blob: 755e9cddac47eebe516751a55929f4339b576c5f [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
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
18#include <linux/compiler.h>
19#include <linux/time.h>
20#include <linux/cache.h>
21
22#include <asm/atomic.h>
23#include <asm/types.h>
24#include <linux/spinlock.h>
25#include <linux/mm.h>
26#include <linux/highmem.h>
27#include <linux/poll.h>
28#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070029#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include <net/checksum.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070031#include <linux/dmaengine.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032
33#define HAVE_ALLOC_SKB /* For the drivers to know */
34#define HAVE_ALIGNABLE_SKB /* Ditto 8) */
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
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;
Benjamin LaHaise4947d3e2006-01-03 14:06:50 -0800136 unsigned short nr_frags;
Herbert Xu79671682006-06-22 02:40:14 -0700137 unsigned short gso_size;
138 /* Warning: this field is not always filled in (UFO)! */
139 unsigned short gso_segs;
140 unsigned short gso_type;
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700141 unsigned int ip6_frag_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 struct sk_buff *frag_list;
143 skb_frag_t frags[MAX_SKB_FRAGS];
144};
145
146/* We divide dataref into two halves. The higher 16 bits hold references
147 * to the payload part of skb->data. The lower 16 bits hold references to
148 * the entire skb->data. It is up to the users of the skb to agree on
149 * where the payload starts.
150 *
151 * All users must obey the rule that the skb->data reference count must be
152 * greater than or equal to the payload reference count.
153 *
154 * Holding a reference to the payload part means that the user does not
155 * care about modifications to the header part of skb->data.
156 */
157#define SKB_DATAREF_SHIFT 16
158#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
159
Patrick McHardya61bbcf2005-08-14 17:24:31 -0700160struct skb_timeval {
161 u32 off_sec;
162 u32 off_usec;
163};
164
David S. Millerd179cd12005-08-17 14:57:30 -0700165
166enum {
167 SKB_FCLONE_UNAVAILABLE,
168 SKB_FCLONE_ORIG,
169 SKB_FCLONE_CLONE,
170};
171
Herbert Xu79671682006-06-22 02:40:14 -0700172enum {
173 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700174 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700175
176 /* This indicates the skb is from an untrusted source. */
177 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700178
179 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700180 SKB_GSO_TCP_ECN = 1 << 3,
181
182 SKB_GSO_TCPV6 = 1 << 4,
Herbert Xu79671682006-06-22 02:40:14 -0700183};
184
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185/**
186 * struct sk_buff - socket buffer
187 * @next: Next buffer in list
188 * @prev: Previous buffer in list
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 * @sk: Socket we are owned by
Herbert Xu325ed822005-10-03 13:57:23 -0700190 * @tstamp: Time we arrived
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 * @dev: Device we arrived on/are leaving by
192 * @input_dev: Device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 * @h: Transport layer header
194 * @nh: Network layer header
195 * @mac: Link layer header
Martin Waitz67be2dd2005-05-01 08:59:26 -0700196 * @dst: destination entry
197 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 * @cb: Control buffer. Free for use by every layer. Put private vars here
199 * @len: Length of actual data
200 * @data_len: Data length
201 * @mac_len: Length of link layer header
202 * @csum: Checksum
Martin Waitz67be2dd2005-05-01 08:59:26 -0700203 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 * @cloned: Head may be cloned (check refcnt to be sure)
205 * @nohdr: Payload reference only, must not modify header
206 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700207 * @fclone: skbuff clone status
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 * @ip_summed: Driver fed us an IP checksum
209 * @priority: Packet queueing priority
210 * @users: User count - see {datagram,tcp}.c
211 * @protocol: Packet protocol from driver
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212 * @truesize: Buffer size
213 * @head: Head of buffer
214 * @data: Data head pointer
215 * @tail: Tail pointer
216 * @end: End pointer
217 * @destructor: Destruct function
218 * @nfmark: Can be used for communication between hooks
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 * @nfct: Associated connection, if any
Randy Dunlapc83c2482005-10-18 22:07:41 -0700220 * @ipvs_property: skbuff is owned by ipvs
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 * @nfctinfo: Relationship of this skb to the connection
Randy Dunlap461ddf32005-11-20 21:25:15 -0800222 * @nfct_reasm: netfilter conntrack re-assembly pointer
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 * @tc_index: Traffic control index
225 * @tc_verd: traffic control verdict
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700226 * @dma_cookie: a cookie to one of several possible DMA operations
227 * done by skb DMA functions
James Morris984bc162006-06-09 00:29:17 -0700228 * @secmark: security marking
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 */
230
231struct sk_buff {
232 /* These two members must be first. */
233 struct sk_buff *next;
234 struct sk_buff *prev;
235
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 struct sock *sk;
Patrick McHardya61bbcf2005-08-14 17:24:31 -0700237 struct skb_timeval tstamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 struct net_device *dev;
239 struct net_device *input_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240
241 union {
242 struct tcphdr *th;
243 struct udphdr *uh;
244 struct icmphdr *icmph;
245 struct igmphdr *igmph;
246 struct iphdr *ipiph;
247 struct ipv6hdr *ipv6h;
248 unsigned char *raw;
249 } h;
250
251 union {
252 struct iphdr *iph;
253 struct ipv6hdr *ipv6h;
254 struct arphdr *arph;
255 unsigned char *raw;
256 } nh;
257
258 union {
259 unsigned char *raw;
260 } mac;
261
262 struct dst_entry *dst;
263 struct sec_path *sp;
264
265 /*
266 * This is the control buffer. It is free to use for every
267 * layer. Please put your private variables there. If you
268 * want to keep them across layers you have to do a skb_clone()
269 * first. This is owned by whoever has the skb queued ATM.
