ia64/xen-unstable

view linux-2.6-xen-sparse/include/linux/skbuff.h @ 8534:da7873110bbb

Tiny bootstrap cleanup.

Signed-off-by: Keir Fraser <keir@xensource.com>
author kaf24@firebug.cl.cam.ac.uk
date Mon Jan 09 19:46:46 2006 +0100 (2006-01-09)
parents 06d84bf87159
children fd9b2c1bb577
line source
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 */
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
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>
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>
32 #define HAVE_ALLOC_SKB /* For the drivers to know */
33 #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
34 #define SLAB_SKB /* Slabified skbuffs */
36 #define CHECKSUM_NONE 0
37 #define CHECKSUM_HW 1
38 #define CHECKSUM_UNNECESSARY 2
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))
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 */
86 struct net_device;
88 #ifdef CONFIG_NETFILTER
89 struct nf_conntrack {
90 atomic_t use;
91 void (*destroy)(struct nf_conntrack *);
92 };
94 #ifdef CONFIG_BRIDGE_NETFILTER
95 struct 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
107 #endif
109 struct sk_buff_head {
110 /* These two members must be first. */
111 struct sk_buff *next;
112 struct sk_buff *prev;
114 __u32 qlen;
115 spinlock_t lock;
116 };
118 struct sk_buff;
120 /* To allow 64K frame to be packed as single skb without frag_list */
121 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
123 typedef struct skb_frag_struct skb_frag_t;
125 struct skb_frag_struct {
126 struct page *page;
127 __u16 page_offset;
128 __u16 size;
129 };
131 /* This data is invariant across clones and lives at
132 * the end of the header data, ie. at skb->end.
133 */
134 struct 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 };
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)
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
170 * @dst: destination entry
171 * @sp: the security path, used for xfrm
172 * @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
177 * @local_df: allow local fragmentation
178 * @cloned: Head may be cloned (check refcnt to be sure)
179 * @nohdr: Payload reference only, must not modify header
180 * @proto_csum_valid: Protocol csum validated since arriving at localhost
181 * @proto_csum_blank: Protocol csum must be added before leaving localhost
182 * @pkt_type: Packet class
183 * @ip_summed: Driver fed us an IP checksum
184 * @priority: Packet queueing priority
185 * @users: User count - see {datagram,tcp}.c
186 * @protocol: Packet protocol from driver
187 * @security: Security level of packet
188 * @truesize: Buffer size
189 * @head: Head of buffer
190 * @data: Data head pointer
191 * @tail: Tail pointer
192 * @end: End pointer
193 * @destructor: Destruct function
194 * @nfmark: Can be used for communication between hooks
195 * @nfcache: Cache info
196 * @nfct: Associated connection, if any
197 * @nfctinfo: Relationship of this skb to the connection
198 * @nf_debug: Netfilter debugging
199 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
200 * @private: Data which is private to the HIPPI implementation
201 * @tc_index: Traffic control index
202 * @tc_verd: traffic control verdict
203 * @tc_classid: traffic control classid
204 */
206 struct sk_buff {
207 /* These two members must be first. */
208 struct sk_buff *next;
209 struct sk_buff *prev;
211 struct sk_buff_head *list;
212 struct sock *sk;
213 struct timeval stamp;
214 struct net_device *dev;
215 struct net_device *input_dev;
216 struct net_device *real_dev;
218 union {
219 struct tcphdr *th;
220 struct udphdr *uh;
221 struct icmphdr *icmph;
222 struct igmphdr *igmph;
223 struct iphdr *ipiph;
224 struct ipv6hdr *ipv6h;
225 unsigned char *raw;
226 } h;
228 union {
229 struct iphdr *iph;
230 struct ipv6hdr *ipv6h;
231 struct arphdr *arph;
232 unsigned char *raw;
233 } nh;
235 union {
236 unsigned char *raw;
237 } mac;
239 struct dst_entry *dst;
240 struct sec_path *sp;
242 /*
243 * This is the control buffer. It is free to use for every
244 * layer. Please put your private variables there. If you
245 * want to keep them across layers you have to do a skb_clone()
246 * first. This is owned by whoever has the skb queued ATM.
