blob: cc04f5cd2286953a16a66dbfd89fe96211b0026c [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>
30#include <net/checksum.h>
31
32#define HAVE_ALLOC_SKB /* For the drivers to know */
33#define HAVE_ALIGNABLE_SKB /* Ditto 8) */
34#define SLAB_SKB /* Slabified skbuffs */
35
36#define CHECKSUM_NONE 0
37#define CHECKSUM_HW 1
38#define CHECKSUM_UNNECESSARY 2
39
40#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
41 ~(SMP_CACHE_BYTES - 1))
42#define SKB_MAX_ORDER(X, ORDER) (((PAGE_SIZE << (ORDER)) - (X) - \
43 sizeof(struct skb_shared_info)) & \
44 ~(SMP_CACHE_BYTES - 1))
45#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
46#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
47
48/* A. Checksumming of received packets by device.
49 *
50 * NONE: device failed to checksum this packet.
51 * skb->csum is undefined.
52 *
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)
58 *
59 * HW: 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 HW,
63 * not UNNECESSARY.
64 *
65 * B. Checksumming on output.
66 *
67 * NONE: skb is checksummed by protocol or csum is not required.
68 *
69 * HW: device is required to csum packet as seen by hard_start_xmit
70 * from skb->h.raw to the end and to record the checksum
71 * at skb->h.raw+skb->csum.
72 *
73 * Device must show its capabilities in dev->features, set
74 * at device setup time.
75 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
76 * everything.
77 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
78 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
79 * TCP/UDP over IPv4. Sigh. Vendors like this
80 * way by an unknown reason. Though, see comment above
81 * about CHECKSUM_UNNECESSARY. 8)
82 *
83 * Any questions? No questions, good. --ANK
84 */
85
Linus Torvalds1da177e2005-04-16 15:20:36 -070086struct net_device;
87
88#ifdef CONFIG_NETFILTER
89struct nf_conntrack {
90 atomic_t use;
91 void (*destroy)(struct nf_conntrack *);
92};
93
94#ifdef CONFIG_BRIDGE_NETFILTER
95struct nf_bridge_info {
96 atomic_t use;
97 struct net_device *physindev;
98 struct net_device *physoutdev;
99#if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
100 struct net_device *netoutdev;
101#endif
102 unsigned int mask;
103 unsigned long data[32 / sizeof(unsigned long)];
104};
105#endif
106
107#endif
108
109struct sk_buff_head {
110 /* These two members must be first. */
111 struct sk_buff *next;
112 struct sk_buff *prev;
113
114 __u32 qlen;
115 spinlock_t lock;
116};
117
118struct sk_buff;
119
120/* To allow 64K frame to be packed as single skb without frag_list */
121#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
122
123typedef struct skb_frag_struct skb_frag_t;
124
125struct skb_frag_struct {
126 struct page *page;
127 __u16 page_offset;
128 __u16 size;
129};
130
131/* This data is invariant across clones and lives at
132 * the end of the header data, ie. at skb->end.
133 */
134struct skb_shared_info {
135 atomic_t dataref;
136 unsigned int nr_frags;
137 unsigned short tso_size;
138 unsigned short tso_segs;
139 struct sk_buff *frag_list;
140 skb_frag_t frags[MAX_SKB_FRAGS];
141};
142
143/* We divide dataref into two halves. The higher 16 bits hold references
144 * to the payload part of skb->data. The lower 16 bits hold references to
145 * the entire skb->data. It is up to the users of the skb to agree on
146 * where the payload starts.
147 *
148 * All users must obey the rule that the skb->data reference count must be
149 * greater than or equal to the payload reference count.
150 *
151 * Holding a reference to the payload part means that the user does not
152 * care about modifications to the header part of skb->data.
