printk removal, receive reordering bugfix
[cor_2_6_31.git] / include / linux / skbuff.h
blobf2c69a2cca172aff841e6ce97656d5fa9cb412a9
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 * @do_not_encrypt: set to prevent encryption of this frame
308 * @dma_cookie: a cookie to one of several possible DMA operations
309 * done by skb DMA functions
310 * @secmark: security marking
311 * @vlan_tci: vlan tag control information
314 struct sk_buff {
315 /* These two members must be first. */
316 struct sk_buff *next;
317 struct sk_buff *prev;
319 struct sock *sk;
320 ktime_t tstamp;
321 struct net_device *dev;
323 unsigned long _skb_dst;
324 #ifdef CONFIG_XFRM
325 struct sec_path *sp;
326 #endif
328 * This is the control buffer. It is free to use for every
329 * layer. Please put your private variables there. If you
330 * want to keep them across layers you have to do a skb_clone()
331 * first. This is owned by whoever has the skb queued ATM.
333 char cb[48];
335 unsigned int len,
336 data_len;
337 __u16 mac_len,
338 hdr_len;
339 union {
340 __wsum csum;
341 struct {
342 __u16 csum_start;
343 __u16 csum_offset;
346 __u32 priority;
347 kmemcheck_bitfield_begin(flags1);
348 __u8 local_df:1,
349 cloned:1,
350 ip_summed:2,
351 nohdr:1,
352 nfctinfo:3;
353 __u8 pkt_type:3,
354 fclone:2,
355 ipvs_property:1,
356 peeked:1,
357 nf_trace:1;
358 kmemcheck_bitfield_end(flags1);
359 __be16 protocol;
361 void (*destructor)(struct sk_buff *skb);
362 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
363 struct nf_conntrack *nfct;
364 struct sk_buff *nfct_reasm;
365 #endif
366 #ifdef CONFIG_BRIDGE_NETFILTER
367 struct nf_bridge_info *nf_bridge;
368 #endif
370 int iif;
371 __u16 queue_mapping;
372 #ifdef CONFIG_NET_SCHED
373 __u16 tc_index; /* traffic control index */
374 #ifdef CONFIG_NET_CLS_ACT
375 __u16 tc_verd; /* traffic control verdict */
376 #endif
377 #endif
379 kmemcheck_bitfield_begin(flags2);
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 #endif
386 kmemcheck_bitfield_end(flags2);
388 /* 0/13/14 bit hole */
390 #ifdef CONFIG_NET_DMA
391 dma_cookie_t dma_cookie;
392 #endif
393 #ifdef CONFIG_NETWORK_SECMARK
394 __u32 secmark;
395 #endif
397 __u32 mark;
399 __u16 vlan_tci;
401 sk_buff_data_t transport_header;
402 sk_buff_data_t network_header;
403 sk_buff_data_t mac_header;
404 /* These elements must be at the end, see alloc_skb() for details. */
405 sk_buff_data_t tail;
406 sk_buff_data_t end;
407 unsigned char *head,
408 *data;
409 unsigned int truesize;
410 atomic_t users;
413 #ifdef __KERNEL__
415 * Handling routines are only of interest to the kernel
417 #include <linux/slab.h>
419 #include <asm/system.h>
421 #ifdef CONFIG_HAS_DMA
422 #include <linux/dma-mapping.h>
423 extern int skb_dma_map(struct device *dev, struct sk_buff *skb,
424 enum dma_data_direction dir);
425 extern void skb_dma_unmap(struct device *dev, struct sk_buff *skb,
426 enum dma_data_direction dir);
427 #endif
429 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
431 return (struct dst_entry *)skb->_skb_dst;
434 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
436 skb->_skb_dst = (unsigned long)dst;
439 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
441 return (struct rtable *)skb_dst(skb);
444 extern void kfree_skb(struct sk_buff *skb);
445 extern void consume_skb(struct sk_buff *skb);
446 extern void __kfree_skb(struct sk_buff *skb);
447 extern struct sk_buff *__alloc_skb(unsigned int size,
448 gfp_t priority, int fclone, int node);
449 static inline struct sk_buff *alloc_skb(unsigned int size,
450 gfp_t priority)
452 return __alloc_skb(size, priority, 0, -1);
455 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
456 gfp_t priority)
458 return __alloc_skb(size, priority, 1, -1);
461 extern int skb_recycle_check(struct sk_buff *skb, int skb_size);
463 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
464 extern struct sk_buff *skb_clone(struct sk_buff *skb,
465 gfp_t priority);
466 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
467 gfp_t priority);
468 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
469 gfp_t gfp_mask);
470 extern int pskb_expand_head(struct sk_buff *skb,
471 int nhead, int ntail,
472 gfp_t gfp_mask);
473 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
474 unsigned int headroom);
475 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
476 int newheadroom, int newtailroom,
477 gfp_t priority);
478 extern int skb_to_sgvec(struct sk_buff *skb,
479 struct scatterlist *sg, int offset,
480 int len);
481 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
482 struct sk_buff **trailer);
483 extern int skb_pad(struct sk_buff *skb, int pad);
484 #define dev_kfree_skb(a) consume_skb(a)
485 #define dev_consume_skb(a) kfree_skb_clean(a)
486 extern void skb_over_panic(struct sk_buff *skb, int len,
487 void *here);
488 extern void skb_under_panic(struct sk_buff *skb, int len,
489 void *here);
491 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
492 int getfrag(void *from, char *to, int offset,
493 int len,int odd, struct sk_buff *skb),
494 void *from, int length);
496 struct skb_seq_state
498 __u32 lower_offset;
499 __u32 upper_offset;
500 __u32 frag_idx;
501 __u32 stepped_offset;
502 struct sk_buff *root_skb;
503 struct sk_buff *cur_skb;
504 __u8 *frag_data;
507 extern void skb_prepare_seq_read(struct sk_buff *skb,
508 unsigned int from, unsigned int to,
509 struct skb_seq_state *st);
510 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
511 struct skb_seq_state *st);
512 extern void skb_abort_seq_read(struct skb_seq_state *st);
514 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
515 unsigned int to, struct ts_config *config,
516 struct ts_state *state);
518 #ifdef NET_SKBUFF_DATA_USES_OFFSET
519 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
521 return skb->head + skb->end;
523 #else
524 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
526 return skb->end;
528 #endif
530 /* Internal */
531 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
533 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
535 return &skb_shinfo(skb)->hwtstamps;
538 static inline union skb_shared_tx *skb_tx(struct sk_buff *skb)
540 return &skb_shinfo(skb)->tx_flags;
544 * skb_queue_empty - check if a queue is empty
545 * @list: queue head
547 * Returns true if the queue is empty, false otherwise.
