netfilter: fix description of expected checkentry return code on xt_target
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / include / linux / skbuff.h
blobc9525bce80f60b7882a3a643a44e249c819fc658
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 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 struct sk_buff *frag_list;
198 struct skb_shared_hwtstamps hwtstamps;
201 * Warning : all fields before dataref are cleared in __alloc_skb()
203 atomic_t dataref;
205 skb_frag_t frags[MAX_SKB_FRAGS];
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 * @_skb_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 * @skb_iif: ifindex of device we arrived on
302 * @rxhash: the packet hash computed on receive
303 * @queue_mapping: Queue mapping for multiqueue devices
304 * @tc_index: Traffic control index
305 * @tc_verd: traffic control verdict
306 * @ndisc_nodetype: router type (from link layer)
307 * @dma_cookie: a cookie to one of several possible DMA operations
308 * done by skb DMA functions
309 * @secmark: security marking
310 * @vlan_tci: vlan tag control information
313 struct sk_buff {
314 /* These two members must be first. */
315 struct sk_buff *next;
316 struct sk_buff *prev;
318 ktime_t tstamp;
320 struct sock *sk;
321 struct net_device *dev;
324 * This is the control buffer. It is free to use for every
325 * layer. Please put your private variables there. If you
326 * want to keep them across layers you have to do a skb_clone()
327 * first. This is owned by whoever has the skb queued ATM.
329 char cb[48] __aligned(8);
331 unsigned long _skb_dst;
332 #ifdef CONFIG_XFRM
333 struct sec_path *sp;
334 #endif
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 skb_iif;
371 #ifdef CONFIG_NET_SCHED
372 __u16 tc_index; /* traffic control index */
373 #ifdef CONFIG_NET_CLS_ACT
374 __u16 tc_verd; /* traffic control verdict */
375 #endif
376 #endif
378 __u32 rxhash;
380 kmemcheck_bitfield_begin(flags2);
381 __u16 queue_mapping:16;
382 #ifdef CONFIG_IPV6_NDISC_NODETYPE
383 __u8 ndisc_nodetype:2;
384 #endif
385 kmemcheck_bitfield_end(flags2);
387 /* 0/14 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
395 union {
396 __u32 mark;
397 __u32 dropcount;
400 __u16 vlan_tci;
402 sk_buff_data_t transport_header;
403 sk_buff_data_t network_header;
404 sk_buff_data_t mac_header;
405 /* These elements must be at the end, see alloc_skb() for details. */
406 sk_buff_data_t tail;
407 sk_buff_data_t end;
408 unsigned char *head,
409 *data;
410 unsigned int truesize;
411 atomic_t users;
414 #ifdef __KERNEL__
416 * Handling routines are only of interest to the kernel
418 #include <linux/slab.h>
420 #include <asm/system.h>
422 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
424 return (struct dst_entry *)skb->_skb_dst;
427 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
429 skb->_skb_dst = (unsigned long)dst;
432 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
434 return (struct rtable *)skb_dst(skb);
437 extern void kfree_skb(struct sk_buff *skb);
438 extern void consume_skb(struct sk_buff *skb);
439 extern void __kfree_skb(struct sk_buff *skb);
440 extern struct sk_buff *__alloc_skb(unsigned int size,
441 gfp_t priority, int fclone, int node);
442 static inline struct sk_buff *alloc_skb(unsigned int size,
443 gfp_t priority)
445 return __alloc_skb(size, priority, 0, -1);
448 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
449 gfp_t priority)
451 return __alloc_skb(size, priority, 1, -1);
454 extern int skb_recycle_check(struct sk_buff *skb, int skb_size);
456 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
457 extern struct sk_buff *skb_clone(struct sk_buff *skb,
458 gfp_t priority);
459 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
460 gfp_t priority);
461 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
462 gfp_t gfp_mask);
463 extern int pskb_expand_head(struct sk_buff *skb,
464 int nhead, int ntail,
465 gfp_t gfp_mask);
466 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
467 unsigned int headroom);
468 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
469 int newheadroom, int newtailroom,
470 gfp_t priority);
471 extern int skb_to_sgvec(struct sk_buff *skb,
472 struct scatterlist *sg, int offset,
473 int len);
474 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
475 struct sk_buff **trailer);
476 extern int skb_pad(struct sk_buff *skb, int pad);
477 #define dev_kfree_skb(a) consume_skb(a)
479 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
480 int getfrag(void *from, char *to, int offset,
481 int len,int odd, struct sk_buff *skb),
482 void *from, int length);
484 struct skb_seq_state {
485 __u32 lower_offset;
486 __u32 upper_offset;
487 __u32 frag_idx;
488 __u32 stepped_offset;
489 struct sk_buff *root_skb;
490 struct sk_buff *cur_skb;
491 __u8 *frag_data;
494 extern void skb_prepare_seq_read(struct sk_buff *skb,
495 unsigned int from, unsigned int to,
496 struct skb_seq_state *st);
497 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
498 struct skb_seq_state *st);
499 extern void skb_abort_seq_read(struct skb_seq_state *st);
501 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
502 unsigned int to, struct ts_config *config,
503 struct ts_state *state);
505 #ifdef NET_SKBUFF_DATA_USES_OFFSET
506 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
508 return skb->head + skb->end;
510 #else
511 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
513 return skb->end;
515 #endif
517 /* Internal */
518 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
520 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
522 return &skb_shinfo(skb)->hwtstamps;
525 static inline union skb_shared_tx *skb_tx(struct sk_buff *skb)
527 return &skb_shinfo(skb)->tx_flags;
531 * skb_queue_empty - check if a queue is empty
532 * @list: queue head
534 * Returns true if the queue is empty, false otherwise.