247 */
248 char cb[40];
250 unsigned int len,
251 data_len,
252 mac_len,
253 csum;
254 unsigned char local_df,
255 cloned:1,
256 nohdr:1,
257 proto_csum_valid:1,
258 proto_csum_blank:1,
259 pkt_type,
260 ip_summed;
261 __u32 priority;
262 unsigned short protocol,
263 security;
265 void (*destructor)(struct sk_buff *skb);
266 #ifdef CONFIG_NETFILTER
267 unsigned long nfmark;
268 __u32 nfcache;
269 __u32 nfctinfo;
270 struct nf_conntrack *nfct;
271 #ifdef CONFIG_NETFILTER_DEBUG
272 unsigned int nf_debug;
273 #endif
274 #ifdef CONFIG_BRIDGE_NETFILTER
275 struct nf_bridge_info *nf_bridge;
276 #endif
277 #endif /* CONFIG_NETFILTER */
278 #if defined(CONFIG_HIPPI)
279 union {
280 __u32 ifield;
281 } private;
282 #endif
283 #ifdef CONFIG_NET_SCHED
284 __u32 tc_index; /* traffic control index */
285 #ifdef CONFIG_NET_CLS_ACT
286 __u32 tc_verd; /* traffic control verdict */
287 __u32 tc_classid; /* traffic control classid */
288 #endif
290 #endif
293 /* These elements must be at the end, see alloc_skb() for details. */
294 unsigned int truesize;
295 atomic_t users;
296 unsigned char *head,
297 *data,
298 *tail,
299 *end;
300 };
302 #ifdef __KERNEL__
303 /*
304 * Handling routines are only of interest to the kernel
305 */
306 #include <linux/slab.h>
308 #include <asm/system.h>
310 extern void __kfree_skb(struct sk_buff *skb);
311 extern struct sk_buff *alloc_skb(unsigned int size, int priority);
312 extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
313 unsigned int size, int priority);
314 extern void kfree_skbmem(struct sk_buff *skb);
315 extern struct sk_buff *skb_clone(struct sk_buff *skb, int priority);
316 extern struct sk_buff *skb_copy(const struct sk_buff *skb, int priority);
317 extern struct sk_buff *pskb_copy(struct sk_buff *skb, int gfp_mask);
318 extern int pskb_expand_head(struct sk_buff *skb,
319 int nhead, int ntail, int gfp_mask);
320 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
321 unsigned int headroom);
322 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
323 int newheadroom, int newtailroom,
324 int priority);
325 extern struct sk_buff * skb_pad(struct sk_buff *skb, int pad);
326 #define dev_kfree_skb(a) kfree_skb(a)
327 extern void skb_over_panic(struct sk_buff *skb, int len,
328 void *here);
329 extern void skb_under_panic(struct sk_buff *skb, int len,
330 void *here);
332 /* Internal */
333 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
335 /**
336 * skb_queue_empty - check if a queue is empty
337 * @list: queue head
338 *
339 * Returns true if the queue is empty, false otherwise.
340 */
341 static inline int skb_queue_empty(const struct sk_buff_head *list)
342 {
343 return list->next == (struct sk_buff *)list;
344 }
346 /**
347 * skb_get - reference buffer
348 * @skb: buffer to reference
349 *
350 * Makes another reference to a socket buffer and returns a pointer
351 * to the buffer.
352 */
353 static inline struct sk_buff *skb_get(struct sk_buff *skb)
354 {
355 atomic_inc(&skb->users);
356 return skb;
357 }
359 /*
360 * If users == 1, we are the only owner and are can avoid redundant
361 * atomic change.
362 */
364 /**
365 * kfree_skb - free an sk_buff
366 * @skb: buffer to free
367 *
368 * Drop a reference to the buffer and free it if the usage count has
369 * hit zero.