549 static inline int skb_queue_empty(const struct sk_buff_head *list)
551 return list->next == (struct sk_buff *)list;
555 * skb_queue_is_last - check if skb is the last entry in the queue
556 * @list: queue head
557 * @skb: buffer
559 * Returns true if @skb is the last buffer on the list.
561 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
562 const struct sk_buff *skb)
564 return (skb->next == (struct sk_buff *) list);
568 * skb_queue_is_first - check if skb is the first entry in the queue
569 * @list: queue head
570 * @skb: buffer
572 * Returns true if @skb is the first buffer on the list.
574 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
575 const struct sk_buff *skb)
577 return (skb->prev == (struct sk_buff *) list);
581 * skb_queue_next - return the next packet in the queue
582 * @list: queue head
583 * @skb: current buffer
585 * Return the next packet in @list after @skb. It is only valid to
586 * call this if skb_queue_is_last() evaluates to false.
588 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
589 const struct sk_buff *skb)
591 /* This BUG_ON may seem severe, but if we just return then we
592 * are going to dereference garbage.
594 BUG_ON(skb_queue_is_last(list, skb));
595 return skb->next;
599 * skb_queue_prev - return the prev packet in the queue
600 * @list: queue head
601 * @skb: current buffer
603 * Return the prev packet in @list before @skb. It is only valid to
604 * call this if skb_queue_is_first() evaluates to false.
606 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
607 const struct sk_buff *skb)
609 /* This BUG_ON may seem severe, but if we just return then we
610 * are going to dereference garbage.
612 BUG_ON(skb_queue_is_first(list, skb));
613 return skb->prev;
617 * skb_get - reference buffer
618 * @skb: buffer to reference
620 * Makes another reference to a socket buffer and returns a pointer
621 * to the buffer.
623 static inline struct sk_buff *skb_get(struct sk_buff *skb)
625 atomic_inc(&skb->users);
626 return skb;
630 * If users == 1, we are the only owner and are can avoid redundant
631 * atomic change.
635 * skb_cloned - is the buffer a clone
636 * @skb: buffer to check
638 * Returns true if the buffer was generated with skb_clone() and is
639 * one of multiple shared copies of the buffer. Cloned buffers are
640 * shared data so must not be written to under normal circumstances.
642 static inline int skb_cloned(const struct sk_buff *skb)
644 return skb->cloned &&
645 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
649 * skb_header_cloned - is the header a clone
650 * @skb: buffer to check
652 * Returns true if modifying the header part of the buffer requires
653 * the data to be copied.
655 static inline int skb_header_cloned(const struct sk_buff *skb)
657 int dataref;
659 if (!skb->cloned)
660 return 0;
662 dataref = atomic_read(&skb_shinfo(skb)->dataref);
663 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
664 return dataref != 1;
668 * skb_header_release - release reference to header
669 * @skb: buffer to operate on
671 * Drop a reference to the header part of the buffer. This is done
672 * by acquiring a payload reference. You must not read from the header
673 * part of skb->data after this.
675 static inline void skb_header_release(struct sk_buff *skb)
677 BUG_ON(skb->nohdr);
678 skb->nohdr = 1;
679 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
683 * skb_shared - is the buffer shared
684 * @skb: buffer to check
686 * Returns true if more than one person has a reference to this
687 * buffer.
689 static inline int skb_shared(const struct sk_buff *skb)
691 return atomic_read(&skb->users) != 1;
695 * skb_share_check - check if buffer is shared and if so clone it
696 * @skb: buffer to check
697 * @pri: priority for memory allocation
699 * If the buffer is shared the buffer is cloned and the old copy
700 * drops a reference. A new clone with a single reference is returned.
701 * If the buffer is not shared the original buffer is returned. When
702 * being called from interrupt status or with spinlocks held pri must
703 * be GFP_ATOMIC.
705 * NULL is returned on a memory allocation failure.