536 static inline int skb_queue_empty(const struct sk_buff_head *list)
538 return list->next == (struct sk_buff *)list;
542 * skb_queue_is_last - check if skb is the last entry in the queue
543 * @list: queue head
544 * @skb: buffer
546 * Returns true if @skb is the last buffer on the list.
548 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
549 const struct sk_buff *skb)
551 return (skb->next == (struct sk_buff *) list);
555 * skb_queue_is_first - check if skb is the first entry in the queue
556 * @list: queue head
557 * @skb: buffer
559 * Returns true if @skb is the first buffer on the list.
561 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
562 const struct sk_buff *skb)
564 return (skb->prev == (struct sk_buff *) list);
568 * skb_queue_next - return the next packet in the queue
569 * @list: queue head
570 * @skb: current buffer
572 * Return the next packet in @list after @skb. It is only valid to
573 * call this if skb_queue_is_last() evaluates to false.
575 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
576 const struct sk_buff *skb)
578 /* This BUG_ON may seem severe, but if we just return then we
579 * are going to dereference garbage.
581 BUG_ON(skb_queue_is_last(list, skb));
582 return skb->next;
586 * skb_queue_prev - return the prev packet in the queue
587 * @list: queue head
588 * @skb: current buffer
590 * Return the prev packet in @list before @skb. It is only valid to
591 * call this if skb_queue_is_first() evaluates to false.
593 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
594 const struct sk_buff *skb)
596 /* This BUG_ON may seem severe, but if we just return then we
597 * are going to dereference garbage.
599 BUG_ON(skb_queue_is_first(list, skb));
600 return skb->prev;
604 * skb_get - reference buffer
605 * @skb: buffer to reference
607 * Makes another reference to a socket buffer and returns a pointer
608 * to the buffer.
610 static inline struct sk_buff *skb_get(struct sk_buff *skb)
612 atomic_inc(&skb->users);
613 return skb;
617 * If users == 1, we are the only owner and are can avoid redundant
618 * atomic change.
622 * skb_cloned - is the buffer a clone
623 * @skb: buffer to check
625 * Returns true if the buffer was generated with skb_clone() and is
626 * one of multiple shared copies of the buffer. Cloned buffers are
627 * shared data so must not be written to under normal circumstances.
629 static inline int skb_cloned(const struct sk_buff *skb)
631 return skb->cloned &&
632 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
636 * skb_header_cloned - is the header a clone
637 * @skb: buffer to check
639 * Returns true if modifying the header part of the buffer requires
640 * the data to be copied.
642 static inline int skb_header_cloned(const struct sk_buff *skb)
644 int dataref;
646 if (!skb->cloned)
647 return 0;
649 dataref = atomic_read(&skb_shinfo(skb)->dataref);
650 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
651 return dataref != 1;
655 * skb_header_release - release reference to header
656 * @skb: buffer to operate on
658 * Drop a reference to the header part of the buffer. This is done
659 * by acquiring a payload reference. You must not read from the header
660 * part of skb->data after this.
662 static inline void skb_header_release(struct sk_buff *skb)
664 BUG_ON(skb->nohdr);
665 skb->nohdr = 1;
666 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
670 * skb_shared - is the buffer shared
671 * @skb: buffer to check
673 * Returns true if more than one person has a reference to this
674 * buffer.
676 static inline int skb_shared(const struct sk_buff *skb)
678 return atomic_read(&skb->users) != 1;
682 * skb_share_check - check if buffer is shared and if so clone it
683 * @skb: buffer to check
684 * @pri: priority for memory allocation
686 * If the buffer is shared the buffer is cloned and the old copy
687 * drops a reference. A new clone with a single reference is returned.
688 * If the buffer is not shared the original buffer is returned. When
689 * being called from interrupt status or with spinlocks held pri must
690 * be GFP_ATOMIC.
692 * NULL is returned on a memory allocation failure.
694 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
695 gfp_t pri)
697 might_sleep_if(pri & __GFP_WAIT);
698 if (skb_shared(skb)) {
699 struct sk_buff *nskb = skb_clone(skb, pri);
700 kfree_skb(skb);
701 skb = nskb;
703 return skb;
707 * Copy shared buffers into a new sk_buff. We effectively do COW on
708 * packets to handle cases where we have a local reader and forward
709 * and a couple of other messy ones. The normal one is tcpdumping
710 * a packet thats being forwarded.
714 * skb_unshare - make a copy of a shared buffer
715 * @skb: buffer to check
716 * @pri: priority for memory allocation
718 * If the socket buffer is a clone then this function creates a new
719 * copy of the data, drops a reference count on the old copy and returns
720 * the new copy with the reference count at 1. If the buffer is not a clone
721 * the original buffer is returned. When called with a spinlock held or
722 * from interrupt state @pri must be %GFP_ATOMIC
724 * %NULL is returned on a memory allocation failure.