270 */
Patrick McHardy3e3850e2006-01-06 23:04:54 -0800271 char cb[48];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272
273 unsigned int len,
274 data_len,
275 mac_len,
276 csum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277 __u32 priority;
Thomas Graf1cbb3382005-07-05 14:13:41 -0700278 __u8 local_df:1,
279 cloned:1,
280 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700281 nohdr:1,
282 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700283 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800284 fclone:2,
285 ipvs_property:1;
Alexey Dobriyana0d3bea2005-08-11 16:05:50 -0700286 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287
288 void (*destructor)(struct sk_buff *skb);
289#ifdef CONFIG_NETFILTER
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 struct nf_conntrack *nfct;
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800291#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
292 struct sk_buff *nfct_reasm;
293#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294#ifdef CONFIG_BRIDGE_NETFILTER
295 struct nf_bridge_info *nf_bridge;
296#endif
Patrick McHardy77d2ca32006-03-20 17:12:12 -0800297 __u32 nfmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298#endif /* CONFIG_NETFILTER */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700300 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700302 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304#endif
Chris Leech97fc2f02006-05-23 17:55:33 -0700305#ifdef CONFIG_NET_DMA
306 dma_cookie_t dma_cookie;
307#endif
James Morris984bc162006-06-09 00:29:17 -0700308#ifdef CONFIG_NETWORK_SECMARK
309 __u32 secmark;
310#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311
312
313 /* These elements must be at the end, see alloc_skb() for details. */
314 unsigned int truesize;
315 atomic_t users;
316 unsigned char *head,
317 *data,
318 *tail,
319 *end;
320};
321
322#ifdef __KERNEL__
323/*
324 * Handling routines are only of interest to the kernel
325 */
326#include <linux/slab.h>
327
328#include <asm/system.h>
329
Jörn Engel231d06a2006-03-20 21:28:35 -0800330extern void kfree_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331extern void __kfree_skb(struct sk_buff *skb);
David S. Millerd179cd12005-08-17 14:57:30 -0700332extern struct sk_buff *__alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100333 gfp_t priority, int fclone);
David S. Millerd179cd12005-08-17 14:57:30 -0700334static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100335 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700336{
337 return __alloc_skb(size, priority, 0);
338}
339
340static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100341 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700342{
343 return __alloc_skb(size, priority, 1);
344}
345
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700347 unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100348 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349extern void kfree_skbmem(struct sk_buff *skb);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700350extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100351 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700352extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100353 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700354extern struct sk_buff *pskb_copy(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100355 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700357 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100358 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
360 unsigned int headroom);
361extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
362 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100363 gfp_t priority);
Herbert Xu5b057c62006-06-23 02:06:41 -0700364extern int skb_pad(struct sk_buff *skb, int pad);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365#define dev_kfree_skb(a) kfree_skb(a)
366extern void skb_over_panic(struct sk_buff *skb, int len,
367 void *here);
368extern void skb_under_panic(struct sk_buff *skb, int len,
369 void *here);
David S. Millerdc6de332006-04-20 00:10:50 -0700370extern void skb_truesize_bug(struct sk_buff *skb);
371
372static inline void skb_truesize_check(struct sk_buff *skb)
373{
374 if (unlikely((int)skb->truesize < sizeof(struct sk_buff) + skb->len))
375 skb_truesize_bug(skb);
376}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700378extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
379 int getfrag(void *from, char *to, int offset,
380 int len,int odd, struct sk_buff *skb),
381 void *from, int length);
382
Thomas Graf677e90e2005-06-23 20:59:51 -0700383struct skb_seq_state
384{
385 __u32 lower_offset;
386 __u32 upper_offset;
387 __u32 frag_idx;
388 __u32 stepped_offset;
389 struct sk_buff *root_skb;
390 struct sk_buff *cur_skb;
391 __u8 *frag_data;
392};
393
394extern void skb_prepare_seq_read(struct sk_buff *skb,
395 unsigned int from, unsigned int to,
396 struct skb_seq_state *st);
397extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
398 struct skb_seq_state *st);
399extern void skb_abort_seq_read(struct skb_seq_state *st);
400
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700401extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
402 unsigned int to, struct ts_config *config,
403 struct ts_state *state);
404
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405/* Internal */
406#define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
407
408/**
409 * skb_queue_empty - check if a queue is empty
410 * @list: queue head
411 *
412 * Returns true if the queue is empty, false otherwise.
413 */
414static inline int skb_queue_empty(const struct sk_buff_head *list)
415{
416 return list->next == (struct sk_buff *)list;
417}
418
419/**
420 * skb_get - reference buffer
421 * @skb: buffer to reference
422 *
423 * Makes another reference to a socket buffer and returns a pointer
424 * to the buffer.
425 */
426static inline struct sk_buff *skb_get(struct sk_buff *skb)
427{
428 atomic_inc(&skb->users);
429 return skb;
430}
431
432/*
433 * If users == 1, we are the only owner and are can avoid redundant
434 * atomic change.