707 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
708 gfp_t pri)
710 might_sleep_if(pri & __GFP_WAIT);
711 if (skb_shared(skb)) {
712 struct sk_buff *nskb = skb_clone(skb, pri);
713 kfree_skb(skb);
714 skb = nskb;
716 return skb;
720 * Copy shared buffers into a new sk_buff. We effectively do COW on
721 * packets to handle cases where we have a local reader and forward
722 * and a couple of other messy ones. The normal one is tcpdumping
723 * a packet thats being forwarded.
727 * skb_unshare - make a copy of a shared buffer
728 * @skb: buffer to check
729 * @pri: priority for memory allocation
731 * If the socket buffer is a clone then this function creates a new
732 * copy of the data, drops a reference count on the old copy and returns
733 * the new copy with the reference count at 1. If the buffer is not a clone
734 * the original buffer is returned. When called with a spinlock held or
735 * from interrupt state @pri must be %GFP_ATOMIC
737 * %NULL is returned on a memory allocation failure.
739 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
740 gfp_t pri)
742 might_sleep_if(pri & __GFP_WAIT);
743 if (skb_cloned(skb)) {
744 struct sk_buff *nskb = skb_copy(skb, pri);
745 kfree_skb(skb); /* Free our shared copy */
746 skb = nskb;
748 return skb;
752 * skb_peek
753 * @list_: list to peek at
755 * Peek an &sk_buff. Unlike most other operations you _MUST_
756 * be careful with this one. A peek leaves the buffer on the
757 * list and someone else may run off with it. You must hold
758 * the appropriate locks or have a private queue to do this.
760 * Returns %NULL for an empty list or a pointer to the head element.
761 * The reference count is not incremented and the reference is therefore
762 * volatile. Use with caution.
764 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
766 struct sk_buff *list = ((struct sk_buff *)list_)->next;
767 if (list == (struct sk_buff *)list_)
768 list = NULL;
769 return list;
773 * skb_peek_tail
774 * @list_: list to peek at
776 * Peek an &sk_buff. Unlike most other operations you _MUST_
777 * be careful with this one. A peek leaves the buffer on the
778 * list and someone else may run off with it. You must hold
779 * the appropriate locks or have a private queue to do this.
781 * Returns %NULL for an empty list or a pointer to the tail element.
782 * The reference count is not incremented and the reference is therefore
783 * volatile. Use with caution.
785 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
787 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
788 if (list == (struct sk_buff *)list_)
789 list = NULL;
790 return list;
794 * skb_queue_len - get queue length
795 * @list_: list to measure
797 * Return the length of an &sk_buff queue.
799 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
801 return list_->qlen;
805 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
806 * @list: queue to initialize
808 * This initializes only the list and queue length aspects of
809 * an sk_buff_head object. This allows to initialize the list
810 * aspects of an sk_buff_head without reinitializing things like
811 * the spinlock. It can also be used for on-stack sk_buff_head
812 * objects where the spinlock is known to not be used.
814 static inline void __skb_queue_head_init(struct sk_buff_head *list)
816 list->prev = list->next = (struct sk_buff *)list;
817 list->qlen = 0;
821 * This function creates a split out lock class for each invocation;
822 * this is needed for now since a whole lot of users of the skb-queue
823 * infrastructure in drivers have different locking usage (in hardirq)
824 * than the networking core (in softirq only). In the long run either the
825 * network layer or drivers should need annotation to consolidate the
826 * main types of usage into 3 classes.
828 static inline void skb_queue_head_init(struct sk_buff_head *list)
830 spin_lock_init(&list->lock);
831 __skb_queue_head_init(list);
834 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
835 struct lock_class_key *class)
837 skb_queue_head_init(list);
838 lockdep_set_class(&list->lock, class);
842 * Insert an sk_buff on a list.
844 * The "__skb_xxxx()" functions are the non-atomic ones that
845 * can only be called with interrupts disabled.
847 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
848 static inline void __skb_insert(struct sk_buff *newsk,
849 struct sk_buff *prev, struct sk_buff *next,
850 struct sk_buff_head *list)
852 newsk->next = next;
853 newsk->prev = prev;
854 next->prev = prev->next = newsk;
855 list->qlen++;
858 static inline void __skb_queue_splice(const struct sk_buff_head *list,
859 struct sk_buff *prev,
860 struct sk_buff *next)
862 struct sk_buff *first = list->next;
863 struct sk_buff *last = list->prev;
865 first->prev = prev;
866 prev->next = first;
868 last->next = next;
869 next->prev = last;
873 * skb_queue_splice - join two skb lists, this is designed for stacks
874 * @list: the new list to add
875 * @head: the place to add it in the first list
877 static inline void skb_queue_splice(const struct sk_buff_head *list,
878 struct sk_buff_head *head)
880 if (!skb_queue_empty(list)) {
881 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
882 head->qlen += list->qlen;
887 * skb_queue_splice - join two skb lists and reinitialise the emptied list
888 * @list: the new list to add
889 * @head: the place to add it in the first list
891 * The list at @list is reinitialised
893 static inline void skb_queue_splice_init(struct sk_buff_head *list,
894 struct sk_buff_head *head)
896 if (!skb_queue_empty(list)) {
897 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
898 head->qlen += list->qlen;
899 __skb_queue_head_init(list);
904 * skb_queue_splice_tail - join two skb lists, each list being a queue
905 * @list: the new list to add
906 * @head: the place to add it in the first list
908 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
909 struct sk_buff_head *head)
911 if (!skb_queue_empty(list)) {
912 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
913 head->qlen += list->qlen;
918 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
919 * @list: the new list to add
920 * @head: the place to add it in the first list
922 * Each of the lists is a queue.
