]> bbs.cooldavid.org Git - net-next-2.6.git/blame - include/linux/skbuff.h
[NETLINK]: Fix "nocast type" warnings
[net-next-2.6.git] / include / linux / skbuff.h
CommitLineData
1da177e4
LT
1/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
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.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
17#include <linux/config.h>
18#include <linux/kernel.h>
19#include <linux/compiler.h>
20#include <linux/time.h>
21#include <linux/cache.h>
22
23#include <asm/atomic.h>
24#include <asm/types.h>
25#include <linux/spinlock.h>
26#include <linux/mm.h>
27#include <linux/highmem.h>
28#include <linux/poll.h>
29#include <linux/net.h>
3fc7e8a6 30#include <linux/textsearch.h>
1da177e4
LT
31#include <net/checksum.h>
32
33#define HAVE_ALLOC_SKB /* For the drivers to know */
34#define HAVE_ALIGNABLE_SKB /* Ditto 8) */
35#define SLAB_SKB /* Slabified skbuffs */
36
37#define CHECKSUM_NONE 0
38#define CHECKSUM_HW 1
39#define CHECKSUM_UNNECESSARY 2
40
41#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
42 ~(SMP_CACHE_BYTES - 1))
43#define SKB_MAX_ORDER(X, ORDER) (((PAGE_SIZE << (ORDER)) - (X) - \
44 sizeof(struct skb_shared_info)) & \
45 ~(SMP_CACHE_BYTES - 1))
46#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
47#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
48
49/* A. Checksumming of received packets by device.
50 *
51 * NONE: device failed to checksum this packet.
52 * skb->csum is undefined.
53 *
54 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
55 * skb->csum is undefined.
56 * It is bad option, but, unfortunately, many of vendors do this.
57 * Apparently with secret goal to sell you new device, when you
58 * will add new protocol to your host. F.e. IPv6. 8)
59 *
60 * HW: the most generic way. Device supplied checksum of _all_
61 * the packet as seen by netif_rx in skb->csum.
62 * NOTE: Even if device supports only some protocols, but
63 * is able to produce some skb->csum, it MUST use HW,
64 * not UNNECESSARY.
65 *
66 * B. Checksumming on output.
67 *
68 * NONE: skb is checksummed by protocol or csum is not required.
69 *
70 * HW: device is required to csum packet as seen by hard_start_xmit
71 * from skb->h.raw to the end and to record the checksum
72 * at skb->h.raw+skb->csum.
73 *
74 * Device must show its capabilities in dev->features, set
75 * at device setup time.
76 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
77 * everything.
78 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
79 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
80 * TCP/UDP over IPv4. Sigh. Vendors like this
81 * way by an unknown reason. Though, see comment above
82 * about CHECKSUM_UNNECESSARY. 8)
83 *
84 * Any questions? No questions, good. --ANK
85 */
86
1da177e4
LT
87struct net_device;
88
89#ifdef CONFIG_NETFILTER
90struct nf_conntrack {
91 atomic_t use;
92 void (*destroy)(struct nf_conntrack *);
93};
94
95#ifdef CONFIG_BRIDGE_NETFILTER
96struct nf_bridge_info {
97 atomic_t use;
98 struct net_device *physindev;
99 struct net_device *physoutdev;
100#if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
101 struct net_device *netoutdev;
102#endif
103 unsigned int mask;
104 unsigned long data[32 / sizeof(unsigned long)];
105};
106#endif
107
108#endif
109
110struct sk_buff_head {
111 /* These two members must be first. */
112 struct sk_buff *next;
113 struct sk_buff *prev;
114
115 __u32 qlen;
116 spinlock_t lock;
117};
118
119struct sk_buff;
120
121/* To allow 64K frame to be packed as single skb without frag_list */
122#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
123
124typedef struct skb_frag_struct skb_frag_t;
125
126struct skb_frag_struct {
127 struct page *page;
128 __u16 page_offset;
129 __u16 size;
130};
131
132/* This data is invariant across clones and lives at
133 * the end of the header data, ie. at skb->end.
134 */
135struct skb_shared_info {
136 atomic_t dataref;
137 unsigned int nr_frags;
138 unsigned short tso_size;
139 unsigned short tso_segs;
140 struct sk_buff *frag_list;
141 skb_frag_t frags[MAX_SKB_FRAGS];
142};
143
144/* We divide dataref into two halves. The higher 16 bits hold references
145 * to the payload part of skb->data. The lower 16 bits hold references to
146 * the entire skb->data. It is up to the users of the skb to agree on
147 * where the payload starts.
148 *
149 * All users must obey the rule that the skb->data reference count must be
150 * greater than or equal to the payload reference count.
151 *
152 * Holding a reference to the payload part means that the user does not
153 * care about modifications to the header part of skb->data.