435 */
436
437/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 * skb_cloned - is the buffer a clone
439 * @skb: buffer to check
440 *
441 * Returns true if the buffer was generated with skb_clone() and is
442 * one of multiple shared copies of the buffer. Cloned buffers are
443 * shared data so must not be written to under normal circumstances.
444 */
445static inline int skb_cloned(const struct sk_buff *skb)
446{
447 return skb->cloned &&
448 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
449}
450
451/**
452 * skb_header_cloned - is the header a clone
453 * @skb: buffer to check
454 *
455 * Returns true if modifying the header part of the buffer requires
456 * the data to be copied.
457 */
458static inline int skb_header_cloned(const struct sk_buff *skb)
459{
460 int dataref;
461
462 if (!skb->cloned)
463 return 0;
464
465 dataref = atomic_read(&skb_shinfo(skb)->dataref);
466 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
467 return dataref != 1;
468}
469
470/**
471 * skb_header_release - release reference to header
472 * @skb: buffer to operate on
473 *
474 * Drop a reference to the header part of the buffer. This is done
475 * by acquiring a payload reference. You must not read from the header
476 * part of skb->data after this.
477 */
478static inline void skb_header_release(struct sk_buff *skb)
479{
480 BUG_ON(skb->nohdr);
481 skb->nohdr = 1;
482 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
483}
484
485/**
486 * skb_shared - is the buffer shared
487 * @skb: buffer to check
488 *
489 * Returns true if more than one person has a reference to this
490 * buffer.
491 */
492static inline int skb_shared(const struct sk_buff *skb)
493{
494 return atomic_read(&skb->users) != 1;
495}
496
497/**
498 * skb_share_check - check if buffer is shared and if so clone it
499 * @skb: buffer to check
500 * @pri: priority for memory allocation
501 *
502 * If the buffer is shared the buffer is cloned and the old copy
503 * drops a reference. A new clone with a single reference is returned.
504 * If the buffer is not shared the original buffer is returned. When
505 * being called from interrupt status or with spinlocks held pri must
506 * be GFP_ATOMIC.
507 *
508 * NULL is returned on a memory allocation failure.
509 */
Victor Fusco86a76ca2005-07-08 14:57:47 -0700510static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100511 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512{
513 might_sleep_if(pri & __GFP_WAIT);
514 if (skb_shared(skb)) {
515 struct sk_buff *nskb = skb_clone(skb, pri);
516 kfree_skb(skb);
517 skb = nskb;
518 }
519 return skb;
520}
521
522/*
523 * Copy shared buffers into a new sk_buff. We effectively do COW on
524 * packets to handle cases where we have a local reader and forward
525 * and a couple of other messy ones. The normal one is tcpdumping
526 * a packet thats being forwarded.
527 */
528
529/**
530 * skb_unshare - make a copy of a shared buffer
531 * @skb: buffer to check
532 * @pri: priority for memory allocation
533 *
534 * If the socket buffer is a clone then this function creates a new
535 * copy of the data, drops a reference count on the old copy and returns
536 * the new copy with the reference count at 1. If the buffer is not a clone
537 * the original buffer is returned. When called with a spinlock held or
538 * from interrupt state @pri must be %GFP_ATOMIC
539 *
540 * %NULL is returned on a memory allocation failure.
541 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700542static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100543 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544{
545 might_sleep_if(pri & __GFP_WAIT);
546 if (skb_cloned(skb)) {
547 struct sk_buff *nskb = skb_copy(skb, pri);
548 kfree_skb(skb); /* Free our shared copy */
549 skb = nskb;
550 }
551 return skb;
552}
553
554/**
555 * skb_peek
556 * @list_: list to peek at
557 *
558 * Peek an &sk_buff. Unlike most other operations you _MUST_
559 * be careful with this one. A peek leaves the buffer on the
560 * list and someone else may run off with it. You must hold
561 * the appropriate locks or have a private queue to do this.
562 *
563 * Returns %NULL for an empty list or a pointer to the head element.
564 * The reference count is not incremented and the reference is therefore
565 * volatile. Use with caution.
566 */
567static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
568{
569 struct sk_buff *list = ((struct sk_buff *)list_)->next;
570 if (list == (struct sk_buff *)list_)
571 list = NULL;
572 return list;
573}
574
575/**
576 * skb_peek_tail
577 * @list_: list to peek at
578 *
579 * Peek an &sk_buff. Unlike most other operations you _MUST_
580 * be careful with this one. A peek leaves the buffer on the
581 * list and someone else may run off with it. You must hold
582 * the appropriate locks or have a private queue to do this.
583 *
584 * Returns %NULL for an empty list or a pointer to the tail element.
585 * The reference count is not incremented and the reference is therefore
586 * volatile. Use with caution.
587 */
588static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
589{
590 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
591 if (list == (struct sk_buff *)list_)
592 list = NULL;
593 return list;
594}
595
596/**
597 * skb_queue_len - get queue length
598 * @list_: list to measure
599 *
600 * Return the length of an &sk_buff queue.
601 */
602static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
603{
604 return list_->qlen;
605}
606
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700607/*
608 * This function creates a split out lock class for each invocation;
609 * this is needed for now since a whole lot of users of the skb-queue
610 * infrastructure in drivers have different locking usage (in hardirq)
611 * than the networking core (in softirq only). In the long run either the
612 * network layer or drivers should need annotation to consolidate the
613 * main types of usage into 3 classes.