923 * The list at @list is reinitialised
925 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
926 struct sk_buff_head *head)
928 if (!skb_queue_empty(list)) {
929 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
930 head->qlen += list->qlen;
931 __skb_queue_head_init(list);
936 * __skb_queue_after - queue a buffer at the list head
937 * @list: list to use
938 * @prev: place after this buffer
939 * @newsk: buffer to queue
941 * Queue a buffer int the middle of a list. This function takes no locks
942 * and you must therefore hold required locks before calling it.
944 * A buffer cannot be placed on two lists at the same time.
946 static inline void __skb_queue_after(struct sk_buff_head *list,
947 struct sk_buff *prev,
948 struct sk_buff *newsk)
950 __skb_insert(newsk, prev, prev->next, list);
953 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
954 struct sk_buff_head *list);
956 static inline void __skb_queue_before(struct sk_buff_head *list,
957 struct sk_buff *next,
958 struct sk_buff *newsk)
960 __skb_insert(newsk, next->prev, next, list);
964 * __skb_queue_head - queue a buffer at the list head
965 * @list: list to use
966 * @newsk: buffer to queue
968 * Queue a buffer at the start of a list. This function takes no locks
969 * and you must therefore hold required locks before calling it.
971 * A buffer cannot be placed on two lists at the same time.
973 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
974 static inline void __skb_queue_head(struct sk_buff_head *list,
975 struct sk_buff *newsk)
977 __skb_queue_after(list, (struct sk_buff *)list, newsk);
981 * __skb_queue_tail - queue a buffer at the list tail
982 * @list: list to use
983 * @newsk: buffer to queue
985 * Queue a buffer at the end of a list. This function takes no locks
986 * and you must therefore hold required locks before calling it.
988 * A buffer cannot be placed on two lists at the same time.
990 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
991 static inline void __skb_queue_tail(struct sk_buff_head *list,
992 struct sk_buff *newsk)
994 __skb_queue_before(list, (struct sk_buff *)list, newsk);
998 * remove sk_buff from list. _Must_ be called atomically, and with
999 * the list known..
1001 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1002 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1004 struct sk_buff *next, *prev;
1006 list->qlen--;
1007 next = skb->next;
1008 prev = skb->prev;
1009 skb->next = skb->prev = NULL;
1010 next->prev = prev;
1011 prev->next = next;
1015 * __skb_dequeue - remove from the head of the queue
1016 * @list: list to dequeue from
1018 * Remove the head of the list. This function does not take any locks
1019 * so must be used with appropriate locks held only. The head item is
1020 * returned or %NULL if the list is empty.
1022 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1023 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1025 struct sk_buff *skb = skb_peek(list);
1026 if (skb)
1027 __skb_unlink(skb, list);
1028 return skb;
1032 * __skb_dequeue_tail - remove from the tail of the queue
1033 * @list: list to dequeue from
1035 * Remove the tail of the list. This function does not take any locks
1036 * so must be used with appropriate locks held only. The tail item is
1037 * returned or %NULL if the list is empty.
1039 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1040 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1042 struct sk_buff *skb = skb_peek_tail(list);
1043 if (skb)
1044 __skb_unlink(skb, list);
1045 return skb;
1049 static inline int skb_is_nonlinear(const struct sk_buff *skb)
1051 return skb->data_len;
1054 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1056 return skb->len - skb->data_len;
1059 static inline int skb_pagelen(const struct sk_buff *skb)
1061 int i, len = 0;
1063 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1064 len += skb_shinfo(skb)->frags[i].size;
1065 return len + skb_headlen(skb);
1068 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1069 struct page *page, int off, int size)
1071 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1073 frag->page = page;
1074 frag->page_offset = off;
1075 frag->size = size;
1076 skb_shinfo(skb)->nr_frags = i + 1;
1079 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1080 int off, int size);
1082 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1083 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frags(skb))
1084 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1086 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1087 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1089 return skb->head + skb->tail;
1092 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1094 skb->tail = skb->data - skb->head;
1097 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1099 skb_reset_tail_pointer(skb);
1100 skb->tail += offset;
1102 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1103 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1105 return skb->tail;
1108 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1110 skb->tail = skb->data;
1113 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1115 skb->tail = skb->data + offset;
1118 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1121 * Add data to an sk_buff
1123 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1124 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1126 unsigned char *tmp = skb_tail_pointer(skb);
1127 SKB_LINEAR_ASSERT(skb);
1128 skb->tail += len;
1129 skb->len += len;
1130 return tmp;
1133 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1134 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1136 skb->data -= len;
1137 skb->len += len;
1138 return skb->data;
1141 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1142 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1144 skb->len -= len;
1145 BUG_ON(skb->len < skb->data_len);
1146 return skb->data += len;
1149 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1151 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1153 if (len > skb_headlen(skb) &&
1154 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1155 return NULL;
1156 skb->len -= len;
1157 return skb->data += len;
1160 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1162 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1165 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1167 if (likely(len <= skb_headlen(skb)))
1168 return 1;
1169 if (unlikely(len > skb->len))
1170 return 0;
1171 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1175 * skb_headroom - bytes at buffer head
1176 * @skb: buffer to check
1178 * Return the number of bytes of free space at the head of an &sk_buff.
