net: skb->dst accessors
[linux-2.6.git] / include / linux / skbuff.h
blob9ef6eb20247bd831e8b3ea450ece3fc76a6dcf38
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/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 /* Don't change this without changing skb_csum_unnecessary! */
33 #define CHECKSUM_NONE 0
34 #define CHECKSUM_UNNECESSARY 1
35 #define CHECKSUM_COMPLETE 2
36 #define CHECKSUM_PARTIAL 3
38 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
39 ~(SMP_CACHE_BYTES - 1))
40 #define SKB_WITH_OVERHEAD(X) \
41 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
42 #define SKB_MAX_ORDER(X, ORDER) \
43 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
44 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
45 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
47 /* A. Checksumming of received packets by device.
49 * NONE: device failed to checksum this packet.
50 * skb->csum is undefined.
52 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
53 * skb->csum is undefined.
54 * It is bad option, but, unfortunately, many of vendors do this.
55 * Apparently with secret goal to sell you new device, when you
56 * will add new protocol to your host. F.e. IPv6. 8)
58 * COMPLETE: the most generic way. Device supplied checksum of _all_
59 * the packet as seen by netif_rx in skb->csum.
60 * NOTE: Even if device supports only some protocols, but
61 * is able to produce some skb->csum, it MUST use COMPLETE,
62 * not UNNECESSARY.
64 * PARTIAL: identical to the case for output below. This may occur
65 * on a packet received directly from another Linux OS, e.g.,
66 * a virtualised Linux kernel on the same host. The packet can
67 * be treated in the same way as UNNECESSARY except that on
68 * output (i.e., forwarding) the checksum must be filled in
69 * by the OS or the hardware.
71 * B. Checksumming on output.
73 * NONE: skb is checksummed by protocol or csum is not required.
75 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
76 * from skb->csum_start to the end and to record the checksum
77 * at skb->csum_start + skb->csum_offset.
79 * Device must show its capabilities in dev->features, set
80 * at device setup time.
81 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
82 * everything.
83 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
84 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
85 * TCP/UDP over IPv4. Sigh. Vendors like this
86 * way by an unknown reason. Though, see comment above
87 * about CHECKSUM_UNNECESSARY. 8)
88 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
90 * Any questions? No questions, good. --ANK
93 struct net_device;
94 struct scatterlist;
95 struct pipe_inode_info;
97 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
98 struct nf_conntrack {
99 atomic_t use;
101 #endif
103 #ifdef CONFIG_BRIDGE_NETFILTER
104 struct nf_bridge_info {
105 atomic_t use;
106 struct net_device *physindev;
107 struct net_device *physoutdev;
108 unsigned int mask;
109 unsigned long data[32 / sizeof(unsigned long)];
111 #endif
113 struct sk_buff_head {
114 /* These two members must be first. */
115 struct sk_buff *next;
116 struct sk_buff *prev;
118 __u32 qlen;
119 spinlock_t lock;
122 struct sk_buff;
124 /* To allow 64K frame to be packed as single skb without frag_list */
125 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
127 typedef struct skb_frag_struct skb_frag_t;
129 struct skb_frag_struct {
130 struct page *page;
131 __u32 page_offset;
132 __u32 size;
135 #define HAVE_HW_TIME_STAMP
138 * struct skb_shared_hwtstamps - hardware time stamps
139 * @hwtstamp: hardware time stamp transformed into duration
140 * since arbitrary point in time
141 * @syststamp: hwtstamp transformed to system time base
143 * Software time stamps generated by ktime_get_real() are stored in
144 * skb->tstamp. The relation between the different kinds of time
145 * stamps is as follows:
147 * syststamp and tstamp can be compared against each other in
148 * arbitrary combinations. The accuracy of a
149 * syststamp/tstamp/"syststamp from other device" comparison is
150 * limited by the accuracy of the transformation into system time
151 * base. This depends on the device driver and its underlying
152 * hardware.
154 * hwtstamps can only be compared against other hwtstamps from
155 * the same device.
157 * This structure is attached to packets as part of the
158 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
160 struct skb_shared_hwtstamps {
161 ktime_t hwtstamp;
162 ktime_t syststamp;
166 * struct skb_shared_tx - instructions for time stamping of outgoing packets
167 * @hardware: generate hardware time stamp
168 * @software: generate software time stamp
169 * @in_progress: device driver is going to provide
170 * hardware time stamp
171 * @flags: all shared_tx flags
173 * These flags are attached to packets as part of the
174 * &skb_shared_info. Use skb_tx() to get a pointer.
176 union skb_shared_tx {
177 struct {
178 __u8 hardware:1,
179 software:1,
180 in_progress:1;
182 __u8 flags;
185 /* This data is invariant across clones and lives at
186 * the end of the header data, ie. at skb->end.
188 struct skb_shared_info {
189 atomic_t dataref;
190 unsigned short nr_frags;
191 unsigned short gso_size;
192 /* Warning: this field is not always filled in (UFO)! */
193 unsigned short gso_segs;
194 unsigned short gso_type;
195 __be32 ip6_frag_id;
196 union skb_shared_tx tx_flags;
197 #ifdef CONFIG_HAS_DMA
198 unsigned int num_dma_maps;
199 #endif
200 struct sk_buff *frag_list;
201 struct skb_shared_hwtstamps hwtstamps;
202 skb_frag_t frags[MAX_SKB_FRAGS];
203 #ifdef CONFIG_HAS_DMA
204 dma_addr_t dma_maps[MAX_SKB_FRAGS + 1];
205 #endif
206 /* Intermediate layers must ensure that destructor_arg
207 * remains valid until skb destructor */
208 void * destructor_arg;
211 /* We divide dataref into two halves. The higher 16 bits hold references
212 * to the payload part of skb->data. The lower 16 bits hold references to
213 * the entire skb->data. A clone of a headerless skb holds the length of
214 * the header in skb->hdr_len.
216 * All users must obey the rule that the skb->data reference count must be
217 * greater than or equal to the payload reference count.
219 * Holding a reference to the payload part means that the user does not
220 * care about modifications to the header part of skb->data.