153 */
154#define SKB_DATAREF_SHIFT 16
155#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
156
157/**
158 * struct sk_buff - socket buffer
159 * @next: Next buffer in list
160 * @prev: Previous buffer in list
161 * @list: List we are on
162 * @sk: Socket we are owned by
163 * @stamp: Time we arrived
164 * @dev: Device we arrived on/are leaving by
165 * @input_dev: Device we arrived on
166 * @real_dev: The real device we are using
167 * @h: Transport layer header
168 * @nh: Network layer header
169 * @mac: Link layer header
Martin Waitz67be2dd2005-05-01 08:59:26 -0700170 * @dst: destination entry
171 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 * @cb: Control buffer. Free for use by every layer. Put private vars here
173 * @len: Length of actual data
174 * @data_len: Data length
175 * @mac_len: Length of link layer header
176 * @csum: Checksum
Martin Waitz67be2dd2005-05-01 08:59:26 -0700177 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 * @cloned: Head may be cloned (check refcnt to be sure)
179 * @nohdr: Payload reference only, must not modify header
180 * @pkt_type: Packet class
181 * @ip_summed: Driver fed us an IP checksum
182 * @priority: Packet queueing priority
183 * @users: User count - see {datagram,tcp}.c
184 * @protocol: Packet protocol from driver
185 * @security: Security level of packet
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
196 * @nf_debug: Netfilter debugging
197 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
198 * @private: Data which is private to the HIPPI implementation
199 * @tc_index: Traffic control index
200 * @tc_verd: traffic control verdict
201 * @tc_classid: traffic control classid
202 */
203
204struct sk_buff {
205 /* These two members must be first. */
206 struct sk_buff *next;
207 struct sk_buff *prev;
208
209 struct sk_buff_head *list;
210 struct sock *sk;
211 struct timeval stamp;
212 struct net_device *dev;
213 struct net_device *input_dev;
214 struct net_device *real_dev;
215
216 union {
217 struct tcphdr *th;
218 struct udphdr *uh;
219 struct icmphdr *icmph;
220 struct igmphdr *igmph;
221 struct iphdr *ipiph;
222 struct ipv6hdr *ipv6h;
223 unsigned char *raw;
224 } h;
225
226 union {
227 struct iphdr *iph;
228 struct ipv6hdr *ipv6h;
229 struct arphdr *arph;
230 unsigned char *raw;
231 } nh;
232
233 union {
234 unsigned char *raw;
235 } mac;
236
237 struct dst_entry *dst;
238 struct sec_path *sp;
239
240 /*
241 * This is the control buffer. It is free to use for every
242 * layer. Please put your private variables there. If you
243 * want to keep them across layers you have to do a skb_clone()
244 * first. This is owned by whoever has the skb queued ATM.
245 */
246 char cb[40];
247
248 unsigned int len,
249 data_len,
250 mac_len,
251 csum;
252 unsigned char local_df,
253 cloned:1,
254 nohdr:1,
255 pkt_type,
256 ip_summed;
257 __u32 priority;
258 unsigned short protocol,
259 security;
260
261 void (*destructor)(struct sk_buff *skb);
262#ifdef CONFIG_NETFILTER
263 unsigned long nfmark;
264 __u32 nfcache;
265 __u32 nfctinfo;
266 struct nf_conntrack *nfct;
267#ifdef CONFIG_NETFILTER_DEBUG
268 unsigned int nf_debug;
269#endif
270#ifdef CONFIG_BRIDGE_NETFILTER
271 struct nf_bridge_info *nf_bridge;
272#endif
273#endif /* CONFIG_NETFILTER */
274#if defined(CONFIG_HIPPI)
275 union {
276 __u32 ifield;
277 } private;
278#endif
279#ifdef CONFIG_NET_SCHED
280 __u32 tc_index; /* traffic control index */
281#ifdef CONFIG_NET_CLS_ACT
282 __u32 tc_verd; /* traffic control verdict */
283 __u32 tc_classid; /* traffic control classid */
284#endif
285
286#endif
287
288
289 /* These elements must be at the end, see alloc_skb() for details. */
290 unsigned int truesize;
291 atomic_t users;
292 unsigned char *head,
293 *data,
294 *tail,
295 *end;
296};
297
298#ifdef __KERNEL__
299/*
300 * Handling routines are only of interest to the kernel
301 */
302#include <linux/slab.h>
303
304#include <asm/system.h>
305
306extern void __kfree_skb(struct sk_buff *skb);
307extern struct sk_buff *alloc_skb(unsigned int size, int priority);
308extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
309 unsigned int size, int priority);
310extern void kfree_skbmem(struct sk_buff *skb);
311extern struct sk_buff *skb_clone(struct sk_buff *skb, int priority);
312extern struct sk_buff *skb_copy(const struct sk_buff *skb, int priority);
313extern struct sk_buff *pskb_copy(struct sk_buff *skb, int gfp_mask);
314extern int pskb_expand_head(struct sk_buff *skb,
315 int nhead, int ntail, int gfp_mask);
316extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
317 unsigned int headroom);
318extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
319 int newheadroom, int newtailroom,
320 int priority);
321extern struct sk_buff * skb_pad(struct sk_buff *skb, int pad);
322#define dev_kfree_skb(a) kfree_skb(a)
323extern void skb_over_panic(struct sk_buff *skb, int len,
324 void *here);
325extern void skb_under_panic(struct sk_buff *skb, int len,
326 void *here);
327
328/* Internal */
329#define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
330
331/**
332 * skb_queue_empty - check if a queue is empty
333 * @list: queue head
334 *
335 * Returns true if the queue is empty, false otherwise.