1180 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1182 return skb->data - skb->head;
1186 * skb_tailroom - bytes at buffer end
1187 * @skb: buffer to check
1189 * Return the number of bytes of free space at the tail of an sk_buff
1191 static inline int skb_tailroom(const struct sk_buff *skb)
1193 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1197 * skb_reserve - adjust headroom
1198 * @skb: buffer to alter
1199 * @len: bytes to move
1201 * Increase the headroom of an empty &sk_buff by reducing the tail
1202 * room. This is only allowed for an empty buffer.
1204 static inline void skb_reserve(struct sk_buff *skb, int len)
1206 skb->data += len;
1207 skb->tail += len;
1210 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1211 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1213 return skb->head + skb->transport_header;
1216 static inline void skb_reset_transport_header(struct sk_buff *skb)
1218 skb->transport_header = skb->data - skb->head;
1221 static inline void skb_set_transport_header(struct sk_buff *skb,
1222 const int offset)
1224 skb_reset_transport_header(skb);
1225 skb->transport_header += offset;
1228 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1230 return skb->head + skb->network_header;
1233 static inline void skb_reset_network_header(struct sk_buff *skb)
1235 skb->network_header = skb->data - skb->head;
1238 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1240 skb_reset_network_header(skb);
1241 skb->network_header += offset;
1244 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1246 return skb->head + skb->mac_header;
1249 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1251 return skb->mac_header != ~0U;
1254 static inline void skb_reset_mac_header(struct sk_buff *skb)
1256 skb->mac_header = skb->data - skb->head;
1259 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1261 skb_reset_mac_header(skb);
1262 skb->mac_header += offset;
1265 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1267 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1269 return skb->transport_header;
1272 static inline void skb_reset_transport_header(struct sk_buff *skb)
1274 skb->transport_header = skb->data;
1277 static inline void skb_set_transport_header(struct sk_buff *skb,
1278 const int offset)
1280 skb->transport_header = skb->data + offset;
1283 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1285 return skb->network_header;
1288 static inline void skb_reset_network_header(struct sk_buff *skb)
1290 skb->network_header = skb->data;
1293 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1295 skb->network_header = skb->data + offset;
1298 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1300 return skb->mac_header;
1303 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1305 return skb->mac_header != NULL;
1308 static inline void skb_reset_mac_header(struct sk_buff *skb)
1310 skb->mac_header = skb->data;
1313 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1315 skb->mac_header = skb->data + offset;
1317 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1319 static inline int skb_transport_offset(const struct sk_buff *skb)
1321 return skb_transport_header(skb) - skb->data;
1324 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1326 return skb->transport_header - skb->network_header;
1329 static inline int skb_network_offset(const struct sk_buff *skb)
1331 return skb_network_header(skb) - skb->data;
1335 * CPUs often take a performance hit when accessing unaligned memory
1336 * locations. The actual performance hit varies, it can be small if the
1337 * hardware handles it or large if we have to take an exception and fix it
1338 * in software.
1340 * Since an ethernet header is 14 bytes network drivers often end up with
1341 * the IP header at an unaligned offset. The IP header can be aligned by
1342 * shifting the start of the packet by 2 bytes. Drivers should do this
1343 * with:
1345 * skb_reserve(skb, NET_IP_ALIGN);
1347 * The downside to this alignment of the IP header is that the DMA is now
1348 * unaligned. On some architectures the cost of an unaligned DMA is high
1349 * and this cost outweighs the gains made by aligning the IP header.
1351 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1352 * to be overridden.
1354 #ifndef NET_IP_ALIGN
1355 #define NET_IP_ALIGN 2
1356 #endif
1359 * The networking layer reserves some headroom in skb data (via
1360 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1361 * the header has to grow. In the default case, if the header has to grow
1362 * 32 bytes or less we avoid the reallocation.
1364 * Unfortunately this headroom changes the DMA alignment of the resulting
1365 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1366 * on some architectures. An architecture can override this value,
1367 * perhaps setting it to a cacheline in size (since that will maintain
1368 * cacheline alignment of the DMA). It must be a power of 2.
1370 * Various parts of the networking layer expect at least 32 bytes of
1371 * headroom, you should not reduce this.
1373 #ifndef NET_SKB_PAD
1374 #define NET_SKB_PAD 32
1375 #endif
1377 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1379 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1381 if (unlikely(skb->data_len)) {
1382 WARN_ON(1);
1383 return;
1385 skb->len = len;
1386 skb_set_tail_pointer(skb, len);
1389 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1391 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1393 if (skb->data_len)
1394 return ___pskb_trim(skb, len);
1395 __skb_trim(skb, len);
1396 return 0;
1399 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1401 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1405 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1406 * @skb: buffer to alter
1407 * @len: new length
1409 * This is identical to pskb_trim except that the caller knows that
1410 * the skb is not cloned so we should never get an error due to out-
1411 * of-memory.
1413 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1415 int err = pskb_trim(skb, len);
1416 BUG_ON(err);
1420 * skb_orphan - orphan a buffer
1421 * @skb: buffer to orphan
1423 * If a buffer currently has an owner then we call the owner's
1424 * destructor function and make the @skb unowned. The buffer continues
1425 * to exist but is no longer charged to its former owner.