154 */
155#define SKB_DATAREF_SHIFT 16
156#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
157
158/**
159 * struct sk_buff - socket buffer
160 * @next: Next buffer in list
161 * @prev: Previous buffer in list
162 * @list: List we are on
163 * @sk: Socket we are owned by
164 * @stamp: Time we arrived
165 * @dev: Device we arrived on/are leaving by
166 * @input_dev: Device we arrived on
167 * @real_dev: The real device we are using
168 * @h: Transport layer header
169 * @nh: Network layer header
170 * @mac: Link layer header
67be2dd1
MW
171 * @dst: destination entry
172 * @sp: the security path, used for xfrm
1da177e4
LT
173 * @cb: Control buffer. Free for use by every layer. Put private vars here
174 * @len: Length of actual data
175 * @data_len: Data length
176 * @mac_len: Length of link layer header
177 * @csum: Checksum
67be2dd1 178 * @local_df: allow local fragmentation
1da177e4
LT
179 * @cloned: Head may be cloned (check refcnt to be sure)
180 * @nohdr: Payload reference only, must not modify header
181 * @pkt_type: Packet class
182 * @ip_summed: Driver fed us an IP checksum
183 * @priority: Packet queueing priority
184 * @users: User count - see {datagram,tcp}.c
185 * @protocol: Packet protocol from driver
1da177e4
LT
186 * @truesize: Buffer size
187 * @head: Head of buffer
188 * @data: Data head pointer
189 * @tail: Tail pointer
190 * @end: End pointer
191 * @destructor: Destruct function
192 * @nfmark: Can be used for communication between hooks
193 * @nfcache: Cache info
194 * @nfct: Associated connection, if any
195 * @nfctinfo: Relationship of this skb to the connection
1da177e4
LT
196 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
197 * @private: Data which is private to the HIPPI implementation
198 * @tc_index: Traffic control index
199 * @tc_verd: traffic control verdict
200 * @tc_classid: traffic control classid
201 */
202
203struct sk_buff {
204 /* These two members must be first. */
205 struct sk_buff *next;
206 struct sk_buff *prev;
207
208 struct sk_buff_head *list;
209 struct sock *sk;
210 struct timeval stamp;
211 struct net_device *dev;
212 struct net_device *input_dev;
213 struct net_device *real_dev;
214
215 union {
216 struct tcphdr *th;
217 struct udphdr *uh;
218 struct icmphdr *icmph;
219 struct igmphdr *igmph;
220 struct iphdr *ipiph;
221 struct ipv6hdr *ipv6h;
222 unsigned char *raw;
223 } h;
224
225 union {
226 struct iphdr *iph;
227 struct ipv6hdr *ipv6h;
228 struct arphdr *arph;
229 unsigned char *raw;
230 } nh;
231
232 union {
233 unsigned char *raw;
234 } mac;
235
236 struct dst_entry *dst;
237 struct sec_path *sp;
238
239 /*
240 * This is the control buffer. It is free to use for every
241 * layer. Please put your private variables there. If you
242 * want to keep them across layers you have to do a skb_clone()
243 * first. This is owned by whoever has the skb queued ATM.
244 */
245 char cb[40];
246
247 unsigned int len,
248 data_len,
249 mac_len,
250 csum;
1da177e4 251 __u32 priority;
1cbb3380
TG
252 __u8 local_df:1,
253 cloned:1,
254 ip_summed:2,
255 nohdr:1;
256 /* 3 bits spare */
257 __u8 pkt_type;
258 __u16 protocol;
1da177e4
LT
259
260 void (*destructor)(struct sk_buff *skb);
261#ifdef CONFIG_NETFILTER
1cbb3380 262 unsigned long nfmark;
1da177e4
LT
263 __u32 nfcache;
264 __u32 nfctinfo;
265 struct nf_conntrack *nfct;
1da177e4
LT
266#ifdef CONFIG_BRIDGE_NETFILTER
267 struct nf_bridge_info *nf_bridge;
268#endif
269#endif /* CONFIG_NETFILTER */
270#if defined(CONFIG_HIPPI)
271 union {
272 __u32 ifield;
273 } private;
274#endif
275#ifdef CONFIG_NET_SCHED
276 __u32 tc_index; /* traffic control index */
277#ifdef CONFIG_NET_CLS_ACT
278 __u32 tc_verd; /* traffic control verdict */
279 __u32 tc_classid; /* traffic control classid */
280#endif
281
282#endif
283
284
285 /* These elements must be at the end, see alloc_skb() for details. */
286 unsigned int truesize;
287 atomic_t users;
288 unsigned char *head,
289 *data,
290 *tail,
291 *end;
292};
293
294#ifdef __KERNEL__
295/*
296 * Handling routines are only of interest to the kernel
297 */
298#include <linux/slab.h>
299
300#include <asm/system.h>
301
302extern void __kfree_skb(struct sk_buff *skb);
86a76caf
VF
303extern struct sk_buff *alloc_skb(unsigned int size,