336 */
337static inline int skb_queue_empty(const struct sk_buff_head *list)
338{
339 return list->next == (struct sk_buff *)list;
340}
341
342/**
343 * skb_get - reference buffer
344 * @skb: buffer to reference
345 *
346 * Makes another reference to a socket buffer and returns a pointer
347 * to the buffer.
348 */
349static inline struct sk_buff *skb_get(struct sk_buff *skb)
350{
351 atomic_inc(&skb->users);
352 return skb;
353}
354
355/*
356 * If users == 1, we are the only owner and are can avoid redundant
357 * atomic change.
358 */
359
360/**
361 * kfree_skb - free an sk_buff
362 * @skb: buffer to free
363 *
364 * Drop a reference to the buffer and free it if the usage count has
365 * hit zero.
366 */
367static inline void kfree_skb(struct sk_buff *skb)
368{
369 if (likely(atomic_read(&skb->users) == 1))
370 smp_rmb();
371 else if (likely(!atomic_dec_and_test(&skb->users)))
372 return;
373 __kfree_skb(skb);
374}
375
376/**
377 * skb_cloned - is the buffer a clone
378 * @skb: buffer to check
379 *
380 * Returns true if the buffer was generated with skb_clone() and is
381 * one of multiple shared copies of the buffer. Cloned buffers are
382 * shared data so must not be written to under normal circumstances.
383 */
384static inline int skb_cloned(const struct sk_buff *skb)
385{
386 return skb->cloned &&
387 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
388}
389
390/**
391 * skb_header_cloned - is the header a clone
392 * @skb: buffer to check
393 *
394 * Returns true if modifying the header part of the buffer requires
395 * the data to be copied.
396 */
397static inline int skb_header_cloned(const struct sk_buff *skb)
398{
399 int dataref;
400
401 if (!skb->cloned)
402 return 0;
403
404 dataref = atomic_read(&skb_shinfo(skb)->dataref);
405 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
406 return dataref != 1;
407}
408
409/**
410 * skb_header_release - release reference to header
411 * @skb: buffer to operate on
412 *
413 * Drop a reference to the header part of the buffer. This is done
414 * by acquiring a payload reference. You must not read from the header
415 * part of skb->data after this.
416 */
417static inline void skb_header_release(struct sk_buff *skb)
418{
419 BUG_ON(skb->nohdr);
420 skb->nohdr = 1;
421 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
422}
423
424/**
425 * skb_shared - is the buffer shared
426 * @skb: buffer to check
427 *
428 * Returns true if more than one person has a reference to this
429 * buffer.
430 */
431static inline int skb_shared(const struct sk_buff *skb)
432{
433 return atomic_read(&skb->users) != 1;
434}
435
436/**
437 * skb_share_check - check if buffer is shared and if so clone it
438 * @skb: buffer to check
439 * @pri: priority for memory allocation
440 *
441 * If the buffer is shared the buffer is cloned and the old copy
442 * drops a reference. A new clone with a single reference is returned.
443 * If the buffer is not shared the original buffer is returned. When
444 * being called from interrupt status or with spinlocks held pri must
445 * be GFP_ATOMIC.
446 *
447 * NULL is returned on a memory allocation failure.
448 */
449static inline struct sk_buff *skb_share_check(struct sk_buff *skb, int pri)
450{
451 might_sleep_if(pri & __GFP_WAIT);
452 if (skb_shared(skb)) {
453 struct sk_buff *nskb = skb_clone(skb, pri);
454 kfree_skb(skb);
455 skb = nskb;
456 }
457 return skb;
458}
459
460/*
461 * Copy shared buffers into a new sk_buff. We effectively do COW on
462 * packets to handle cases where we have a local reader and forward
463 * and a couple of other messy ones. The normal one is tcpdumping
464 * a packet thats being forwarded.