614 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615static inline void skb_queue_head_init(struct sk_buff_head *list)
616{
617 spin_lock_init(&list->lock);
618 list->prev = list->next = (struct sk_buff *)list;
619 list->qlen = 0;
620}
621
622/*
623 * Insert an sk_buff at the start of a list.
624 *
625 * The "__skb_xxxx()" functions are the non-atomic ones that
626 * can only be called with interrupts disabled.
627 */
628
629/**
Stephen Hemminger300ce172005-10-30 13:47:34 -0800630 * __skb_queue_after - queue a buffer at the list head
631 * @list: list to use
632 * @prev: place after this buffer
633 * @newsk: buffer to queue
634 *
635 * Queue a buffer int the middle of a list. This function takes no locks
636 * and you must therefore hold required locks before calling it.
637 *
638 * A buffer cannot be placed on two lists at the same time.
639 */
640static inline void __skb_queue_after(struct sk_buff_head *list,
641 struct sk_buff *prev,
642 struct sk_buff *newsk)
643{
644 struct sk_buff *next;
645 list->qlen++;
646
647 next = prev->next;
648 newsk->next = next;
649 newsk->prev = prev;
650 next->prev = prev->next = newsk;
651}
652
653/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 * __skb_queue_head - queue a buffer at the list head
655 * @list: list to use
656 * @newsk: buffer to queue
657 *
658 * Queue a buffer at the start of a list. This function takes no locks
659 * and you must therefore hold required locks before calling it.
660 *
661 * A buffer cannot be placed on two lists at the same time.
662 */
663extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
664static inline void __skb_queue_head(struct sk_buff_head *list,
665 struct sk_buff *newsk)
666{
Stephen Hemminger300ce172005-10-30 13:47:34 -0800667 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668}
669
670/**
671 * __skb_queue_tail - queue a buffer at the list tail
672 * @list: list to use
673 * @newsk: buffer to queue
674 *
675 * Queue a buffer at the end of a list. This function takes no locks
676 * and you must therefore hold required locks before calling it.
677 *
678 * A buffer cannot be placed on two lists at the same time.
679 */
680extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
681static inline void __skb_queue_tail(struct sk_buff_head *list,
682 struct sk_buff *newsk)
683{
684 struct sk_buff *prev, *next;
685
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 list->qlen++;
687 next = (struct sk_buff *)list;
688 prev = next->prev;
689 newsk->next = next;
690 newsk->prev = prev;
691 next->prev = prev->next = newsk;
692}
693
694
695/**
696 * __skb_dequeue - remove from the head of the queue
697 * @list: list to dequeue from
698 *
699 * Remove the head of the list. This function does not take any locks
700 * so must be used with appropriate locks held only. The head item is
701 * returned or %NULL if the list is empty.
702 */
703extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
704static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
705{
706 struct sk_buff *next, *prev, *result;
707
708 prev = (struct sk_buff *) list;
709 next = prev->next;
710 result = NULL;
711 if (next != prev) {
712 result = next;
713 next = next->next;
714 list->qlen--;
715 next->prev = prev;
716 prev->next = next;
717 result->next = result->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 }
719 return result;
720}
721
722
723/*
724 * Insert a packet on a list.
725 */
David S. Miller8728b832005-08-09 19:25:21 -0700726extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727static inline void __skb_insert(struct sk_buff *newsk,
728 struct sk_buff *prev, struct sk_buff *next,
729 struct sk_buff_head *list)
730{
731 newsk->next = next;
732 newsk->prev = prev;
733 next->prev = prev->next = newsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 list->qlen++;
735}
736
737/*
738 * Place a packet after a given packet in a list.
739 */
David S. Miller8728b832005-08-09 19:25:21 -0700740extern void skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
741static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742{
David S. Miller8728b832005-08-09 19:25:21 -0700743 __skb_insert(newsk, old, old->next, list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744}
745
746/*
747 * remove sk_buff from list. _Must_ be called atomically, and with
748 * the list known..
749 */
David S. Miller8728b832005-08-09 19:25:21 -0700750extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
752{
753 struct sk_buff *next, *prev;
754
755 list->qlen--;
756 next = skb->next;
757 prev = skb->prev;
758 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 next->prev = prev;
760 prev->next = next;
761}
762
763
764/* XXX: more streamlined implementation */
765
766/**
767 * __skb_dequeue_tail - remove from the tail of the queue
768 * @list: list to dequeue from
769 *
770 * Remove the tail of the list. This function does not take any locks
771 * so must be used with appropriate locks held only. The tail item is
772 * returned or %NULL if the list is empty.