222 #define SKB_DATAREF_SHIFT 16
223 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
226 enum {
227 SKB_FCLONE_UNAVAILABLE,
228 SKB_FCLONE_ORIG,
229 SKB_FCLONE_CLONE,
232 enum {
233 SKB_GSO_TCPV4 = 1 << 0,
234 SKB_GSO_UDP = 1 << 1,
236 /* This indicates the skb is from an untrusted source. */
237 SKB_GSO_DODGY = 1 << 2,
239 /* This indicates the tcp segment has CWR set. */
240 SKB_GSO_TCP_ECN = 1 << 3,
242 SKB_GSO_TCPV6 = 1 << 4,
244 SKB_GSO_FCOE = 1 << 5,
247 #if BITS_PER_LONG > 32
248 #define NET_SKBUFF_DATA_USES_OFFSET 1
249 #endif
251 #ifdef NET_SKBUFF_DATA_USES_OFFSET
252 typedef unsigned int sk_buff_data_t;
253 #else
254 typedef unsigned char *sk_buff_data_t;
255 #endif
257 /**
258 * struct sk_buff - socket buffer
259 * @next: Next buffer in list
260 * @prev: Previous buffer in list
261 * @sk: Socket we are owned by
262 * @tstamp: Time we arrived
263 * @dev: Device we arrived on/are leaving by
264 * @transport_header: Transport layer header
265 * @network_header: Network layer header
266 * @mac_header: Link layer header
267 * @dst: destination entry
268 * @sp: the security path, used for xfrm
269 * @cb: Control buffer. Free for use by every layer. Put private vars here
270 * @len: Length of actual data
271 * @data_len: Data length
272 * @mac_len: Length of link layer header
273 * @hdr_len: writable header length of cloned skb
274 * @csum: Checksum (must include start/offset pair)
275 * @csum_start: Offset from skb->head where checksumming should start
276 * @csum_offset: Offset from csum_start where checksum should be stored
277 * @local_df: allow local fragmentation
278 * @cloned: Head may be cloned (check refcnt to be sure)
279 * @nohdr: Payload reference only, must not modify header
280 * @pkt_type: Packet class
281 * @fclone: skbuff clone status
282 * @ip_summed: Driver fed us an IP checksum
283 * @priority: Packet queueing priority
284 * @users: User count - see {datagram,tcp}.c
285 * @protocol: Packet protocol from driver
286 * @truesize: Buffer size
287 * @head: Head of buffer
288 * @data: Data head pointer
289 * @tail: Tail pointer
290 * @end: End pointer
291 * @destructor: Destruct function
292 * @mark: Generic packet mark
293 * @nfct: Associated connection, if any
294 * @ipvs_property: skbuff is owned by ipvs
295 * @peeked: this packet has been seen already, so stats have been
296 * done for it, don't do them again
297 * @nf_trace: netfilter packet trace flag
298 * @nfctinfo: Relationship of this skb to the connection
299 * @nfct_reasm: netfilter conntrack re-assembly pointer
300 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
301 * @iif: ifindex of device we arrived on
302 * @queue_mapping: Queue mapping for multiqueue devices
303 * @tc_index: Traffic control index
304 * @tc_verd: traffic control verdict
305 * @ndisc_nodetype: router type (from link layer)
306 * @do_not_encrypt: set to prevent encryption of this frame
307 * @requeue: set to indicate that the wireless core should attempt
308 * a software retry on this frame if we failed to
309 * receive an ACK for it
310 * @dma_cookie: a cookie to one of several possible DMA operations
311 * done by skb DMA functions
312 * @secmark: security marking
313 * @vlan_tci: vlan tag control information
316 struct sk_buff {
317 /* These two members must be first. */
318 struct sk_buff *next;
319 struct sk_buff *prev;
321 struct sock *sk;
322 ktime_t tstamp;
323 struct net_device *dev;
325 union {
326 unsigned long _skb_dst;
328 #ifdef CONFIG_XFRM
329 struct sec_path *sp;
330 #endif
332 * This is the control buffer. It is free to use for every
333 * layer. Please put your private variables there. If you
334 * want to keep them across layers you have to do a skb_clone()
335 * first. This is owned by whoever has the skb queued ATM.
337 char cb[48];
339 unsigned int len,
340 data_len;
341 __u16 mac_len,
342 hdr_len;
343 union {
344 __wsum csum;
345 struct {
346 __u16 csum_start;
347 __u16 csum_offset;
350 __u32 priority;
351 __u8 local_df:1,
352 cloned:1,
353 ip_summed:2,
354 nohdr:1,
355 nfctinfo:3;
356 __u8 pkt_type:3,
357 fclone:2,
358 ipvs_property:1,
359 peeked:1,
360 nf_trace:1;
361 __be16 protocol;
363 void (*destructor)(struct sk_buff *skb);
364 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
365 struct nf_conntrack *nfct;
366 struct sk_buff *nfct_reasm;
367 #endif
368 #ifdef CONFIG_BRIDGE_NETFILTER
369 struct nf_bridge_info *nf_bridge;
370 #endif
372 int iif;
373 __u16 queue_mapping;
374 #ifdef CONFIG_NET_SCHED
375 __u16 tc_index; /* traffic control index */
376 #ifdef CONFIG_NET_CLS_ACT
377 __u16 tc_verd; /* traffic control verdict */
378 #endif
379 #endif
380 #ifdef CONFIG_IPV6_NDISC_NODETYPE
381 __u8 ndisc_nodetype:2;
382 #endif
383 #if defined(CONFIG_MAC80211) || defined(CONFIG_MAC80211_MODULE)
384 __u8 do_not_encrypt:1;
385 __u8 requeue:1;
386 #endif
387 /* 0/13/14 bit hole */
389 #ifdef CONFIG_NET_DMA
390 dma_cookie_t dma_cookie;
391 #endif
392 #ifdef CONFIG_NETWORK_SECMARK
393 __u32 secmark;
394 #endif
396 __u32 mark;
398 __u16 vlan_tci;
400 sk_buff_data_t transport_header;
401 sk_buff_data_t network_header;
402 sk_buff_data_t mac_header;
403 /* These elements must be at the end, see alloc_skb() for details. */
404 sk_buff_data_t tail;
405 sk_buff_data_t end;
406 unsigned char *head,
407 *data;
408 unsigned int truesize;
409 atomic_t users;
412 #ifdef __KERNEL__
414 * Handling routines are only of interest to the kernel
416 #include <linux/slab.h>
418 #include <asm/system.h>
420 #ifdef CONFIG_HAS_DMA
421 #include <linux/dma-mapping.h>
422 extern int skb_dma_map(struct device *dev, struct sk_buff *skb,
423 enum dma_data_direction dir);
424 extern void skb_dma_unmap(struct device *dev, struct sk_buff *skb,
425 enum dma_data_direction dir);
426 #endif
428 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
430 return (struct dst_entry *)skb->_skb_dst;
433 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
435 skb->_skb_dst = (unsigned long)dst;
438 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
440 return (struct rtable *)skb_dst(skb);
443 extern void kfree_skb(struct sk_buff *skb);
444 extern void consume_skb(struct sk_buff *skb);
445 extern void __kfree_skb(struct sk_buff *skb);
446 extern struct sk_buff *__alloc_skb(unsigned int size,
447 gfp_t priority, int fclone, int node);
448 static inline struct sk_buff *alloc_skb(unsigned int size,
