Merge branch 'docs-next' of git://git.lwn.net/linux-2.6
[linux-2.6/next.git] / include / linux / skbuff.h
blobdf7b23ac66e6dac7610efe67fa95888f0f4e3ae2
1 /*
2 * Definitions for the 'struct sk_buff' memory handlers.
4 * Authors:
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/kmemcheck.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>
26 #include <linux/net.h>
27 #include <linux/textsearch.h>
28 #include <net/checksum.h>
29 #include <linux/rcupdate.h>
30 #include <linux/dmaengine.h>
31 #include <linux/hrtimer.h>
33 /* Don't change this without changing skb_csum_unnecessary! */
34 #define CHECKSUM_NONE 0
35 #define CHECKSUM_UNNECESSARY 1
36 #define CHECKSUM_COMPLETE 2
37 #define CHECKSUM_PARTIAL 3
39 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
40 ~(SMP_CACHE_BYTES - 1))
41 #define SKB_WITH_OVERHEAD(X) \
42 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
43 #define SKB_MAX_ORDER(X, ORDER) \
44 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
45 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
46 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
48 /* A. Checksumming of received packets by device.
50 * NONE: device failed to checksum this packet.
51 * skb->csum is undefined.
53 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
54 * skb->csum is undefined.
55 * It is bad option, but, unfortunately, many of vendors do this.
56 * Apparently with secret goal to sell you new device, when you
57 * will add new protocol to your host. F.e. IPv6. 8)
59 * COMPLETE: the most generic way. Device supplied checksum of _all_
60 * the packet as seen by netif_rx in skb->csum.
61 * NOTE: Even if device supports only some protocols, but
62 * is able to produce some skb->csum, it MUST use COMPLETE,
63 * not UNNECESSARY.
65 * PARTIAL: identical to the case for output below. This may occur
66 * on a packet received directly from another Linux OS, e.g.,
67 * a virtualised Linux kernel on the same host. The packet can
68 * be treated in the same way as UNNECESSARY except that on
69 * output (i.e., forwarding) the checksum must be filled in
70 * by the OS or the hardware.
72 * B. Checksumming on output.
74 * NONE: skb is checksummed by protocol or csum is not required.
76 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
77 * from skb->csum_start to the end and to record the checksum
78 * at skb->csum_start + skb->csum_offset.
80 * Device must show its capabilities in dev->features, set
81 * at device setup time.
82 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
83 * everything.
84 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
85 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
86 * TCP/UDP over IPv4. Sigh. Vendors like this
87 * way by an unknown reason. Though, see comment above
88 * about CHECKSUM_UNNECESSARY. 8)
89 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
91 * Any questions? No questions, good. --ANK
94 struct net_device;
95 struct scatterlist;
96 struct pipe_inode_info;
98 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
99 struct nf_conntrack {
100 atomic_t use;
102 #endif
104 #ifdef CONFIG_BRIDGE_NETFILTER
105 struct nf_bridge_info {
106 atomic_t use;
107 struct net_device *physindev;
108 struct net_device *physoutdev;
109 unsigned int mask;
110 unsigned long data[32 / sizeof(unsigned long)];
112 #endif
114 struct sk_buff_head {
115 /* These two members must be first. */
116 struct sk_buff *next;
117 struct sk_buff *prev;
119 __u32 qlen;
120 spinlock_t lock;
123 struct sk_buff;
125 /* To allow 64K frame to be packed as single skb without frag_list */
126 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
128 typedef struct skb_frag_struct skb_frag_t;
130 struct skb_frag_struct {
131 struct page *page;
132 __u32 page_offset;
133 __u32 size;
136 #define HAVE_HW_TIME_STAMP
139 * struct skb_shared_hwtstamps - hardware time stamps
140 * @hwtstamp: hardware time stamp transformed into duration
141 * since arbitrary point in time
142 * @syststamp: hwtstamp transformed to system time base
144 * Software time stamps generated by ktime_get_real() are stored in
145 * skb->tstamp. The relation between the different kinds of time
146 * stamps is as follows:
148 * syststamp and tstamp can be compared against each other in
149 * arbitrary combinations. The accuracy of a
150 * syststamp/tstamp/"syststamp from other device" comparison is
151 * limited by the accuracy of the transformation into system time
152 * base. This depends on the device driver and its underlying
153 * hardware.
155 * hwtstamps can only be compared against other hwtstamps from
156 * the same device.
158 * This structure is attached to packets as part of the
159 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
161 struct skb_shared_hwtstamps {
162 ktime_t hwtstamp;
163 ktime_t syststamp;
167 * struct skb_shared_tx - instructions for time stamping of outgoing packets
168 * @hardware: generate hardware time stamp
169 * @software: generate software time stamp
170 * @in_progress: device driver is going to provide
171 * hardware time stamp
172 * @flags: all shared_tx flags
174 * These flags are attached to packets as part of the
175 * &skb_shared_info. Use skb_tx() to get a pointer.
177 union skb_shared_tx {
178 struct {
179 __u8 hardware:1,
180 software:1,
181 in_progress:1;
183 __u8 flags;
186 /* This data is invariant across clones and lives at
187 * the end of the header data, ie. at skb->end.
189 struct skb_shared_info {
190 atomic_t dataref;
191 unsigned short nr_frags;
192 unsigned short gso_size;
193 #ifdef CONFIG_HAS_DMA
194 dma_addr_t dma_head;
195 #endif
196 /* Warning: this field is not always filled in (UFO)! */
197 unsigned short gso_segs;
198 unsigned short gso_type;
199 __be32 ip6_frag_id;
200 union skb_shared_tx tx_flags;
201 struct sk_buff *frag_list;
202 struct skb_shared_hwtstamps hwtstamps;
203 skb_frag_t frags[MAX_SKB_FRAGS];
204 #ifdef CONFIG_HAS_DMA
205 dma_addr_t dma_maps[MAX_SKB_FRAGS];
206 #endif
207 /* Intermediate layers must ensure that destructor_arg
208 * remains valid until skb destructor */
209 void * destructor_arg;
212 /* We divide dataref into two halves. The higher 16 bits hold references
213 * to the payload part of skb->data. The lower 16 bits hold references to
214 * the entire skb->data. A clone of a headerless skb holds the length of
215 * the header in skb->hdr_len.
217 * All users must obey the rule that the skb->data reference count must be
218 * greater than or equal to the payload reference count.
220 * Holding a reference to the payload part means that the user does not
221 * care about modifications to the header part of skb->data.