465 */
466
467/**
468 * skb_unshare - make a copy of a shared buffer
469 * @skb: buffer to check
470 * @pri: priority for memory allocation
471 *
472 * If the socket buffer is a clone then this function creates a new
473 * copy of the data, drops a reference count on the old copy and returns
474 * the new copy with the reference count at 1. If the buffer is not a clone
475 * the original buffer is returned. When called with a spinlock held or
476 * from interrupt state @pri must be %GFP_ATOMIC
477 *
478 * %NULL is returned on a memory allocation failure.
479 */
480static inline struct sk_buff *skb_unshare(struct sk_buff *skb, int pri)
481{
482 might_sleep_if(pri & __GFP_WAIT);
483 if (skb_cloned(skb)) {
484 struct sk_buff *nskb = skb_copy(skb, pri);
485 kfree_skb(skb); /* Free our shared copy */
486 skb = nskb;
487 }
488 return skb;
489}
490
491/**
492 * skb_peek
493 * @list_: list to peek at
494 *
495 * Peek an &sk_buff. Unlike most other operations you _MUST_
496 * be careful with this one. A peek leaves the buffer on the
497 * list and someone else may run off with it. You must hold
498 * the appropriate locks or have a private queue to do this.
499 *
500 * Returns %NULL for an empty list or a pointer to the head element.
501 * The reference count is not incremented and the reference is therefore
502 * volatile. Use with caution.
503 */
504static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
505{
506 struct sk_buff *list = ((struct sk_buff *)list_)->next;
507 if (list == (struct sk_buff *)list_)
508 list = NULL;
509 return list;
510}
511
512/**
513 * skb_peek_tail
514 * @list_: list to peek at
515 *
516 * Peek an &sk_buff. Unlike most other operations you _MUST_
517 * be careful with this one. A peek leaves the buffer on the
518 * list and someone else may run off with it. You must hold
519 * the appropriate locks or have a private queue to do this.
520 *
521 * Returns %NULL for an empty list or a pointer to the tail element.
522 * The reference count is not incremented and the reference is therefore
523 * volatile. Use with caution.
524 */
525static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
526{
527 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
528 if (list == (struct sk_buff *)list_)
529 list = NULL;
530 return list;
531}
532
533/**
534 * skb_queue_len - get queue length
535 * @list_: list to measure
536 *
537 * Return the length of an &sk_buff queue.
538 */
539static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
540{
541 return list_->qlen;
542}
543
544static inline void skb_queue_head_init(struct sk_buff_head *list)
545{
546 spin_lock_init(&list->lock);
547 list->prev = list->next = (struct sk_buff *)list;
548 list->qlen = 0;
549}
550
551/*
552 * Insert an sk_buff at the start of a list.
553 *
554 * The "__skb_xxxx()" functions are the non-atomic ones that
555 * can only be called with interrupts disabled.
556 */
557
558/**
559 * __skb_queue_head - queue a buffer at the list head
560 * @list: list to use
561 * @newsk: buffer to queue
562 *
563 * Queue a buffer at the start of a list. This function takes no locks
564 * and you must therefore hold required locks before calling it.
565 *
566 * A buffer cannot be placed on two lists at the same time.
567 */
568extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
569static inline void __skb_queue_head(struct sk_buff_head *list,
570 struct sk_buff *newsk)
571{
572 struct sk_buff *prev, *next;
573
574 newsk->list = list;
575 list->qlen++;
576 prev = (struct sk_buff *)list;
577 next = prev->next;
578 newsk->next = next;
579 newsk->prev = prev;
580 next->prev = prev->next = newsk;
581}
582
583/**
584 * __skb_queue_tail - queue a buffer at the list tail
585 * @list: list to use
586 * @newsk: buffer to queue
587 *
588 * Queue a buffer at the end 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_tail(struct sk_buff_head *list, struct sk_buff *newsk);
594static inline void __skb_queue_tail(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 next = (struct sk_buff *)list;
602 prev = next->prev;
603 newsk->next = next;
604 newsk->prev = prev;
605 next->prev = prev->next = newsk;
606}
607
608
609/**
610 * __skb_dequeue - remove from the head of the queue
611 * @list: list to dequeue from
612 *
613 * Remove the head of the list. This function does not take any locks
614 * so must be used with appropriate locks held only. The head item is
615 * returned or %NULL if the list is empty.