304 unsigned int __nocast priority);
1da177e4 305extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
86a76caf
VF
306 unsigned int size,
307 unsigned int __nocast priority);
1da177e4 308extern void kfree_skbmem(struct sk_buff *skb);
86a76caf
VF
309extern struct sk_buff *skb_clone(struct sk_buff *skb,
310 unsigned int __nocast priority);
311extern struct sk_buff *skb_copy(const struct sk_buff *skb,
312 unsigned int __nocast priority);
313extern struct sk_buff *pskb_copy(struct sk_buff *skb,
314 unsigned int __nocast gfp_mask);
1da177e4 315extern int pskb_expand_head(struct sk_buff *skb,
86a76caf
VF
316 int nhead, int ntail,
317 unsigned int __nocast gfp_mask);
1da177e4
LT
318extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
319 unsigned int headroom);
320extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
321 int newheadroom, int newtailroom,
86a76caf 322 unsigned int __nocast priority);
1da177e4
LT
323extern struct sk_buff * skb_pad(struct sk_buff *skb, int pad);
324#define dev_kfree_skb(a) kfree_skb(a)
325extern void skb_over_panic(struct sk_buff *skb, int len,
326 void *here);
327extern void skb_under_panic(struct sk_buff *skb, int len,
328 void *here);
329
677e90ed
TG
330struct skb_seq_state
331{
332 __u32 lower_offset;
333 __u32 upper_offset;
334 __u32 frag_idx;
335 __u32 stepped_offset;
336 struct sk_buff *root_skb;
337 struct sk_buff *cur_skb;
338 __u8 *frag_data;
339};
340
341extern void skb_prepare_seq_read(struct sk_buff *skb,
342 unsigned int from, unsigned int to,
343 struct skb_seq_state *st);
344extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
345 struct skb_seq_state *st);
346extern void skb_abort_seq_read(struct skb_seq_state *st);
347
3fc7e8a6
TG
348extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
349 unsigned int to, struct ts_config *config,
350 struct ts_state *state);
351
1da177e4
LT
352/* Internal */
353#define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
354
355/**
356 * skb_queue_empty - check if a queue is empty
357 * @list: queue head
358 *
359 * Returns true if the queue is empty, false otherwise.
360 */
361static inline int skb_queue_empty(const struct sk_buff_head *list)
362{
363 return list->next == (struct sk_buff *)list;
364}
365
366/**
367 * skb_get - reference buffer
368 * @skb: buffer to reference
369 *
370 * Makes another reference to a socket buffer and returns a pointer
371 * to the buffer.
372 */
373static inline struct sk_buff *skb_get(struct sk_buff *skb)
374{
375 atomic_inc(&skb->users);
376 return skb;
377}
378
379/*
380 * If users == 1, we are the only owner and are can avoid redundant
381 * atomic change.
382 */
383
384/**
385 * kfree_skb - free an sk_buff
386 * @skb: buffer to free
387 *
388 * Drop a reference to the buffer and free it if the usage count has
389 * hit zero.
390 */
391static inline void kfree_skb(struct sk_buff *skb)
392{
393 if (likely(atomic_read(&skb->users) == 1))
394 smp_rmb();
395 else if (likely(!atomic_dec_and_test(&skb->users)))
396 return;
397 __kfree_skb(skb);
398}
399
400/**
401 * skb_cloned - is the buffer a clone
402 * @skb: buffer to check
403 *
404 * Returns true if the buffer was generated with skb_clone() and is
405 * one of multiple shared copies of the buffer. Cloned buffers are
406 * shared data so must not be written to under normal circumstances.
407 */
408static inline int skb_cloned(const struct sk_buff *skb)
409{
410 return skb->cloned &&
411 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
412}
413
414/**
415 * skb_header_cloned - is the header a clone
416 * @skb: buffer to check
417 *
418 * Returns true if modifying the header part of the buffer requires
419 * the data to be copied.
420 */
421static inline int skb_header_cloned(const struct sk_buff *skb)
422{
423 int dataref;
424
425 if (!skb->cloned)
426 return 0;
427
428 dataref = atomic_read(&skb_shinfo(skb)->dataref);
429 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
430 return dataref != 1;
431}
432
433/**
434 * skb_header_release - release reference to header
435 * @skb: buffer to operate on
436 *
437 * Drop a reference to the header part of the buffer. This is done
438 * by acquiring a payload reference. You must not read from the header
439 * part of skb->data after this.
