp = skb_tail_pointer(skb);
864 SKB_LINEAR_ASSERT(skb);
871 * skb_put - add data to a buffer
872 * @skb: buffer to use
873 * @len: amount of data to add
875 * This function extends the used data area of the buffer. If this would
876 * exceed the total buffer size the kernel will panic. A pointer to the
877 * first byte of the extra data is returned.
879 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
881 unsigned char *tmp = skb_tail_pointer(skb);
882 SKB_LINEAR_ASSERT(skb);
885 if (unlikely(skb->tail > skb->end))
886 skb_over_panic(skb, len, current_text_addr());
890 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
898 * skb_push - add data to the start of a buffer
899 * @skb: buffer to use
900 * @len: amount of data to add
902 * This function extends the used data area of the buffer at the buffer
903 * start. If this would exceed the total buffer headroom the kernel will
904 * panic. A pointer to the first byte of the extra data is returned.
906 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
910 if (unlikely(skb->data<skb->head))
911 skb_under_panic(skb, len, current_text_addr());
915 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
918 BUG_ON(skb->len < skb->data_len);
919 return skb->data += len;
923 * skb_pull - remove data from the start of a buffer
924 * @skb: buffer to use
925 * @len: amount of data to remove
927 * This function removes data from the start of a buffer, returning
928 * the memory to the headroom. A pointer to the next data in the buffer
929 * is returned. Once the data has been pulled future pushes will overwrite
932 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
934 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
937 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
939 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
941 if (len > skb_headlen(skb) &&
942 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
945 return skb->data += len;
948 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
950 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
953 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
955 if (likely(len <= skb_headlen(skb)))
957 if (unlikely(len > skb->len))
959 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
963 * skb_headroom - bytes at buffer head
964 * @skb: buffer to check
966 * Return the number of bytes of free space at the head of an &sk_buff.
968 static inline int skb_headroom(const struct sk_buff *skb)
970 return skb->data - skb->head;
974 * skb_tailroom - bytes at buffer end
975 * @skb: buffer to check
977 * Return the number of bytes of free space at the tail of an sk_buff
979 static inline int skb_tailroom(const struct sk_buff *skb)
981 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
985 * skb_reserve - adjust headroom
986 * @skb: buffer to alter
987 * @len: bytes to move
989 * Increase the headroom of an empty &sk_buff by reducing the tail
990 * room. This is only allowed for an empty buffer.
992 static inline void skb_reserve(struct sk_buff *skb, int len)
998 #ifdef NET_SKBUFF_DATA_USES_OFFSET
999 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1001 return skb->head + skb->transport_header;
1004 static inline void skb_reset_transport_header(struct sk_buff *skb)
1006 skb->transport_header = skb->data - skb->head;
1009 static inline void skb_set_transport_header(struct sk_buff *skb,
1012 skb_reset_transport_header(skb);
1013 skb->transport_header += offset;
1016 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1018 return skb->head + skb->network_header;
1021 static inline void skb_reset_network_header(struct sk_buff *skb)
1023 skb->network_header = skb->data - skb->head;
1026 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1028 skb_reset_network_header(skb);
1029 skb->network_header += offset;
1032 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1034 return skb->head + skb->mac_header;
1037 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1039 return skb->mac_header != ~0U;
1042 static inline void skb_reset_mac_header(struct sk_buff *skb)
1044 skb->mac_header = skb->data - skb->head;
1047 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1049 skb_reset_mac_header(skb);
1050 skb->mac_header += offset;
1053 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1055 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1057 return skb->transport_header;
1060 static inline void skb_reset_transport_header(struct sk_buff *skb)
1062 skb->transport_header = skb->data;
1065 static inline void skb_set_transport_header(struct sk_buff *skb,
1068 skb->transport_header = skb->data + offset;
1071 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1073 return skb->network_header;
1076 static inline void skb_reset_network_header(struct sk_buff *skb)
1078 skb->network_header = skb->data;
1081 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1083 skb->network_header = skb->data + offset;
1086 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1088 return skb->mac_header;
1091 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1093 return skb->mac_header != NULL;
1096 static inline void skb_reset_mac_header(struct sk_buff *skb)
1098 skb->mac_header = skb->data;
1101 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1103 skb->mac_header = skb->data + offset;
1105 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1107 static inline int skb_transport_offset(const struct sk_buff *skb)
1109 return skb_transport_header(skb) - skb->data;
1112 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1114 return skb->transport_header - skb->network_header;
1117 static inline int skb_network_offset(const struct sk_buff *skb)
1119 return skb_network_header(skb) - skb->data;
1123 * CPUs often take a performance hit when accessing unaligned memory
1124 * locations. The actual performance hit varies, it can be small if the
1125 * hardware handles it or large if we have to take an exception and fix it
1128 * Since an ethernet header is 14 bytes network drivers often end up with
1129 * the IP header at an unaligned offset. The IP header can be aligned by
1130 * shifting the start of the packet by 2 bytes. Drivers should do this
1133 * skb_reserve(NET_IP_ALIGN);
1135 * The downside to this alignment of the IP header is that the DMA is now
1136 * unaligned. On some architectures the cost of an unaligned DMA is high
1137 * and this cost outweighs the gains made by aligning the IP header.