1427 static inline void skb_orphan(struct sk_buff *skb)
1429 if (skb->destructor)
1430 skb->destructor(skb);
1431 skb->destructor = NULL;
1432 skb->sk = NULL;
1436 * __skb_queue_purge - empty a list
1437 * @list: list to empty
1439 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1440 * the list and one reference dropped. This function does not take the
1441 * list lock and the caller must hold the relevant locks to use it.
1443 extern void skb_queue_purge(struct sk_buff_head *list);
1444 static inline void __skb_queue_purge(struct sk_buff_head *list)
1446 struct sk_buff *skb;
1447 while ((skb = __skb_dequeue(list)) != NULL)
1448 kfree_skb(skb);
1452 * __dev_alloc_skb - allocate an skbuff for receiving
1453 * @length: length to allocate
1454 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1456 * Allocate a new &sk_buff and assign it a usage count of one. The
1457 * buffer has unspecified headroom built in. Users should allocate
1458 * the headroom they think they need without accounting for the
1459 * built in space. The built in space is used for optimisations.
1461 * %NULL is returned if there is no free memory.
1463 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1464 gfp_t gfp_mask)
1466 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1467 if (likely(skb))
1468 skb_reserve(skb, NET_SKB_PAD);
1469 return skb;
1472 extern struct sk_buff *dev_alloc_skb(unsigned int length);
1474 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1475 unsigned int length, gfp_t gfp_mask);
1478 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1479 * @dev: network device to receive on
1480 * @length: length to allocate
1482 * Allocate a new &sk_buff and assign it a usage count of one. The
1483 * buffer has unspecified headroom built in. Users should allocate
1484 * the headroom they think they need without accounting for the
1485 * built in space. The built in space is used for optimisations.
1487 * %NULL is returned if there is no free memory. Although this function
1488 * allocates memory it can be called from an interrupt.
1490 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1491 unsigned int length)
1493 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1496 extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1499 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1500 * @dev: network device to receive on
1502 * Allocate a new page node local to the specified device.
1504 * %NULL is returned if there is no free memory.
1506 static inline struct page *netdev_alloc_page(struct net_device *dev)
1508 return __netdev_alloc_page(dev, GFP_ATOMIC);
1511 static inline void netdev_free_page(struct net_device *dev, struct page *page)
1513 __free_page(page);
1517 * skb_clone_writable - is the header of a clone writable
1518 * @skb: buffer to check
1519 * @len: length up to which to write
1521 * Returns true if modifying the header part of the cloned buffer
1522 * does not requires the data to be copied.
1524 static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1526 return !skb_header_cloned(skb) &&
1527 skb_headroom(skb) + len <= skb->hdr_len;
1530 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1531 int cloned)
1533 int delta = 0;
1535 if (headroom < NET_SKB_PAD)
1536 headroom = NET_SKB_PAD;
1537 if (headroom > skb_headroom(skb))
1538 delta = headroom - skb_headroom(skb);
1540 if (delta || cloned)
1541 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1542 GFP_ATOMIC);
1543 return 0;
1547 * skb_cow - copy header of skb when it is required
1548 * @skb: buffer to cow
1549 * @headroom: needed headroom
1551 * If the skb passed lacks sufficient headroom or its data part
1552 * is shared, data is reallocated. If reallocation fails, an error
1553 * is returned and original skb is not changed.
1555 * The result is skb with writable area skb->head...skb->tail
1556 * and at least @headroom of space at head.
1558 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1560 return __skb_cow(skb, headroom, skb_cloned(skb));
1564 * skb_cow_head - skb_cow but only making the head writable
1565 * @skb: buffer to cow
1566 * @headroom: needed headroom
1568 * This function is identical to skb_cow except that we replace the
1569 * skb_cloned check by skb_header_cloned. It should be used when
1570 * you only need to push on some header and do not need to modify
1571 * the data.
1573 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1575 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1579 * skb_padto - pad an skbuff up to a minimal size
1580 * @skb: buffer to pad
1581 * @len: minimal length
1583 * Pads up a buffer to ensure the trailing bytes exist and are
1584 * blanked. If the buffer already contains sufficient data it
1585 * is untouched. Otherwise it is extended. Returns zero on
1586 * success. The skb is freed on error.
1589 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1591 unsigned int size = skb->len;
1592 if (likely(size >= len))
1593 return 0;
1594 return skb_pad(skb, len - size);
1597 static inline int skb_add_data(struct sk_buff *skb,
1598 char __user *from, int copy)
1600 const int off = skb->len;
1602 if (skb->ip_summed == CHECKSUM_NONE) {
1603 int err = 0;
1604 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1605 copy, 0, &err);
1606 if (!err) {
1607 skb->csum = csum_block_add(skb->csum, csum, off);
1608 return 0;
1610 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1611 return 0;
1613 __skb_trim(skb, off);
1614 return -EFAULT;
1617 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1618 struct page *page, int off)
1620 if (i) {
1621 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1623 return page == frag->page &&
1624 off == frag->page_offset + frag->size;
1626 return 0;
1629 static inline int __skb_linearize(struct sk_buff *skb)
1631 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1635 * skb_linearize - convert paged skb to linear one
1636 * @skb: buffer to linarize
1638 * If there is no free memory -ENOMEM is returned, otherwise zero
1639 * is returned and the old skb data released.
1641 static inline int skb_linearize(struct sk_buff *skb)
1643 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1647 * skb_linearize_cow - make sure skb is linear and writable
1648 * @skb: buffer to process
1650 * If there is no free memory -ENOMEM is returned, otherwise zero
1651 * is returned and the old skb data released.