440 */
441static inline void skb_header_release(struct sk_buff *skb)
442{
443 BUG_ON(skb->nohdr);
444 skb->nohdr = 1;
445 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
446}
447
448/**
449 * skb_shared - is the buffer shared
450 * @skb: buffer to check
451 *
452 * Returns true if more than one person has a reference to this
453 * buffer.
454 */
455static inline int skb_shared(const struct sk_buff *skb)
456{
457 return atomic_read(&skb->users) != 1;
458}
459
460/**
461 * skb_share_check - check if buffer is shared and if so clone it
462 * @skb: buffer to check
463 * @pri: priority for memory allocation
464 *
465 * If the buffer is shared the buffer is cloned and the old copy
466 * drops a reference. A new clone with a single reference is returned.
467 * If the buffer is not shared the original buffer is returned. When
468 * being called from interrupt status or with spinlocks held pri must
469 * be GFP_ATOMIC.
470 *
471 * NULL is returned on a memory allocation failure.
472 */
86a76caf
VF
473static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
474 unsigned int __nocast pri)
1da177e4
LT
475{
476 might_sleep_if(pri & __GFP_WAIT);
477 if (skb_shared(skb)) {
478 struct sk_buff *nskb = skb_clone(skb, pri);
479 kfree_skb(skb);
480 skb = nskb;
481 }
482 return skb;
483}
484
485/*
486 * Copy shared buffers into a new sk_buff. We effectively do COW on
487 * packets to handle cases where we have a local reader and forward
488 * and a couple of other messy ones. The normal one is tcpdumping
489 * a packet thats being forwarded.
490 */
491
492/**
493 * skb_unshare - make a copy of a shared buffer
494 * @skb: buffer to check
495 * @pri: priority for memory allocation
496 *
497 * If the socket buffer is a clone then this function creates a new
498 * copy of the data, drops a reference count on the old copy and returns
499 * the new copy with the reference count at 1. If the buffer is not a clone
500 * the original buffer is returned. When called with a spinlock held or
501 * from interrupt state @pri must be %GFP_ATOMIC
502 *
503 * %NULL is returned on a memory allocation failure.
504 */
505static inline struct sk_buff *skb_unshare(struct sk_buff *skb, int pri)
506{
507 might_sleep_if(pri & __GFP_WAIT);
508 if (skb_cloned(skb)) {
509 struct sk_buff *nskb = skb_copy(skb, pri);
510 kfree_skb(skb); /* Free our shared copy */
511 skb = nskb;
512 }
513 return skb;
514}
515
516/**
517 * skb_peek
518 * @list_: list to peek at
519 *
520 * Peek an &sk_buff. Unlike most other operations you _MUST_
521 * be careful with this one. A peek leaves the buffer on the
522 * list and someone else may run off with it. You must hold
523 * the appropriate locks or have a private queue to do this.
524 *
525 * Returns %NULL for an empty list or a pointer to the head element.
526 * The reference count is not incremented and the reference is therefore
527 * volatile. Use with caution.
528 */
529static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
530{
531 struct sk_buff *list = ((struct sk_buff *)list_)->next;
532 if (list == (struct sk_buff *)list_)
533 list = NULL;
534 return list;
535}
536
537/**
538 * skb_peek_tail
539 * @list_: list to peek at
540 *
541 * Peek an &sk_buff. Unlike most other operations you _MUST_
542 * be careful with this one. A peek leaves the buffer on the
543 * list and someone else may run off with it. You must hold
544 * the appropriate locks or have a private queue to do this.
545 *
546 * Returns %NULL for an empty list or a pointer to the tail element.
547 * The reference count is not incremented and the reference is therefore
548 * volatile. Use with caution.
549 */
550static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
551{
552 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
553 if (list == (struct sk_buff *)list_)
554 list = NULL;
555 return list;
556}
557
558/**
559 * skb_queue_len - get queue length
560 * @list_: list to measure
561 *
562 * Return the length of an &sk_buff queue.
563 */
564static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
565{
566 return list_->qlen;
567}
568
569static inline void skb_queue_head_init(struct sk_buff_head *list)
570{
571 spin_lock_init(&list->lock);
572 list->prev = list->next = (struct sk_buff *)list;
573 list->qlen = 0;
574}
575
576/*
577 * Insert an sk_buff at the start of a list.
578 *
579 * The "__skb_xxxx()" functions are the non-atomic ones that
580 * can only be called with interrupts disabled.
581 */
582
583/**
584 * __skb_queue_head - queue a buffer at the list head
585 * @list: list to use
586 * @newsk: buffer to queue
587 *
588 * Queue a buffer at the start of a list. This function takes no locks
589 * and you must therefore hold required locks before calling it.
590 *
591 * A buffer cannot be placed on two lists at the same time.