449 gfp_t priority)
451 return __alloc_skb(size, priority, 0, -1);
454 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
455 gfp_t priority)
457 return __alloc_skb(size, priority, 1, -1);
460 extern int skb_recycle_check(struct sk_buff *skb, int skb_size);
462 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
463 extern struct sk_buff *skb_clone(struct sk_buff *skb,
464 gfp_t priority);
465 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
466 gfp_t priority);
467 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
468 gfp_t gfp_mask);
469 extern int pskb_expand_head(struct sk_buff *skb,
470 int nhead, int ntail,
471 gfp_t gfp_mask);
472 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
473 unsigned int headroom);
474 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
475 int newheadroom, int newtailroom,
476 gfp_t priority);
477 extern int skb_to_sgvec(struct sk_buff *skb,
478 struct scatterlist *sg, int offset,
479 int len);
480 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
481 struct sk_buff **trailer);
482 extern int skb_pad(struct sk_buff *skb, int pad);
483 #define dev_kfree_skb(a) consume_skb(a)
484 #define dev_consume_skb(a) kfree_skb_clean(a)
485 extern void skb_over_panic(struct sk_buff *skb, int len,
486 void *here);
487 extern void skb_under_panic(struct sk_buff *skb, int len,
488 void *here);
490 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
491 int getfrag(void *from, char *to, int offset,
492 int len,int odd, struct sk_buff *skb),
493 void *from, int length);
495 struct skb_seq_state
497 __u32 lower_offset;
498 __u32 upper_offset;
499 __u32 frag_idx;
500 __u32 stepped_offset;
501 struct sk_buff *root_skb;
502 struct sk_buff *cur_skb;
503 __u8 *frag_data;
506 extern void skb_prepare_seq_read(struct sk_buff *skb,
507 unsigned int from, unsigned int to,
508 struct skb_seq_state *st);
509 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
510 struct skb_seq_state *st);
511 extern void skb_abort_seq_read(struct skb_seq_state *st);
513 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
514 unsigned int to, struct ts_config *config,
515 struct ts_state *state);
517 #ifdef NET_SKBUFF_DATA_USES_OFFSET
518 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
520 return skb->head + skb->end;
522 #else
523 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
525 return skb->end;
527 #endif
529 /* Internal */
530 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
532 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
534 return &skb_shinfo(skb)->hwtstamps;
537 static inline union skb_shared_tx *skb_tx(struct sk_buff *skb)
539 return &skb_shinfo(skb)->tx_flags;
543 * skb_queue_empty - check if a queue is empty
544 * @list: queue head
546 * Returns true if the queue is empty, false otherwise.
548 static inline int skb_queue_empty(const struct sk_buff_head *list)
550 return list->next == (struct sk_buff *)list;
554 * skb_queue_is_last - check if skb is the last entry in the queue
555 * @list: queue head
556 * @skb: buffer
558 * Returns true if @skb is the last buffer on the list.
560 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
561 const struct sk_buff *skb)
563 return (skb->next == (struct sk_buff *) list);
567 * skb_queue_is_first - check if skb is the first entry in the queue
568 * @list: queue head
569 * @skb: buffer
571 * Returns true if @skb is the first buffer on the list.
573 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
574 const struct sk_buff *skb)
576 return (skb->prev == (struct sk_buff *) list);
580 * skb_queue_next - return the next packet in the queue
581 * @list: queue head
582 * @skb: current buffer
584 * Return the next packet in @list after @skb. It is only valid to
585 * call this if skb_queue_is_last() evaluates to false.
587 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
588 const struct sk_buff *skb)
590 /* This BUG_ON may seem severe, but if we just return then we
591 * are going to dereference garbage.
593 BUG_ON(skb_queue_is_last(list, skb));
594 return skb->next;
598 * skb_queue_prev - return the prev packet in the queue
599 * @list: queue head
600 * @skb: current buffer
602 * Return the prev packet in @list before @skb. It is only valid to
603 * call this if skb_queue_is_first() evaluates to false.
605 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
606 const struct sk_buff *skb)
608 /* This BUG_ON may seem severe, but if we just return then we
609 * are going to dereference garbage.
611 BUG_ON(skb_queue_is_first(list, skb));
612 return skb->prev;
616 * skb_get - reference buffer
617 * @skb: buffer to reference
619 * Makes another reference to a socket buffer and returns a pointer
620 * to the buffer.
622 static inline struct sk_buff *skb_get(struct sk_buff *skb)
624 atomic_inc(&skb->users);
625 return skb;
629 * If users == 1, we are the only owner and are can avoid redundant
630 * atomic change.
634 * skb_cloned - is the buffer a clone
635 * @skb: buffer to check
637 * Returns true if the buffer was generated with skb_clone() and is
638 * one of multiple shared copies of the buffer. Cloned buffers are
639 * shared data so must not be written to under normal circumstances.
641 static inline int skb_cloned(const struct sk_buff *skb)
643 return skb->cloned &&
644 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
648 * skb_header_cloned - is the header a clone
649 * @skb: buffer to check
651 * Returns true if modifying the header part of the buffer requires
652 * the data to be copied.
654 static inline int skb_header_cloned(const struct sk_buff *skb)
656 int dataref;
658 if (!skb->cloned)
659 return 0;
661 dataref = atomic_read(&skb_shinfo(skb)->dataref);
662 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
663 return dataref != 1;
667 * skb_header_release - release reference to header
668 * @skb: buffer to operate on
670 * Drop a reference to the header part of the buffer. This is done
671 * by acquiring a payload reference. You must not read from the header
672 * part of skb->data after this.
674 static inline void skb_header_release(struct sk_buff *skb)
676 BUG_ON(skb->nohdr);
677 skb->nohdr = 1;
678 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
682 * skb_shared - is the buffer shared
683 * @skb: buffer to check
685 * Returns true if more than one person has a reference to this
686 * buffer.