370 */
371 static inline void kfree_skb(struct sk_buff *skb)
372 {
373 if (likely(atomic_read(&skb->users) == 1))
374 smp_rmb();
375 else if (likely(!atomic_dec_and_test(&skb->users)))
376 return;
377 __kfree_skb(skb);
378 }
380 /**
381 * skb_cloned - is the buffer a clone
382 * @skb: buffer to check
383 *
384 * Returns true if the buffer was generated with skb_clone() and is
385 * one of multiple shared copies of the buffer. Cloned buffers are
386 * shared data so must not be written to under normal circumstances.
387 */
388 static inline int skb_cloned(const struct sk_buff *skb)
389 {
390 return skb->cloned &&
391 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
392 }
394 /**
395 * skb_header_cloned - is the header a clone
396 * @skb: buffer to check
397 *
398 * Returns true if modifying the header part of the buffer requires
399 * the data to be copied.
400 */
401 static inline int skb_header_cloned(const struct sk_buff *skb)
402 {
403 int dataref;
405 if (!skb->cloned)
406 return 0;
408 dataref = atomic_read(&skb_shinfo(skb)->dataref);
409 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
410 return dataref != 1;
411 }
413 /**
414 * skb_header_release - release reference to header
415 * @skb: buffer to operate on
416 *
417 * Drop a reference to the header part of the buffer. This is done
418 * by acquiring a payload reference. You must not read from the header
419 * part of skb->data after this.
420 */
421 static inline void skb_header_release(struct sk_buff *skb)
422 {
423 BUG_ON(skb->nohdr);
424 skb->nohdr = 1;
425 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
426 }
428 /**
429 * skb_shared - is the buffer shared
430 * @skb: buffer to check
431 *
432 * Returns true if more than one person has a reference to this
433 * buffer.
434 */
435 static inline int skb_shared(const struct sk_buff *skb)
436 {
437 return atomic_read(&skb->users) != 1;
438 }
440 /**
441 * skb_share_check - check if buffer is shared and if so clone it
442 * @skb: buffer to check
443 * @pri: priority for memory allocation
444 *
445 * If the buffer is shared the buffer is cloned and the old copy
446 * drops a reference. A new clone with a single reference is returned.
447 * If the buffer is not shared the original buffer is returned. When
448 * being called from interrupt status or with spinlocks held pri must
449 * be GFP_ATOMIC.
450 *
451 * NULL is returned on a memory allocation failure.
452 */
453 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, int pri)
454 {
455 might_sleep_if(pri & __GFP_WAIT);
456 if (skb_shared(skb)) {
457 struct sk_buff *nskb = skb_clone(skb, pri);
458 kfree_skb(skb);
459 skb = nskb;
460 }
461 return skb;
462 }
464 /*
465 * Copy shared buffers into a new sk_buff. We effectively do COW on
466 * packets to handle cases where we have a local reader and forward
467 * and a couple of other messy ones. The normal one is tcpdumping
468 * a packet thats being forwarded.
469 */
471 /**
472 * skb_unshare - make a copy of a shared buffer
473 * @skb: buffer to check
474 * @pri: priority for memory allocation
475 *
476 * If the socket buffer is a clone then this function creates a new
477 * copy of the data, drops a reference count on the old copy and returns
478 * the new copy with the reference count at 1. If the buffer is not a clone
479 * the original buffer is returned. When called with a spinlock held or
480 * from interrupt state @pri must be %GFP_ATOMIC
481 *
482 * %NULL is returned on a memory allocation failure.
483 */
484 static inline struct sk_buff *skb_unshare(struct sk_buff *skb, int pri)
485 {
486 might_sleep_if(pri & __GFP_WAIT);
487 if (skb_cloned(skb)) {
488 struct sk_buff *nskb = skb_copy(skb, pri);
489 kfree_skb(skb); /* Free our shared copy */
490 skb = nskb;
491 }
492 return skb;
493 }
495 /**
496 * skb_peek
497 * @list_: list to peek at
498 *
499 * Peek an &sk_buff. Unlike most other operations you _MUST_
500 * be careful with this one. A peek leaves the buffer on the
501 * list and someone else may run off with it. You must hold
502 * the appropriate locks or have a private queue to do this.