1653 static inline int skb_linearize_cow(struct sk_buff *skb)
1655 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1656 __skb_linearize(skb) : 0;
1660 * skb_postpull_rcsum - update checksum for received skb after pull
1661 * @skb: buffer to update
1662 * @start: start of data before pull
1663 * @len: length of data pulled
1665 * After doing a pull on a received packet, you need to call this to
1666 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1667 * CHECKSUM_NONE so that it can be recomputed from scratch.
1670 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1671 const void *start, unsigned int len)
1673 if (skb->ip_summed == CHECKSUM_COMPLETE)
1674 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1677 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1680 * pskb_trim_rcsum - trim received skb and update checksum
1681 * @skb: buffer to trim
1682 * @len: new length
1684 * This is exactly the same as pskb_trim except that it ensures the
1685 * checksum of received packets are still valid after the operation.
1688 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1690 if (likely(len >= skb->len))
1691 return 0;
1692 if (skb->ip_summed == CHECKSUM_COMPLETE)
1693 skb->ip_summed = CHECKSUM_NONE;
1694 return __pskb_trim(skb, len);
1697 #define skb_queue_walk(queue, skb) \
1698 for (skb = (queue)->next; \
1699 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1700 skb = skb->next)
1702 #define skb_queue_walk_safe(queue, skb, tmp) \
1703 for (skb = (queue)->next, tmp = skb->next; \
1704 skb != (struct sk_buff *)(queue); \
1705 skb = tmp, tmp = skb->next)
1707 #define skb_queue_walk_from(queue, skb) \
1708 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1709 skb = skb->next)
1711 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1712 for (tmp = skb->next; \
1713 skb != (struct sk_buff *)(queue); \
1714 skb = tmp, tmp = skb->next)
1716 #define skb_queue_reverse_walk(queue, skb) \
1717 for (skb = (queue)->prev; \
1718 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1719 skb = skb->prev)
1722 static inline bool skb_has_frags(const struct sk_buff *skb)
1724 return skb_shinfo(skb)->frag_list != NULL;
1727 static inline void skb_frag_list_init(struct sk_buff *skb)
1729 skb_shinfo(skb)->frag_list = NULL;
1732 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1734 frag->next = skb_shinfo(skb)->frag_list;
1735 skb_shinfo(skb)->frag_list = frag;
1738 #define skb_walk_frags(skb, iter) \
1739 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1741 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1742 int *peeked, int *err);
1743 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1744 int noblock, int *err);
1745 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1746 struct poll_table_struct *wait);
1747 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1748 int offset, struct iovec *to,
1749 int size);
1750 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1751 int hlen,
1752 struct iovec *iov);
1753 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1754 int offset,
1755 const struct iovec *from,
1756 int from_offset,
1757 int len);
1758 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1759 int offset,
1760 const struct iovec *to,
1761 int to_offset,
1762 int size);
1763 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1764 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1765 unsigned int flags);
1766 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1767 int len, __wsum csum);
1768 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1769 void *to, int len);
1770 extern int skb_store_bits(struct sk_buff *skb, int offset,
1771 const void *from, int len);
1772 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1773 int offset, u8 *to, int len,
1774 __wsum csum);
1775 extern int skb_splice_bits(struct sk_buff *skb,
1776 unsigned int offset,
1777 struct pipe_inode_info *pipe,
1778 unsigned int len,
1779 unsigned int flags);
1780 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1781 extern void skb_split(struct sk_buff *skb,
1782 struct sk_buff *skb1, const u32 len);
1783 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1784 int shiftlen);
1786 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1788 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1789 int len, void *buffer)
1791 int hlen = skb_headlen(skb);
1793 if (hlen - offset >= len)
1794 return skb->data + offset;
1796 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1797 return NULL;
1799 return buffer;
1802 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1803 void *to,
1804 const unsigned int len)
1806 memcpy(to, skb->data, len);
1809 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1810 const int offset, void *to,
1811 const unsigned int len)
1813 memcpy(to, skb->data + offset, len);
1816 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1817 const void *from,
1818 const unsigned int len)
1820 memcpy(skb->data, from, len);
1823 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1824 const int offset,
1825 const void *from,
1826 const unsigned int len)
1828 memcpy(skb->data + offset, from, len);
1831 extern void skb_init(void);
1833 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1835 return skb->tstamp;
1839 * skb_get_timestamp - get timestamp from a skb
1840 * @skb: skb to get stamp from
1841 * @stamp: pointer to struct timeval to store stamp in
1843 * Timestamps are stored in the skb as offsets to a base timestamp.
1844 * This function converts the offset back to a struct timeval and stores
1845 * it in stamp.
1847 static inline void skb_get_timestamp(const struct sk_buff *skb,
1848 struct timeval *stamp)
1850 *stamp = ktime_to_timeval(skb->tstamp);
1853 static inline void skb_get_timestampns(const struct sk_buff *skb,
1854 struct timespec *stamp)
1856 *stamp = ktime_to_timespec(skb->tstamp);
1859 static inline void __net_timestamp(struct sk_buff *skb)
1861 skb->tstamp = ktime_get_real();
1864 static inline ktime_t net_timedelta(ktime_t t)
1866 return ktime_sub(ktime_get_real(), t);
1869 static inline ktime_t net_invalid_timestamp(void)
1871 return ktime_set(0, 0);
1875 * skb_tstamp_tx - queue clone of skb with send time stamps
1876 * @orig_skb: the original outgoing packet
1877 * @hwtstamps: hardware time stamps, may be NULL if not available
1879 * If the skb has a socket associated, then this function clones the
1880 * skb (thus sharing the actual data and optional structures), stores
1881 * the optional hardware time stamping information (if non NULL) or
1882 * generates a software time stamp (otherwise), then queues the clone
1883 * to the error queue of the socket. Errors are silently ignored.