616 */
617extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
618static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
619{
620 struct sk_buff *next, *prev, *result;
621
622 prev = (struct sk_buff *) list;
623 next = prev->next;
624 result = NULL;
625 if (next != prev) {
626 result = next;
627 next = next->next;
628 list->qlen--;
629 next->prev = prev;
630 prev->next = next;
631 result->next = result->prev = NULL;
632 result->list = NULL;
633 }
634 return result;
635}
636
637
638/*
639 * Insert a packet on a list.
640 */
641extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk);
642static inline void __skb_insert(struct sk_buff *newsk,
643 struct sk_buff *prev, struct sk_buff *next,
644 struct sk_buff_head *list)
645{
646 newsk->next = next;
647 newsk->prev = prev;
648 next->prev = prev->next = newsk;
649 newsk->list = list;
650 list->qlen++;
651}
652
653/*
654 * Place a packet after a given packet in a list.
655 */
656extern void skb_append(struct sk_buff *old, struct sk_buff *newsk);
657static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk)
658{
659 __skb_insert(newsk, old, old->next, old->list);
660}
661
662/*
663 * remove sk_buff from list. _Must_ be called atomically, and with
664 * the list known..
665 */
666extern void skb_unlink(struct sk_buff *skb);
667static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
668{
669 struct sk_buff *next, *prev;
670
671 list->qlen--;
672 next = skb->next;
673 prev = skb->prev;
674 skb->next = skb->prev = NULL;
675 skb->list = NULL;
676 next->prev = prev;
677 prev->next = next;
678}
679
680
681/* XXX: more streamlined implementation */
682
683/**
684 * __skb_dequeue_tail - remove from the tail of the queue
685 * @list: list to dequeue from
686 *
687 * Remove the tail of the list. This function does not take any locks
688 * so must be used with appropriate locks held only. The tail item is
689 * returned or %NULL if the list is empty.
690 */
691extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
692static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
693{
694 struct sk_buff *skb = skb_peek_tail(list);
695 if (skb)
696 __skb_unlink(skb, list);
697 return skb;
698}
699
700
701static inline int skb_is_nonlinear(const struct sk_buff *skb)
702{
703 return skb->data_len;
704}
705
706static inline unsigned int skb_headlen(const struct sk_buff *skb)
707{
708 return skb->len - skb->data_len;
709}
710
711static inline int skb_pagelen(const struct sk_buff *skb)
712{
713 int i, len = 0;
714
715 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
716 len += skb_shinfo(skb)->frags[i].size;
717 return len + skb_headlen(skb);
718}
719
720static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
721 struct page *page, int off, int size)
722{
723 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
724
725 frag->page = page;
726 frag->page_offset = off;
727 frag->size = size;
728 skb_shinfo(skb)->nr_frags = i + 1;
729}
730
731#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
732#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
733#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
734
735/*
736 * Add data to an sk_buff
737 */
738static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
739{
740 unsigned char *tmp = skb->tail;
741 SKB_LINEAR_ASSERT(skb);
742 skb->tail += len;
743 skb->len += len;
744 return tmp;
745}
746
747/**
748 * skb_put - add data to a buffer
749 * @skb: buffer to use
750 * @len: amount of data to add
751 *
752 * This function extends the used data area of the buffer. If this would
753 * exceed the total buffer size the kernel will panic. A pointer to the
754 * first byte of the extra data is returned.
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 if (unlikely(skb->tail>skb->end))
763 skb_over_panic(skb, len, current_text_addr());
764 return tmp;
765}
766
767static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
768{
769 skb->data -= len;
770 skb->len += len;
771 return skb->data;
772}
773
774/**
775 * skb_push - add data to the start of a buffer
776 * @skb: buffer to use
777 * @len: amount of data to add
778 *
779 * This function extends the used data area of the buffer at the buffer
780 * start. If this would exceed the total buffer headroom the kernel will
781 * panic. A pointer to the first byte of the extra data is returned.