503 *
504 * Returns %NULL for an empty list or a pointer to the head element.
505 * The reference count is not incremented and the reference is therefore
506 * volatile. Use with caution.
507 */
508 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
509 {
510 struct sk_buff *list = ((struct sk_buff *)list_)->next;
511 if (list == (struct sk_buff *)list_)
512 list = NULL;
513 return list;
514 }
516 /**
517 * skb_peek_tail
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 tail element.
526 * The reference count is not incremented and the reference is therefore
527 * volatile. Use with caution.
528 */
529 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
530 {
531 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
532 if (list == (struct sk_buff *)list_)
533 list = NULL;
534 return list;
535 }
537 /**
538 * skb_queue_len - get queue length
539 * @list_: list to measure
540 *
541 * Return the length of an &sk_buff queue.
542 */
543 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
544 {
545 return list_->qlen;
546 }
548 static inline void skb_queue_head_init(struct sk_buff_head *list)
549 {
550 spin_lock_init(&list->lock);
551 list->prev = list->next = (struct sk_buff *)list;
552 list->qlen = 0;
553 }
555 /*
556 * Insert an sk_buff at the start of a list.
557 *
558 * The "__skb_xxxx()" functions are the non-atomic ones that
559 * can only be called with interrupts disabled.
560 */
562 /**
563 * __skb_queue_head - queue a buffer at the list head
564 * @list: list to use
565 * @newsk: buffer to queue
566 *
567 * Queue a buffer at the start of a list. This function takes no locks
568 * and you must therefore hold required locks before calling it.
569 *
570 * A buffer cannot be placed on two lists at the same time.
571 */
572 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
573 static inline void __skb_queue_head(struct sk_buff_head *list,
574 struct sk_buff *newsk)
575 {
576 struct sk_buff *prev, *next;
578 newsk->list = list;
579 list->qlen++;
580 prev = (struct sk_buff *)list;
581 next = prev->next;
582 newsk->next = next;
583 newsk->prev = prev;
584 next->prev = prev->next = newsk;
585 }
587 /**
588 * __skb_queue_tail - queue a buffer at the list tail
589 * @list: list to use
590 * @newsk: buffer to queue
591 *
592 * Queue a buffer at the end of a list. This function takes no locks
593 * and you must therefore hold required locks before calling it.
594 *
595 * A buffer cannot be placed on two lists at the same time.
596 */
597 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
598 static inline void __skb_queue_tail(struct sk_buff_head *list,
599 struct sk_buff *newsk)
600 {
601 struct sk_buff *prev, *next;
603 newsk->list = list;
604 list->qlen++;
605 next = (struct sk_buff *)list;
606 prev = next->prev;
607 newsk->next = next;
608 newsk->prev = prev;
609 next->prev = prev->next = newsk;
610 }
613 /**
614 * __skb_dequeue - remove from the head of the queue
615 * @list: list to dequeue from
616 *
617 * Remove the head of the list. This function does not take any locks
618 * so must be used with appropriate locks held only. The head item is
619 * returned or %NULL if the list is empty.
620 */
621 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
622 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
623 {
624 struct sk_buff *next, *prev, *result;
626 prev = (struct sk_buff *) list;
627 next = prev->next;
628 result = NULL;
629 if (next != prev) {
630 result = next;
631 next = next->next;
632 list->qlen--;
633 next->prev = prev;
634 prev->next = next;
635 result->next = result->prev = NULL;
636 result->list = NULL;
637 }
638 return result;
639 }
642 /*
643 * Insert a packet on a list.
644 */
645 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk);
646 static inline void __skb_insert(struct sk_buff *newsk,
647 struct sk_buff *prev, struct sk_buff *next,
648 struct sk_buff_head *list)
649 {
650 newsk->next = next;
651 newsk->prev = prev;
652 next->prev = prev->next = newsk;
653 newsk->list = list;
654 list->qlen++;
655 }
657 /*
658 * Place a packet after a given packet in a list.