688 static inline int skb_shared(const struct sk_buff *skb)
690 return atomic_read(&skb->users) != 1;
694 * skb_share_check - check if buffer is shared and if so clone it
695 * @skb: buffer to check
696 * @pri: priority for memory allocation
698 * If the buffer is shared the buffer is cloned and the old copy
699 * drops a reference. A new clone with a single reference is returned.
700 * If the buffer is not shared the original buffer is returned. When
701 * being called from interrupt status or with spinlocks held pri must
702 * be GFP_ATOMIC.
704 * NULL is returned on a memory allocation failure.
706 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
707 gfp_t pri)
709 might_sleep_if(pri & __GFP_WAIT);
710 if (skb_shared(skb)) {
711 struct sk_buff *nskb = skb_clone(skb, pri);
712 kfree_skb(skb);
713 skb = nskb;
715 return skb;
719 * Copy shared buffers into a new sk_buff. We effectively do COW on
720 * packets to handle cases where we have a local reader and forward
721 * and a couple of other messy ones. The normal one is tcpdumping
722 * a packet thats being forwarded.
726 * skb_unshare - make a copy of a shared buffer
727 * @skb: buffer to check
728 * @pri: priority for memory allocation
730 * If the socket buffer is a clone then this function creates a new
731 * copy of the data, drops a reference count on the old copy and returns
732 * the new copy with the reference count at 1. If the buffer is not a clone
733 * the original buffer is returned. When called with a spinlock held or
734 * from interrupt state @pri must be %GFP_ATOMIC
736 * %NULL is returned on a memory allocation failure.
738 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
739 gfp_t pri)
741 might_sleep_if(pri & __GFP_WAIT);
742 if (skb_cloned(skb)) {
743 struct sk_buff *nskb = skb_copy(skb, pri);
744 kfree_skb(skb); /* Free our shared copy */
745 skb = nskb;
747 return skb;
751 * skb_peek
752 * @list_: list to peek at
754 * Peek an &sk_buff. Unlike most other operations you _MUST_
755 * be careful with this one. A peek leaves the buffer on the
756 * list and someone else may run off with it. You must hold
757 * the appropriate locks or have a private queue to do this.
759 * Returns %NULL for an empty list or a pointer to the head element.
760 * The reference count is not incremented and the reference is therefore
761 * volatile. Use with caution.
763 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
765 struct sk_buff *list = ((struct sk_buff *)list_)->next;
766 if (list == (struct sk_buff *)list_)
767 list = NULL;
768 return list;
772 * skb_peek_tail
773 * @list_: list to peek at
775 * Peek an &sk_buff. Unlike most other operations you _MUST_
776 * be careful with this one. A peek leaves the buffer on the
777 * list and someone else may run off with it. You must hold
778 * the appropriate locks or have a private queue to do this.
780 * Returns %NULL for an empty list or a pointer to the tail element.
781 * The reference count is not incremented and the reference is therefore
782 * volatile. Use with caution.
784 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
786 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
787 if (list == (struct sk_buff *)list_)
788 list = NULL;
789 return list;
793 * skb_queue_len - get queue length
794 * @list_: list to measure
796 * Return the length of an &sk_buff queue.
798 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
800 return list_->qlen;
804 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
805 * @list: queue to initialize
807 * This initializes only the list and queue length aspects of
808 * an sk_buff_head object. This allows to initialize the list
809 * aspects of an sk_buff_head without reinitializing things like
810 * the spinlock. It can also be used for on-stack sk_buff_head
811 * objects where the spinlock is known to not be used.
813 static inline void __skb_queue_head_init(struct sk_buff_head *list)
815 list->prev = list->next = (struct sk_buff *)list;
816 list->qlen = 0;
820 * This function creates a split out lock class for each invocation;
821 * this is needed for now since a whole lot of users of the skb-queue
822 * infrastructure in drivers have different locking usage (in hardirq)
823 * than the networking core (in softirq only). In the long run either the
824 * network layer or drivers should need annotation to consolidate the
825 * main types of usage into 3 classes.
827 static inline void skb_queue_head_init(struct sk_buff_head *list)
829 spin_lock_init(&list->lock);
830 __skb_queue_head_init(list);
833 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
834 struct lock_class_key *class)
836 skb_queue_head_init(list);
837 lockdep_set_class(&list->lock, class);
841 * Insert an sk_buff on a list.
843 * The "__skb_xxxx()" functions are the non-atomic ones that
844 * can only be called with interrupts disabled.
846 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
847 static inline void __skb_insert(struct sk_buff *newsk,
848 struct sk_buff *prev, struct sk_buff *next,
849 struct sk_buff_head *list)
851 newsk->next = next;
852 newsk->prev = prev;
853 next->prev = prev->next = newsk;
854 list->qlen++;
857 static inline void __skb_queue_splice(const struct sk_buff_head *list,
858 struct sk_buff *prev,
859 struct sk_buff *next)
861 struct sk_buff *first = list->next;
862 struct sk_buff *last = list->prev;
864 first->prev = prev;
865 prev->next = first;
867 last->next = next;
868 next->prev = last;
872 * skb_queue_splice - join two skb lists, this is designed for stacks
873 * @list: the new list to add
874 * @head: the place to add it in the first list
876 static inline void skb_queue_splice(const struct sk_buff_head *list,
877 struct sk_buff_head *head)
879 if (!skb_queue_empty(list)) {
880 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
881 head->qlen += list->qlen;
886 * skb_queue_splice - join two skb lists and reinitialise the emptied list
887 * @list: the new list to add
888 * @head: the place to add it in the first list
890 * The list at @list is reinitialised
892 static inline void skb_queue_splice_init(struct sk_buff_head *list,
893 struct sk_buff_head *head)
895 if (!skb_queue_empty(list)) {
896 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
897 head->qlen += list->qlen;
898 __skb_queue_head_init(list);
903 * skb_queue_splice_tail - join two skb lists, each list being a queue
904 * @list: the new list to add
905 * @head: the place to add it in the first list
907 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
908 struct sk_buff_head *head)
910 if (!skb_queue_empty(list)) {
911 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
912 head->qlen += list->qlen;
917 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
918 * @list: the new list to add
919 * @head: the place to add it in the first list
921 * Each of the lists is a queue.
922 * The list at @list is reinitialised
924 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
925 struct sk_buff_head *head)
927 if (!skb_queue_empty(list)) {
928 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
929 head->qlen += list->qlen;
930 __skb_queue_head_init(list);
935 * __skb_queue_after - queue a buffer at the list head
936 * @list: list to use
937 * @prev: place after this buffer
938 * @newsk: buffer to queue
940 * Queue a buffer int the middle of a list. This function takes no locks
941 * and you must therefore hold required locks before calling it.