726 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
727 gfp_t pri)
729 might_sleep_if(pri & __GFP_WAIT);
730 if (skb_cloned(skb)) {
731 struct sk_buff *nskb = skb_copy(skb, pri);
732 kfree_skb(skb); /* Free our shared copy */
733 skb = nskb;
735 return skb;
739 * skb_peek - peek at the head of an &sk_buff_head
740 * @list_: list to peek at
742 * Peek an &sk_buff. Unlike most other operations you _MUST_
743 * be careful with this one. A peek leaves the buffer on the
744 * list and someone else may run off with it. You must hold
745 * the appropriate locks or have a private queue to do this.
747 * Returns %NULL for an empty list or a pointer to the head element.
748 * The reference count is not incremented and the reference is therefore
749 * volatile. Use with caution.
751 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
753 struct sk_buff *list = ((struct sk_buff *)list_)->next;
754 if (list == (struct sk_buff *)list_)
755 list = NULL;
756 return list;
760 * skb_peek_tail - peek at the tail of an &sk_buff_head
761 * @list_: list to peek at
763 * Peek an &sk_buff. Unlike most other operations you _MUST_
764 * be careful with this one. A peek leaves the buffer on the
765 * list and someone else may run off with it. You must hold
766 * the appropriate locks or have a private queue to do this.
768 * Returns %NULL for an empty list or a pointer to the tail element.
769 * The reference count is not incremented and the reference is therefore
770 * volatile. Use with caution.
772 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
774 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
775 if (list == (struct sk_buff *)list_)
776 list = NULL;
777 return list;
781 * skb_queue_len - get queue length
782 * @list_: list to measure
784 * Return the length of an &sk_buff queue.
786 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
788 return list_->qlen;
792 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
793 * @list: queue to initialize
795 * This initializes only the list and queue length aspects of
796 * an sk_buff_head object. This allows to initialize the list
797 * aspects of an sk_buff_head without reinitializing things like
798 * the spinlock. It can also be used for on-stack sk_buff_head
799 * objects where the spinlock is known to not be used.
801 static inline void __skb_queue_head_init(struct sk_buff_head *list)
803 list->prev = list->next = (struct sk_buff *)list;
804 list->qlen = 0;
808 * This function creates a split out lock class for each invocation;
809 * this is needed for now since a whole lot of users of the skb-queue
810 * infrastructure in drivers have different locking usage (in hardirq)
811 * than the networking core (in softirq only). In the long run either the
812 * network layer or drivers should need annotation to consolidate the
813 * main types of usage into 3 classes.
815 static inline void skb_queue_head_init(struct sk_buff_head *list)
817 spin_lock_init(&list->lock);
818 __skb_queue_head_init(list);
821 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
822 struct lock_class_key *class)
824 skb_queue_head_init(list);
825 lockdep_set_class(&list->lock, class);
829 * Insert an sk_buff on a list.
831 * The "__skb_xxxx()" functions are the non-atomic ones that
832 * can only be called with interrupts disabled.
834 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
835 static inline void __skb_insert(struct sk_buff *newsk,
836 struct sk_buff *prev, struct sk_buff *next,
837 struct sk_buff_head *list)
839 newsk->next = next;
840 newsk->prev = prev;
841 next->prev = prev->next = newsk;
842 list->qlen++;
845 static inline void __skb_queue_splice(const struct sk_buff_head *list,
846 struct sk_buff *prev,
847 struct sk_buff *next)
849 struct sk_buff *first = list->next;
850 struct sk_buff *last = list->prev;
852 first->prev = prev;
853 prev->next = first;
855 last->next = next;
856 next->prev = last;
860 * skb_queue_splice - join two skb lists, this is designed for stacks
861 * @list: the new list to add
862 * @head: the place to add it in the first list
864 static inline void skb_queue_splice(const struct sk_buff_head *list,
865 struct sk_buff_head *head)
867 if (!skb_queue_empty(list)) {
868 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
869 head->qlen += list->qlen;
874 * skb_queue_splice - join two skb lists and reinitialise the emptied list
875 * @list: the new list to add
876 * @head: the place to add it in the first list
878 * The list at @list is reinitialised
880 static inline void skb_queue_splice_init(struct sk_buff_head *list,
881 struct sk_buff_head *head)
883 if (!skb_queue_empty(list)) {
884 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
885 head->qlen += list->qlen;
886 __skb_queue_head_init(list);
891 * skb_queue_splice_tail - join two skb lists, each list being a queue
892 * @list: the new list to add
893 * @head: the place to add it in the first list
895 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
896 struct sk_buff_head *head)
898 if (!skb_queue_empty(list)) {
899 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
900 head->qlen += list->qlen;
905 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
906 * @list: the new list to add
907 * @head: the place to add it in the first list
909 * Each of the lists is a queue.
910 * The list at @list is reinitialised
912 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
913 struct sk_buff_head *head)
915 if (!skb_queue_empty(list)) {
916 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
917 head->qlen += list->qlen;
918 __skb_queue_head_init(list);
923 * __skb_queue_after - queue a buffer at the list head
924 * @list: list to use
925 * @prev: place after this buffer
926 * @newsk: buffer to queue
928 * Queue a buffer int the middle of a list. This function takes no locks
929 * and you must therefore hold required locks before calling it.
931 * A buffer cannot be placed on two lists at the same time.
