blob: 14b9504134954d2b624494d1d8590e66c92081eb [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070030#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#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
Linus Torvalds1da177e2005-04-16 15:20:36 -070087struct 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
Martin Waitz67be2dd2005-05-01 08:59:26 -0700171 * @dst: destination entry
172 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 * @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
Martin Waitz67be2dd2005-05-01 08:59:26 -0700178 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 * @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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 * @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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196 * @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;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 __u32 priority;
Thomas Graf1cbb3382005-07-05 14:13:41 -0700252 __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;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259
260 void (*destructor)(struct sk_buff *skb);
261#ifdef CONFIG_NETFILTER
Thomas Graf1cbb3382005-07-05 14:13:41 -0700262 unsigned long nfmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 __u32 nfcache;
264 __u32 nfctinfo;
265 struct nf_conntrack *nfct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266#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);
303extern struct sk_buff *alloc_skb(unsigned int size, int priority);
304extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
305 unsigned int size, int priority);
306extern void kfree_skbmem(struct sk_buff *skb);
307extern struct sk_buff *skb_clone(struct sk_buff *skb, int priority);
308extern struct sk_buff *skb_copy(const struct sk_buff *skb, int priority);
309extern struct sk_buff *pskb_copy(struct sk_buff *skb, int gfp_mask);
310extern int pskb_expand_head(struct sk_buff *skb,
311 int nhead, int ntail, int gfp_mask);
312extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
313 unsigned int headroom);
314extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
315 int newheadroom, int newtailroom,
316 int priority);
317extern struct sk_buff * skb_pad(struct sk_buff *skb, int pad);
318#define dev_kfree_skb(a) kfree_skb(a)
319extern void skb_over_panic(struct sk_buff *skb, int len,
320 void *here);
321extern void skb_under_panic(struct sk_buff *skb, int len,
322 void *here);
323
Thomas Graf677e90e2005-06-23 20:59:51 -0700324struct skb_seq_state
325{
326 __u32 lower_offset;
327 __u32 upper_offset;
328 __u32 frag_idx;
329 __u32 stepped_offset;
330 struct sk_buff *root_skb;
331 struct sk_buff *cur_skb;
332 __u8 *frag_data;
333};
334
335extern void skb_prepare_seq_read(struct sk_buff *skb,
336 unsigned int from, unsigned int to,
337 struct skb_seq_state *st);
338extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
339 struct skb_seq_state *st);
340extern void skb_abort_seq_read(struct skb_seq_state *st);
341
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700342extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
343 unsigned int to, struct ts_config *config,
344 struct ts_state *state);
345
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346/* Internal */
347#define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
348
349/**
350 * skb_queue_empty - check if a queue is empty
351 * @list: queue head
352 *
353 * Returns true if the queue is empty, false otherwise.
354 */
355static inline int skb_queue_empty(const struct sk_buff_head *list)
356{
357 return list->next == (struct sk_buff *)list;
358}
359
360/**
361 * skb_get - reference buffer
362 * @skb: buffer to reference
363 *
364 * Makes another reference to a socket buffer and returns a pointer
365 * to the buffer.
366 */
367static inline struct sk_buff *skb_get(struct sk_buff *skb)
368{
369 atomic_inc(&skb->users);
370 return skb;
371}
372
373/*
374 * If users == 1, we are the only owner and are can avoid redundant
375 * atomic change.
376 */
377
378/**
379 * kfree_skb - free an sk_buff
380 * @skb: buffer to free
381 *
382 * Drop a reference to the buffer and free it if the usage count has
383 * hit zero.
384 */
385static inline void kfree_skb(struct sk_buff *skb)
386{
387 if (likely(atomic_read(&skb->users) == 1))
388 smp_rmb();
389 else if (likely(!atomic_dec_and_test(&skb->users)))
390 return;
391 __kfree_skb(skb);
392}
393
394/**
395 * skb_cloned - is the buffer a clone
396 * @skb: buffer to check
397 *
398 * Returns true if the buffer was generated with skb_clone() and is
399 * one of multiple shared copies of the buffer. Cloned buffers are
400 * shared data so must not be written to under normal circumstances.