659 */
660 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk);
661 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk)
662 {
663 __skb_insert(newsk, old, old->next, old->list);
664 }
666 /*
667 * remove sk_buff from list. _Must_ be called atomically, and with
668 * the list known..
669 */
670 extern void skb_unlink(struct sk_buff *skb);
671 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
672 {
673 struct sk_buff *next, *prev;
675 list->qlen--;
676 next = skb->next;
677 prev = skb->prev;
678 skb->next = skb->prev = NULL;
679 skb->list = NULL;
680 next->prev = prev;
681 prev->next = next;
682 }
685 /* XXX: more streamlined implementation */
687 /**
688 * __skb_dequeue_tail - remove from the tail of the queue
689 * @list: list to dequeue from
690 *
691 * Remove the tail of the list. This function does not take any locks
692 * so must be used with appropriate locks held only. The tail item is
693 * returned or %NULL if the list is empty.
694 */
695 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
696 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
697 {
698 struct sk_buff *skb = skb_peek_tail(list);
699 if (skb)
700 __skb_unlink(skb, list);
701 return skb;
702 }
705 static inline int skb_is_nonlinear(const struct sk_buff *skb)
706 {
707 return skb->data_len;
708 }
710 static inline unsigned int skb_headlen(const struct sk_buff *skb)
711 {
712 return skb->len - skb->data_len;
713 }
715 static inline int skb_pagelen(const struct sk_buff *skb)
716 {
717 int i, len = 0;
719 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
720 len += skb_shinfo(skb)->frags[i].size;
721 return len + skb_headlen(skb);
722 }
724 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
725 struct page *page, int off, int size)
726 {
727 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
729 frag->page = page;
730 frag->page_offset = off;
731 frag->size = size;
732 skb_shinfo(skb)->nr_frags = i + 1;
733 }
735 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
736 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
737 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
739 /*
740 * Add data to an sk_buff
741 */
742 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
743 {
744 unsigned char *tmp = skb->tail;
745 SKB_LINEAR_ASSERT(skb);
746 skb->tail += len;
747 skb->len += len;
748 return tmp;
749 }
751 /**
752 * skb_put - add data to a buffer
753 * @skb: buffer to use
754 * @len: amount of data to add
755 *
756 * This function extends the used data area of the buffer. If this would
757 * exceed the total buffer size the kernel will panic. A pointer to the
758 * first byte of the extra data is returned.
759 */
760 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
761 {
762 unsigned char *tmp = skb->tail;
763 SKB_LINEAR_ASSERT(skb);
764 skb->tail += len;
765 skb->len += len;
766 if (unlikely(skb->tail>skb->end))
767 skb_over_panic(skb, len, current_text_addr());
768 return tmp;
769 }
771 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
772 {
773 skb->data -= len;
774 skb->len += len;
775 return skb->data;
776 }
778 /**
779 * skb_push - add data to the start of a buffer
780 * @skb: buffer to use
781 * @len: amount of data to add
782 *
783 * This function extends the used data area of the buffer at the buffer
784 * start. If this would exceed the total buffer headroom the kernel will
785 * panic. A pointer to the first byte of the extra data is returned.
786 */
787 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
788 {
789 skb->data -= len;
790 skb->len += len;
791 if (unlikely(skb->data<skb->head))
792 skb_under_panic(skb, len, current_text_addr());
793 return skb->data;
794 }
796 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
797 {
798 skb->len -= len;
799 BUG_ON(skb->len < skb->data_len);
800 return skb->data += len;
801 }
803 /**
804 * skb_pull - remove data from the start of a buffer
805 * @skb: buffer to use
806 * @len: amount of data to remove
807 *
808 * This function removes data from the start of a buffer, returning
809 * the memory to the headroom. A pointer to the next data in the buffer
810 * is returned. Once the data has been pulled future pushes will overwrite
811 * the old data.