592 */
593extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
594static inline void __skb_queue_head(struct sk_buff_head *list,
595 struct sk_buff *newsk)
596{
597 struct sk_buff *prev, *next;
598
599 newsk->list = list;
600 list->qlen++;
601 prev = (struct sk_buff *)list;
602 next = prev->next;
603 newsk->next = next;
604 newsk->prev = prev;
605 next->prev = prev->next = newsk;
606}
607
608/**
609 * __skb_queue_tail - queue a buffer at the list tail
610 * @list: list to use
611 * @newsk: buffer to queue
612 *
613 * Queue a buffer at the end of a list. This function takes no locks
614 * and you must therefore hold required locks before calling it.
615 *
616 * A buffer cannot be placed on two lists at the same time.
617 */
618extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
619static inline void __skb_queue_tail(struct sk_buff_head *list,
620 struct sk_buff *newsk)
621{
622 struct sk_buff *prev, *next;
623
624 newsk->list = list;
625 list->qlen++;
626 next = (struct sk_buff *)list;
627 prev = next->prev;
628 newsk->next = next;
629 newsk->prev = prev;
630 next->prev = prev->next = newsk;
631}
632
633
634/**
635 * __skb_dequeue - remove from the head of the queue
636 * @list: list to dequeue from
637 *
638 * Remove the head of the list. This function does not take any locks
639 * so must be used with appropriate locks held only. The head item is
640 * returned or %NULL if the list is empty.
641 */
642extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
643static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
644{
645 struct sk_buff *next, *prev, *result;
646
647 prev = (struct sk_buff *) list;
648 next = prev->next;
649 result = NULL;
650 if (next != prev) {
651 result = next;
652 next = next->next;
653 list->qlen--;
654 next->prev = prev;
655 prev->next = next;
656 result->next = result->prev = NULL;
657 result->list = NULL;
658 }
659 return result;
660}
661
662
663/*
664 * Insert a packet on a list.
665 */
666extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk);
667static inline void __skb_insert(struct sk_buff *newsk,
668 struct sk_buff *prev, struct sk_buff *next,
669 struct sk_buff_head *list)
670{
671 newsk->next = next;
672 newsk->prev = prev;
673 next->prev = prev->next = newsk;
674 newsk->list = list;
675 list->qlen++;
676}
677
678/*
679 * Place a packet after a given packet in a list.
680 */
681extern void skb_append(struct sk_buff *old, struct sk_buff *newsk);
682static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk)
683{
684 __skb_insert(newsk, old, old->next, old->list);
685}
686
687/*
688 * remove sk_buff from list. _Must_ be called atomically, and with
689 * the list known..
690 */
691extern void skb_unlink(struct sk_buff *skb);
692static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
693{
694 struct sk_buff *next, *prev;
695
696 list->qlen--;
697 next = skb->next;
698 prev = skb->prev;
699 skb->next = skb->prev = NULL;
700 skb->list = NULL;
701 next->prev = prev;
702 prev->next = next;
703}
704
705
706/* XXX: more streamlined implementation */
707
708/**
709 * __skb_dequeue_tail - remove from the tail of the queue
710 * @list: list to dequeue from
711 *
712 * Remove the tail of the list. This function does not take any locks
713 * so must be used with appropriate locks held only. The tail item is
714 * returned or %NULL if the list is empty.
715 */
716extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
717static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
718{
719 struct sk_buff *skb = skb_peek_tail(list);
720 if (skb)
721 __skb_unlink(skb, list);
722 return skb;
723}
724
725
726static inline int skb_is_nonlinear(const struct sk_buff *skb)
727{
728 return skb->data_len;
729}
730
731static inline unsigned int skb_headlen(const struct sk_buff *skb)
732{
733 return skb->len - skb->data_len;
734}
735
736static inline int skb_pagelen(const struct sk_buff *skb)
737{
738 int i, len = 0;
739
740 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
741 len += skb_shinfo(skb)->frags[i].size;
742 return len + skb_headlen(skb);
743}
744
745static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
746 struct page *page, int off, int size)
747{
748 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
749
750 frag->page = page;
751 frag->page_offset = off;
752 frag->size = size;
753 skb_shinfo(skb)->nr_frags = i + 1;
754}
755
756#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
757#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
758#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
759
760/*
761 * Add data to an sk_buff
762 */
763static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
764{
765 unsigned char *tmp = skb->tail;
766 SKB_LINEAR_ASSERT(skb);
767 skb->tail += len;
768 skb->len += len;
769 return tmp;
770}
771
772/**
773 * skb_put - add data to a buffer
774 * @skb: buffer to use
775 * @len: amount of data to add
776 *
777 * This function extends the used data area of the buffer. If this would
778 * exceed the total buffer size the kernel will panic. A pointer to the
779 * first byte of the extra data is returned.