773 */
774extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
775static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
776{
777 struct sk_buff *skb = skb_peek_tail(list);
778 if (skb)
779 __skb_unlink(skb, list);
780 return skb;
781}
782
783
784static inline int skb_is_nonlinear(const struct sk_buff *skb)
785{
786 return skb->data_len;
787}
788
789static inline unsigned int skb_headlen(const struct sk_buff *skb)
790{
791 return skb->len - skb->data_len;
792}
793
794static inline int skb_pagelen(const struct sk_buff *skb)
795{
796 int i, len = 0;
797
798 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
799 len += skb_shinfo(skb)->frags[i].size;
800 return len + skb_headlen(skb);
801}
802
803static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
804 struct page *page, int off, int size)
805{
806 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
807
808 frag->page = page;
809 frag->page_offset = off;
810 frag->size = size;
811 skb_shinfo(skb)->nr_frags = i + 1;
812}
813
814#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
815#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
816#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
817
818/*
819 * Add data to an sk_buff
820 */
821static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
822{
823 unsigned char *tmp = skb->tail;
824 SKB_LINEAR_ASSERT(skb);
825 skb->tail += len;
826 skb->len += len;
827 return tmp;
828}
829
830/**
831 * skb_put - add data to a buffer
832 * @skb: buffer to use
833 * @len: amount of data to add
834 *
835 * This function extends the used data area of the buffer. If this would
836 * exceed the total buffer size the kernel will panic. A pointer to the
837 * first byte of the extra data is returned.
838 */
839static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
840{
841 unsigned char *tmp = skb->tail;
842 SKB_LINEAR_ASSERT(skb);
843 skb->tail += len;
844 skb->len += len;
845 if (unlikely(skb->tail>skb->end))
846 skb_over_panic(skb, len, current_text_addr());
847 return tmp;
848}
849
850static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
851{
852 skb->data -= len;
853 skb->len += len;
854 return skb->data;
855}
856
857/**
858 * skb_push - add data to the start of a buffer
859 * @skb: buffer to use
860 * @len: amount of data to add
861 *
862 * This function extends the used data area of the buffer at the buffer
863 * start. If this would exceed the total buffer headroom the kernel will
864 * panic. A pointer to the first byte of the extra data is returned.
865 */
866static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
867{
868 skb->data -= len;
869 skb->len += len;
870 if (unlikely(skb->data<skb->head))
871 skb_under_panic(skb, len, current_text_addr());
872 return skb->data;
873}
874
875static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
876{
877 skb->len -= len;
878 BUG_ON(skb->len < skb->data_len);
879 return skb->data += len;
880}
881
882/**
883 * skb_pull - remove data from the start of a buffer
884 * @skb: buffer to use
885 * @len: amount of data to remove
886 *
887 * This function removes data from the start of a buffer, returning
888 * the memory to the headroom. A pointer to the next data in the buffer
889 * is returned. Once the data has been pulled future pushes will overwrite
890 * the old data.
891 */
892static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
893{
894 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
895}
896
897extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
898
899static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
900{
901 if (len > skb_headlen(skb) &&
902 !__pskb_pull_tail(skb, len-skb_headlen(skb)))
903 return NULL;
904 skb->len -= len;
905 return skb->data += len;
906}
907
908static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
909{
910 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
911}
912
913static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
914{
915 if (likely(len <= skb_headlen(skb)))
916 return 1;
917 if (unlikely(len > skb->len))
918 return 0;
919 return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
920}
921
922/**
923 * skb_headroom - bytes at buffer head
924 * @skb: buffer to check
925 *
926 * Return the number of bytes of free space at the head of an &sk_buff.
927 */
928static inline int skb_headroom(const struct sk_buff *skb)
929{
930 return skb->data - skb->head;
931}
932
933/**
934 * skb_tailroom - bytes at buffer end
935 * @skb: buffer to check
936 *
937 * Return the number of bytes of free space at the tail of an sk_buff
938 */
939static inline int skb_tailroom(const struct sk_buff *skb)
940{
941 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
942}
943
944/**
945 * skb_reserve - adjust headroom
946 * @skb: buffer to alter
947 * @len: bytes to move
948 *
949 * Increase the headroom of an empty &sk_buff by reducing the tail
950 * room. This is only allowed for an empty buffer.
951 */
David S. Miller82431262006-01-17 02:54:21 -0800952static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953{
954 skb->data += len;
955 skb->tail += len;
956}
957
958/*
959 * CPUs often take a performance hit when accessing unaligned memory
960 * locations. The actual performance hit varies, it can be small if the
961 * hardware handles it or large if we have to take an exception and fix it
962 * in software.
963 *
964 * Since an ethernet header is 14 bytes network drivers often end up with
965 * the IP header at an unaligned offset. The IP header can be aligned by
966 * shifting the start of the packet by 2 bytes. Drivers should do this
967 * with:
968 *
969 * skb_reserve(NET_IP_ALIGN);
970 *
971 * The downside to this alignment of the IP header is that the DMA is now
972 * unaligned. On some architectures the cost of an unaligned DMA is high
973 * and this cost outweighs the gains made by aligning the IP header.
974 *
975 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
976 * to be overridden.
977 */
978#ifndef NET_IP_ALIGN
979#define NET_IP_ALIGN 2
980#endif
981
Anton Blanchard025be812006-03-31 02:27:06 -0800982/*
983 * The networking layer reserves some headroom in skb data (via
984 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
985 * the header has to grow. In the default case, if the header has to grow
986 * 16 bytes or less we avoid the reallocation.
987 *
988 * Unfortunately this headroom changes the DMA alignment of the resulting
989 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
990 * on some architectures. An architecture can override this value,
991 * perhaps setting it to a cacheline in size (since that will maintain
992 * cacheline alignment of the DMA). It must be a power of 2.
993 *
994 * Various parts of the networking layer expect at least 16 bytes of
995 * headroom, you should not reduce this.