223 #define SKB_DATAREF_SHIFT 16
224 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
227 enum {
228 SKB_FCLONE_UNAVAILABLE,
229 SKB_FCLONE_ORIG,
230 SKB_FCLONE_CLONE,
233 enum {
234 SKB_GSO_TCPV4 = 1 << 0,
235 SKB_GSO_UDP = 1 << 1,
237 /* This indicates the skb is from an untrusted source. */
238 SKB_GSO_DODGY = 1 << 2,
240 /* This indicates the tcp segment has CWR set. */
241 SKB_GSO_TCP_ECN = 1 << 3,
243 SKB_GSO_TCPV6 = 1 << 4,
245 SKB_GSO_FCOE = 1 << 5,
248 #if BITS_PER_LONG > 32
249 #define NET_SKBUFF_DATA_USES_OFFSET 1
250 #endif
252 #ifdef NET_SKBUFF_DATA_USES_OFFSET
253 typedef unsigned int sk_buff_data_t;
254 #else
255 typedef unsigned char *sk_buff_data_t;
256 #endif
258 /**
259 * struct sk_buff - socket buffer
260 * @next: Next buffer in list
261 * @prev: Previous buffer in list
262 * @sk: Socket we are owned by
263 * @tstamp: Time we arrived
264 * @dev: Device we arrived on/are leaving by
265 * @transport_header: Transport layer header
266 * @network_header: Network layer header
267 * @mac_header: Link layer header
268 * @_skb_dst: destination entry
269 * @sp: the security path, used for xfrm
270 * @cb: Control buffer. Free for use by every layer. Put private vars here
271 * @len: Length of actual data
272 * @data_len: Data length
273 * @mac_len: Length of link layer header
274 * @hdr_len: writable header length of cloned skb
275 * @csum: Checksum (must include start/offset pair)
276 * @csum_start: Offset from skb->head where checksumming should start
277 * @csum_offset: Offset from csum_start where checksum should be stored
278 * @local_df: allow local fragmentation
279 * @cloned: Head may be cloned (check refcnt to be sure)
280 * @nohdr: Payload reference only, must not modify header
281 * @pkt_type: Packet class
282 * @fclone: skbuff clone status
283 * @ip_summed: Driver fed us an IP checksum
284 * @priority: Packet queueing priority
285 * @users: User count - see {datagram,tcp}.c
286 * @protocol: Packet protocol from driver
287 * @truesize: Buffer size
288 * @head: Head of buffer
289 * @data: Data head pointer
290 * @tail: Tail pointer
291 * @end: End pointer
292 * @destructor: Destruct function
293 * @mark: Generic packet mark
294 * @nfct: Associated connection, if any
295 * @ipvs_property: skbuff is owned by ipvs
296 * @peeked: this packet has been seen already, so stats have been
297 * done for it, don't do them again
298 * @nf_trace: netfilter packet trace flag
299 * @nfctinfo: Relationship of this skb to the connection
300 * @nfct_reasm: netfilter conntrack re-assembly pointer
301 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
302 * @iif: ifindex of device we arrived on
303 * @queue_mapping: Queue mapping for multiqueue devices
304 * @tc_index: Traffic control index
305 * @tc_verd: traffic control verdict
306 * @ndisc_nodetype: router type (from link layer)
307 * @dma_cookie: a cookie to one of several possible DMA operations
308 * done by skb DMA functions
309 * @secmark: security marking
310 * @vlan_tci: vlan tag control information
313 struct sk_buff {
314 /* These two members must be first. */
315 struct sk_buff *next;
316 struct sk_buff *prev;
318 struct sock *sk;
319 ktime_t tstamp;
320 struct net_device *dev;
322 unsigned long _skb_dst;
323 #ifdef CONFIG_XFRM
324 struct sec_path *sp;
325 #endif
327 * This is the control buffer. It is free to use for every
328 * layer. Please put your private variables there. If you
329 * want to keep them across layers you have to do a skb_clone()
330 * first. This is owned by whoever has the skb queued ATM.
332 char cb[48];
334 unsigned int len,
335 data_len;
336 __u16 mac_len,
337 hdr_len;
338 union {
339 __wsum csum;
340 struct {
341 __u16 csum_start;
342 __u16 csum_offset;
345 __u32 priority;
346 kmemcheck_bitfield_begin(flags1);
347 __u8 local_df:1,
348 cloned:1,
349 ip_summed:2,
350 nohdr:1,
351 nfctinfo:3;
352 __u8 pkt_type:3,
353 fclone:2,
354 ipvs_property:1,
355 peeked:1,
356 nf_trace:1;
357 kmemcheck_bitfield_end(flags1);
358 __be16 protocol;
360 void (*destructor)(struct sk_buff *skb);
361 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
362 struct nf_conntrack *nfct;
363 struct sk_buff *nfct_reasm;
364 #endif
365 #ifdef CONFIG_BRIDGE_NETFILTER
366 struct nf_bridge_info *nf_bridge;
367 #endif
369 int iif;
370 __u16 queue_mapping;
371 #ifdef CONFIG_NET_SCHED
372 __u16 tc_index; /* traffic control index */
373 #ifdef CONFIG_NET_CLS_ACT
374 __u16 tc_verd; /* traffic control verdict */
375 #endif
376 #endif
378 kmemcheck_bitfield_begin(flags2);
379 #ifdef CONFIG_IPV6_NDISC_NODETYPE
380 __u8 ndisc_nodetype:2;
381 #endif
382 kmemcheck_bitfield_end(flags2);
384 /* 0/14 bit hole */
386 #ifdef CONFIG_NET_DMA
387 dma_cookie_t dma_cookie;
388 #endif
389 #ifdef CONFIG_NETWORK_SECMARK
390 __u32 secmark;
391 #endif
393 __u32 mark;
395 __u16 vlan_tci;
397 sk_buff_data_t transport_header;
398 sk_buff_data_t network_header;
399 sk_buff_data_t mac_header;
400 /* These elements must be at the end, see alloc_skb() for details. */
401 sk_buff_data_t tail;
402 sk_buff_data_t end;
403 unsigned char *head,
404 *data;
405 unsigned int truesize;
406 atomic_t users;
409 #ifdef __KERNEL__
411 * Handling routines are only of interest to the kernel
413 #include <linux/slab.h>
415 #include <asm/system.h>
417 #ifdef CONFIG_HAS_DMA
418 #include <linux/dma-mapping.h>
419 extern int skb_dma_map(struct device *dev, struct sk_buff *skb,
420 enum dma_data_direction dir);
421 extern void skb_dma_unmap(struct device *dev, struct sk_buff *skb,
422 enum dma_data_direction dir);
423 #endif
425 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
427 return (struct dst_entry *)skb->_skb_dst;
430 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
432 skb->_skb_dst = (unsigned long)dst;
435 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
437 return (struct rtable *)skb_dst(skb);
440 extern void kfree_skb(struct sk_buff *skb);
441 extern void consume_skb(struct sk_buff *skb);
442 extern void __kfree_skb(struct sk_buff *skb);
443 extern struct sk_buff *__alloc_skb(unsigned int size,
444 gfp_t priority, int fclone, int node);
445 static inline struct sk_buff *alloc_skb(unsigned int size,
446 gfp_t priority)
448 return __alloc_skb(size, priority, 0, -1);