1139 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1142 #ifndef NET_IP_ALIGN
1143 #define NET_IP_ALIGN 2
1147 * The networking layer reserves some headroom in skb data (via
1148 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1149 * the header has to grow. In the default case, if the header has to grow
1150 * 16 bytes or less we avoid the reallocation.
1152 * Unfortunately this headroom changes the DMA alignment of the resulting
1153 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1154 * on some architectures. An architecture can override this value,
1155 * perhaps setting it to a cacheline in size (since that will maintain
1156 * cacheline alignment of the DMA). It must be a power of 2.
1158 * Various parts of the networking layer expect at least 16 bytes of
1159 * headroom, you should not reduce this.
1162 #define NET_SKB_PAD 16
1165 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1167 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1169 if (unlikely(skb->data_len)) {
1174 skb_set_tail_pointer(skb, len);
1178 * skb_trim - remove end from a buffer
1179 * @skb: buffer to alter
1182 * Cut the length of a buffer down by removing data from the tail. If
1183 * the buffer is already under the length specified it is not modified.
1184 * The skb must be linear.
1186 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
1189 __skb_trim(skb, len);
1193 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1196 return ___pskb_trim(skb, len);
1197 __skb_trim(skb, len);
1201 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1203 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1207 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1208 * @skb: buffer to alter
1211 * This is identical to pskb_trim except that the caller knows that
1212 * the skb is not cloned so we should never get an error due to out-
1215 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1217 int err = pskb_trim(skb, len);
1222 * skb_orphan - orphan a buffer
1223 * @skb: buffer to orphan
1225 * If a buffer currently has an owner then we call the owner's
1226 * destructor function and make the @skb unowned. The buffer continues
1227 * to exist but is no longer charged to its former owner.
1229 static inline void skb_orphan(struct sk_buff *skb)
1231 if (skb->destructor)
1232 skb->destructor(skb);
1233 skb->destructor = NULL;
1238 * __skb_queue_purge - empty a list
1239 * @list: list to empty
1241 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1242 * the list and one reference dropped. This function does not take the
1243 * list lock and the caller must hold the relevant locks to use it.
1245 extern void skb_queue_purge(struct sk_buff_head *list);
1246 static inline void __skb_queue_purge(struct sk_buff_head *list)
1248 struct sk_buff *skb;
1249 while ((skb = __skb_dequeue(list)) != NULL)
1254 * __dev_alloc_skb - allocate an skbuff for receiving
1255 * @length: length to allocate
1256 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1258 * Allocate a new &sk_buff and assign it a usage count of one. The
1259 * buffer has unspecified headroom built in. Users should allocate
1260 * the headroom they think they need without accounting for the
1261 * built in space. The built in space is used for optimisations.
1263 * %NULL is returned if there is no free memory.
1265 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1268 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1270 skb_reserve(skb, NET_SKB_PAD);
1275 * dev_alloc_skb - allocate an skbuff for receiving
1276 * @length: length to allocate
1278 * Allocate a new &sk_buff and assign it a usage count of one. The
1279 * buffer has unspecified headroom built in. Users should allocate
1280 * the headroom they think they need without accounting for the
1281 * built in space. The built in space is used for optimisations.