996 */
997#ifndef NET_SKB_PAD
998#define NET_SKB_PAD 16
999#endif
1000
Herbert Xu3cc0e872006-06-09 16:13:38 -07001001extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002
1003static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1004{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001005 if (unlikely(skb->data_len)) {
1006 WARN_ON(1);
1007 return;
1008 }
1009 skb->len = len;
1010 skb->tail = skb->data + len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011}
1012
1013/**
1014 * skb_trim - remove end from a buffer
1015 * @skb: buffer to alter
1016 * @len: new length
1017 *
1018 * Cut the length of a buffer down by removing data from the tail. If
1019 * the buffer is already under the length specified it is not modified.
Herbert Xu3cc0e872006-06-09 16:13:38 -07001020 * The skb must be linear.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 */
1022static inline void skb_trim(struct sk_buff *skb, unsigned int len)
1023{
1024 if (skb->len > len)
1025 __skb_trim(skb, len);
1026}
1027
1028
1029static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1030{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001031 if (skb->data_len)
1032 return ___pskb_trim(skb, len);
1033 __skb_trim(skb, len);
1034 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035}
1036
1037static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1038{
1039 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1040}
1041
1042/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001043 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1044 * @skb: buffer to alter
1045 * @len: new length
1046 *
1047 * This is identical to pskb_trim except that the caller knows that
1048 * the skb is not cloned so we should never get an error due to out-
1049 * of-memory.
1050 */
1051static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1052{
1053 int err = pskb_trim(skb, len);
1054 BUG_ON(err);
1055}
1056
1057/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 * skb_orphan - orphan a buffer
1059 * @skb: buffer to orphan
1060 *
1061 * If a buffer currently has an owner then we call the owner's
1062 * destructor function and make the @skb unowned. The buffer continues
1063 * to exist but is no longer charged to its former owner.
1064 */
1065static inline void skb_orphan(struct sk_buff *skb)
1066{
1067 if (skb->destructor)
1068 skb->destructor(skb);
1069 skb->destructor = NULL;
1070 skb->sk = NULL;
1071}
1072
1073/**
1074 * __skb_queue_purge - empty a list
1075 * @list: list to empty
1076 *
1077 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1078 * the list and one reference dropped. This function does not take the
1079 * list lock and the caller must hold the relevant locks to use it.
1080 */
1081extern void skb_queue_purge(struct sk_buff_head *list);
1082static inline void __skb_queue_purge(struct sk_buff_head *list)
1083{
1084 struct sk_buff *skb;
1085 while ((skb = __skb_dequeue(list)) != NULL)
1086 kfree_skb(skb);
1087}
1088
1089/**
Christoph Hellwigb4e54de2006-07-24 15:31:14 -07001090 * __dev_alloc_skb - allocate an skbuff for receiving
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 * @length: length to allocate
1092 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1093 *
1094 * Allocate a new &sk_buff and assign it a usage count of one. The
1095 * buffer has unspecified headroom built in. Users should allocate
1096 * the headroom they think they need without accounting for the
1097 * built in space. The built in space is used for optimisations.
1098 *
Christoph Hellwig766ea8c2006-08-07 15:49:53 -07001099 * %NULL is returned if there is no free memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
Al Virodd0fc662005-10-07 07:46:04 +01001102 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103{
Anton Blanchard025be812006-03-31 02:27:06 -08001104 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 if (likely(skb))
Anton Blanchard025be812006-03-31 02:27:06 -08001106 skb_reserve(skb, NET_SKB_PAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 return skb;
1108}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109
1110/**
Christoph Hellwigb4e54de2006-07-24 15:31:14 -07001111 * dev_alloc_skb - allocate an skbuff for receiving
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 * @length: length to allocate
1113 *
1114 * Allocate a new &sk_buff and assign it a usage count of one. The
1115 * buffer has unspecified headroom built in. Users should allocate
1116 * the headroom they think they need without accounting for the
1117 * built in space. The built in space is used for optimisations.
1118 *
Christoph Hellwig766ea8c2006-08-07 15:49:53 -07001119 * %NULL is returned if there is no free memory. Although this function
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 * allocates memory it can be called from an interrupt.
1121 */
1122static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1123{
1124 return __dev_alloc_skb(length, GFP_ATOMIC);
1125}
1126
Christoph Hellwig8af27452006-07-31 22:35:23 -07001127extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1128 unsigned int length, gfp_t gfp_mask);
1129
1130/**
1131 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1132 * @dev: network device to receive on
1133 * @length: length to allocate
1134 *
1135 * Allocate a new &sk_buff and assign it a usage count of one. The
1136 * buffer has unspecified headroom built in. Users should allocate
1137 * the headroom they think they need without accounting for the
1138 * built in space. The built in space is used for optimisations.
1139 *
1140 * %NULL is returned if there is no free memory. Although this function
1141 * allocates memory it can be called from an interrupt.
1142 */
1143static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1144 unsigned int length)
1145{
1146 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1147}
1148
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149/**
1150 * skb_cow - copy header of skb when it is required
1151 * @skb: buffer to cow
1152 * @headroom: needed headroom
1153 *
1154 * If the skb passed lacks sufficient headroom or its data part
1155 * is shared, data is reallocated. If reallocation fails, an error
1156 * is returned and original skb is not changed.