451 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
452 gfp_t priority)
454 return __alloc_skb(size, priority, 1, -1);
457 extern int skb_recycle_check(struct sk_buff *skb, int skb_size);
459 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
460 extern struct sk_buff *skb_clone(struct sk_buff *skb,
461 gfp_t priority);
462 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
463 gfp_t priority);
464 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
465 gfp_t gfp_mask);
466 extern int pskb_expand_head(struct sk_buff *skb,
467 int nhead, int ntail,
468 gfp_t gfp_mask);
469 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
470 unsigned int headroom);
471 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
472 int newheadroom, int newtailroom,
473 gfp_t priority);
474 extern int skb_to_sgvec(struct sk_buff *skb,
475 struct scatterlist *sg, int offset,
476 int len);
477 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
478 struct sk_buff **trailer);
479 extern int skb_pad(struct sk_buff *skb, int pad);
480 #define dev_kfree_skb(a) consume_skb(a)
481 #define dev_consume_skb(a) kfree_skb_clean(a)
482 extern void skb_over_panic(struct sk_buff *skb, int len,
483 void *here);
484 extern void skb_under_panic(struct sk_buff *skb, int len,
485 void *here);
487 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
488 int getfrag(void *from, char *to, int offset,
489 int len,int odd, struct sk_buff *skb),
490 void *from, int length);
492 struct skb_seq_state
494 __u32 lower_offset;
495 __u32 upper_offset;
496 __u32 frag_idx;
497 __u32 stepped_offset;
498 struct sk_buff *root_skb;
499 struct sk_buff *cur_skb;
500 __u8 *frag_data;
503 extern void skb_prepare_seq_read(struct sk_buff *skb,
504 unsigned int from, unsigned int to,
505 struct skb_seq_state *st);
506 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
507 struct skb_seq_state *st);
508 extern void skb_abort_seq_read(struct skb_seq_state *st);
510 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
511 unsigned int to, struct ts_config *config,
512 struct ts_state *state);
514 #ifdef NET_SKBUFF_DATA_USES_OFFSET
515 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
517 return skb->head + skb->end;
519 #else
520 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
522 return skb->end;
524 #endif
526 /* Internal */
527 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
529 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
531 return &skb_shinfo(skb)->hwtstamps;
534 static inline union skb_shared_tx *skb_tx(struct sk_buff *skb)
536 return &skb_shinfo(skb)->tx_flags;
540 * skb_queue_empty - check if a queue is empty
541 * @list: queue head
543 * Returns true if the queue is empty, false otherwise.
545 static inline int skb_queue_empty(const struct sk_buff_head *list)
547 return list->next == (struct sk_buff *)list;
551 * skb_queue_is_last - check if skb is the last entry in the queue
552 * @list: queue head
553 * @skb: buffer
555 * Returns true if @skb is the last buffer on the list.
557 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
558 const struct sk_buff *skb)
560 return (skb->next == (struct sk_buff *) list);
564 * skb_queue_is_first - check if skb is the first entry in the queue
565 * @list: queue head
566 * @skb: buffer
568 * Returns true if @skb is the first buffer on the list.
570 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
571 const struct sk_buff *skb)
573 return (skb->prev == (struct sk_buff *) list);
577 * skb_queue_next - return the next packet in the queue
578 * @list: queue head
579 * @skb: current buffer
581 * Return the next packet in @list after @skb. It is only valid to
582 * call this if skb_queue_is_last() evaluates to false.
584 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
585 const struct sk_buff *skb)
587 /* This BUG_ON may seem severe, but if we just return then we
588 * are going to dereference garbage.
590 BUG_ON(skb_queue_is_last(list, skb));
591 return skb->next;
595 * skb_queue_prev - return the prev packet in the queue
596 * @list: queue head
597 * @skb: current buffer
599 * Return the prev packet in @list before @skb. It is only valid to
600 * call this if skb_queue_is_first() evaluates to false.
602 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
603 const struct sk_buff *skb)
605 /* This BUG_ON may seem severe, but if we just return then we
606 * are going to dereference garbage.
608 BUG_ON(skb_queue_is_first(list, skb));
609 return skb->prev;
613 * skb_get - reference buffer
614 * @skb: buffer to reference
616 * Makes another reference to a socket buffer and returns a pointer
617 * to the buffer.
619 static inline struct sk_buff *skb_get(struct sk_buff *skb)
621 atomic_inc(&skb->users);
622 return skb;
626 * If users == 1, we are the only owner and are can avoid redundant
627 * atomic change.
631 * skb_cloned - is the buffer a clone
632 * @skb: buffer to check
634 * Returns true if the buffer was generated with skb_clone() and is
635 * one of multiple shared copies of the buffer. Cloned buffers are
636 * shared data so must not be written to under normal circumstances.
638 static inline int skb_cloned(const struct sk_buff *skb)
640 return skb->cloned &&
641 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
645 * skb_header_cloned - is the header a clone
646 * @skb: buffer to check
648 * Returns true if modifying the header part of the buffer requires
649 * the data to be copied.
651 static inline int skb_header_cloned(const struct sk_buff *skb)
653 int dataref;
655 if (!skb->cloned)
656 return 0;
658 dataref = atomic_read(&skb_shinfo(skb)->dataref);
659 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
660 return dataref != 1;
664 * skb_header_release - release reference to header
665 * @skb: buffer to operate on
667 * Drop a reference to the header part of the buffer. This is done
668 * by acquiring a payload reference. You must not read from the header
669 * part of skb->data after this.
671 static inline void skb_header_release(struct sk_buff *skb)
673 BUG_ON(skb->nohdr);
674 skb->nohdr = 1;
675 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
679 * skb_shared - is the buffer shared
680 * @skb: buffer to check
682 * Returns true if more than one person has a reference to this
683 * buffer.