1885 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
1886 struct skb_shared_hwtstamps *hwtstamps);
1888 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
1889 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
1891 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
1893 return skb->ip_summed & CHECKSUM_UNNECESSARY;
1897 * skb_checksum_complete - Calculate checksum of an entire packet
1898 * @skb: packet to process
1900 * This function calculates the checksum over the entire packet plus
1901 * the value of skb->csum. The latter can be used to supply the
1902 * checksum of a pseudo header as used by TCP/UDP. It returns the
1903 * checksum.
1905 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1906 * this function can be used to verify that checksum on received
1907 * packets. In that case the function should return zero if the
1908 * checksum is correct. In particular, this function will return zero
1909 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1910 * hardware has already verified the correctness of the checksum.
1912 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
1914 return skb_csum_unnecessary(skb) ?
1915 0 : __skb_checksum_complete(skb);
1918 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1919 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
1920 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1922 if (nfct && atomic_dec_and_test(&nfct->use))
1923 nf_conntrack_destroy(nfct);
1925 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1927 if (nfct)
1928 atomic_inc(&nfct->use);
1930 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1932 if (skb)
1933 atomic_inc(&skb->users);
1935 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1937 if (skb)
1938 kfree_skb(skb);
1940 #endif
1941 #ifdef CONFIG_BRIDGE_NETFILTER
1942 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1944 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1945 kfree(nf_bridge);
1947 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1949 if (nf_bridge)
1950 atomic_inc(&nf_bridge->use);
1952 #endif /* CONFIG_BRIDGE_NETFILTER */
1953 static inline void nf_reset(struct sk_buff *skb)
1955 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1956 nf_conntrack_put(skb->nfct);
1957 skb->nfct = NULL;
1958 nf_conntrack_put_reasm(skb->nfct_reasm);
1959 skb->nfct_reasm = NULL;
1960 #endif
1961 #ifdef CONFIG_BRIDGE_NETFILTER
1962 nf_bridge_put(skb->nf_bridge);
1963 skb->nf_bridge = NULL;
1964 #endif
1967 /* Note: This doesn't put any conntrack and bridge info in dst. */
1968 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1970 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1971 dst->nfct = src->nfct;
1972 nf_conntrack_get(src->nfct);
1973 dst->nfctinfo = src->nfctinfo;
1974 dst->nfct_reasm = src->nfct_reasm;
1975 nf_conntrack_get_reasm(src->nfct_reasm);
1976 #endif
1977 #ifdef CONFIG_BRIDGE_NETFILTER
1978 dst->nf_bridge = src->nf_bridge;
1979 nf_bridge_get(src->nf_bridge);
1980 #endif
1983 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1985 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1986 nf_conntrack_put(dst->nfct);
1987 nf_conntrack_put_reasm(dst->nfct_reasm);
1988 #endif
1989 #ifdef CONFIG_BRIDGE_NETFILTER
1990 nf_bridge_put(dst->nf_bridge);
1991 #endif
1992 __nf_copy(dst, src);
1995 #ifdef CONFIG_NETWORK_SECMARK
1996 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1998 to->secmark = from->secmark;
2001 static inline void skb_init_secmark(struct sk_buff *skb)
2003 skb->secmark = 0;
2005 #else
2006 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2009 static inline void skb_init_secmark(struct sk_buff *skb)
2011 #endif
2013 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2015 skb->queue_mapping = queue_mapping;
2018 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2020 return skb->queue_mapping;
2023 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2025 to->queue_mapping = from->queue_mapping;
2028 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2030 skb->queue_mapping = rx_queue + 1;
2033 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2035 return skb->queue_mapping - 1;
2038 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2040 return (skb->queue_mapping != 0);
2043 extern u16 skb_tx_hash(const struct net_device *dev,
2044 const struct sk_buff *skb);
2046 #ifdef CONFIG_XFRM
2047 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2049 return skb->sp;
2051 #else
2052 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2054 return NULL;
2056 #endif
2058 static inline int skb_is_gso(const struct sk_buff *skb)
2060 return skb_shinfo(skb)->gso_size;
2063 static inline int skb_is_gso_v6(const struct sk_buff *skb)
2065 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2068 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2070 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2072 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2073 * wanted then gso_type will be set. */
2074 struct skb_shared_info *shinfo = skb_shinfo(skb);
2075 if (shinfo->gso_size != 0 && unlikely(shinfo->gso_type == 0)) {
2076 __skb_warn_lro_forwarding(skb);
2077 return true;
2079 return false;
2082 static inline void skb_forward_csum(struct sk_buff *skb)
2084 /* Unfortunately we don't support this one. Any brave souls? */
2085 if (skb->ip_summed == CHECKSUM_COMPLETE)
2086 skb->ip_summed = CHECKSUM_NONE;
2089 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2090 #endif /* __KERNEL__ */
2091 #endif /* _LINUX_SKBUFF_H */