401 */
402static inline int skb_cloned(const struct sk_buff *skb)
403{
404 return skb->cloned &&
405 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
406}
407
408/**
409 * skb_header_cloned - is the header a clone
410 * @skb: buffer to check
411 *
412 * Returns true if modifying the header part of the buffer requires
413 * the data to be copied.
414 */
415static inline int skb_header_cloned(const struct sk_buff *skb)
416{
417 int dataref;
418
419 if (!skb->cloned)
420 return 0;
421
422 dataref = atomic_read(&skb_shinfo(skb)->dataref);
423 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
424 return dataref != 1;
425}
426
427/**
428 * skb_header_release - release reference to header
429 * @skb: buffer to operate on
430 *
431 * Drop a reference to the header part of the buffer. This is done
432 * by acquiring a payload reference. You must not read from the header
433 * part of skb->data after this.
434 */
435static inline void skb_header_release(struct sk_buff *skb)
436{
437 BUG_ON(skb->nohdr);
438 skb->nohdr = 1;
439 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
440}
441
442/**
443 * skb_shared - is the buffer shared
444 * @skb: buffer to check
445 *
446 * Returns true if more than one person has a reference to this
447 * buffer.
448 */
449static inline int skb_shared(const struct sk_buff *skb)
450{
451 return atomic_read(&skb->users) != 1;
452}
453
454/**
455 * skb_share_check - check if buffer is shared and if so clone it
456 * @skb: buffer to check
457 * @pri: priority for memory allocation
458 *
459 * If the buffer is shared the buffer is cloned and the old copy
460 * drops a reference. A new clone with a single reference is returned.
461 * If the buffer is not shared the original buffer is returned. When
462 * being called from interrupt status or with spinlocks held pri must
463 * be GFP_ATOMIC.
464 *
465 * NULL is returned on a memory allocation failure.
466 */
467static inline struct sk_buff *skb_share_check(struct sk_buff *skb, int pri)
468{
469 might_sleep_if(pri & __GFP_WAIT);
470 if (skb_shared(skb)) {
471 struct sk_buff *nskb = skb_clone(skb, pri);
472 kfree_skb(skb);
473 skb = nskb;
474 }
475 return skb;
476}
477
478/*
479 * Copy shared buffers into a new sk_buff. We effectively do COW on
480 * packets to handle cases where we have a local reader and forward
481 * and a couple of other messy ones. The normal one is tcpdumping
482 * a packet thats being forwarded.
483 */
484
485/**
486 * skb_unshare - make a copy of a shared buffer
487 * @skb: buffer to check
488 * @pri: priority for memory allocation
489 *
490 * If the socket buffer is a clone then this function creates a new
491 * copy of the data, drops a reference count on the old copy and returns
492 * the new copy with the reference count at 1. If the buffer is not a clone
493 * the original buffer is returned. When called with a spinlock held or
494 * from interrupt state @pri must be %GFP_ATOMIC
495 *
496 * %NULL is returned on a memory allocation failure.
497 */
498static inline struct sk_buff *skb_unshare(struct sk_buff *skb, int pri)
499{
500 might_sleep_if(pri & __GFP_WAIT);
501 if (skb_cloned(skb)) {
502 struct sk_buff *nskb = skb_copy(skb, pri);
503 kfree_skb(skb); /* Free our shared copy */
504 skb = nskb;
505 }
506 return skb;
507}
508
509/**
510 * skb_peek
511 * @list_: list to peek at
512 *
513 * Peek an &sk_buff. Unlike most other operations you _MUST_
514 * be careful with this one. A peek leaves the buffer on the
515 * list and someone else may run off with it. You must hold
516 * the appropriate locks or have a private queue to do this.
517 *
518 * Returns %NULL for an empty list or a pointer to the head element.
519 * The reference count is not incremented and the reference is therefore
520 * volatile. Use with caution.