933 static inline void __skb_queue_after(struct sk_buff_head *list,
934 struct sk_buff *prev,
935 struct sk_buff *newsk)
937 __skb_insert(newsk, prev, prev->next, list);
940 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
941 struct sk_buff_head *list);
943 static inline void __skb_queue_before(struct sk_buff_head *list,
944 struct sk_buff *next,
945 struct sk_buff *newsk)
947 __skb_insert(newsk, next->prev, next, list);
951 * __skb_queue_head - queue a buffer at the list head
952 * @list: list to use
953 * @newsk: buffer to queue
955 * Queue a buffer at the start of a list. This function takes no locks
956 * and you must therefore hold required locks before calling it.
958 * A buffer cannot be placed on two lists at the same time.
960 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
961 static inline void __skb_queue_head(struct sk_buff_head *list,
962 struct sk_buff *newsk)
964 __skb_queue_after(list, (struct sk_buff *)list, newsk);
968 * __skb_queue_tail - queue a buffer at the list tail
969 * @list: list to use
970 * @newsk: buffer to queue
972 * Queue a buffer at the end of a list. This function takes no locks
973 * and you must therefore hold required locks before calling it.
975 * A buffer cannot be placed on two lists at the same time.
977 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
978 static inline void __skb_queue_tail(struct sk_buff_head *list,
979 struct sk_buff *newsk)
981 __skb_queue_before(list, (struct sk_buff *)list, newsk);
985 * remove sk_buff from list. _Must_ be called atomically, and with
986 * the list known..
988 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
989 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
991 struct sk_buff *next, *prev;
993 list->qlen--;
994 next = skb->next;
995 prev = skb->prev;
996 skb->next = skb->prev = NULL;
997 next->prev = prev;
998 prev->next = next;
1002 * __skb_dequeue - remove from the head of the queue
1003 * @list: list to dequeue from
1005 * Remove the head of the list. This function does not take any locks
1006 * so must be used with appropriate locks held only. The head item is
1007 * returned or %NULL if the list is empty.
1009 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1010 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1012 struct sk_buff *skb = skb_peek(list);
1013 if (skb)
1014 __skb_unlink(skb, list);
1015 return skb;
1019 * __skb_dequeue_tail - remove from the tail of the queue
1020 * @list: list to dequeue from
1022 * Remove the tail of the list. This function does not take any locks
1023 * so must be used with appropriate locks held only. The tail item is
1024 * returned or %NULL if the list is empty.
1026 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1027 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1029 struct sk_buff *skb = skb_peek_tail(list);
1030 if (skb)
1031 __skb_unlink(skb, list);
1032 return skb;
1036 static inline int skb_is_nonlinear(const struct sk_buff *skb)
1038 return skb->data_len;
1041 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1043 return skb->len - skb->data_len;
1046 static inline int skb_pagelen(const struct sk_buff *skb)
1048 int i, len = 0;
1050 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1051 len += skb_shinfo(skb)->frags[i].size;
1052 return len + skb_headlen(skb);
1055 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1056 struct page *page, int off, int size)
1058 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1060 frag->page = page;
1061 frag->page_offset = off;
1062 frag->size = size;
1063 skb_shinfo(skb)->nr_frags = i + 1;
1066 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1067 int off, int size);
1069 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1070 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frags(skb))
1071 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1073 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1074 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1076 return skb->head + skb->tail;
1079 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1081 skb->tail = skb->data - skb->head;
1084 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1086 skb_reset_tail_pointer(skb);
1087 skb->tail += offset;
1089 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1090 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1092 return skb->tail;
1095 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1097 skb->tail = skb->data;
1100 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1102 skb->tail = skb->data + offset;
1105 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1108 * Add data to an sk_buff
1110 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1111 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1113 unsigned char *tmp = skb_tail_pointer(skb);
1114 SKB_LINEAR_ASSERT(skb);
1115 skb->tail += len;
1116 skb->len += len;
1117 return tmp;
1120 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1121 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1123 skb->data -= len;
1124 skb->len += len;
1125 return skb->data;
1128 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1129 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1131 skb->len -= len;
1132 BUG_ON(skb->len < skb->data_len);
1133 return skb->data += len;
1136 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1138 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1141 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1143 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1145 if (len > skb_headlen(skb) &&
1146 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1147 return NULL;
1148 skb->len -= len;
1149 return skb->data += len;
1152 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1154 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1157 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1159 if (likely(len <= skb_headlen(skb)))
1160 return 1;
1161 if (unlikely(len > skb->len))
1162 return 0;
1163 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1167 * skb_headroom - bytes at buffer head
1168 * @skb: buffer to check
1170 * Return the number of bytes of free space at the head of an &sk_buff.
1172 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1174 return skb->data - skb->head;
1178 * skb_tailroom - bytes at buffer end
1179 * @skb: buffer to check
1181 * Return the number of bytes of free space at the tail of an sk_buff
1183 static inline int skb_tailroom(const struct sk_buff *skb)
1185 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1189 * skb_reserve - adjust headroom
1190 * @skb: buffer to alter
1191 * @len: bytes to move
1193 * Increase the headroom of an empty &sk_buff by reducing the tail
1194 * room. This is only allowed for an empty buffer.