685 static inline int skb_shared(const struct sk_buff *skb)
687 return atomic_read(&skb->users) != 1;
691 * skb_share_check - check if buffer is shared and if so clone it
692 * @skb: buffer to check
693 * @pri: priority for memory allocation
695 * If the buffer is shared the buffer is cloned and the old copy
696 * drops a reference. A new clone with a single reference is returned.
697 * If the buffer is not shared the original buffer is returned. When
698 * being called from interrupt status or with spinlocks held pri must
699 * be GFP_ATOMIC.
701 * NULL is returned on a memory allocation failure.
703 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
704 gfp_t pri)
706 might_sleep_if(pri & __GFP_WAIT);
707 if (skb_shared(skb)) {
708 struct sk_buff *nskb = skb_clone(skb, pri);
709 kfree_skb(skb);
710 skb = nskb;
712 return skb;
716 * Copy shared buffers into a new sk_buff. We effectively do COW on
717 * packets to handle cases where we have a local reader and forward
718 * and a couple of other messy ones. The normal one is tcpdumping
719 * a packet thats being forwarded.
723 * skb_unshare - make a copy of a shared buffer
724 * @skb: buffer to check
725 * @pri: priority for memory allocation
727 * If the socket buffer is a clone then this function creates a new
728 * copy of the data, drops a reference count on the old copy and returns
729 * the new copy with the reference count at 1. If the buffer is not a clone
730 * the original buffer is returned. When called with a spinlock held or
731 * from interrupt state @pri must be %GFP_ATOMIC
733 * %NULL is returned on a memory allocation failure.
735 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
736 gfp_t pri)
738 might_sleep_if(pri & __GFP_WAIT);
739 if (skb_cloned(skb)) {
740 struct sk_buff *nskb = skb_copy(skb, pri);
741 kfree_skb(skb); /* Free our shared copy */
742 skb = nskb;
744 return skb;
748 * skb_peek
749 * @list_: list to peek at
751 * Peek an &sk_buff. Unlike most other operations you _MUST_
752 * be careful with this one. A peek leaves the buffer on the
753 * list and someone else may run off with it. You must hold
754 * the appropriate locks or have a private queue to do this.
756 * Returns %NULL for an empty list or a pointer to the head element.
757 * The reference count is not incremented and the reference is therefore
758 * volatile. Use with caution.
760 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
762 struct sk_buff *list = ((struct sk_buff *)list_)->next;
763 if (list == (struct sk_buff *)list_)
764 list = NULL;
765 return list;
769 * skb_peek_tail
770 * @list_: list to peek at
772 * Peek an &sk_buff. Unlike most other operations you _MUST_
773 * be careful with this one. A peek leaves the buffer on the
774 * list and someone else may run off with it. You must hold
775 * the appropriate locks or have a private queue to do this.
777 * Returns %NULL for an empty list or a pointer to the tail element.
778 * The reference count is not incremented and the reference is therefore
779 * volatile. Use with caution.
781 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
783 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
784 if (list == (struct sk_buff *)list_)
785 list = NULL;
786 return list;
790 * skb_queue_len - get queue length
791 * @list_: list to measure
793 * Return the length of an &sk_buff queue.
795 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
797 return list_->qlen;
801 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
802 * @list: queue to initialize
804 * This initializes only the list and queue length aspects of
805 * an sk_buff_head object. This allows to initialize the list
806 * aspects of an sk_buff_head without reinitializing things like
807 * the spinlock. It can also be used for on-stack sk_buff_head
808 * objects where the spinlock is known to not be used.
810 static inline void __skb_queue_head_init(struct sk_buff_head *list)
812 list->prev = list->next = (struct sk_buff *)list;
813 list->qlen = 0;
817 * This function creates a split out lock class for each invocation;
818 * this is needed for now since a whole lot of users of the skb-queue
819 * infrastructure in drivers have different locking usage (in hardirq)
820 * than the networking core (in softirq only). In the long run either the
821 * network layer or drivers should need annotation to consolidate the
822 * main types of usage into 3 classes.
824 static inline void skb_queue_head_init(struct sk_buff_head *list)
826 spin_lock_init(&list->lock);
827 __skb_queue_head_init(list);
830 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
831 struct lock_class_key *class)
833 skb_queue_head_init(list);
834 lockdep_set_class(&list->lock, class);
838 * Insert an sk_buff on a list.
840 * The "__skb_xxxx()" functions are the non-atomic ones that
841 * can only be called with interrupts disabled.
843 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
844 static inline void __skb_insert(struct sk_buff *newsk,
845 struct sk_buff *prev, struct sk_buff *next,
846 struct sk_buff_head *list)
848 newsk->next = next;
849 newsk->prev = prev;
850 next->prev = prev->next = newsk;
851 list->qlen++;
854 static inline void __skb_queue_splice(const struct sk_buff_head *list,
855 struct sk_buff *prev,
856 struct sk_buff *next)
858 struct sk_buff *first = list->next;
859 struct sk_buff *last = list->prev;
861 first->prev = prev;
862 prev->next = first;
864 last->next = next;
865 next->prev = last;
869 * skb_queue_splice - join two skb lists, this is designed for stacks
870 * @list: the new list to add
871 * @head: the place to add it in the first list
873 static inline void skb_queue_splice(const struct sk_buff_head *list,
874 struct sk_buff_head *head)
876 if (!skb_queue_empty(list)) {
877 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
878 head->qlen += list->qlen;
883 * skb_queue_splice - join two skb lists and reinitialise the emptied list
884 * @list: the new list to add
885 * @head: the place to add it in the first list
887 * The list at @list is reinitialised
889 static inline void skb_queue_splice_init(struct sk_buff_head *list,
890 struct sk_buff_head *head)
892 if (!skb_queue_empty(list)) {
893 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
894 head->qlen += list->qlen;
895 __skb_queue_head_init(list);
900 * skb_queue_splice_tail - join two skb lists, each list being a queue
901 * @list: the new list to add
902 * @head: the place to add it in the first list
904 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
905 struct sk_buff_head *head)
907 if (!skb_queue_empty(list)) {
908 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
909 head->qlen += list->qlen;
914 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
915 * @list: the new list to add
916 * @head: the place to add it in the first list
918 * Each of the lists is a queue.
919 * The list at @list is reinitialised
921 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
922 struct sk_buff_head *head)
924 if (!skb_queue_empty(list)) {
925 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
926 head->qlen += list->qlen;
927 __skb_queue_head_init(list);
932 * __skb_queue_after - queue a buffer at the list head
933 * @list: list to use
934 * @prev: place after this buffer
935 * @newsk: buffer to queue
937 * Queue a buffer int the middle of a list. This function takes no locks
938 * and you must therefore hold required locks before calling it.