521 */
522static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
523{
524 struct sk_buff *list = ((struct sk_buff *)list_)->next;
525 if (list == (struct sk_buff *)list_)
526 list = NULL;
527 return list;
528}
529
530/**
531 * skb_peek_tail
532 * @list_: list to peek at
533 *
534 * Peek an &sk_buff. Unlike most other operations you _MUST_
535 * be careful with this one. A peek leaves the buffer on the
536 * list and someone else may run off with it. You must hold
537 * the appropriate locks or have a private queue to do this.
538 *
539 * Returns %NULL for an empty list or a pointer to the tail element.
540 * The reference count is not incremented and the reference is therefore
541 * volatile. Use with caution.
542 */
543static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
544{
545 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
546 if (list == (struct sk_buff *)list_)
547 list = NULL;
548 return list;
549}
550
551/**
552 * skb_queue_len - get queue length
553 * @list_: list to measure
554 *
555 * Return the length of an &sk_buff queue.
556 */
557static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
558{
559 return list_->qlen;
560}
561
562static inline void skb_queue_head_init(struct sk_buff_head *list)
563{
564 spin_lock_init(&list->lock);
565 list->prev = list->next = (struct sk_buff *)list;
566 list->qlen = 0;
567}
568
569/*
570 * Insert an sk_buff at the start of a list.
571 *
572 * The "__skb_xxxx()" functions are the non-atomic ones that
573 * can only be called with interrupts disabled.
574 */
575
576/**
577 * __skb_queue_head - queue a buffer at the list head
578 * @list: list to use
579 * @newsk: buffer to queue
580 *
581 * Queue a buffer at the start of a list. This function takes no locks
582 * and you must therefore hold required locks before calling it.
583 *
584 * A buffer cannot be placed on two lists at the same time.
585 */
586extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
587static inline void __skb_queue_head(struct sk_buff_head *list,
588 struct sk_buff *newsk)
589{
590 struct sk_buff *prev, *next;
591
592 newsk->list = list;
593 list->qlen++;
594 prev = (struct sk_buff *)list;
595 next = prev->next;
596 newsk->next = next;
597 newsk->prev = prev;
598 next->prev = prev->next = newsk;
599}
600
601/**
602 * __skb_queue_tail - queue a buffer at the list tail
603 * @list: list to use
604 * @newsk: buffer to queue
605 *
606 * Queue a buffer at the end of a list. This function takes no locks
607 * and you must therefore hold required locks before calling it.
608 *
609 * A buffer cannot be placed on two lists at the same time.
610 */
611extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
612static inline void __skb_queue_tail(struct sk_buff_head *list,
613 struct sk_buff *newsk)
614{
615 struct sk_buff *prev, *next;
616
617 newsk->list = list;
618 list->qlen++;
619 next = (struct sk_buff *)list;
620 prev = next->prev;
621 newsk->next = next;
622 newsk->prev = prev;
623 next->prev = prev->next = newsk;
624}
625
626
627/**
628 * __skb_dequeue - remove from the head of the queue
629 * @list: list to dequeue from
630 *
631 * Remove the head of the list. This function does not take any locks
632 * so must be used with appropriate locks held only. The head item is
633 * returned or %NULL if the list is empty.
634 */
635extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
636static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
637{
638 struct sk_buff *next, *prev, *result;
639
640 prev = (struct sk_buff *) list;
641 next = prev->next;
642 result = NULL;
643 if (next != prev) {
644 result = next;
645 next = next->next;
646 list->qlen--;
647 next->prev = prev;
648 prev->next = next;
649 result->next = result->prev = NULL;
650 result->list = NULL;
651 }
652 return result;
653}
654
655
656/*
657 * Insert a packet on a list.
658 */
659extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk);
660static inline void __skb_insert(struct sk_buff *newsk,
661 struct sk_buff *prev, struct sk_buff *next,
662 struct sk_buff_head *list)
663{
664 newsk->next = next;
665 newsk->prev = prev;
666 next->prev = prev->next = newsk;
667 newsk->list = list;
668 list->qlen++;
669}
670
671/*
672 * Place a packet after a given packet in a list.