1196 static inline void skb_reserve(struct sk_buff *skb, int len)
1198 skb->data += len;
1199 skb->tail += len;
1202 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1203 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1205 return skb->head + skb->transport_header;
1208 static inline void skb_reset_transport_header(struct sk_buff *skb)
1210 skb->transport_header = skb->data - skb->head;
1213 static inline void skb_set_transport_header(struct sk_buff *skb,
1214 const int offset)
1216 skb_reset_transport_header(skb);
1217 skb->transport_header += offset;
1220 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1222 return skb->head + skb->network_header;
1225 static inline void skb_reset_network_header(struct sk_buff *skb)
1227 skb->network_header = skb->data - skb->head;
1230 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1232 skb_reset_network_header(skb);
1233 skb->network_header += offset;
1236 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1238 return skb->head + skb->mac_header;
1241 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1243 return skb->mac_header != ~0U;
1246 static inline void skb_reset_mac_header(struct sk_buff *skb)
1248 skb->mac_header = skb->data - skb->head;
1251 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1253 skb_reset_mac_header(skb);
1254 skb->mac_header += offset;
1257 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1259 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1261 return skb->transport_header;
1264 static inline void skb_reset_transport_header(struct sk_buff *skb)
1266 skb->transport_header = skb->data;
1269 static inline void skb_set_transport_header(struct sk_buff *skb,
1270 const int offset)
1272 skb->transport_header = skb->data + offset;
1275 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1277 return skb->network_header;
1280 static inline void skb_reset_network_header(struct sk_buff *skb)
1282 skb->network_header = skb->data;
1285 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1287 skb->network_header = skb->data + offset;
1290 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1292 return skb->mac_header;
1295 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1297 return skb->mac_header != NULL;
1300 static inline void skb_reset_mac_header(struct sk_buff *skb)
1302 skb->mac_header = skb->data;
1305 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1307 skb->mac_header = skb->data + offset;
1309 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1311 static inline int skb_transport_offset(const struct sk_buff *skb)
1313 return skb_transport_header(skb) - skb->data;
1316 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1318 return skb->transport_header - skb->network_header;
1321 static inline int skb_network_offset(const struct sk_buff *skb)
1323 return skb_network_header(skb) - skb->data;
1327 * CPUs often take a performance hit when accessing unaligned memory
1328 * locations. The actual performance hit varies, it can be small if the
1329 * hardware handles it or large if we have to take an exception and fix it
1330 * in software.
1332 * Since an ethernet header is 14 bytes network drivers often end up with
1333 * the IP header at an unaligned offset. The IP header can be aligned by
1334 * shifting the start of the packet by 2 bytes. Drivers should do this
1335 * with:
1337 * skb_reserve(skb, NET_IP_ALIGN);
1339 * The downside to this alignment of the IP header is that the DMA is now
1340 * unaligned. On some architectures the cost of an unaligned DMA is high
1341 * and this cost outweighs the gains made by aligning the IP header.
1343 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1344 * to be overridden.
1346 #ifndef NET_IP_ALIGN
1347 #define NET_IP_ALIGN 2
1348 #endif
1351 * The networking layer reserves some headroom in skb data (via
1352 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1353 * the header has to grow. In the default case, if the header has to grow
1354 * 32 bytes or less we avoid the reallocation.
1356 * Unfortunately this headroom changes the DMA alignment of the resulting
1357 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1358 * on some architectures. An architecture can override this value,
1359 * perhaps setting it to a cacheline in size (since that will maintain
1360 * cacheline alignment of the DMA). It must be a power of 2.
1362 * Various parts of the networking layer expect at least 32 bytes of
1363 * headroom, you should not reduce this.
1364 * With RPS, we raised NET_SKB_PAD to 64 so that get_rps_cpus() fetches span
1365 * a 64 bytes aligned block to fit modern (>= 64 bytes) cache line sizes
1366 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1368 #ifndef NET_SKB_PAD
1369 #define NET_SKB_PAD 64
1370 #endif
1372 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1374 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1376 if (unlikely(skb->data_len)) {
1377 WARN_ON(1);
1378 return;
1380 skb->len = len;
1381 skb_set_tail_pointer(skb, len);
1384 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1386 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1388 if (skb->data_len)
1389 return ___pskb_trim(skb, len);
1390 __skb_trim(skb, len);
1391 return 0;
1394 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1396 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1400 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1401 * @skb: buffer to alter
1402 * @len: new length
1404 * This is identical to pskb_trim except that the caller knows that
1405 * the skb is not cloned so we should never get an error due to out-
1406 * of-memory.
1408 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1410 int err = pskb_trim(skb, len);
1411 BUG_ON(err);
1415 * skb_orphan - orphan a buffer
1416 * @skb: buffer to orphan
1418 * If a buffer currently has an owner then we call the owner's
1419 * destructor function and make the @skb unowned. The buffer continues
1420 * to exist but is no longer charged to its former owner.
1422 static inline void skb_orphan(struct sk_buff *skb)
1424 if (skb->destructor)
1425 skb->destructor(skb);
1426 skb->destructor = NULL;
1427 skb->sk = NULL;
1431 * __skb_queue_purge - empty a list
1432 * @list: list to empty
1434 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1435 * the list and one reference dropped. This function does not take the
1436 * list lock and the caller must hold the relevant locks to use it.