782 */
783static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
784{
785 skb->data -= len;
786 skb->len += len;
787 if (unlikely(skb->data<skb->head))
788 skb_under_panic(skb, len, current_text_addr());
789 return skb->data;
790}
791
792static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
793{
794 skb->len -= len;
795 BUG_ON(skb->len < skb->data_len);
796 return skb->data += len;
797}
798
799/**
800 * skb_pull - remove data from the start of a buffer
801 * @skb: buffer to use
802 * @len: amount of data to remove
803 *
804 * This function removes data from the start of a buffer, returning
805 * the memory to the headroom. A pointer to the next data in the buffer
806 * is returned. Once the data has been pulled future pushes will overwrite
807 * the old data.
808 */
809static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
810{
811 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
812}
813
814extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
815
816static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
817{
818 if (len > skb_headlen(skb) &&
819 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
820 return NULL;
821 skb->len -= len;
822 return skb->data += len;
823}
824
825static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
826{
827 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
828}
829
830static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
831{
832 if (likely(len <= skb_headlen(skb)))
833 return 1;
834 if (unlikely(len > skb->len))
835 return 0;
836 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
837}
838
839/**
840 * skb_headroom - bytes at buffer head
841 * @skb: buffer to check
842 *
843 * Return the number of bytes of free space at the head of an &sk_buff.
844 */
845static inline int skb_headroom(const struct sk_buff *skb)
846{
847 return skb->data - skb->head;
848}
849
850/**
851 * skb_tailroom - bytes at buffer end
852 * @skb: buffer to check
853 *
854 * Return the number of bytes of free space at the tail of an sk_buff
855 */
856static inline int skb_tailroom(const struct sk_buff *skb)
857{
858 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
859}
860
861/**
862 * skb_reserve - adjust headroom
863 * @skb: buffer to alter
864 * @len: bytes to move
865 *
866 * Increase the headroom of an empty &sk_buff by reducing the tail
867 * room. This is only allowed for an empty buffer.
868 */
869static inline void skb_reserve(struct sk_buff *skb, unsigned int len)
870{
871 skb->data += len;
872 skb->tail += len;
873}
874
875/*
876 * CPUs often take a performance hit when accessing unaligned memory
877 * locations. The actual performance hit varies, it can be small if the
878 * hardware handles it or large if we have to take an exception and fix it
879 * in software.
880 *
881 * Since an ethernet header is 14 bytes network drivers often end up with
882 * the IP header at an unaligned offset. The IP header can be aligned by
883 * shifting the start of the packet by 2 bytes. Drivers should do this
884 * with:
885 *
886 * skb_reserve(NET_IP_ALIGN);
887 *
888 * The downside to this alignment of the IP header is that the DMA is now
889 * unaligned. On some architectures the cost of an unaligned DMA is high
890 * and this cost outweighs the gains made by aligning the IP header.
891 *
892 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
893 * to be overridden.
894 */
895#ifndef NET_IP_ALIGN
896#define NET_IP_ALIGN 2
897#endif
898
899extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
900
901static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
902{
903 if (!skb->data_len) {
904 skb->len = len;
905 skb->tail = skb->data + len;
906 } else
907 ___pskb_trim(skb, len, 0);
908}
909
910/**
911 * skb_trim - remove end from a buffer
912 * @skb: buffer to alter
913 * @len: new length
914 *
915 * Cut the length of a buffer down by removing data from the tail. If
916 * the buffer is already under the length specified it is not modified.
917 */
918static inline void skb_trim(struct sk_buff *skb, unsigned int len)
919{
920 if (skb->len > len)
921 __skb_trim(skb, len);
922}
923
924
925static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
926{
927 if (!skb->data_len) {
928 skb->len = len;
929 skb->tail = skb->data+len;
930 return 0;
931 }
932 return ___pskb_trim(skb, len, 1);
933}
934
935static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
936{
937 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
938}
939
940/**
941 * skb_orphan - orphan a buffer
942 * @skb: buffer to orphan
943 *
944 * If a buffer currently has an owner then we call the owner's
945 * destructor function and make the @skb unowned. The buffer continues
946 * to exist but is no longer charged to its former owner.
947 */
948static inline void skb_orphan(struct sk_buff *skb)
949{
950 if (skb->destructor)
951 skb->destructor(skb);
952 skb->destructor = NULL;
953 skb->sk = NULL;
954}
955
956/**
957 * __skb_queue_purge - empty a list
958 * @list: list to empty
959 *
960 * Delete all buffers on an &sk_buff list. Each buffer is removed from
961 * the list and one reference dropped. This function does not take the
962 * list lock and the caller must hold the relevant locks to use it.