812 */
813 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
814 {
815 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
816 }
818 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
820 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
821 {
822 if (len > skb_headlen(skb) &&
823 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
824 return NULL;
825 skb->len -= len;
826 return skb->data += len;
827 }
829 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
830 {
831 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
832 }
834 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
835 {
836 if (likely(len <= skb_headlen(skb)))
837 return 1;
838 if (unlikely(len > skb->len))
839 return 0;
840 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
841 }
843 /**
844 * skb_headroom - bytes at buffer head
845 * @skb: buffer to check
846 *
847 * Return the number of bytes of free space at the head of an &sk_buff.
848 */
849 static inline int skb_headroom(const struct sk_buff *skb)
850 {
851 return skb->data - skb->head;
852 }
854 /**
855 * skb_tailroom - bytes at buffer end
856 * @skb: buffer to check
857 *
858 * Return the number of bytes of free space at the tail of an sk_buff
859 */
860 static inline int skb_tailroom(const struct sk_buff *skb)
861 {
862 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
863 }
865 /**
866 * skb_reserve - adjust headroom
867 * @skb: buffer to alter
868 * @len: bytes to move
869 *
870 * Increase the headroom of an empty &sk_buff by reducing the tail
871 * room. This is only allowed for an empty buffer.
872 */
873 static inline void skb_reserve(struct sk_buff *skb, unsigned int len)
874 {
875 skb->data += len;
876 skb->tail += len;
877 }
879 /*
880 * CPUs often take a performance hit when accessing unaligned memory
881 * locations. The actual performance hit varies, it can be small if the
882 * hardware handles it or large if we have to take an exception and fix it
883 * in software.
884 *
885 * Since an ethernet header is 14 bytes network drivers often end up with
886 * the IP header at an unaligned offset. The IP header can be aligned by
887 * shifting the start of the packet by 2 bytes. Drivers should do this
888 * with:
889 *
890 * skb_reserve(NET_IP_ALIGN);
891 *
892 * The downside to this alignment of the IP header is that the DMA is now
893 * unaligned. On some architectures the cost of an unaligned DMA is high
894 * and this cost outweighs the gains made by aligning the IP header.
895 *
896 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
897 * to be overridden.
898 */
899 #ifndef NET_IP_ALIGN
900 #define NET_IP_ALIGN 2
901 #endif
903 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
905 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
906 {
907 if (!skb->data_len) {
908 skb->len = len;
909 skb->tail = skb->data + len;
910 } else
911 ___pskb_trim(skb, len, 0);
912 }
914 /**
915 * skb_trim - remove end from a buffer
916 * @skb: buffer to alter
917 * @len: new length
918 *
919 * Cut the length of a buffer down by removing data from the tail. If
920 * the buffer is already under the length specified it is not modified.
921 */
922 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
923 {
924 if (skb->len > len)
925 __skb_trim(skb, len);
926 }
929 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
930 {
931 if (!skb->data_len) {
932 skb->len = len;
933 skb->tail = skb->data+len;
934 return 0;
935 }
936 return ___pskb_trim(skb, len, 1);
937 }
939 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
940 {
941 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
942 }
944 /**
945 * skb_orphan - orphan a buffer
946 * @skb: buffer to orphan
947 *
948 * If a buffer currently has an owner then we call the owner's
949 * destructor function and make the @skb unowned. The buffer continues
950 * to exist but is no longer charged to its former owner.
951 */
952 static inline void skb_orphan(struct sk_buff *skb)
953 {
954 if (skb->destructor)
955 skb->destructor(skb);
956 skb->destructor = NULL;
957 skb->sk = NULL;
958 }
960 /**
961 * __skb_queue_purge - empty a list
962 * @list: list to empty
963 *
964 * Delete all buffers on an &sk_buff list. Each buffer is removed from
965 * the list and one reference dropped. This function does not take the
966 * list lock and the caller must hold the relevant locks to use it.