940 * A buffer cannot be placed on two lists at the same time.
942 static inline void __skb_queue_after(struct sk_buff_head *list,
943 struct sk_buff *prev,
944 struct sk_buff *newsk)
946 __skb_insert(newsk, prev, prev->next, list);
949 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
950 struct sk_buff_head *list);
952 static inline void __skb_queue_before(struct sk_buff_head *list,
953 struct sk_buff *next,
954 struct sk_buff *newsk)
956 __skb_insert(newsk, next->prev, next, list);
960 * __skb_queue_head - queue a buffer at the list head
961 * @list: list to use
962 * @newsk: buffer to queue
964 * Queue a buffer at the start of a list. This function takes no locks
965 * and you must therefore hold required locks before calling it.
967 * A buffer cannot be placed on two lists at the same time.
969 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
970 static inline void __skb_queue_head(struct sk_buff_head *list,
971 struct sk_buff *newsk)
973 __skb_queue_after(list, (struct sk_buff *)list, newsk);
977 * __skb_queue_tail - queue a buffer at the list tail
978 * @list: list to use
979 * @newsk: buffer to queue
981 * Queue a buffer at the end of a list. This function takes no locks
982 * and you must therefore hold required locks before calling it.
984 * A buffer cannot be placed on two lists at the same time.
986 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
987 static inline void __skb_queue_tail(struct sk_buff_head *list,
988 struct sk_buff *newsk)
990 __skb_queue_before(list, (struct sk_buff *)list, newsk);
994 * remove sk_buff from list. _Must_ be called atomically, and with
995 * the list known..
997 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
998 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1000 struct sk_buff *next, *prev;
1002 list->qlen--;
1003 next = skb->next;
1004 prev = skb->prev;
1005 skb->next = skb->prev = NULL;
1006 next->prev = prev;
1007 prev->next = next;
1011 * __skb_dequeue - remove from the head of the queue
1012 * @list: list to dequeue from
1014 * Remove the head of the list. This function does not take any locks
1015 * so must be used with appropriate locks held only. The head item is
1016 * returned or %NULL if the list is empty.
1018 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1019 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1021 struct sk_buff *skb = skb_peek(list);
1022 if (skb)
1023 __skb_unlink(skb, list);
1024 return skb;
1028 * __skb_dequeue_tail - remove from the tail of the queue
1029 * @list: list to dequeue from
1031 * Remove the tail of the list. This function does not take any locks
1032 * so must be used with appropriate locks held only. The tail item is
1033 * returned or %NULL if the list is empty.
1035 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1036 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1038 struct sk_buff *skb = skb_peek_tail(list);
1039 if (skb)
1040 __skb_unlink(skb, list);
1041 return skb;
1045 static inline int skb_is_nonlinear(const struct sk_buff *skb)
1047 return skb->data_len;
1050 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1052 return skb->len - skb->data_len;
1055 static inline int skb_pagelen(const struct sk_buff *skb)
1057 int i, len = 0;
1059 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1060 len += skb_shinfo(skb)->frags[i].size;
1061 return len + skb_headlen(skb);
1064 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1065 struct page *page, int off, int size)
1067 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1069 frag->page = page;
1070 frag->page_offset = off;
1071 frag->size = size;
1072 skb_shinfo(skb)->nr_frags = i + 1;
1075 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1076 int off, int size);
1078 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1079 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frags(skb))
1080 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1082 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1083 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1085 return skb->head + skb->tail;
1088 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1090 skb->tail = skb->data - skb->head;
1093 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1095 skb_reset_tail_pointer(skb);
1096 skb->tail += offset;
1098 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1099 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1101 return skb->tail;
1104 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1106 skb->tail = skb->data;
1109 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1111 skb->tail = skb->data + offset;
1114 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1117 * Add data to an sk_buff
1119 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1120 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1122 unsigned char *tmp = skb_tail_pointer(skb);
1123 SKB_LINEAR_ASSERT(skb);
1124 skb->tail += len;
1125 skb->len += len;
1126 return tmp;
1129 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1130 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1132 skb->data -= len;
1133 skb->len += len;
1134 return skb->data;
1137 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1138 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1140 skb->len -= len;
1141 BUG_ON(skb->len < skb->data_len);
1142 return skb->data += len;
1145 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1147 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1149 if (len > skb_headlen(skb) &&
1150 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1151 return NULL;
1152 skb->len -= len;
1153 return skb->data += len;
1156 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1158 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1161 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1163 if (likely(len <= skb_headlen(skb)))
1164 return 1;
1165 if (unlikely(len > skb->len))
1166 return 0;
1167 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1171 * skb_headroom - bytes at buffer head
1172 * @skb: buffer to check
1174 * Return the number of bytes of free space at the head of an &sk_buff.
1176 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1178 return skb->data - skb->head;
1182 * skb_tailroom - bytes at buffer end
1183 * @skb: buffer to check
1185 * Return the number of bytes of free space at the tail of an sk_buff
1187 static inline int skb_tailroom(const struct sk_buff *skb)
1189 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1193 * skb_reserve - adjust headroom
1194 * @skb: buffer to alter
1195 * @len: bytes to move
1197 * Increase the headroom of an empty &sk_buff by reducing the tail
1198 * room. This is only allowed for an empty buffer.
1200 static inline void skb_reserve(struct sk_buff *skb, int len)
1202 skb->data += len;
1203 skb->tail += len;
1206 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1207 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1209 return skb->head + skb->transport_header;
1212 static inline void skb_reset_transport_header(struct sk_buff *skb)
1214 skb->transport_header = skb->data - skb->head;
1217 static inline void skb_set_transport_header(struct sk_buff *skb,
1218 const int offset)
1220 skb_reset_transport_header(skb);
1221 skb->transport_header += offset;
1224 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1226 return skb->head + skb->network_header;
1229 static inline void skb_reset_network_header(struct sk_buff *skb)
1231 skb->network_header = skb->data - skb->head;
1234 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1236 skb_reset_network_header(skb);
1237 skb->network_header += offset;
1240 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1242 return skb->head + skb->mac_header;
1245 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1247 return skb->mac_header != ~0U;
1250 static inline void skb_reset_mac_header(struct sk_buff *skb)
1252 skb->mac_header = skb->data - skb->head;
1255 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1257 skb_reset_mac_header(skb);
1258 skb->mac_header += offset;
1261 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1263 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1265 return skb->transport_header;
1268 static inline void skb_reset_transport_header(struct sk_buff *skb)
1270 skb->transport_header = skb->data;
1273 static inline void skb_set_transport_header(struct sk_buff *skb,
1274 const int offset)
1276 skb->transport_header = skb->data + offset;
1279 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1281 return skb->network_header;
1284 static inline void skb_reset_network_header(struct sk_buff *skb)
1286 skb->network_header = skb->data;
1289 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1291 skb->network_header = skb->data + offset;
1294 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1296 return skb->mac_header;
1299 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1301 return skb->mac_header != NULL;
1304 static inline void skb_reset_mac_header(struct sk_buff *skb)
1306 skb->mac_header = skb->data;
1309 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1311 skb->mac_header = skb->data + offset;
1313 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1315 static inline int skb_transport_offset(const struct sk_buff *skb)
1317 return skb_transport_header(skb) - skb->data;
1320 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1322 return skb->transport_header - skb->network_header;
1325 static inline int skb_network_offset(const struct sk_buff *skb)
1327 return skb_network_header(skb) - skb->data;
1331 * CPUs often take a performance hit when accessing unaligned memory
1332 * locations. The actual performance hit varies, it can be small if the
1333 * hardware handles it or large if we have to take an exception and fix it
1334 * in software.