943 * A buffer cannot be placed on two lists at the same time.
945 static inline void __skb_queue_after(struct sk_buff_head *list,
946 struct sk_buff *prev,
947 struct sk_buff *newsk)
949 __skb_insert(newsk, prev, prev->next, list);
952 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
953 struct sk_buff_head *list);
955 static inline void __skb_queue_before(struct sk_buff_head *list,
956 struct sk_buff *next,
957 struct sk_buff *newsk)
959 __skb_insert(newsk, next->prev, next, list);
963 * __skb_queue_head - queue a buffer at the list head
964 * @list: list to use
965 * @newsk: buffer to queue
967 * Queue a buffer at the start of a list. This function takes no locks
968 * and you must therefore hold required locks before calling it.
970 * A buffer cannot be placed on two lists at the same time.
972 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
973 static inline void __skb_queue_head(struct sk_buff_head *list,
974 struct sk_buff *newsk)
976 __skb_queue_after(list, (struct sk_buff *)list, newsk);
980 * __skb_queue_tail - queue a buffer at the list tail
981 * @list: list to use
982 * @newsk: buffer to queue
984 * Queue a buffer at the end of a list. This function takes no locks
985 * and you must therefore hold required locks before calling it.
987 * A buffer cannot be placed on two lists at the same time.
989 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
990 static inline void __skb_queue_tail(struct sk_buff_head *list,
991 struct sk_buff *newsk)
993 __skb_queue_before(list, (struct sk_buff *)list, newsk);
997 * remove sk_buff from list. _Must_ be called atomically, and with
998 * the list known..
1000 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1001 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1003 struct sk_buff *next, *prev;
1005 list->qlen--;
1006 next = skb->next;
1007 prev = skb->prev;
1008 skb->next = skb->prev = NULL;
1009 next->prev = prev;
1010 prev->next = next;
1014 * __skb_dequeue - remove from the head of the queue
1015 * @list: list to dequeue from
1017 * Remove the head of the list. This function does not take any locks
1018 * so must be used with appropriate locks held only. The head item is
1019 * returned or %NULL if the list is empty.
1021 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1022 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1024 struct sk_buff *skb = skb_peek(list);
1025 if (skb)
1026 __skb_unlink(skb, list);
1027 return skb;
1031 * __skb_dequeue_tail - remove from the tail of the queue
1032 * @list: list to dequeue from
1034 * Remove the tail of the list. This function does not take any locks
1035 * so must be used with appropriate locks held only. The tail item is
1036 * returned or %NULL if the list is empty.
1038 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1039 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1041 struct sk_buff *skb = skb_peek_tail(list);
1042 if (skb)
1043 __skb_unlink(skb, list);
1044 return skb;
1048 static inline int skb_is_nonlinear(const struct sk_buff *skb)
1050 return skb->data_len;
1053 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1055 return skb->len - skb->data_len;
1058 static inline int skb_pagelen(const struct sk_buff *skb)
1060 int i, len = 0;
1062 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1063 len += skb_shinfo(skb)->frags[i].size;
1064 return len + skb_headlen(skb);
1067 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1068 struct page *page, int off, int size)
1070 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1072 frag->page = page;
1073 frag->page_offset = off;
1074 frag->size = size;
1075 skb_shinfo(skb)->nr_frags = i + 1;
1078 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1079 int off, int size);
1081 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1082 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
1083 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1085 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1086 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1088 return skb->head + skb->tail;
1091 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1093 skb->tail = skb->data - skb->head;
1096 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1098 skb_reset_tail_pointer(skb);
1099 skb->tail += offset;
1101 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1102 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1104 return skb->tail;
1107 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1109 skb->tail = skb->data;
1112 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1114 skb->tail = skb->data + offset;
1117 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1120 * Add data to an sk_buff
1122 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1123 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1125 unsigned char *tmp = skb_tail_pointer(skb);
1126 SKB_LINEAR_ASSERT(skb);
1127 skb->tail += len;
1128 skb->len += len;
1129 return tmp;
1132 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1133 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1135 skb->data -= len;
1136 skb->len += len;
1137 return skb->data;
1140 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1141 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1143 skb->len -= len;
1144 BUG_ON(skb->len < skb->data_len);
1145 return skb->data += len;
1148 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1150 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1152 if (len > skb_headlen(skb) &&
1153 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1154 return NULL;
1155 skb->len -= len;
1156 return skb->data += len;
1159 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1161 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1164 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1166 if (likely(len <= skb_headlen(skb)))
1167 return 1;
1168 if (unlikely(len > skb->len))
1169 return 0;
1170 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1174 * skb_headroom - bytes at buffer head
1175 * @skb: buffer to check
1177 * Return the number of bytes of free space at the head of an &sk_buff.
1179 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1181 return skb->data - skb->head;
1185 * skb_tailroom - bytes at buffer end
1186 * @skb: buffer to check
1188 * Return the number of bytes of free space at the tail of an sk_buff
1190 static inline int skb_tailroom(const struct sk_buff *skb)
1192 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1196 * skb_reserve - adjust headroom
1197 * @skb: buffer to alter
1198 * @len: bytes to move
1200 * Increase the headroom of an empty &sk_buff by reducing the tail
1201 * room. This is only allowed for an empty buffer.