963 */
964extern void skb_queue_purge(struct sk_buff_head *list);
965static inline void __skb_queue_purge(struct sk_buff_head *list)
966{
967 struct sk_buff *skb;
968 while ((skb = __skb_dequeue(list)) != NULL)
969 kfree_skb(skb);
970}
971
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700972#ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973/**
974 * __dev_alloc_skb - allocate an skbuff for sending
975 * @length: length to allocate
976 * @gfp_mask: get_free_pages mask, passed to alloc_skb
977 *
978 * Allocate a new &sk_buff and assign it a usage count of one. The
979 * buffer has unspecified headroom built in. Users should allocate
980 * the headroom they think they need without accounting for the
981 * built in space. The built in space is used for optimisations.
982 *
983 * %NULL is returned in there is no free memory.
984 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
986 int gfp_mask)
987{
988 struct sk_buff *skb = alloc_skb(length + 16, gfp_mask);
989 if (likely(skb))
990 skb_reserve(skb, 16);
991 return skb;
992}
993#else
994extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
995#endif
996
997/**
998 * dev_alloc_skb - allocate an skbuff for sending
999 * @length: length to allocate
1000 *
1001 * Allocate a new &sk_buff and assign it a usage count of one. The
1002 * buffer has unspecified headroom built in. Users should allocate
1003 * the headroom they think they need without accounting for the
1004 * built in space. The built in space is used for optimisations.
1005 *
1006 * %NULL is returned in there is no free memory. Although this function
1007 * allocates memory it can be called from an interrupt.
1008 */
1009static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1010{
1011 return __dev_alloc_skb(length, GFP_ATOMIC);
1012}
1013
1014/**
1015 * skb_cow - copy header of skb when it is required
1016 * @skb: buffer to cow
1017 * @headroom: needed headroom
1018 *
1019 * If the skb passed lacks sufficient headroom or its data part
1020 * is shared, data is reallocated. If reallocation fails, an error
1021 * is returned and original skb is not changed.
1022 *
1023 * The result is skb with writable area skb->head...skb->tail
1024 * and at least @headroom of space at head.
1025 */
1026static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1027{
1028 int delta = (headroom > 16 ? headroom : 16) - skb_headroom(skb);
1029
1030 if (delta < 0)
1031 delta = 0;
1032
1033 if (delta || skb_cloned(skb))
1034 return pskb_expand_head(skb, (delta + 15) & ~15, 0, GFP_ATOMIC);
1035 return 0;
1036}
1037
1038/**
1039 * skb_padto - pad an skbuff up to a minimal size
1040 * @skb: buffer to pad
1041 * @len: minimal length
1042 *
1043 * Pads up a buffer to ensure the trailing bytes exist and are
1044 * blanked. If the buffer already contains sufficient data it
1045 * is untouched. Returns the buffer, which may be a replacement
1046 * for the original, or NULL for out of memory - in which case
1047 * the original buffer is still freed.
1048 */
1049
1050static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
1051{
1052 unsigned int size = skb->len;
1053 if (likely(size >= len))
1054 return skb;
1055 return skb_pad(skb, len-size);
1056}
1057
1058static inline int skb_add_data(struct sk_buff *skb,
1059 char __user *from, int copy)
1060{
1061 const int off = skb->len;
1062
1063 if (skb->ip_summed == CHECKSUM_NONE) {
1064 int err = 0;
1065 unsigned int csum = csum_and_copy_from_user(from,
1066 skb_put(skb, copy),
1067 copy, 0, &err);
1068 if (!err) {
1069 skb->csum = csum_block_add(skb->csum, csum, off);
1070 return 0;
1071 }
1072 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1073 return 0;
1074
1075 __skb_trim(skb, off);
1076 return -EFAULT;
1077}
1078
1079static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1080 struct page *page, int off)
1081{
1082 if (i) {
1083 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1084
1085 return page == frag->page &&
1086 off == frag->page_offset + frag->size;
1087 }
1088 return 0;
1089}
1090
1091/**
1092 * skb_linearize - convert paged skb to linear one
1093 * @skb: buffer to linarize
1094 * @gfp: allocation mode
1095 *
1096 * If there is no free memory -ENOMEM is returned, otherwise zero
1097 * is returned and the old skb data released.