1336 * Since an ethernet header is 14 bytes network drivers often end up with
1337 * the IP header at an unaligned offset. The IP header can be aligned by
1338 * shifting the start of the packet by 2 bytes. Drivers should do this
1339 * with:
1341 * skb_reserve(skb, NET_IP_ALIGN);
1343 * The downside to this alignment of the IP header is that the DMA is now
1344 * unaligned. On some architectures the cost of an unaligned DMA is high
1345 * and this cost outweighs the gains made by aligning the IP header.
1347 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1348 * to be overridden.
1350 #ifndef NET_IP_ALIGN
1351 #define NET_IP_ALIGN 2
1352 #endif
1355 * The networking layer reserves some headroom in skb data (via
1356 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1357 * the header has to grow. In the default case, if the header has to grow
1358 * 32 bytes or less we avoid the reallocation.
1360 * Unfortunately this headroom changes the DMA alignment of the resulting
1361 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1362 * on some architectures. An architecture can override this value,
1363 * perhaps setting it to a cacheline in size (since that will maintain
1364 * cacheline alignment of the DMA). It must be a power of 2.
1366 * Various parts of the networking layer expect at least 32 bytes of
1367 * headroom, you should not reduce this.
1369 #ifndef NET_SKB_PAD
1370 #define NET_SKB_PAD 32
1371 #endif
1373 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1375 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1377 if (unlikely(skb->data_len)) {
1378 WARN_ON(1);
1379 return;
1381 skb->len = len;
1382 skb_set_tail_pointer(skb, len);
1385 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1387 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1389 if (skb->data_len)
1390 return ___pskb_trim(skb, len);
1391 __skb_trim(skb, len);
1392 return 0;
1395 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1397 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1401 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1402 * @skb: buffer to alter
1403 * @len: new length
1405 * This is identical to pskb_trim except that the caller knows that
1406 * the skb is not cloned so we should never get an error due to out-
1407 * of-memory.
1409 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1411 int err = pskb_trim(skb, len);
1412 BUG_ON(err);
1416 * skb_orphan - orphan a buffer
1417 * @skb: buffer to orphan
1419 * If a buffer currently has an owner then we call the owner's
1420 * destructor function and make the @skb unowned. The buffer continues
1421 * to exist but is no longer charged to its former owner.
1423 static inline void skb_orphan(struct sk_buff *skb)
1425 if (skb->destructor)
1426 skb->destructor(skb);
1427 skb->destructor = NULL;
1428 skb->sk = NULL;
1432 * __skb_queue_purge - empty a list
1433 * @list: list to empty
1435 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1436 * the list and one reference dropped. This function does not take the
1437 * list lock and the caller must hold the relevant locks to use it.
1439 extern void skb_queue_purge(struct sk_buff_head *list);
1440 static inline void __skb_queue_purge(struct sk_buff_head *list)
1442 struct sk_buff *skb;
1443 while ((skb = __skb_dequeue(list)) != NULL)
1444 kfree_skb(skb);
1448 * __dev_alloc_skb - allocate an skbuff for receiving
1449 * @length: length to allocate
1450 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1452 * Allocate a new &sk_buff and assign it a usage count of one. The
1453 * buffer has unspecified headroom built in. Users should allocate
1454 * the headroom they think they need without accounting for the
1455 * built in space. The built in space is used for optimisations.
1457 * %NULL is returned if there is no free memory.
1459 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1460 gfp_t gfp_mask)
1462 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1463 if (likely(skb))
1464 skb_reserve(skb, NET_SKB_PAD);
1465 return skb;
1468 extern struct sk_buff *dev_alloc_skb(unsigned int length);
1470 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1471 unsigned int length, gfp_t gfp_mask);
1474 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1475 * @dev: network device to receive on
1476 * @length: length to allocate
1478 * Allocate a new &sk_buff and assign it a usage count of one. The
1479 * buffer has unspecified headroom built in. Users should allocate
1480 * the headroom they think they need without accounting for the
1481 * built in space. The built in space is used for optimisations.
1483 * %NULL is returned if there is no free memory. Although this function
1484 * allocates memory it can be called from an interrupt.
1486 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1487 unsigned int length)
1489 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1492 extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1495 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1496 * @dev: network device to receive on
1498 * Allocate a new page node local to the specified device.
1500 * %NULL is returned if there is no free memory.
1502 static inline struct page *netdev_alloc_page(struct net_device *dev)
1504 return __netdev_alloc_page(dev, GFP_ATOMIC);
1507 static inline void netdev_free_page(struct net_device *dev, struct page *page)
1509 __free_page(page);
1513 * skb_clone_writable - is the header of a clone writable
1514 * @skb: buffer to check
1515 * @len: length up to which to write
1517 * Returns true if modifying the header part of the cloned buffer
1518 * does not requires the data to be copied.
1520 static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1522 return !skb_header_cloned(skb) &&
1523 skb_headroom(skb) + len <= skb->hdr_len;
1526 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1527 int cloned)
1529 int delta = 0;
1531 if (headroom < NET_SKB_PAD)
1532 headroom = NET_SKB_PAD;
1533 if (headroom > skb_headroom(skb))
1534 delta = headroom - skb_headroom(skb);
1536 if (delta || cloned)
1537 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1538 GFP_ATOMIC);
1539 return 0;
1543 * skb_cow - copy header of skb when it is required
1544 * @skb: buffer to cow
1545 * @headroom: needed headroom
1547 * If the skb passed lacks sufficient headroom or its data part
1548 * is shared, data is reallocated. If reallocation fails, an error
1549 * is returned and original skb is not changed.