1438 extern void skb_queue_purge(struct sk_buff_head *list);
1439 static inline void __skb_queue_purge(struct sk_buff_head *list)
1441 struct sk_buff *skb;
1442 while ((skb = __skb_dequeue(list)) != NULL)
1443 kfree_skb(skb);
1447 * __dev_alloc_skb - allocate an skbuff for receiving
1448 * @length: length to allocate
1449 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1451 * Allocate a new &sk_buff and assign it a usage count of one. The
1452 * buffer has unspecified headroom built in. Users should allocate
1453 * the headroom they think they need without accounting for the
1454 * built in space. The built in space is used for optimisations.
1456 * %NULL is returned if there is no free memory.
1458 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1459 gfp_t gfp_mask)
1461 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1462 if (likely(skb))
1463 skb_reserve(skb, NET_SKB_PAD);
1464 return skb;
1467 extern struct sk_buff *dev_alloc_skb(unsigned int length);
1469 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1470 unsigned int length, gfp_t gfp_mask);
1473 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1474 * @dev: network device to receive on
1475 * @length: length to allocate
1477 * Allocate a new &sk_buff and assign it a usage count of one. The
1478 * buffer has unspecified headroom built in. Users should allocate
1479 * the headroom they think they need without accounting for the
1480 * built in space. The built in space is used for optimisations.
1482 * %NULL is returned if there is no free memory. Although this function
1483 * allocates memory it can be called from an interrupt.
1485 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1486 unsigned int length)
1488 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1491 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1492 unsigned int length)
1494 struct sk_buff *skb = netdev_alloc_skb(dev, length + NET_IP_ALIGN);
1496 if (NET_IP_ALIGN && skb)
1497 skb_reserve(skb, NET_IP_ALIGN);
1498 return skb;
1501 extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1504 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1505 * @dev: network device to receive on
1507 * Allocate a new page node local to the specified device.
1509 * %NULL is returned if there is no free memory.
1511 static inline struct page *netdev_alloc_page(struct net_device *dev)
1513 return __netdev_alloc_page(dev, GFP_ATOMIC);
1516 static inline void netdev_free_page(struct net_device *dev, struct page *page)
1518 __free_page(page);
1522 * skb_clone_writable - is the header of a clone writable
1523 * @skb: buffer to check
1524 * @len: length up to which to write
1526 * Returns true if modifying the header part of the cloned buffer
1527 * does not requires the data to be copied.
1529 static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1531 return !skb_header_cloned(skb) &&
1532 skb_headroom(skb) + len <= skb->hdr_len;
1535 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1536 int cloned)
1538 int delta = 0;
1540 if (headroom < NET_SKB_PAD)
1541 headroom = NET_SKB_PAD;
1542 if (headroom > skb_headroom(skb))
1543 delta = headroom - skb_headroom(skb);
1545 if (delta || cloned)
1546 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1547 GFP_ATOMIC);
1548 return 0;
1552 * skb_cow - copy header of skb when it is required
1553 * @skb: buffer to cow
1554 * @headroom: needed headroom
1556 * If the skb passed lacks sufficient headroom or its data part
1557 * is shared, data is reallocated. If reallocation fails, an error
1558 * is returned and original skb is not changed.
1560 * The result is skb with writable area skb->head...skb->tail
1561 * and at least @headroom of space at head.
1563 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1565 return __skb_cow(skb, headroom, skb_cloned(skb));
1569 * skb_cow_head - skb_cow but only making the head writable
1570 * @skb: buffer to cow
1571 * @headroom: needed headroom
1573 * This function is identical to skb_cow except that we replace the
1574 * skb_cloned check by skb_header_cloned. It should be used when
1575 * you only need to push on some header and do not need to modify
1576 * the data.
1578 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1580 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1584 * skb_padto - pad an skbuff up to a minimal size
1585 * @skb: buffer to pad
1586 * @len: minimal length
1588 * Pads up a buffer to ensure the trailing bytes exist and are
1589 * blanked. If the buffer already contains sufficient data it
1590 * is untouched. Otherwise it is extended. Returns zero on
1591 * success. The skb is freed on error.
1594 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1596 unsigned int size = skb->len;
1597 if (likely(size >= len))
1598 return 0;
1599 return skb_pad(skb, len - size);
1602 static inline int skb_add_data(struct sk_buff *skb,
1603 char __user *from, int copy)
1605 const int off = skb->len;
1607 if (skb->ip_summed == CHECKSUM_NONE) {
1608 int err = 0;
1609 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1610 copy, 0, &err);
1611 if (!err) {
1612 skb->csum = csum_block_add(skb->csum, csum, off);
1613 return 0;
1615 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1616 return 0;
1618 __skb_trim(skb, off);
1619 return -EFAULT;
1622 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1623 struct page *page, int off)
1625 if (i) {
1626 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1628 return page == frag->page &&
1629 off == frag->page_offset + frag->size;
1631 return 0;
1634 static inline int __skb_linearize(struct sk_buff *skb)
1636 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1640 * skb_linearize - convert paged skb to linear one
1641 * @skb: buffer to linarize
1643 * If there is no free memory -ENOMEM is returned, otherwise zero
1644 * is returned and the old skb data released.
1646 static inline int skb_linearize(struct sk_buff *skb)
1648 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1652 * skb_linearize_cow - make sure skb is linear and writable
1653 * @skb: buffer to process
1655 * If there is no free memory -ENOMEM is returned, otherwise zero
1656 * is returned and the old skb data released.