1157 *
1158 * The result is skb with writable area skb->head...skb->tail
1159 * and at least @headroom of space at head.
1160 */
1161static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1162{
Anton Blanchard025be812006-03-31 02:27:06 -08001163 int delta = (headroom > NET_SKB_PAD ? headroom : NET_SKB_PAD) -
1164 skb_headroom(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165
1166 if (delta < 0)
1167 delta = 0;
1168
1169 if (delta || skb_cloned(skb))
Anton Blanchard025be812006-03-31 02:27:06 -08001170 return pskb_expand_head(skb, (delta + (NET_SKB_PAD-1)) &
1171 ~(NET_SKB_PAD-1), 0, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 return 0;
1173}
1174
1175/**
1176 * skb_padto - pad an skbuff up to a minimal size
1177 * @skb: buffer to pad
1178 * @len: minimal length
1179 *
1180 * Pads up a buffer to ensure the trailing bytes exist and are
1181 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07001182 * is untouched. Otherwise it is extended. Returns zero on
1183 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184 */
1185
Herbert Xu5b057c62006-06-23 02:06:41 -07001186static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187{
1188 unsigned int size = skb->len;
1189 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07001190 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191 return skb_pad(skb, len-size);
1192}
1193
1194static inline int skb_add_data(struct sk_buff *skb,
1195 char __user *from, int copy)
1196{
1197 const int off = skb->len;
1198
1199 if (skb->ip_summed == CHECKSUM_NONE) {
1200 int err = 0;
1201 unsigned int csum = csum_and_copy_from_user(from,
1202 skb_put(skb, copy),
1203 copy, 0, &err);
1204 if (!err) {
1205 skb->csum = csum_block_add(skb->csum, csum, off);
1206 return 0;
1207 }
1208 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1209 return 0;
1210
1211 __skb_trim(skb, off);
1212 return -EFAULT;
1213}
1214
1215static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1216 struct page *page, int off)
1217{
1218 if (i) {
1219 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1220
1221 return page == frag->page &&
1222 off == frag->page_offset + frag->size;
1223 }
1224 return 0;
1225}
1226
Herbert Xu364c6ba2006-06-09 16:10:40 -07001227static inline int __skb_linearize(struct sk_buff *skb)
1228{
1229 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1230}
1231
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232/**
1233 * skb_linearize - convert paged skb to linear one
1234 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 *
1236 * If there is no free memory -ENOMEM is returned, otherwise zero
1237 * is returned and the old skb data released.
1238 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07001239static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240{
Herbert Xu364c6ba2006-06-09 16:10:40 -07001241 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1242}
1243
1244/**
1245 * skb_linearize_cow - make sure skb is linear and writable
1246 * @skb: buffer to process
1247 *
1248 * If there is no free memory -ENOMEM is returned, otherwise zero
1249 * is returned and the old skb data released.
1250 */
1251static inline int skb_linearize_cow(struct sk_buff *skb)
1252{
1253 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1254 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255}
1256
1257/**
1258 * skb_postpull_rcsum - update checksum for received skb after pull
1259 * @skb: buffer to update
1260 * @start: start of data before pull
1261 * @len: length of data pulled
1262 *
1263 * After doing a pull on a received packet, you need to call this to
1264 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1265 * so that it can be recomputed from scratch.
1266 */
1267
1268static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08001269 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270{
1271 if (skb->ip_summed == CHECKSUM_HW)
1272 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1273}
1274
Herbert Xucbb042f2006-03-20 22:43:56 -08001275unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1276
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277/**
1278 * pskb_trim_rcsum - trim received skb and update checksum
1279 * @skb: buffer to trim
1280 * @len: new length
1281 *
1282 * This is exactly the same as pskb_trim except that it ensures the
1283 * checksum of received packets are still valid after the operation.
1284 */
1285
1286static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1287{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07001288 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 return 0;
1290 if (skb->ip_summed == CHECKSUM_HW)
1291 skb->ip_summed = CHECKSUM_NONE;
1292 return __pskb_trim(skb, len);
1293}
1294
1295static inline void *kmap_skb_frag(const skb_frag_t *frag)
1296{
1297#ifdef CONFIG_HIGHMEM
1298 BUG_ON(in_irq());
1299
1300 local_bh_disable();
1301#endif
1302 return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1303}
1304
1305static inline void kunmap_skb_frag(void *vaddr)
1306{
1307 kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1308#ifdef CONFIG_HIGHMEM
1309 local_bh_enable();
1310#endif
1311}
1312
1313#define skb_queue_walk(queue, skb) \
1314 for (skb = (queue)->next; \
1315 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1316 skb = skb->next)
1317
Stephen Hemminger300ce172005-10-30 13:47:34 -08001318#define skb_queue_reverse_walk(queue, skb) \
1319 for (skb = (queue)->prev; \
1320 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1321 skb = skb->prev)
1322
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323
1324extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1325 int noblock, int *err);
1326extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1327 struct poll_table_struct *wait);
1328extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1329 int offset, struct iovec *to,
1330 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08001331extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 int hlen,
1333 struct iovec *iov);
1334extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Herbert Xu3305b802005-12-13 23:16:37 -08001335extern void skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1336 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
1338 int len, unsigned int csum);
1339extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1340 void *to, int len);
Herbert Xu357b40a2005-04-19 22:30:14 -07001341extern int skb_store_bits(const struct sk_buff *skb, int offset,
1342 void *from, int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
1344 int offset, u8 *to, int len,
1345 unsigned int csum);
1346extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1347extern void skb_split(struct sk_buff *skb,
1348 struct sk_buff *skb1, const u32 len);
1349
Herbert Xu576a30e2006-06-27 13:22:38 -07001350extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001351
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1353 int len, void *buffer)
1354{
1355 int hlen = skb_headlen(skb);
1356
Patrick McHardy55820ee2005-07-05 14:08:10 -07001357 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 return skb->data + offset;
1359
1360 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1361 return NULL;
1362
1363 return buffer;
1364}
1365
1366extern void skb_init(void);
1367extern void skb_add_mtu(int mtu);
1368
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001369/**
1370 * skb_get_timestamp - get timestamp from a skb
1371 * @skb: skb to get stamp from
1372 * @stamp: pointer to struct timeval to store stamp in
1373 *
1374 * Timestamps are stored in the skb as offsets to a base timestamp.