673 */
674extern void skb_append(struct sk_buff *old, struct sk_buff *newsk);
675static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk)
676{
677 __skb_insert(newsk, old, old->next, old->list);
678}
679
680/*
681 * remove sk_buff from list. _Must_ be called atomically, and with
682 * the list known..
683 */
684extern void skb_unlink(struct sk_buff *skb);
685static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
686{
687 struct sk_buff *next, *prev;
688
689 list->qlen--;
690 next = skb->next;
691 prev = skb->prev;
692 skb->next = skb->prev = NULL;
693 skb->list = NULL;
694 next->prev = prev;
695 prev->next = next;
696}
697
698
699/* XXX: more streamlined implementation */
700
701/**
702 * __skb_dequeue_tail - remove from the tail of the queue
703 * @list: list to dequeue from
704 *
705 * Remove the tail of the list. This function does not take any locks
706 * so must be used with appropriate locks held only. The tail item is
707 * returned or %NULL if the list is empty.
708 */
709extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
710static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
711{
712 struct sk_buff *skb = skb_peek_tail(list);
713 if (skb)
714 __skb_unlink(skb, list);
715 return skb;
716}
717
718
719static inline int skb_is_nonlinear(const struct sk_buff *skb)
720{
721 return skb->data_len;
722}
723
724static inline unsigned int skb_headlen(const struct sk_buff *skb)
725{
726 return skb->len - skb->data_len;
727}
728
729static inline int skb_pagelen(const struct sk_buff *skb)
730{
731 int i, len = 0;
732
733 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
734 len += skb_shinfo(skb)->frags[i].size;
735 return len + skb_headlen(skb);
736}
737
738static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
739 struct page *page, int off, int size)
740{
741 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
742
743 frag->page = page;
744 frag->page_offset = off;
745 frag->size = size;
746 skb_shinfo(skb)->nr_frags = i + 1;
747}
748
749#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
750#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
751#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
752
753/*
754 * Add data to an sk_buff
755 */
756static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
757{
758 unsigned char *tmp = skb->tail;
759 SKB_LINEAR_ASSERT(skb);
760 skb->tail += len;
761 skb->len += len;
762 return tmp;
763}
764
765/**
766 * skb_put - add data to a buffer
767 * @skb: buffer to use
768 * @len: amount of data to add
769 *
770 * This function extends the used data area of the buffer. If this would
771 * exceed the total buffer size the kernel will panic. A pointer to the
772 * first byte of the extra data is returned.
773 */
774static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
775{
776 unsigned char *tmp = skb->tail;
777 SKB_LINEAR_ASSERT(skb);
778 skb->tail += len;
779 skb->len += len;
780 if (unlikely(skb->tail>skb->end))
781 skb_over_panic(skb, len, current_text_addr());
782 return tmp;
783}
784
785static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
786{
787 skb->data -= len;
788 skb->len += len;
789 return skb->data;
790}
791
792/**
793 * skb_push - add data to the start of a buffer
794 * @skb: buffer to use
795 * @len: amount of data to add
796 *
797 * This function extends the used data area of the buffer at the buffer
798 * start. If this would exceed the total buffer headroom the kernel will
799 * panic. A pointer to the first byte of the extra data is returned.
800 */
801static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
802{
803 skb->data -= len;
804 skb->len += len;
805 if (unlikely(skb->data<skb->head))
806 skb_under_panic(skb, len, current_text_addr());
807 return skb->data;
808}
809
810static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
811{
812 skb->len -= len;
813 BUG_ON(skb->len < skb->data_len);
814 return skb->data += len;
815}
816
817/**
818 * skb_pull - remove data from the start of a buffer
819 * @skb: buffer to use
820 * @len: amount of data to remove
821 *
822 * This function removes data from the start of a buffer, returning
823 * the memory to the headroom. A pointer to the next data in the buffer
824 * is returned. Once the data has been pulled future pushes will overwrite
825 * the old data.