1203 static inline void skb_reserve(struct sk_buff *skb, int len)
1205 skb->data += len;
1206 skb->tail += len;
1209 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1210 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1212 return skb->head + skb->transport_header;
1215 static inline void skb_reset_transport_header(struct sk_buff *skb)
1217 skb->transport_header = skb->data - skb->head;
1220 static inline void skb_set_transport_header(struct sk_buff *skb,
1221 const int offset)
1223 skb_reset_transport_header(skb);
1224 skb->transport_header += offset;
1227 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1229 return skb->head + skb->network_header;
1232 static inline void skb_reset_network_header(struct sk_buff *skb)
1234 skb->network_header = skb->data - skb->head;
1237 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1239 skb_reset_network_header(skb);
1240 skb->network_header += offset;
1243 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1245 return skb->head + skb->mac_header;
1248 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1250 return skb->mac_header != ~0U;
1253 static inline void skb_reset_mac_header(struct sk_buff *skb)
1255 skb->mac_header = skb->data - skb->head;
1258 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1260 skb_reset_mac_header(skb);
1261 skb->mac_header += offset;
1264 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1266 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1268 return skb->transport_header;
1271 static inline void skb_reset_transport_header(struct sk_buff *skb)
1273 skb->transport_header = skb->data;
1276 static inline void skb_set_transport_header(struct sk_buff *skb,
1277 const int offset)
1279 skb->transport_header = skb->data + offset;
1282 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1284 return skb->network_header;
1287 static inline void skb_reset_network_header(struct sk_buff *skb)
1289 skb->network_header = skb->data;
1292 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1294 skb->network_header = skb->data + offset;
1297 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1299 return skb->mac_header;
1302 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1304 return skb->mac_header != NULL;
1307 static inline void skb_reset_mac_header(struct sk_buff *skb)
1309 skb->mac_header = skb->data;
1312 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1314 skb->mac_header = skb->data + offset;
1316 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1318 static inline int skb_transport_offset(const struct sk_buff *skb)
1320 return skb_transport_header(skb) - skb->data;
1323 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1325 return skb->transport_header - skb->network_header;
1328 static inline int skb_network_offset(const struct sk_buff *skb)
1330 return skb_network_header(skb) - skb->data;
1334 * CPUs often take a performance hit when accessing unaligned memory
1335 * locations. The actual performance hit varies, it can be small if the
1336 * hardware handles it or large if we have to take an exception and fix it
1337 * in software.
1339 * Since an ethernet header is 14 bytes network drivers often end up with
1340 * the IP header at an unaligned offset. The IP header can be aligned by
1341 * shifting the start of the packet by 2 bytes. Drivers should do this
1342 * with:
1344 * skb_reserve(NET_IP_ALIGN);
1346 * The downside to this alignment of the IP header is that the DMA is now
1347 * unaligned. On some architectures the cost of an unaligned DMA is high
1348 * and this cost outweighs the gains made by aligning the IP header.
1350 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1351 * to be overridden.
1353 #ifndef NET_IP_ALIGN
1354 #define NET_IP_ALIGN 2
1355 #endif
1358 * The networking layer reserves some headroom in skb data (via
1359 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1360 * the header has to grow. In the default case, if the header has to grow
1361 * 32 bytes or less we avoid the reallocation.
1363 * Unfortunately this headroom changes the DMA alignment of the resulting
1364 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1365 * on some architectures. An architecture can override this value,
1366 * perhaps setting it to a cacheline in size (since that will maintain
1367 * cacheline alignment of the DMA). It must be a power of 2.
1369 * Various parts of the networking layer expect at least 32 bytes of
1370 * headroom, you should not reduce this.
1372 #ifndef NET_SKB_PAD
1373 #define NET_SKB_PAD 32
1374 #endif
1376 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1378 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1380 if (unlikely(skb->data_len)) {
1381 WARN_ON(1);
1382 return;
1384 skb->len = len;
1385 skb_set_tail_pointer(skb, len);
1388 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1390 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1392 if (skb->data_len)
1393 return ___pskb_trim(skb, len);
1394 __skb_trim(skb, len);
1395 return 0;
1398 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1400 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1404 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1405 * @skb: buffer to alter
1406 * @len: new length
1408 * This is identical to pskb_trim except that the caller knows that
1409 * the skb is not cloned so we should never get an error due to out-
1410 * of-memory.
1412 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1414 int err = pskb_trim(skb, len);
1415 BUG_ON(err);
1419 * skb_orphan - orphan a buffer
1420 * @skb: buffer to orphan
1422 * If a buffer currently has an owner then we call the owner's
1423 * destructor function and make the @skb unowned. The buffer continues
1424 * to exist but is no longer charged to its former owner.
1426 static inline void skb_orphan(struct sk_buff *skb)
1428 if (skb->destructor)
1429 skb->destructor(skb);
1430 skb->destructor = NULL;
1431 skb->sk = NULL;
1435 * __skb_queue_purge - empty a list
1436 * @list: list to empty
1438 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1439 * the list and one reference dropped. This function does not take the
1440 * list lock and the caller must hold the relevant locks to use it.
1442 extern void skb_queue_purge(struct sk_buff_head *list);
1443 static inline void __skb_queue_purge(struct sk_buff_head *list)
1445 struct sk_buff *skb;
1446 while ((skb = __skb_dequeue(list)) != NULL)
1447 kfree_skb(skb);
1451 * __dev_alloc_skb - allocate an skbuff for receiving
1452 * @length: length to allocate
1453 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1455 * Allocate a new &sk_buff and assign it a usage count of one. The
1456 * buffer has unspecified headroom built in. Users should allocate
1457 * the headroom they think they need without accounting for the
1458 * built in space. The built in space is used for optimisations.
1460 * %NULL is returned if there is no free memory.
1462 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1463 gfp_t gfp_mask)
1465 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1466 if (likely(skb))
1467 skb_reserve(skb, NET_SKB_PAD);
1468 return skb;
1471 extern struct sk_buff *dev_alloc_skb(unsigned int length);
1473 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1474 unsigned int length, gfp_t gfp_mask);
1477 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1478 * @dev: network device to receive on
1479 * @length: length to allocate
1481 * Allocate a new &sk_buff and assign it a usage count of one. The
1482 * buffer has unspecified headroom built in. Users should allocate
1483 * the headroom they think they need without accounting for the
1484 * built in space. The built in space is used for optimisations.
1486 * %NULL is returned if there is no free memory. Although this function
1487 * allocates memory it can be called from an interrupt.
1489 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1490 unsigned int length)
1492 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1495 extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1498 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1499 * @dev: network device to receive on
1501 * Allocate a new page node local to the specified device.
1503 * %NULL is returned if there is no free memory.
1505 static inline struct page *netdev_alloc_page(struct net_device *dev)
1507 return __netdev_alloc_page(dev, GFP_ATOMIC);
1510 static inline void netdev_free_page(struct net_device *dev, struct page *page)
1512 __free_page(page);
1516 * skb_clone_writable - is the header of a clone writable
1517 * @skb: buffer to check
1518 * @len: length up to which to write
1520 * Returns true if modifying the header part of the cloned buffer
1521 * does not requires the data to be copied.