1551 * The result is skb with writable area skb->head...skb->tail
1552 * and at least @headroom of space at head.
1554 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1556 return __skb_cow(skb, headroom, skb_cloned(skb));
1560 * skb_cow_head - skb_cow but only making the head writable
1561 * @skb: buffer to cow
1562 * @headroom: needed headroom
1564 * This function is identical to skb_cow except that we replace the
1565 * skb_cloned check by skb_header_cloned. It should be used when
1566 * you only need to push on some header and do not need to modify
1567 * the data.
1569 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1571 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1575 * skb_padto - pad an skbuff up to a minimal size
1576 * @skb: buffer to pad
1577 * @len: minimal length
1579 * Pads up a buffer to ensure the trailing bytes exist and are
1580 * blanked. If the buffer already contains sufficient data it
1581 * is untouched. Otherwise it is extended. Returns zero on
1582 * success. The skb is freed on error.
1585 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1587 unsigned int size = skb->len;
1588 if (likely(size >= len))
1589 return 0;
1590 return skb_pad(skb, len - size);
1593 static inline int skb_add_data(struct sk_buff *skb,
1594 char __user *from, int copy)
1596 const int off = skb->len;
1598 if (skb->ip_summed == CHECKSUM_NONE) {
1599 int err = 0;
1600 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1601 copy, 0, &err);
1602 if (!err) {
1603 skb->csum = csum_block_add(skb->csum, csum, off);
1604 return 0;
1606 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1607 return 0;
1609 __skb_trim(skb, off);
1610 return -EFAULT;
1613 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1614 struct page *page, int off)
1616 if (i) {
1617 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1619 return page == frag->page &&
1620 off == frag->page_offset + frag->size;
1622 return 0;
1625 static inline int __skb_linearize(struct sk_buff *skb)
1627 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1631 * skb_linearize - convert paged skb to linear one
1632 * @skb: buffer to linarize
1634 * If there is no free memory -ENOMEM is returned, otherwise zero
1635 * is returned and the old skb data released.
1637 static inline int skb_linearize(struct sk_buff *skb)
1639 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1643 * skb_linearize_cow - make sure skb is linear and writable
1644 * @skb: buffer to process
1646 * If there is no free memory -ENOMEM is returned, otherwise zero
1647 * is returned and the old skb data released.
1649 static inline int skb_linearize_cow(struct sk_buff *skb)
1651 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1652 __skb_linearize(skb) : 0;
1656 * skb_postpull_rcsum - update checksum for received skb after pull
1657 * @skb: buffer to update
1658 * @start: start of data before pull
1659 * @len: length of data pulled
1661 * After doing a pull on a received packet, you need to call this to
1662 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1663 * CHECKSUM_NONE so that it can be recomputed from scratch.
1666 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1667 const void *start, unsigned int len)
1669 if (skb->ip_summed == CHECKSUM_COMPLETE)
1670 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1673 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1676 * pskb_trim_rcsum - trim received skb and update checksum
1677 * @skb: buffer to trim
1678 * @len: new length
1680 * This is exactly the same as pskb_trim except that it ensures the
1681 * checksum of received packets are still valid after the operation.
1684 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1686 if (likely(len >= skb->len))
1687 return 0;
1688 if (skb->ip_summed == CHECKSUM_COMPLETE)
1689 skb->ip_summed = CHECKSUM_NONE;
1690 return __pskb_trim(skb, len);
1693 #define skb_queue_walk(queue, skb) \
1694 for (skb = (queue)->next; \
1695 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1696 skb = skb->next)
1698 #define skb_queue_walk_safe(queue, skb, tmp) \
1699 for (skb = (queue)->next, tmp = skb->next; \
1700 skb != (struct sk_buff *)(queue); \
1701 skb = tmp, tmp = skb->next)
1703 #define skb_queue_walk_from(queue, skb) \
1704 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1705 skb = skb->next)
1707 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1708 for (tmp = skb->next; \
1709 skb != (struct sk_buff *)(queue); \
1710 skb = tmp, tmp = skb->next)
1712 #define skb_queue_reverse_walk(queue, skb) \
1713 for (skb = (queue)->prev; \
1714 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1715 skb = skb->prev)
1718 static inline bool skb_has_frags(const struct sk_buff *skb)
1720 return skb_shinfo(skb)->frag_list != NULL;
1723 static inline void skb_frag_list_init(struct sk_buff *skb)
1725 skb_shinfo(skb)->frag_list = NULL;
1728 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1730 frag->next = skb_shinfo(skb)->frag_list;
1731 skb_shinfo(skb)->frag_list = frag;
1734 #define skb_walk_frags(skb, iter) \
1735 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1737 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1738 int *peeked, int *err);
1739 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1740 int noblock, int *err);
1741 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1742 struct poll_table_struct *wait);
1743 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1744 int offset, struct iovec *to,
1745 int size);
1746 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1747 int hlen,
1748 struct iovec *iov);
1749 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1750 int offset,
1751 const struct iovec *from,
1752 int from_offset,
1753 int len);
1754 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1755 int offset,
1756 const struct iovec *to,
1757 int to_offset,
1758 int size);
1759 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1760 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1761 unsigned int flags);
1762 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1763 int len, __wsum csum);
1764 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1765 void *to, int len);
1766 extern int skb_store_bits(struct sk_buff *skb, int offset,
1767 const void *from, int len);
1768 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1769 int offset, u8 *to, int len,
1770 __wsum csum);
1771 extern int skb_splice_bits(struct sk_buff *skb,
1772 unsigned int offset,
1773 struct pipe_inode_info *pipe,
1774 unsigned int len,
1775 unsigned int flags);
1776 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1777 extern void skb_split(struct sk_buff *skb,
1778 struct sk_buff *skb1, const u32 len);
1779 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1780 int shiftlen);
1782 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1784 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1785 int len, void *buffer)
1787 int hlen = skb_headlen(skb);
1789 if (hlen - offset >= len)
1790 return skb->data + offset;
1792 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1793 return NULL;
1795 return buffer;
1798 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1799 void *to,
1800 const unsigned int len)
1802 memcpy(to, skb->data, len);
1805 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1806 const int offset, void *to,
1807 const unsigned int len)
1809 memcpy(to, skb->data + offset, len);
1812 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1813 const void *from,
1814 const unsigned int len)
1816 memcpy(skb->data, from, len);
1819 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1820 const int offset,
1821 const void *from,
1822 const unsigned int len)
1824 memcpy(skb->data + offset, from, len);
1827 extern void skb_init(void);
1829 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1831 return skb->tstamp;
1835 * skb_get_timestamp - get timestamp from a skb
1836 * @skb: skb to get stamp from
1837 * @stamp: pointer to struct timeval to store stamp in
1839 * Timestamps are stored in the skb as offsets to a base timestamp.