826 */
827static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
828{
829 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
830}
831
832extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
833
834static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
835{
836 if (len > skb_headlen(skb) &&
837 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
838 return NULL;
839 skb->len -= len;
840 return skb->data += len;
841}
842
843static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
844{
845 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
846}
847
848static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
849{
850 if (likely(len <= skb_headlen(skb)))
851 return 1;
852 if (unlikely(len > skb->len))
853 return 0;
854 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
855}
856
857/**
858 * skb_headroom - bytes at buffer head
859 * @skb: buffer to check
860 *
861 * Return the number of bytes of free space at the head of an &sk_buff.
862 */
863static inline int skb_headroom(const struct sk_buff *skb)
864{
865 return skb->data - skb->head;
866}
867
868/**
869 * skb_tailroom - bytes at buffer end
870 * @skb: buffer to check
871 *
872 * Return the number of bytes of free space at the tail of an sk_buff
873 */
874static inline int skb_tailroom(const struct sk_buff *skb)
875{
876 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
877}
878
879/**
880 * skb_reserve - adjust headroom
881 * @skb: buffer to alter
882 * @len: bytes to move
883 *
884 * Increase the headroom of an empty &sk_buff by reducing the tail
885 * room. This is only allowed for an empty buffer.
886 */
887static inline void skb_reserve(struct sk_buff *skb, unsigned int len)
888{
889 skb->data += len;
890 skb->tail += len;
891}
892
893/*
894 * CPUs often take a performance hit when accessing unaligned memory
895 * locations. The actual performance hit varies, it can be small if the
896 * hardware handles it or large if we have to take an exception and fix it
897 * in software.
898 *
899 * Since an ethernet header is 14 bytes network drivers often end up with
900 * the IP header at an unaligned offset. The IP header can be aligned by
901 * shifting the start of the packet by 2 bytes. Drivers should do this
902 * with:
903 *
904 * skb_reserve(NET_IP_ALIGN);
905 *
906 * The downside to this alignment of the IP header is that the DMA is now
907 * unaligned. On some architectures the cost of an unaligned DMA is high
908 * and this cost outweighs the gains made by aligning the IP header.
909 *
910 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
911 * to be overridden.
912 */
913#ifndef NET_IP_ALIGN
914#define NET_IP_ALIGN 2
915#endif
916
917extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
918
919static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
920{
921 if (!skb->data_len) {
922 skb->len = len;
923 skb->tail = skb->data + len;
924 } else
925 ___pskb_trim(skb, len, 0);
926}
927
928/**
929 * skb_trim - remove end from a buffer
930 * @skb: buffer to alter
931 * @len: new length
932 *
933 * Cut the length of a buffer down by removing data from the tail. If
934 * the buffer is already under the length specified it is not modified.
935 */
936static inline void skb_trim(struct sk_buff *skb, unsigned int len)
937{
938 if (skb->len > len)
939 __skb_trim(skb, len);
940}
941
942
943static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
944{
945 if (!skb->data_len) {
946 skb->len = len;
947 skb->tail = skb->data+len;
948 return 0;
949 }
950 return ___pskb_trim(skb, len, 1);
951}
952
953static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
954{
955 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
956}
957
958/**
959 * skb_orphan - orphan a buffer
960 * @skb: buffer to orphan
961 *
962 * If a buffer currently has an owner then we call the owner's
963 * destructor function and make the @skb unowned. The buffer continues
964 * to exist but is no longer charged to its former owner.
965 */
966static inline void skb_orphan(struct sk_buff *skb)
967{
968 if (skb->destructor)
969 skb->destructor(skb);
970 skb->destructor = NULL;
971 skb->sk = NULL;
972}
973
974/**
975 * __skb_queue_purge - empty a list
976 * @list: list to empty
977 *
978 * Delete all buffers on an &sk_buff list. Each buffer is removed from
979 * the list and one reference dropped. This function does not take the
980 * list lock and the caller must hold the relevant locks to use it.