1283 * %NULL is returned if there is no free memory. Although this function
1284 * allocates memory it can be called from an interrupt.
1286 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1288 return __dev_alloc_skb(length, GFP_ATOMIC);
1291 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1292 unsigned int length, gfp_t gfp_mask);
1295 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1296 * @dev: network device to receive on
1297 * @length: length to allocate
1299 * Allocate a new &sk_buff and assign it a usage count of one. The
1300 * buffer has unspecified headroom built in. Users should allocate
1301 * the headroom they think they need without accounting for the
1302 * built in space. The built in space is used for optimisations.
1304 * %NULL is returned if there is no free memory. Although this function
1305 * allocates memory it can be called from an interrupt.
1307 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1308 unsigned int length)
1310 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1314 * skb_cow - copy header of skb when it is required
1315 * @skb: buffer to cow
1316 * @headroom: needed headroom
1318 * If the skb passed lacks sufficient headroom or its data part
1319 * is shared, data is reallocated. If reallocation fails, an error
1320 * is returned and original skb is not changed.
1322 * The result is skb with writable area skb->head...skb->tail
1323 * and at least @headroom of space at head.
1325 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1327 int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) -
1333 if (delta || skb_cloned(skb))
1334 return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) &
1335 ~(NET_SKB_PAD-1), 0, GFP_ATOMIC);
1340 * skb_padto - pad an skbuff up to a minimal size
1341 * @skb: buffer to pad
1342 * @len: minimal length
1344 * Pads up a buffer to ensure the trailing bytes exist and are
1345 * blanked. If the buffer already contains sufficient data it
1346 * is untouched. Otherwise it is extended. Returns zero on
1347 * success. The skb is freed on error.
1350 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1352 unsigned int size = skb->len;
1353 if (likely(size >= len))
1355 return skb_pad(skb, len-size);
1358 static inline int skb_add_data(struct sk_buff *skb,
1359 char __user *from, int copy)
1361 const int off = skb->len;
1363 if (skb->ip_summed == CHECKSUM_NONE) {
1365 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1368 skb->csum = csum_block_add(skb->csum, csum, off);
1371 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1374 __skb_trim(skb, off);
1378 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1379 struct page *page, int off)
1382 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1384 return page == frag->page &&
1385 off == frag->page_offset + frag->size;
1390 static inline int __skb_linearize(struct sk_buff *skb)
1392 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1396 * skb_linearize - convert paged skb to linear one
1397 * @skb: buffer to linarize
1399 * If there is no free memory -ENOMEM is returned, otherwise zero
1400 * is returned and the old skb data released.
1402 static inline int skb_linearize(struct sk_buff *skb)
1404 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1408 * skb_linearize_cow - make sure skb is linear and writable
1409 * @skb: buffer to process
1411 * If there is no free memory -ENOMEM is returned, otherwise zero
1412 * is returned and the old skb data released.
1414 static inline int skb_linearize_cow(struct sk_buff *skb)
1416 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1417 __skb_linearize(skb) : 0;
1421 * skb_postpull_rcsum - update checksum for received skb after pull
1422 * @skb: buffer to update
1423 * @start: start of data before pull
1424 * @len: length of data pulled
1426 * After doing a pull on a received packet, you need to call this to
1427 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1428 * CHECKSUM_NONE so that it can be recomputed from scratch.
1431 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1432 const void *start, unsigned int len)
1434 if (skb->ip_summed == CHECKSUM_COMPLETE)
1435 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1438 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1441 * pskb_trim_rcsum - trim received skb and update checksum
1442 * @skb: buffer to trim
1445 * This is exactly the same as pskb_trim except that it ensures the
1446 * checksum of received packets are still valid after the operation.