1658 static inline int skb_linearize_cow(struct sk_buff *skb)
1660 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1661 __skb_linearize(skb) : 0;
1665 * skb_postpull_rcsum - update checksum for received skb after pull
1666 * @skb: buffer to update
1667 * @start: start of data before pull
1668 * @len: length of data pulled
1670 * After doing a pull on a received packet, you need to call this to
1671 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1672 * CHECKSUM_NONE so that it can be recomputed from scratch.
1675 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1676 const void *start, unsigned int len)
1678 if (skb->ip_summed == CHECKSUM_COMPLETE)
1679 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1682 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1685 * pskb_trim_rcsum - trim received skb and update checksum
1686 * @skb: buffer to trim
1687 * @len: new length
1689 * This is exactly the same as pskb_trim except that it ensures the
1690 * checksum of received packets are still valid after the operation.
1693 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1695 if (likely(len >= skb->len))
1696 return 0;
1697 if (skb->ip_summed == CHECKSUM_COMPLETE)
1698 skb->ip_summed = CHECKSUM_NONE;
1699 return __pskb_trim(skb, len);
1702 #define skb_queue_walk(queue, skb) \
1703 for (skb = (queue)->next; \
1704 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1705 skb = skb->next)
1707 #define skb_queue_walk_safe(queue, skb, tmp) \
1708 for (skb = (queue)->next, tmp = skb->next; \
1709 skb != (struct sk_buff *)(queue); \
1710 skb = tmp, tmp = skb->next)
1712 #define skb_queue_walk_from(queue, skb) \
1713 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1714 skb = skb->next)
1716 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1717 for (tmp = skb->next; \
1718 skb != (struct sk_buff *)(queue); \
1719 skb = tmp, tmp = skb->next)
1721 #define skb_queue_reverse_walk(queue, skb) \
1722 for (skb = (queue)->prev; \
1723 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1724 skb = skb->prev)
1727 static inline bool skb_has_frags(const struct sk_buff *skb)
1729 return skb_shinfo(skb)->frag_list != NULL;
1732 static inline void skb_frag_list_init(struct sk_buff *skb)
1734 skb_shinfo(skb)->frag_list = NULL;
1737 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1739 frag->next = skb_shinfo(skb)->frag_list;
1740 skb_shinfo(skb)->frag_list = frag;
1743 #define skb_walk_frags(skb, iter) \
1744 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1746 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1747 int *peeked, int *err);
1748 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1749 int noblock, int *err);
1750 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1751 struct poll_table_struct *wait);
1752 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1753 int offset, struct iovec *to,
1754 int size);
1755 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1756 int hlen,
1757 struct iovec *iov);
1758 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1759 int offset,
1760 const struct iovec *from,
1761 int from_offset,
1762 int len);
1763 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1764 int offset,
1765 const struct iovec *to,
1766 int to_offset,
1767 int size);
1768 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1769 extern void skb_free_datagram_locked(struct sock *sk,
1770 struct sk_buff *skb);
1771 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1772 unsigned int flags);
1773 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1774 int len, __wsum csum);
1775 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1776 void *to, int len);
1777 extern int skb_store_bits(struct sk_buff *skb, int offset,
1778 const void *from, int len);
1779 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1780 int offset, u8 *to, int len,
1781 __wsum csum);
1782 extern int skb_splice_bits(struct sk_buff *skb,
1783 unsigned int offset,
1784 struct pipe_inode_info *pipe,
1785 unsigned int len,
1786 unsigned int flags);
1787 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1788 extern void skb_split(struct sk_buff *skb,
1789 struct sk_buff *skb1, const u32 len);
1790 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1791 int shiftlen);
1793 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1795 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1796 int len, void *buffer)
1798 int hlen = skb_headlen(skb);
1800 if (hlen - offset >= len)
1801 return skb->data + offset;
1803 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1804 return NULL;
1806 return buffer;
1809 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1810 void *to,
1811 const unsigned int len)
1813 memcpy(to, skb->data, len);
1816 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1817 const int offset, void *to,
1818 const unsigned int len)
1820 memcpy(to, skb->data + offset, len);
1823 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1824 const void *from,
1825 const unsigned int len)
1827 memcpy(skb->data, from, len);
1830 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1831 const int offset,
1832 const void *from,
1833 const unsigned int len)
1835 memcpy(skb->data + offset, from, len);
1838 extern void skb_init(void);
1840 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1842 return skb->tstamp;
1846 * skb_get_timestamp - get timestamp from a skb
1847 * @skb: skb to get stamp from
1848 * @stamp: pointer to struct timeval to store stamp in
1850 * Timestamps are stored in the skb as offsets to a base timestamp.
1851 * This function converts the offset back to a struct timeval and stores
1852 * it in stamp.
1854 static inline void skb_get_timestamp(const struct sk_buff *skb,
1855 struct timeval *stamp)
1857 *stamp = ktime_to_timeval(skb->tstamp);
1860 static inline void skb_get_timestampns(const struct sk_buff *skb,
1861 struct timespec *stamp)
1863 *stamp = ktime_to_timespec(skb->tstamp);
1866 static inline void __net_timestamp(struct sk_buff *skb)
1868 skb->tstamp = ktime_get_real();
1871 static inline ktime_t net_timedelta(ktime_t t)
1873 return ktime_sub(ktime_get_real(), t);
1876 static inline ktime_t net_invalid_timestamp(void)
1878 return ktime_set(0, 0);
1882 * skb_tstamp_tx - queue clone of skb with send time stamps
1883 * @orig_skb: the original outgoing packet
1884 * @hwtstamps: hardware time stamps, may be NULL if not available
1886 * If the skb has a socket associated, then this function clones the
1887 * skb (thus sharing the actual data and optional structures), stores
1888 * the optional hardware time stamping information (if non NULL) or
1889 * generates a software time stamp (otherwise), then queues the clone
1890 * to the error queue of the socket. Errors are silently ignored.