981 */
982extern void skb_queue_purge(struct sk_buff_head *list);
983static inline void __skb_queue_purge(struct sk_buff_head *list)
984{
985 struct sk_buff *skb;
986 while ((skb = __skb_dequeue(list)) != NULL)
987 kfree_skb(skb);
988}
989
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700990#ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991/**
992 * __dev_alloc_skb - allocate an skbuff for sending
993 * @length: length to allocate
994 * @gfp_mask: get_free_pages mask, passed to alloc_skb
995 *
996 * Allocate a new &sk_buff and assign it a usage count of one. The
997 * buffer has unspecified headroom built in. Users should allocate
998 * the headroom they think they need without accounting for the
999 * built in space. The built in space is used for optimisations.
1000 *
1001 * %NULL is returned in there is no free memory.
1002 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1004 int gfp_mask)
1005{
1006 struct sk_buff *skb = alloc_skb(length + 16, gfp_mask);
1007 if (likely(skb))
1008 skb_reserve(skb, 16);
1009 return skb;
1010}
1011#else
1012extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
1013#endif
1014
1015/**
1016 * dev_alloc_skb - allocate an skbuff for sending
1017 * @length: length to allocate
1018 *
1019 * Allocate a new &sk_buff and assign it a usage count of one. The
1020 * buffer has unspecified headroom built in. Users should allocate
1021 * the headroom they think they need without accounting for the
1022 * built in space. The built in space is used for optimisations.
1023 *
1024 * %NULL is returned in there is no free memory. Although this function
1025 * allocates memory it can be called from an interrupt.
1026 */
1027static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1028{
1029 return __dev_alloc_skb(length, GFP_ATOMIC);
1030}
1031
1032/**
1033 * skb_cow - copy header of skb when it is required
1034 * @skb: buffer to cow
1035 * @headroom: needed headroom
1036 *
1037 * If the skb passed lacks sufficient headroom or its data part
1038 * is shared, data is reallocated. If reallocation fails, an error
1039 * is returned and original skb is not changed.
1040 *
1041 * The result is skb with writable area skb->head...skb->tail
1042 * and at least @headroom of space at head.
1043 */
1044static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1045{
1046 int delta = (headroom > 16 ? headroom : 16) - skb_headroom(skb);
1047
1048 if (delta < 0)
1049 delta = 0;
1050
1051 if (delta || skb_cloned(skb))
1052 return pskb_expand_head(skb, (delta + 15) & ~15, 0, GFP_ATOMIC);
1053 return 0;
1054}
1055
1056/**
1057 * skb_padto - pad an skbuff up to a minimal size
1058 * @skb: buffer to pad
1059 * @len: minimal length
1060 *
1061 * Pads up a buffer to ensure the trailing bytes exist and are
1062 * blanked. If the buffer already contains sufficient data it
1063 * is untouched. Returns the buffer, which may be a replacement
1064 * for the original, or NULL for out of memory - in which case
1065 * the original buffer is still freed.
1066 */
1067
1068static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
1069{
1070 unsigned int size = skb->len;
1071 if (likely(size >= len))
1072 return skb;
1073 return skb_pad(skb, len-size);
1074}
1075
1076static inline int skb_add_data(struct sk_buff *skb,
1077 char __user *from, int copy)
1078{
1079 const int off = skb->len;
1080
1081 if (skb->ip_summed == CHECKSUM_NONE) {
1082 int err = 0;
1083 unsigned int csum = csum_and_copy_from_user(from,
1084 skb_put(skb, copy),
1085 copy, 0, &err);
1086 if (!err) {
1087 skb->csum = csum_block_add(skb->csum, csum, off);
1088 return 0;
1089 }
1090 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1091 return 0;
1092
1093 __skb_trim(skb, off);
1094 return -EFAULT;
1095}
1096
1097static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1098 struct page *page, int off)
1099{
1100 if (i) {
1101 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1102
1103 return page == frag->page &&
1104 off == frag->page_offset + frag->size;
1105 }
1106 return 0;
1107}
1108
1109/**
1110 * skb_linearize - convert paged skb to linear one
1111 * @skb: buffer to linarize
1112 * @gfp: allocation mode
1113 *
1114 * If there is no free memory -ENOMEM is returned, otherwise zero
1115 * is returned and the old skb data released.