1449 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1451 if (likely(len >= skb->len))
1453 if (skb->ip_summed == CHECKSUM_COMPLETE)
1454 skb->ip_summed = CHECKSUM_NONE;
1455 return __pskb_trim(skb, len);
1458 #define skb_queue_walk(queue, skb) \
1459 for (skb = (queue)->next; \
1460 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1463 #define skb_queue_reverse_walk(queue, skb) \
1464 for (skb = (queue)->prev; \
1465 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1469 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1470 int noblock, int *err);
1471 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1472 struct poll_table_struct *wait);
1473 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1474 int offset, struct iovec *to,
1476 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1479 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1480 extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1481 unsigned int flags);
1482 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1483 int len, __wsum csum);
1484 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1486 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1487 void *from, int len);
1488 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1489 int offset, u8 *to, int len,
1491 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1492 extern void skb_split(struct sk_buff *skb,
1493 struct sk_buff *skb1, const u32 len);
1495 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1497 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1498 int len, void *buffer)
1500 int hlen = skb_headlen(skb);
1502 if (hlen - offset >= len)
1503 return skb->data + offset;
1505 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1511 extern void skb_init(void);
1514 * skb_get_timestamp - get timestamp from a skb
1515 * @skb: skb to get stamp from
1516 * @stamp: pointer to struct timeval to store stamp in
1518 * Timestamps are stored in the skb as offsets to a base timestamp.
1519 * This function converts the offset back to a struct timeval and stores
1522 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1524 *stamp = ktime_to_timeval(skb->tstamp);
1527 static inline void __net_timestamp(struct sk_buff *skb)
1529 skb->tstamp = ktime_get_real();
1533 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
1534 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
1537 * skb_checksum_complete - Calculate checksum of an entire packet
1538 * @skb: packet to process
1540 * This function calculates the checksum over the entire packet plus
1541 * the value of skb->csum. The latter can be used to supply the
1542 * checksum of a pseudo header as used by TCP/UDP. It returns the
1545 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1546 * this function can be used to verify that checksum on received
1547 * packets. In that case the function should return zero if the
1548 * checksum is correct. In particular, this function will return zero
1549 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1550 * hardware has already verified the correctness of the checksum.
1552 static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1554 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1555 __skb_checksum_complete(skb);
1558 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1559 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1561 if (nfct && atomic_dec_and_test(&nfct->use))
1562 nfct->destroy(nfct);
1564 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1567 atomic_inc(&nfct->use);
1569 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1572 atomic_inc(&skb->users);
1574 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1580 #ifdef CONFIG_BRIDGE_NETFILTER
1581 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1583 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1586 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1589 atomic_inc(&nf_bridge->use);
1591 #endif /* CONFIG_BRIDGE_NETFILTER */
1592 static inline void nf_reset(struct sk_buff *skb)
1594 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1595 nf_conntrack_put(skb->nfct);
1597 nf_conntrack_put_reasm(skb->nfct_reasm);
1598 skb->nfct_reasm = NULL;
1600 #ifdef CONFIG_BRIDGE_NETFILTER
1601 nf_bridge_put(skb->nf_bridge);
1602 skb->nf_bridge = NULL;
1606 /* Note: This doesn't put any conntrack and bridge info in dst. */
1607 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1609 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1610 dst->nfct = src->nfct;
1611 nf_conntrack_get(src->nfct);
1612 dst->nfctinfo = src->nfctinfo;
1613 dst->nfct_reasm = src->nfct_reasm;
1614 nf_conntrack_get_reasm(src->nfct_reasm);
1616 #ifdef CONFIG_BRIDGE_NETFILTER
1617 dst->nf_bridge = src->nf_bridge;
1618 nf_bridge_get(src->nf_bridge);
1622 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1624 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1625 nf_conntrack_put(dst->nfct);
1626 nf_conntrack_put_reasm(dst->nfct_reasm);
1628 #ifdef CONFIG_BRIDGE_NETFILTER
1629 nf_bridge_put(dst->nf_bridge);
1631 __nf_copy(dst, src);
1634 #ifdef CONFIG_NETWORK_SECMARK
1635 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1637 to->secmark = from->secmark;
1640 static inline void skb_init_secmark(struct sk_buff *skb)
1645 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1648 static inline void skb_init_secmark(struct sk_buff *skb)
1652 static inline int skb_is_gso(const struct sk_buff *skb)
1654 return skb_shinfo(skb)->gso_size;
1657 #endif /* __KERNEL__ */
1658 #endif /* _LINUX_SKBUFF_H */