1523 static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1525 return !skb_header_cloned(skb) &&
1526 skb_headroom(skb) + len <= skb->hdr_len;
1529 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1530 int cloned)
1532 int delta = 0;
1534 if (headroom < NET_SKB_PAD)
1535 headroom = NET_SKB_PAD;
1536 if (headroom > skb_headroom(skb))
1537 delta = headroom - skb_headroom(skb);
1539 if (delta || cloned)
1540 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1541 GFP_ATOMIC);
1542 return 0;
1546 * skb_cow - copy header of skb when it is required
1547 * @skb: buffer to cow
1548 * @headroom: needed headroom
1550 * If the skb passed lacks sufficient headroom or its data part
1551 * is shared, data is reallocated. If reallocation fails, an error
1552 * is returned and original skb is not changed.
1554 * The result is skb with writable area skb->head...skb->tail
1555 * and at least @headroom of space at head.
1557 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1559 return __skb_cow(skb, headroom, skb_cloned(skb));
1563 * skb_cow_head - skb_cow but only making the head writable
1564 * @skb: buffer to cow
1565 * @headroom: needed headroom
1567 * This function is identical to skb_cow except that we replace the
1568 * skb_cloned check by skb_header_cloned. It should be used when
1569 * you only need to push on some header and do not need to modify
1570 * the data.
1572 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1574 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1578 * skb_padto - pad an skbuff up to a minimal size
1579 * @skb: buffer to pad
1580 * @len: minimal length
1582 * Pads up a buffer to ensure the trailing bytes exist and are
1583 * blanked. If the buffer already contains sufficient data it
1584 * is untouched. Otherwise it is extended. Returns zero on
1585 * success. The skb is freed on error.
1588 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1590 unsigned int size = skb->len;
1591 if (likely(size >= len))
1592 return 0;
1593 return skb_pad(skb, len - size);
1596 static inline int skb_add_data(struct sk_buff *skb,
1597 char __user *from, int copy)
1599 const int off = skb->len;
1601 if (skb->ip_summed == CHECKSUM_NONE) {
1602 int err = 0;
1603 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1604 copy, 0, &err);
1605 if (!err) {
1606 skb->csum = csum_block_add(skb->csum, csum, off);
1607 return 0;
1609 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1610 return 0;
1612 __skb_trim(skb, off);
1613 return -EFAULT;
1616 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1617 struct page *page, int off)
1619 if (i) {
1620 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1622 return page == frag->page &&
1623 off == frag->page_offset + frag->size;
1625 return 0;
1628 static inline int __skb_linearize(struct sk_buff *skb)
1630 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1634 * skb_linearize - convert paged skb to linear one
1635 * @skb: buffer to linarize
1637 * If there is no free memory -ENOMEM is returned, otherwise zero
1638 * is returned and the old skb data released.
1640 static inline int skb_linearize(struct sk_buff *skb)
1642 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1646 * skb_linearize_cow - make sure skb is linear and writable
1647 * @skb: buffer to process
1649 * If there is no free memory -ENOMEM is returned, otherwise zero
1650 * is returned and the old skb data released.
1652 static inline int skb_linearize_cow(struct sk_buff *skb)
1654 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1655 __skb_linearize(skb) : 0;
1659 * skb_postpull_rcsum - update checksum for received skb after pull
1660 * @skb: buffer to update
1661 * @start: start of data before pull
1662 * @len: length of data pulled
1664 * After doing a pull on a received packet, you need to call this to
1665 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1666 * CHECKSUM_NONE so that it can be recomputed from scratch.
1669 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1670 const void *start, unsigned int len)
1672 if (skb->ip_summed == CHECKSUM_COMPLETE)
1673 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1676 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1679 * pskb_trim_rcsum - trim received skb and update checksum
1680 * @skb: buffer to trim
1681 * @len: new length
1683 * This is exactly the same as pskb_trim except that it ensures the
1684 * checksum of received packets are still valid after the operation.
1687 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1689 if (likely(len >= skb->len))
1690 return 0;
1691 if (skb->ip_summed == CHECKSUM_COMPLETE)
1692 skb->ip_summed = CHECKSUM_NONE;
1693 return __pskb_trim(skb, len);
1696 #define skb_queue_walk(queue, skb) \
1697 for (skb = (queue)->next; \
1698 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1699 skb = skb->next)
1701 #define skb_queue_walk_safe(queue, skb, tmp) \
1702 for (skb = (queue)->next, tmp = skb->next; \
1703 skb != (struct sk_buff *)(queue); \
1704 skb = tmp, tmp = skb->next)
1706 #define skb_queue_walk_from(queue, skb) \
1707 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1708 skb = skb->next)
1710 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1711 for (tmp = skb->next; \
1712 skb != (struct sk_buff *)(queue); \
1713 skb = tmp, tmp = skb->next)
1715 #define skb_queue_reverse_walk(queue, skb) \
1716 for (skb = (queue)->prev; \
1717 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1718 skb = skb->prev)
1721 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1722 int *peeked, int *err);
1723 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1724 int noblock, int *err);
1725 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1726 struct poll_table_struct *wait);
1727 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1728 int offset, struct iovec *to,
1729 int size);
1730 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1731 int hlen,
1732 struct iovec *iov);
1733 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1734 int offset,
1735 const struct iovec *from,
1736 int from_offset,
1737 int len);
1738 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1739 int offset,
1740 const struct iovec *to,
1741 int to_offset,
1742 int size);
1743 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1744 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1745 unsigned int flags);
1746 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1747 int len, __wsum csum);
1748 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1749 void *to, int len);
1750 extern int skb_store_bits(struct sk_buff *skb, int offset,
1751 const void *from, int len);
1752 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1753 int offset, u8 *to, int len,
1754 __wsum csum);
1755 extern int skb_splice_bits(struct sk_buff *skb,
1756 unsigned int offset,
1757 struct pipe_inode_info *pipe,
1758 unsigned int len,
1759 unsigned int flags);
1760 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1761 extern void skb_split(struct sk_buff *skb,
1762 struct sk_buff *skb1, const u32 len);
1763 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1764 int shiftlen);
1766 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1768 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1769 int len, void *buffer)
1771 int hlen = skb_headlen(skb);
1773 if (hlen - offset >= len)
1774 return skb->data + offset;
1776 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1777 return NULL;
1779 return buffer;
1782 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1783 void *to,
1784 const unsigned int len)
1786 memcpy(to, skb->data, len);
1789 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1790 const int offset, void *to,
1791 const unsigned int len)
1793 memcpy(to, skb->data + offset, len);
1796 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1797 const void *from,
1798 const unsigned int len)
1800 memcpy(skb->data, from, len);
1803 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1804 const int offset,
1805 const void *from,
1806 const unsigned int len)
1808 memcpy(skb->data + offset, from, len);
1811 extern void skb_init(void);
1813 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1815 return skb->tstamp;
1819 * skb_get_timestamp - get timestamp from a skb
1820 * @skb: skb to get stamp from
1821 * @stamp: pointer to struct timeval to store stamp in
1823 * Timestamps are stored in the skb as offsets to a base timestamp.