1098 */
1099extern int __skb_linearize(struct sk_buff *skb, int gfp);
1100static inline int skb_linearize(struct sk_buff *skb, int gfp)
1101{
1102 return __skb_linearize(skb, gfp);
1103}
1104
1105/**
1106 * skb_postpull_rcsum - update checksum for received skb after pull
1107 * @skb: buffer to update
1108 * @start: start of data before pull
1109 * @len: length of data pulled
1110 *
1111 * After doing a pull on a received packet, you need to call this to
1112 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1113 * so that it can be recomputed from scratch.
1114 */
1115
1116static inline void skb_postpull_rcsum(struct sk_buff *skb,
1117 const void *start, int len)
1118{
1119 if (skb->ip_summed == CHECKSUM_HW)
1120 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1121}
1122
1123/**
1124 * pskb_trim_rcsum - trim received skb and update checksum
1125 * @skb: buffer to trim
1126 * @len: new length
1127 *
1128 * This is exactly the same as pskb_trim except that it ensures the
1129 * checksum of received packets are still valid after the operation.
1130 */
1131
1132static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1133{
1134 if (len >= skb->len)
1135 return 0;
1136 if (skb->ip_summed == CHECKSUM_HW)
1137 skb->ip_summed = CHECKSUM_NONE;
1138 return __pskb_trim(skb, len);
1139}
1140
1141static inline void *kmap_skb_frag(const skb_frag_t *frag)
1142{
1143#ifdef CONFIG_HIGHMEM
1144 BUG_ON(in_irq());
1145
1146 local_bh_disable();
1147#endif
1148 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1149}
1150
1151static inline void kunmap_skb_frag(void *vaddr)
1152{
1153 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1154#ifdef CONFIG_HIGHMEM
1155 local_bh_enable();
1156#endif
1157}
1158
1159#define skb_queue_walk(queue, skb) \
1160 for (skb = (queue)->next; \
1161 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1162 skb = skb->next)
1163
1164
1165extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1166 int noblock, int *err);
1167extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1168 struct poll_table_struct *wait);
1169extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1170 int offset, struct iovec *to,
1171 int size);
1172extern int skb_copy_and_csum_datagram_iovec(const
1173 struct sk_buff *skb,
1174 int hlen,
1175 struct iovec *iov);
1176extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1177extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1178 int len, unsigned int csum);
1179extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1180 void *to, int len);
Herbert Xu357b40a2005-04-19 22:30:14 -07001181extern int skb_store_bits(const struct sk_buff *skb, int offset,
1182 void *from, int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1184 int offset, u8 *to, int len,
1185 unsigned int csum);
1186extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1187extern void skb_split(struct sk_buff *skb,
1188 struct sk_buff *skb1, const u32 len);
1189
1190static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1191 int len, void *buffer)
1192{
1193 int hlen = skb_headlen(skb);
1194
1195 if (offset + len <= hlen)
1196 return skb->data + offset;
1197
1198 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1199 return NULL;
1200
1201 return buffer;
1202}
1203
1204extern void skb_init(void);
1205extern void skb_add_mtu(int mtu);
1206
1207#ifdef CONFIG_NETFILTER
1208static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1209{
1210 if (nfct && atomic_dec_and_test(&nfct->use))
1211 nfct->destroy(nfct);
1212}
1213static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1214{
1215 if (nfct)
1216 atomic_inc(&nfct->use);
1217}
1218static inline void nf_reset(struct sk_buff *skb)
1219{
1220 nf_conntrack_put(skb->nfct);
1221 skb->nfct = NULL;
1222#ifdef CONFIG_NETFILTER_DEBUG
1223 skb->nf_debug = 0;
1224#endif
1225}
1226static inline void nf_reset_debug(struct sk_buff *skb)
1227{
1228#ifdef CONFIG_NETFILTER_DEBUG
1229 skb->nf_debug = 0;
1230#endif
1231}
1232
1233#ifdef CONFIG_BRIDGE_NETFILTER
1234static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1235{
1236 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1237 kfree(nf_bridge);
1238}
1239static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1240{
1241 if (nf_bridge)
1242 atomic_inc(&nf_bridge->use);
1243}
1244#endif /* CONFIG_BRIDGE_NETFILTER */
1245#else /* CONFIG_NETFILTER */
1246static inline void nf_reset(struct sk_buff *skb) {}
1247#endif /* CONFIG_NETFILTER */
1248
1249#endif /* __KERNEL__ */
1250#endif /* _LINUX_SKBUFF_H */