967 */
968 extern void skb_queue_purge(struct sk_buff_head *list);
969 static inline void __skb_queue_purge(struct sk_buff_head *list)
970 {
971 struct sk_buff *skb;
972 while ((skb = __skb_dequeue(list)) != NULL)
973 kfree_skb(skb);
974 }
976 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
977 /**
978 * __dev_alloc_skb - allocate an skbuff for sending
979 * @length: length to allocate
980 * @gfp_mask: get_free_pages mask, passed to alloc_skb
981 *
982 * Allocate a new &sk_buff and assign it a usage count of one. The
983 * buffer has unspecified headroom built in. Users should allocate
984 * the headroom they think they need without accounting for the
985 * built in space. The built in space is used for optimisations.
986 *
987 * %NULL is returned in there is no free memory.
988 */
989 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
990 int gfp_mask)
991 {
992 struct sk_buff *skb = alloc_skb(length + 16, gfp_mask);
993 if (likely(skb))
994 skb_reserve(skb, 16);
995 return skb;
996 }
997 #else
998 extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
999 #endif
1001 /**
1002 * dev_alloc_skb - allocate an skbuff for sending
1003 * @length: length to allocate
1005 * Allocate a new &sk_buff and assign it a usage count of one. The
1006 * buffer has unspecified headroom built in. Users should allocate
1007 * the headroom they think they need without accounting for the
1008 * built in space. The built in space is used for optimisations.
1010 * %NULL is returned in there is no free memory. Although this function
1011 * allocates memory it can be called from an interrupt.
1012 */
1013 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1015 return __dev_alloc_skb(length, GFP_ATOMIC);
1018 /**
1019 * skb_cow - copy header of skb when it is required
1020 * @skb: buffer to cow
1021 * @headroom: needed headroom
1023 * If the skb passed lacks sufficient headroom or its data part
1024 * is shared, data is reallocated. If reallocation fails, an error
1025 * is returned and original skb is not changed.
1027 * The result is skb with writable area skb->head...skb->tail
1028 * and at least @headroom of space at head.
1029 */
1030 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1032 int delta = (headroom > 16 ? headroom : 16) - skb_headroom(skb);
1034 if (delta < 0)
1035 delta = 0;
1037 if (delta || skb_cloned(skb))
1038 return pskb_expand_head(skb, (delta + 15) & ~15, 0, GFP_ATOMIC);
1039 return 0;
1042 /**
1043 * skb_padto - pad an skbuff up to a minimal size
1044 * @skb: buffer to pad
1045 * @len: minimal length
1047 * Pads up a buffer to ensure the trailing bytes exist and are
1048 * blanked. If the buffer already contains sufficient data it
1049 * is untouched. Returns the buffer, which may be a replacement
1050 * for the original, or NULL for out of memory - in which case
1051 * the original buffer is still freed.
1052 */
1054 static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
1056 unsigned int size = skb->len;
1057 if (likely(size >= len))
1058 return skb;
1059 return skb_pad(skb, len-size);
1062 static inline int skb_add_data(struct sk_buff *skb,
1063 char __user *from, int copy)
1065 const int off = skb->len;
1067 if (skb->ip_summed == CHECKSUM_NONE) {
1068 int err = 0;
1069 unsigned int csum = csum_and_copy_from_user(from,
1070 skb_put(skb, copy),
1071 copy, 0, &err);
1072 if (!err) {
1073 skb->csum = csum_block_add(skb->csum, csum, off);
1074 return 0;
1076 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1077 return 0;
1079 __skb_trim(skb, off);
1080 return -EFAULT;
1083 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1084 struct page *page, int off)
1086 if (i) {
1087 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1089 return page == frag->page &&
1090 off == frag->page_offset + frag->size;
1092 return 0;
1095 /**
1096 * skb_linearize - convert paged skb to linear one
1097 * @skb: buffer to linarize
1098 * @gfp: allocation mode
1100 * If there is no free memory -ENOMEM is returned, otherwise zero
1101 * is returned and the old skb data released.