1824 * This function converts the offset back to a struct timeval and stores
1825 * it in stamp.
1827 static inline void skb_get_timestamp(const struct sk_buff *skb,
1828 struct timeval *stamp)
1830 *stamp = ktime_to_timeval(skb->tstamp);
1833 static inline void skb_get_timestampns(const struct sk_buff *skb,
1834 struct timespec *stamp)
1836 *stamp = ktime_to_timespec(skb->tstamp);
1839 static inline void __net_timestamp(struct sk_buff *skb)
1841 skb->tstamp = ktime_get_real();
1844 static inline ktime_t net_timedelta(ktime_t t)
1846 return ktime_sub(ktime_get_real(), t);
1849 static inline ktime_t net_invalid_timestamp(void)
1851 return ktime_set(0, 0);
1855 * skb_tstamp_tx - queue clone of skb with send time stamps
1856 * @orig_skb: the original outgoing packet
1857 * @hwtstamps: hardware time stamps, may be NULL if not available
1859 * If the skb has a socket associated, then this function clones the
1860 * skb (thus sharing the actual data and optional structures), stores
1861 * the optional hardware time stamping information (if non NULL) or
1862 * generates a software time stamp (otherwise), then queues the clone
1863 * to the error queue of the socket. Errors are silently ignored.
1865 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
1866 struct skb_shared_hwtstamps *hwtstamps);
1868 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
1869 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
1871 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
1873 return skb->ip_summed & CHECKSUM_UNNECESSARY;
1877 * skb_checksum_complete - Calculate checksum of an entire packet
1878 * @skb: packet to process
1880 * This function calculates the checksum over the entire packet plus
1881 * the value of skb->csum. The latter can be used to supply the
1882 * checksum of a pseudo header as used by TCP/UDP. It returns the
1883 * checksum.
1885 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1886 * this function can be used to verify that checksum on received
1887 * packets. In that case the function should return zero if the
1888 * checksum is correct. In particular, this function will return zero
1889 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1890 * hardware has already verified the correctness of the checksum.
1892 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
1894 return skb_csum_unnecessary(skb) ?
1895 0 : __skb_checksum_complete(skb);
1898 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1899 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
1900 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1902 if (nfct && atomic_dec_and_test(&nfct->use))
1903 nf_conntrack_destroy(nfct);
1905 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1907 if (nfct)
1908 atomic_inc(&nfct->use);
1910 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1912 if (skb)
1913 atomic_inc(&skb->users);
1915 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1917 if (skb)
1918 kfree_skb(skb);
1920 #endif
1921 #ifdef CONFIG_BRIDGE_NETFILTER
1922 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1924 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1925 kfree(nf_bridge);
1927 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1929 if (nf_bridge)
1930 atomic_inc(&nf_bridge->use);
1932 #endif /* CONFIG_BRIDGE_NETFILTER */
1933 static inline void nf_reset(struct sk_buff *skb)
1935 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1936 nf_conntrack_put(skb->nfct);
1937 skb->nfct = NULL;
1938 nf_conntrack_put_reasm(skb->nfct_reasm);
1939 skb->nfct_reasm = NULL;
1940 #endif
1941 #ifdef CONFIG_BRIDGE_NETFILTER
1942 nf_bridge_put(skb->nf_bridge);
1943 skb->nf_bridge = NULL;
1944 #endif
1947 /* Note: This doesn't put any conntrack and bridge info in dst. */
1948 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1950 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1951 dst->nfct = src->nfct;
1952 nf_conntrack_get(src->nfct);
1953 dst->nfctinfo = src->nfctinfo;
1954 dst->nfct_reasm = src->nfct_reasm;
1955 nf_conntrack_get_reasm(src->nfct_reasm);
1956 #endif
1957 #ifdef CONFIG_BRIDGE_NETFILTER
1958 dst->nf_bridge = src->nf_bridge;
1959 nf_bridge_get(src->nf_bridge);
1960 #endif
1963 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1965 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1966 nf_conntrack_put(dst->nfct);
1967 nf_conntrack_put_reasm(dst->nfct_reasm);
1968 #endif
1969 #ifdef CONFIG_BRIDGE_NETFILTER
1970 nf_bridge_put(dst->nf_bridge);
1971 #endif
1972 __nf_copy(dst, src);
1975 #ifdef CONFIG_NETWORK_SECMARK
1976 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1978 to->secmark = from->secmark;
1981 static inline void skb_init_secmark(struct sk_buff *skb)
1983 skb->secmark = 0;
1985 #else
1986 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1989 static inline void skb_init_secmark(struct sk_buff *skb)
1991 #endif
1993 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
1995 skb->queue_mapping = queue_mapping;
1998 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2000 return skb->queue_mapping;
2003 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2005 to->queue_mapping = from->queue_mapping;
2008 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2010 skb->queue_mapping = rx_queue + 1;
2013 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2015 return skb->queue_mapping - 1;
2018 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2020 return (skb->queue_mapping != 0);
2023 extern u16 skb_tx_hash(const struct net_device *dev,
2024 const struct sk_buff *skb);
2026 #ifdef CONFIG_XFRM
2027 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2029 return skb->sp;
2031 #else
2032 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2034 return NULL;
2036 #endif
2038 static inline int skb_is_gso(const struct sk_buff *skb)
2040 return skb_shinfo(skb)->gso_size;
2043 static inline int skb_is_gso_v6(const struct sk_buff *skb)
2045 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2048 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2050 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2052 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2053 * wanted then gso_type will be set. */
2054 struct skb_shared_info *shinfo = skb_shinfo(skb);
2055 if (shinfo->gso_size != 0 && unlikely(shinfo->gso_type == 0)) {
2056 __skb_warn_lro_forwarding(skb);
2057 return true;
2059 return false;
2062 static inline void skb_forward_csum(struct sk_buff *skb)
2064 /* Unfortunately we don't support this one. Any brave souls? */
2065 if (skb->ip_summed == CHECKSUM_COMPLETE)
2066 skb->ip_summed = CHECKSUM_NONE;
2069 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2070 #endif /* __KERNEL__ */
2071 #endif /* _LINUX_SKBUFF_H */