780 */
781static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
782{
783 unsigned char *tmp = skb->tail;
784 SKB_LINEAR_ASSERT(skb);
785 skb->tail += len;
786 skb->len += len;
787 if (unlikely(skb->tail>skb->end))
788 skb_over_panic(skb, len, current_text_addr());
789 return tmp;
790}
791
792static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
793{
794 skb->data -= len;
795 skb->len += len;
796 return skb->data;
797}
798
799/**
800 * skb_push - add data to the start of a buffer
801 * @skb: buffer to use
802 * @len: amount of data to add
803 *
804 * This function extends the used data area of the buffer at the buffer
805 * start. If this would exceed the total buffer headroom the kernel will
806 * panic. A pointer to the first byte of the extra data is returned.
807 */
808static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
809{
810 skb->data -= len;
811 skb->len += len;
812 if (unlikely(skb->data<skb->head))
813 skb_under_panic(skb, len, current_text_addr());
814 return skb->data;
815}
816
817static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
818{
819 skb->len -= len;
820 BUG_ON(skb->len < skb->data_len);
821 return skb->data += len;
822}
823
824/**
825 * skb_pull - remove data from the start of a buffer
826 * @skb: buffer to use
827 * @len: amount of data to remove
828 *
829 * This function removes data from the start of a buffer, returning
830 * the memory to the headroom. A pointer to the next data in the buffer
831 * is returned. Once the data has been pulled future pushes will overwrite
832 * the old data.
833 */
834static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
835{
836 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
837}
838
839extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
840
841static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
842{
843 if (len > skb_headlen(skb) &&
844 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
845 return NULL;
846 skb->len -= len;
847 return skb->data += len;
848}
849
850static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
851{
852 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
853}
854
855static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
856{
857 if (likely(len <= skb_headlen(skb)))
858 return 1;
859 if (unlikely(len > skb->len))
860 return 0;
861 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
862}
863
864/**
865 * skb_headroom - bytes at buffer head
866 * @skb: buffer to check
867 *
868 * Return the number of bytes of free space at the head of an &sk_buff.
869 */
870static inline int skb_headroom(const struct sk_buff *skb)
871{
872 return skb->data - skb->head;
873}
874
875/**
876 * skb_tailroom - bytes at buffer end
877 * @skb: buffer to check
878 *
879 * Return the number of bytes of free space at the tail of an sk_buff
880 */
881static inline int skb_tailroom(const struct sk_buff *skb)
882{
883 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
884}
885
886/**
887 * skb_reserve - adjust headroom
888 * @skb: buffer to alter
889 * @len: bytes to move
890 *
891 * Increase the headroom of an empty &sk_buff by reducing the tail
892 * room. This is only allowed for an empty buffer.
893 */
894static inline void skb_reserve(struct sk_buff *skb, unsigned int len)
895{
896 skb->data += len;
897 skb->tail += len;
898}
899
900/*
901 * CPUs often take a performance hit when accessing unaligned memory
902 * locations. The actual performance hit varies, it can be small if the
903 * hardware handles it or large if we have to take an exception and fix it
904 * in software.
905 *
906 * Since an ethernet header is 14 bytes network drivers often end up with
907 * the IP header at an unaligned offset. The IP header can be aligned by
908 * shifting the start of the packet by 2 bytes. Drivers should do this
909 * with:
910 *
911 * skb_reserve(NET_IP_ALIGN);
912 *
913 * The downside to this alignment of the IP header is that the DMA is now
914 * unaligned. On some architectures the cost of an unaligned DMA is high
915 * and this cost outweighs the gains made by aligning the IP header.
916 *
917 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
918 * to be overridden.
919 */
920#ifndef NET_IP_ALIGN
921#define NET_IP_ALIGN 2
922#endif
923
924extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
925
926static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
927{
928 if (!skb->data_len) {
929 skb->len = len;
930 skb->tail = skb->data + len;
931 } else
932 ___pskb_trim(skb, len, 0);
933}
934
935/**
936 * skb_trim - remove end from a buffer
937 * @skb: buffer to alter
938 * @len: new length
939 *
940 * Cut the length of a buffer down by removing data from the tail. If
941 * the buffer is already under the length specified it is not modified.
942 */
943static inline void skb_trim(struct sk_buff *skb, unsigned int len)
944{
945 if (skb->len > len)
946 __skb_trim(skb, len);
947}
948
949
950static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
951{
952 if (!skb->data_len) {
953 skb->len = len;
954 skb->tail = skb->data+len;
955 return 0;
956 }
957 return ___pskb_trim(skb, len, 1);
958}
959
960static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
961{
962 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
963}
964
965/**
966 * skb_orphan - orphan a buffer
967 * @skb: buffer to orphan
968 *
969 * If a buffer currently has an owner then we call the owner's
970 * destructor function and make the @skb unowned. The buffer continues
971 * to exist but is no longer charged to its former owner.