1102 */
1103 extern int __skb_linearize(struct sk_buff *skb, int gfp);
1104 static inline int skb_linearize(struct sk_buff *skb, int gfp)
1106 return __skb_linearize(skb, gfp);
1109 /**
1110 * skb_postpull_rcsum - update checksum for received skb after pull
1111 * @skb: buffer to update
1112 * @start: start of data before pull
1113 * @len: length of data pulled
1115 * After doing a pull on a received packet, you need to call this to
1116 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1117 * so that it can be recomputed from scratch.
1118 */
1120 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1121 const void *start, int len)
1123 if (skb->ip_summed == CHECKSUM_HW)
1124 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1127 /**
1128 * pskb_trim_rcsum - trim received skb and update checksum
1129 * @skb: buffer to trim
1130 * @len: new length
1132 * This is exactly the same as pskb_trim except that it ensures the
1133 * checksum of received packets are still valid after the operation.
1134 */
1136 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1138 if (len >= skb->len)
1139 return 0;
1140 if (skb->ip_summed == CHECKSUM_HW)
1141 skb->ip_summed = CHECKSUM_NONE;
1142 return __pskb_trim(skb, len);
1145 static inline void *kmap_skb_frag(const skb_frag_t *frag)
1147 #ifdef CONFIG_HIGHMEM
1148 BUG_ON(in_irq());
1150 local_bh_disable();
1151 #endif
1152 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1155 static inline void kunmap_skb_frag(void *vaddr)
1157 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1158 #ifdef CONFIG_HIGHMEM
1159 local_bh_enable();
1160 #endif
1163 #define skb_queue_walk(queue, skb) \
1164 for (skb = (queue)->next; \
1165 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1166 skb = skb->next)
1169 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1170 int noblock, int *err);
1171 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1172 struct poll_table_struct *wait);
1173 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1174 int offset, struct iovec *to,
1175 int size);
1176 extern int skb_copy_and_csum_datagram_iovec(const
1177 struct sk_buff *skb,
1178 int hlen,
1179 struct iovec *iov);
1180 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1181 extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1182 int len, unsigned int csum);
1183 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1184 void *to, int len);
1185 extern int skb_store_bits(const struct sk_buff *skb, int offset,
1186 void *from, int len);
1187 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1188 int offset, u8 *to, int len,
1189 unsigned int csum);
1190 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1191 extern void skb_split(struct sk_buff *skb,
1192 struct sk_buff *skb1, const u32 len);
1194 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1195 int len, void *buffer)
1197 int hlen = skb_headlen(skb);
1199 if (offset + len <= hlen)
1200 return skb->data + offset;
1202 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1203 return NULL;
1205 return buffer;
1208 extern void skb_init(void);
1209 extern void skb_add_mtu(int mtu);
1211 #ifdef CONFIG_NETFILTER
1212 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1214 if (nfct && atomic_dec_and_test(&nfct->use))
1215 nfct->destroy(nfct);
1217 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1219 if (nfct)
1220 atomic_inc(&nfct->use);
1222 static inline void nf_reset(struct sk_buff *skb)
1224 nf_conntrack_put(skb->nfct);
1225 skb->nfct = NULL;
1226 #ifdef CONFIG_NETFILTER_DEBUG
1227 skb->nf_debug = 0;
1228 #endif
1230 static inline void nf_reset_debug(struct sk_buff *skb)
1232 #ifdef CONFIG_NETFILTER_DEBUG
1233 skb->nf_debug = 0;
1234 #endif
1237 #ifdef CONFIG_BRIDGE_NETFILTER
1238 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1240 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1241 kfree(nf_bridge);
1243 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1245 if (nf_bridge)
1246 atomic_inc(&nf_bridge->use);
1248 #endif /* CONFIG_BRIDGE_NETFILTER */
1249 #else /* CONFIG_NETFILTER */
1250 static inline void nf_reset(struct sk_buff *skb) {}
1251 #endif /* CONFIG_NETFILTER */
1253 #endif /* __KERNEL__ */
1254 #endif /* _LINUX_SKBUFF_H */