1892 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
1893 struct skb_shared_hwtstamps *hwtstamps);
1895 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
1896 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
1898 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
1900 return skb->ip_summed & CHECKSUM_UNNECESSARY;
1904 * skb_checksum_complete - Calculate checksum of an entire packet
1905 * @skb: packet to process
1907 * This function calculates the checksum over the entire packet plus
1908 * the value of skb->csum. The latter can be used to supply the
1909 * checksum of a pseudo header as used by TCP/UDP. It returns the
1910 * checksum.
1912 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1913 * this function can be used to verify that checksum on received
1914 * packets. In that case the function should return zero if the
1915 * checksum is correct. In particular, this function will return zero
1916 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1917 * hardware has already verified the correctness of the checksum.
1919 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
1921 return skb_csum_unnecessary(skb) ?
1922 0 : __skb_checksum_complete(skb);
1925 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1926 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
1927 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1929 if (nfct && atomic_dec_and_test(&nfct->use))
1930 nf_conntrack_destroy(nfct);
1932 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1934 if (nfct)
1935 atomic_inc(&nfct->use);
1937 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1939 if (skb)
1940 atomic_inc(&skb->users);
1942 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1944 if (skb)
1945 kfree_skb(skb);
1947 #endif
1948 #ifdef CONFIG_BRIDGE_NETFILTER
1949 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1951 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1952 kfree(nf_bridge);
1954 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1956 if (nf_bridge)
1957 atomic_inc(&nf_bridge->use);
1959 #endif /* CONFIG_BRIDGE_NETFILTER */
1960 static inline void nf_reset(struct sk_buff *skb)
1962 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1963 nf_conntrack_put(skb->nfct);
1964 skb->nfct = NULL;
1965 nf_conntrack_put_reasm(skb->nfct_reasm);
1966 skb->nfct_reasm = NULL;
1967 #endif
1968 #ifdef CONFIG_BRIDGE_NETFILTER
1969 nf_bridge_put(skb->nf_bridge);
1970 skb->nf_bridge = NULL;
1971 #endif
1974 /* Note: This doesn't put any conntrack and bridge info in dst. */
1975 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1977 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1978 dst->nfct = src->nfct;
1979 nf_conntrack_get(src->nfct);
1980 dst->nfctinfo = src->nfctinfo;
1981 dst->nfct_reasm = src->nfct_reasm;
1982 nf_conntrack_get_reasm(src->nfct_reasm);
1983 #endif
1984 #ifdef CONFIG_BRIDGE_NETFILTER
1985 dst->nf_bridge = src->nf_bridge;
1986 nf_bridge_get(src->nf_bridge);
1987 #endif
1990 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1992 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1993 nf_conntrack_put(dst->nfct);
1994 nf_conntrack_put_reasm(dst->nfct_reasm);
1995 #endif
1996 #ifdef CONFIG_BRIDGE_NETFILTER
1997 nf_bridge_put(dst->nf_bridge);
1998 #endif
1999 __nf_copy(dst, src);
2002 #ifdef CONFIG_NETWORK_SECMARK
2003 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2005 to->secmark = from->secmark;
2008 static inline void skb_init_secmark(struct sk_buff *skb)
2010 skb->secmark = 0;
2012 #else
2013 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2016 static inline void skb_init_secmark(struct sk_buff *skb)
2018 #endif
2020 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2022 skb->queue_mapping = queue_mapping;
2025 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2027 return skb->queue_mapping;
2030 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2032 to->queue_mapping = from->queue_mapping;
2035 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2037 skb->queue_mapping = rx_queue + 1;
2040 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2042 return skb->queue_mapping - 1;
2045 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2047 return (skb->queue_mapping != 0);
2050 extern u16 skb_tx_hash(const struct net_device *dev,
2051 const struct sk_buff *skb);
2053 #ifdef CONFIG_XFRM
2054 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2056 return skb->sp;
2058 #else
2059 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2061 return NULL;
2063 #endif
2065 static inline int skb_is_gso(const struct sk_buff *skb)
2067 return skb_shinfo(skb)->gso_size;
2070 static inline int skb_is_gso_v6(const struct sk_buff *skb)
2072 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2075 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2077 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2079 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2080 * wanted then gso_type will be set. */
2081 struct skb_shared_info *shinfo = skb_shinfo(skb);
2082 if (shinfo->gso_size != 0 && unlikely(shinfo->gso_type == 0)) {
2083 __skb_warn_lro_forwarding(skb);
2084 return true;
2086 return false;
2089 static inline void skb_forward_csum(struct sk_buff *skb)
2091 /* Unfortunately we don't support this one. Any brave souls? */
2092 if (skb->ip_summed == CHECKSUM_COMPLETE)
2093 skb->ip_summed = CHECKSUM_NONE;
2096 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2097 #endif /* __KERNEL__ */
2098 #endif /* _LINUX_SKBUFF_H */