1375 * This function converts the offset back to a struct timeval and stores
1376 * it in stamp.
1377 */
Stephen Hemmingerf2c38392005-09-06 15:48:03 -07001378static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001379{
1380 stamp->tv_sec = skb->tstamp.off_sec;
1381 stamp->tv_usec = skb->tstamp.off_usec;
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001382}
1383
1384/**
1385 * skb_set_timestamp - set timestamp of a skb
1386 * @skb: skb to set stamp of
1387 * @stamp: pointer to struct timeval to get stamp from
1388 *
1389 * Timestamps are stored in the skb as offsets to a base timestamp.
1390 * This function converts a struct timeval to an offset and stores
1391 * it in the skb.
1392 */
Stephen Hemmingerf2c38392005-09-06 15:48:03 -07001393static inline void skb_set_timestamp(struct sk_buff *skb, const struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001394{
Herbert Xu325ed822005-10-03 13:57:23 -07001395 skb->tstamp.off_sec = stamp->tv_sec;
1396 skb->tstamp.off_usec = stamp->tv_usec;
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001397}
1398
1399extern void __net_timestamp(struct sk_buff *skb);
1400
Herbert Xufb286bb2005-11-10 13:01:24 -08001401extern unsigned int __skb_checksum_complete(struct sk_buff *skb);
1402
1403/**
1404 * skb_checksum_complete - Calculate checksum of an entire packet
1405 * @skb: packet to process
1406 *
1407 * This function calculates the checksum over the entire packet plus
1408 * the value of skb->csum. The latter can be used to supply the
1409 * checksum of a pseudo header as used by TCP/UDP. It returns the
1410 * checksum.
1411 *
1412 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1413 * this function can be used to verify that checksum on received
1414 * packets. In that case the function should return zero if the
1415 * checksum is correct. In particular, this function will return zero
1416 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1417 * hardware has already verified the correctness of the checksum.
1418 */
1419static inline unsigned int skb_checksum_complete(struct sk_buff *skb)
1420{
1421 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
1422 __skb_checksum_complete(skb);
1423}
1424
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425#ifdef CONFIG_NETFILTER
1426static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1427{
1428 if (nfct && atomic_dec_and_test(&nfct->use))
1429 nfct->destroy(nfct);
1430}
1431static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1432{
1433 if (nfct)
1434 atomic_inc(&nfct->use);
1435}
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08001436#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1437static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1438{
1439 if (skb)
1440 atomic_inc(&skb->users);
1441}
1442static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1443{
1444 if (skb)
1445 kfree_skb(skb);
1446}
1447#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448#ifdef CONFIG_BRIDGE_NETFILTER
1449static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1450{
1451 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1452 kfree(nf_bridge);
1453}
1454static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1455{
1456 if (nf_bridge)
1457 atomic_inc(&nf_bridge->use);
1458}
1459#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08001460static inline void nf_reset(struct sk_buff *skb)
1461{
1462 nf_conntrack_put(skb->nfct);
1463 skb->nfct = NULL;
1464#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1465 nf_conntrack_put_reasm(skb->nfct_reasm);
1466 skb->nfct_reasm = NULL;
1467#endif
1468#ifdef CONFIG_BRIDGE_NETFILTER
1469 nf_bridge_put(skb->nf_bridge);
1470 skb->nf_bridge = NULL;
1471#endif
1472}
1473
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474#else /* CONFIG_NETFILTER */
1475static inline void nf_reset(struct sk_buff *skb) {}
1476#endif /* CONFIG_NETFILTER */
1477
James Morris984bc162006-06-09 00:29:17 -07001478#ifdef CONFIG_NETWORK_SECMARK
1479static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1480{
1481 to->secmark = from->secmark;
1482}
1483
1484static inline void skb_init_secmark(struct sk_buff *skb)
1485{
1486 skb->secmark = 0;
1487}
1488#else
1489static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1490{ }
1491
1492static inline void skb_init_secmark(struct sk_buff *skb)
1493{ }
1494#endif
1495
Herbert Xu89114af2006-07-08 13:34:32 -07001496static inline int skb_is_gso(const struct sk_buff *skb)
1497{
1498 return skb_shinfo(skb)->gso_size;
1499}
1500
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501#endif /* __KERNEL__ */
1502#endif /* _LINUX_SKBUFF_H */