972 */
973static inline void skb_orphan(struct sk_buff *skb)
974{
975 if (skb->destructor)
976 skb->destructor(skb);
977 skb->destructor = NULL;
978 skb->sk = NULL;
979}
980
981/**
982 * __skb_queue_purge - empty a list
983 * @list: list to empty
984 *
985 * Delete all buffers on an &sk_buff list. Each buffer is removed from
986 * the list and one reference dropped. This function does not take the
987 * list lock and the caller must hold the relevant locks to use it.
988 */
989extern void skb_queue_purge(struct sk_buff_head *list);
990static inline void __skb_queue_purge(struct sk_buff_head *list)
991{
992 struct sk_buff *skb;
993 while ((skb = __skb_dequeue(list)) != NULL)
994 kfree_skb(skb);
995}
996
4dc3b16b 997#ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
1da177e4
LT
998/**
999 * __dev_alloc_skb - allocate an skbuff for sending
1000 * @length: length to allocate
1001 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1002 *
1003 * Allocate a new &sk_buff and assign it a usage count of one. The
1004 * buffer has unspecified headroom built in. Users should allocate
1005 * the headroom they think they need without accounting for the
1006 * built in space. The built in space is used for optimisations.
1007 *
1008 * %NULL is returned in there is no free memory.
1009 */
1da177e4 1010static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
86a76caf 1011 unsigned int __nocast gfp_mask)
1da177e4
LT
1012{
1013 struct sk_buff *skb = alloc_skb(length + 16, gfp_mask);
1014 if (likely(skb))
1015 skb_reserve(skb, 16);
1016 return skb;
1017}
1018#else
1019extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
1020#endif
1021
1022/**
1023 * dev_alloc_skb - allocate an skbuff for sending
1024 * @length: length to allocate
1025 *
1026 * Allocate a new &sk_buff and assign it a usage count of one. The
1027 * buffer has unspecified headroom built in. Users should allocate
1028 * the headroom they think they need without accounting for the
1029 * built in space. The built in space is used for optimisations.
1030 *
1031 * %NULL is returned in there is no free memory. Although this function
1032 * allocates memory it can be called from an interrupt.
1033 */
1034static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1035{
1036 return __dev_alloc_skb(length, GFP_ATOMIC);
1037}
1038
1039/**
1040 * skb_cow - copy header of skb when it is required
1041 * @skb: buffer to cow
1042 * @headroom: needed headroom
1043 *
1044 * If the skb passed lacks sufficient headroom or its data part
1045 * is shared, data is reallocated. If reallocation fails, an error
1046 * is returned and original skb is not changed.
1047 *
1048 * The result is skb with writable area skb->head...skb->tail
1049 * and at least @headroom of space at head.
1050 */
1051static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1052{
1053 int delta = (headroom > 16 ? headroom : 16) - skb_headroom(skb);
1054
1055 if (delta < 0)
1056 delta = 0;
1057
1058 if (delta || skb_cloned(skb))
1059 return pskb_expand_head(skb, (delta + 15) & ~15, 0, GFP_ATOMIC);
1060 return 0;
1061}
1062
1063/**
1064 * skb_padto - pad an skbuff up to a minimal size
1065 * @skb: buffer to pad
1066 * @len: minimal length
1067 *
1068 * Pads up a buffer to ensure the trailing bytes exist and are
1069 * blanked. If the buffer already contains sufficient data it
1070 * is untouched. Returns the buffer, which may be a replacement
1071 * for the original, or NULL for out of memory - in which case
1072 * the original buffer is still freed.
1073 */
1074
1075static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
1076{
1077 unsigned int size = skb->len;
1078 if (likely(size >= len))
1079 return skb;
1080 return skb_pad(skb, len-size);
1081}
1082
1083static inline int skb_add_data(struct sk_buff *skb,
1084 char __user *from, int copy)
1085{
1086 const int off = skb->len;
1087
1088 if (skb->ip_summed == CHECKSUM_NONE) {
1089 int err = 0;
1090 unsigned int csum = csum_and_copy_from_user(from,
1091 skb_put(skb, copy),
1092 copy, 0, &err);
1093 if (!err) {
1094 skb->csum = csum_block_add(skb->csum, csum, off);
1095 return 0;
1096 }
1097 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1098 return 0;
1099
1100 __skb_trim(skb, off);
1101 return -EFAULT;
1102}
1103
1104static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1105 struct page *page, int off)
1106{
1107 if (i) {
1108 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1109
1110 return page == frag->page &&
1111 off == frag->page_offset + frag->size;
1112 }
1113 return 0;
1114}
1115
1116/**
1117 * skb_linearize - convert paged skb to linear one
1118 * @skb: buffer to linarize
1119 * @gfp: allocation mode
1120 *
1121 * If there is no free memory -ENOMEM is returned, otherwise zero
1122 * is returned and the old skb data released.