1840 * This function converts the offset back to a struct timeval and stores
1841 * it in stamp.
1843 static inline void skb_get_timestamp(const struct sk_buff *skb,
1844 struct timeval *stamp)
1846 *stamp = ktime_to_timeval(skb->tstamp);
1849 static inline void skb_get_timestampns(const struct sk_buff *skb,
1850 struct timespec *stamp)
1852 *stamp = ktime_to_timespec(skb->tstamp);
1855 static inline void __net_timestamp(struct sk_buff *skb)
1857 skb->tstamp = ktime_get_real();
1860 static inline ktime_t net_timedelta(ktime_t t)
1862 return ktime_sub(ktime_get_real(), t);
1865 static inline ktime_t net_invalid_timestamp(void)
1867 return ktime_set(0, 0);
1871 * skb_tstamp_tx - queue clone of skb with send time stamps
1872 * @orig_skb: the original outgoing packet
1873 * @hwtstamps: hardware time stamps, may be NULL if not available
1875 * If the skb has a socket associated, then this function clones the
1876 * skb (thus sharing the actual data and optional structures), stores
1877 * the optional hardware time stamping information (if non NULL) or
1878 * generates a software time stamp (otherwise), then queues the clone
1879 * to the error queue of the socket. Errors are silently ignored.
1881 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
1882 struct skb_shared_hwtstamps *hwtstamps);
1884 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
1885 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
1887 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
1889 return skb->ip_summed & CHECKSUM_UNNECESSARY;
1893 * skb_checksum_complete - Calculate checksum of an entire packet
1894 * @skb: packet to process
1896 * This function calculates the checksum over the entire packet plus
1897 * the value of skb->csum. The latter can be used to supply the
1898 * checksum of a pseudo header as used by TCP/UDP. It returns the
1899 * checksum.
1901 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1902 * this function can be used to verify that checksum on received
1903 * packets. In that case the function should return zero if the
1904 * checksum is correct. In particular, this function will return zero
1905 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1906 * hardware has already verified the correctness of the checksum.
1908 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
1910 return skb_csum_unnecessary(skb) ?
1911 0 : __skb_checksum_complete(skb);
1914 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1915 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
1916 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1918 if (nfct && atomic_dec_and_test(&nfct->use))
1919 nf_conntrack_destroy(nfct);
1921 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1923 if (nfct)
1924 atomic_inc(&nfct->use);
1926 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1928 if (skb)
1929 atomic_inc(&skb->users);
1931 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1933 if (skb)
1934 kfree_skb(skb);
1936 #endif
1937 #ifdef CONFIG_BRIDGE_NETFILTER
1938 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1940 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1941 kfree(nf_bridge);
1943 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1945 if (nf_bridge)
1946 atomic_inc(&nf_bridge->use);
1948 #endif /* CONFIG_BRIDGE_NETFILTER */
1949 static inline void nf_reset(struct sk_buff *skb)
1951 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1952 nf_conntrack_put(skb->nfct);
1953 skb->nfct = NULL;
1954 nf_conntrack_put_reasm(skb->nfct_reasm);
1955 skb->nfct_reasm = NULL;
1956 #endif
1957 #ifdef CONFIG_BRIDGE_NETFILTER
1958 nf_bridge_put(skb->nf_bridge);
1959 skb->nf_bridge = NULL;
1960 #endif
1963 /* Note: This doesn't put any conntrack and bridge info in dst. */
1964 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1966 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1967 dst->nfct = src->nfct;
1968 nf_conntrack_get(src->nfct);
1969 dst->nfctinfo = src->nfctinfo;
1970 dst->nfct_reasm = src->nfct_reasm;
1971 nf_conntrack_get_reasm(src->nfct_reasm);
1972 #endif
1973 #ifdef CONFIG_BRIDGE_NETFILTER
1974 dst->nf_bridge = src->nf_bridge;
1975 nf_bridge_get(src->nf_bridge);
1976 #endif
1979 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1981 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1982 nf_conntrack_put(dst->nfct);
1983 nf_conntrack_put_reasm(dst->nfct_reasm);
1984 #endif
1985 #ifdef CONFIG_BRIDGE_NETFILTER
1986 nf_bridge_put(dst->nf_bridge);
1987 #endif
1988 __nf_copy(dst, src);
1991 #ifdef CONFIG_NETWORK_SECMARK
1992 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1994 to->secmark = from->secmark;
1997 static inline void skb_init_secmark(struct sk_buff *skb)
1999 skb->secmark = 0;
2001 #else
2002 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2005 static inline void skb_init_secmark(struct sk_buff *skb)
2007 #endif
2009 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2011 skb->queue_mapping = queue_mapping;
2014 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2016 return skb->queue_mapping;
2019 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2021 to->queue_mapping = from->queue_mapping;
2024 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2026 skb->queue_mapping = rx_queue + 1;
2029 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2031 return skb->queue_mapping - 1;
2034 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2036 return (skb->queue_mapping != 0);
2039 extern u16 skb_tx_hash(const struct net_device *dev,
2040 const struct sk_buff *skb);
2042 #ifdef CONFIG_XFRM
2043 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2045 return skb->sp;
2047 #else
2048 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2050 return NULL;
2052 #endif
2054 static inline int skb_is_gso(const struct sk_buff *skb)
2056 return skb_shinfo(skb)->gso_size;
2059 static inline int skb_is_gso_v6(const struct sk_buff *skb)
2061 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2064 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2066 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2068 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2069 * wanted then gso_type will be set. */
2070 struct skb_shared_info *shinfo = skb_shinfo(skb);
2071 if (shinfo->gso_size != 0 && unlikely(shinfo->gso_type == 0)) {
2072 __skb_warn_lro_forwarding(skb);
2073 return true;
2075 return false;
2078 static inline void skb_forward_csum(struct sk_buff *skb)
2080 /* Unfortunately we don't support this one. Any brave souls? */
2081 if (skb->ip_summed == CHECKSUM_COMPLETE)
2082 skb->ip_summed = CHECKSUM_NONE;
2085 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2086 #endif /* __KERNEL__ */
2087 #endif /* _LINUX_SKBUFF_H */