1116 */
1117extern int __skb_linearize(struct sk_buff *skb, int gfp);
1118static inline int skb_linearize(struct sk_buff *skb, int gfp)
1119{
1120 return __skb_linearize(skb, gfp);
1121}
1122
1123/**
1124 * skb_postpull_rcsum - update checksum for received skb after pull
1125 * @skb: buffer to update
1126 * @start: start of data before pull
1127 * @len: length of data pulled
1128 *
1129 * After doing a pull on a received packet, you need to call this to
1130 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1131 * so that it can be recomputed from scratch.
1132 */
1133
1134static inline void skb_postpull_rcsum(struct sk_buff *skb,
1135 const void *start, int len)
1136{
1137 if (skb->ip_summed == CHECKSUM_HW)
1138 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1139}
1140
1141/**
1142 * pskb_trim_rcsum - trim received skb and update checksum
1143 * @skb: buffer to trim
1144 * @len: new length
1145 *
1146 * This is exactly the same as pskb_trim except that it ensures the
1147 * checksum of received packets are still valid after the operation.
1148 */
1149
1150static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1151{
1152 if (len >= skb->len)
1153 return 0;
1154 if (skb->ip_summed == CHECKSUM_HW)
1155 skb->ip_summed = CHECKSUM_NONE;
1156 return __pskb_trim(skb, len);
1157}
1158
1159static inline void *kmap_skb_frag(const skb_frag_t *frag)
1160{
1161#ifdef CONFIG_HIGHMEM
1162 BUG_ON(in_irq());
1163
1164 local_bh_disable();
1165#endif
1166 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1167}
1168
1169static inline void kunmap_skb_frag(void *vaddr)
1170{
1171 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1172#ifdef CONFIG_HIGHMEM
1173 local_bh_enable();
1174#endif
1175}
1176
1177#define skb_queue_walk(queue, skb) \
1178 for (skb = (queue)->next; \
1179 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1180 skb = skb->next)
1181
1182
1183extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1184 int noblock, int *err);
1185extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1186 struct poll_table_struct *wait);
1187extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1188 int offset, struct iovec *to,
1189 int size);
1190extern int skb_copy_and_csum_datagram_iovec(const
1191 struct sk_buff *skb,
1192 int hlen,
1193 struct iovec *iov);
1194extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1195extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1196 int len, unsigned int csum);
1197extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1198 void *to, int len);
Herbert Xu357b40a2005-04-19 22:30:14 -07001199extern int skb_store_bits(const struct sk_buff *skb, int offset,
1200 void *from, int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1202 int offset, u8 *to, int len,
1203 unsigned int csum);
1204extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1205extern void skb_split(struct sk_buff *skb,
1206 struct sk_buff *skb1, const u32 len);
1207
1208static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1209 int len, void *buffer)
1210{
1211 int hlen = skb_headlen(skb);
1212
Patrick McHardy55820ee2005-07-05 14:08:10 -07001213 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214 return skb->data + offset;
1215
1216 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1217 return NULL;
1218
1219 return buffer;
1220}
1221
1222extern void skb_init(void);
1223extern void skb_add_mtu(int mtu);
1224
1225#ifdef CONFIG_NETFILTER
1226static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1227{
1228 if (nfct && atomic_dec_and_test(&nfct->use))
1229 nfct->destroy(nfct);
1230}
1231static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1232{
1233 if (nfct)
1234 atomic_inc(&nfct->use);
1235}
1236static inline void nf_reset(struct sk_buff *skb)
1237{
1238 nf_conntrack_put(skb->nfct);
1239 skb->nfct = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240}
1241
1242#ifdef CONFIG_BRIDGE_NETFILTER
1243static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1244{
1245 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1246 kfree(nf_bridge);
1247}
1248static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1249{
1250 if (nf_bridge)
1251 atomic_inc(&nf_bridge->use);
1252}
1253#endif /* CONFIG_BRIDGE_NETFILTER */
1254#else /* CONFIG_NETFILTER */
1255static inline void nf_reset(struct sk_buff *skb) {}
1256#endif /* CONFIG_NETFILTER */
1257
1258#endif /* __KERNEL__ */
1259#endif /* _LINUX_SKBUFF_H */