1123 */
86a76caf
VF
1124extern int __skb_linearize(struct sk_buff *skb, unsigned int __nocast gfp);
1125static inline int skb_linearize(struct sk_buff *skb, unsigned int __nocast gfp)
1da177e4
LT
1126{
1127 return __skb_linearize(skb, gfp);
1128}
1129
1130/**
1131 * skb_postpull_rcsum - update checksum for received skb after pull
1132 * @skb: buffer to update
1133 * @start: start of data before pull
1134 * @len: length of data pulled
1135 *
1136 * After doing a pull on a received packet, you need to call this to
1137 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1138 * so that it can be recomputed from scratch.
1139 */
1140
1141static inline void skb_postpull_rcsum(struct sk_buff *skb,
1142 const void *start, int len)
1143{
1144 if (skb->ip_summed == CHECKSUM_HW)
1145 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1146}
1147
1148/**
1149 * pskb_trim_rcsum - trim received skb and update checksum
1150 * @skb: buffer to trim
1151 * @len: new length
1152 *
1153 * This is exactly the same as pskb_trim except that it ensures the
1154 * checksum of received packets are still valid after the operation.
1155 */
1156
1157static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1158{
1159 if (len >= skb->len)
1160 return 0;
1161 if (skb->ip_summed == CHECKSUM_HW)
1162 skb->ip_summed = CHECKSUM_NONE;
1163 return __pskb_trim(skb, len);
1164}
1165
1166static inline void *kmap_skb_frag(const skb_frag_t *frag)
1167{
1168#ifdef CONFIG_HIGHMEM
1169 BUG_ON(in_irq());
1170
1171 local_bh_disable();
1172#endif
1173 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1174}
1175
1176static inline void kunmap_skb_frag(void *vaddr)
1177{
1178 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1179#ifdef CONFIG_HIGHMEM
1180 local_bh_enable();
1181#endif
1182}
1183
1184#define skb_queue_walk(queue, skb) \
1185 for (skb = (queue)->next; \
1186 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1187 skb = skb->next)
1188
1189
1190extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1191 int noblock, int *err);
1192extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1193 struct poll_table_struct *wait);
1194extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1195 int offset, struct iovec *to,
1196 int size);
1197extern int skb_copy_and_csum_datagram_iovec(const
1198 struct sk_buff *skb,
1199 int hlen,
1200 struct iovec *iov);
1201extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1202extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1203 int len, unsigned int csum);
1204extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1205 void *to, int len);
357b40a1
HX
1206extern int skb_store_bits(const struct sk_buff *skb, int offset,
1207 void *from, int len);
1da177e4
LT
1208extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1209 int offset, u8 *to, int len,
1210 unsigned int csum);
1211extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1212extern void skb_split(struct sk_buff *skb,
1213 struct sk_buff *skb1, const u32 len);
1214
1215static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1216 int len, void *buffer)
1217{
1218 int hlen = skb_headlen(skb);
1219
55820ee2 1220 if (hlen - offset >= len)
1da177e4
LT
1221 return skb->data + offset;
1222
1223 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1224 return NULL;
1225
1226 return buffer;
1227}
1228
1229extern void skb_init(void);
1230extern void skb_add_mtu(int mtu);
1231
1232#ifdef CONFIG_NETFILTER
1233static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1234{
1235 if (nfct && atomic_dec_and_test(&nfct->use))
1236 nfct->destroy(nfct);
1237}
1238static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1239{
1240 if (nfct)
1241 atomic_inc(&nfct->use);
1242}
1243static inline void nf_reset(struct sk_buff *skb)
1244{
1245 nf_conntrack_put(skb->nfct);
1246 skb->nfct = NULL;
1da177e4
LT
1247}
1248
1249#ifdef CONFIG_BRIDGE_NETFILTER
1250static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1251{
1252 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1253 kfree(nf_bridge);
1254}
1255static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1256{
1257 if (nf_bridge)
1258 atomic_inc(&nf_bridge->use);
1259}
1260#endif /* CONFIG_BRIDGE_NETFILTER */
1261#else /* CONFIG_NETFILTER */
1262static inline void nf_reset(struct sk_buff *skb) {}
1263#endif /* CONFIG_NETFILTER */
1264
1265#endif /* __KERNEL__ */
1266#endif /* _LINUX_SKBUFF_H */