]> bbs.cooldavid.org Git - net-next-2.6.git/blame_incremental - include/linux/skbuff.h
bridge: Fix build error when IGMP_SNOOPING is not enabled
[net-next-2.6.git] / include / linux / skbuff.h
... / ...
CommitLineData
1/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
17#include <linux/kernel.h>
18#include <linux/kmemcheck.h>
19#include <linux/compiler.h>
20#include <linux/time.h>
21#include <linux/cache.h>
22
23#include <asm/atomic.h>
24#include <asm/types.h>
25#include <linux/spinlock.h>
26#include <linux/net.h>
27#include <linux/textsearch.h>
28#include <net/checksum.h>
29#include <linux/rcupdate.h>
30#include <linux/dmaengine.h>
31#include <linux/hrtimer.h>
32
33/* Don't change this without changing skb_csum_unnecessary! */
34#define CHECKSUM_NONE 0
35#define CHECKSUM_UNNECESSARY 1
36#define CHECKSUM_COMPLETE 2
37#define CHECKSUM_PARTIAL 3
38
39#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
40 ~(SMP_CACHE_BYTES - 1))
41#define SKB_WITH_OVERHEAD(X) \
42 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
43#define SKB_MAX_ORDER(X, ORDER) \
44 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
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 * COMPLETE: 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 COMPLETE,
63 * not UNNECESSARY.
64 *
65 * PARTIAL: identical to the case for output below. This may occur
66 * on a packet received directly from another Linux OS, e.g.,
67 * a virtualised Linux kernel on the same host. The packet can
68 * be treated in the same way as UNNECESSARY except that on
69 * output (i.e., forwarding) the checksum must be filled in
70 * by the OS or the hardware.
71 *
72 * B. Checksumming on output.
73 *
74 * NONE: skb is checksummed by protocol or csum is not required.
75 *
76 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
77 * from skb->csum_start to the end and to record the checksum
78 * at skb->csum_start + skb->csum_offset.
79 *
80 * Device must show its capabilities in dev->features, set
81 * at device setup time.
82 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
83 * everything.
84 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
85 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
86 * TCP/UDP over IPv4. Sigh. Vendors like this
87 * way by an unknown reason. Though, see comment above
88 * about CHECKSUM_UNNECESSARY. 8)
89 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
90 *
91 * Any questions? No questions, good. --ANK
92 */
93
94struct net_device;
95struct scatterlist;
96struct pipe_inode_info;
97
98#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
99struct nf_conntrack {
100 atomic_t use;
101};
102#endif
103
104#ifdef CONFIG_BRIDGE_NETFILTER
105struct nf_bridge_info {
106 atomic_t use;
107 struct net_device *physindev;
108 struct net_device *physoutdev;
109 unsigned int mask;
110 unsigned long data[32 / sizeof(unsigned long)];
111};
112#endif
113
114struct sk_buff_head {
115 /* These two members must be first. */
116 struct sk_buff *next;
117 struct sk_buff *prev;
118
119 __u32 qlen;
120 spinlock_t lock;
121};
122
123struct sk_buff;
124
125/* To allow 64K frame to be packed as single skb without frag_list */
126#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
127
128typedef struct skb_frag_struct skb_frag_t;
129
130struct skb_frag_struct {
131 struct page *page;
132 __u32 page_offset;
133 __u32 size;
134};
135
136#define HAVE_HW_TIME_STAMP
137
138/**
139 * struct skb_shared_hwtstamps - hardware time stamps
140 * @hwtstamp: hardware time stamp transformed into duration
141 * since arbitrary point in time
142 * @syststamp: hwtstamp transformed to system time base
143 *
144 * Software time stamps generated by ktime_get_real() are stored in
145 * skb->tstamp. The relation between the different kinds of time
146 * stamps is as follows:
147 *
148 * syststamp and tstamp can be compared against each other in
149 * arbitrary combinations. The accuracy of a
150 * syststamp/tstamp/"syststamp from other device" comparison is
151 * limited by the accuracy of the transformation into system time
152 * base. This depends on the device driver and its underlying
153 * hardware.
154 *
155 * hwtstamps can only be compared against other hwtstamps from
156 * the same device.
157 *
158 * This structure is attached to packets as part of the
159 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
160 */
161struct skb_shared_hwtstamps {
162 ktime_t hwtstamp;
163 ktime_t syststamp;
164};
165
166/**
167 * struct skb_shared_tx - instructions for time stamping of outgoing packets
168 * @hardware: generate hardware time stamp
169 * @software: generate software time stamp
170 * @in_progress: device driver is going to provide
171 * hardware time stamp
172 * @flags: all shared_tx flags
173 *
174 * These flags are attached to packets as part of the
175 * &skb_shared_info. Use skb_tx() to get a pointer.
176 */
177union skb_shared_tx {
178 struct {
179 __u8 hardware:1,
180 software:1,
181 in_progress:1;
182 };
183 __u8 flags;
184};
185
186/* This data is invariant across clones and lives at
187 * the end of the header data, ie. at skb->end.
188 */
189struct skb_shared_info {
190 atomic_t dataref;
191 unsigned short nr_frags;
192 unsigned short gso_size;
193#ifdef CONFIG_HAS_DMA
194 dma_addr_t dma_head;
195#endif
196 /* Warning: this field is not always filled in (UFO)! */
197 unsigned short gso_segs;
198 unsigned short gso_type;
199 __be32 ip6_frag_id;
200 union skb_shared_tx tx_flags;
201 struct sk_buff *frag_list;
202 struct skb_shared_hwtstamps hwtstamps;
203 skb_frag_t frags[MAX_SKB_FRAGS];
204#ifdef CONFIG_HAS_DMA
205 dma_addr_t dma_maps[MAX_SKB_FRAGS];
206#endif
207 /* Intermediate layers must ensure that destructor_arg
208 * remains valid until skb destructor */
209 void * destructor_arg;
210};
211
212/* We divide dataref into two halves. The higher 16 bits hold references
213 * to the payload part of skb->data. The lower 16 bits hold references to
214 * the entire skb->data. A clone of a headerless skb holds the length of
215 * the header in skb->hdr_len.
216 *
217 * All users must obey the rule that the skb->data reference count must be
218 * greater than or equal to the payload reference count.
219 *
220 * Holding a reference to the payload part means that the user does not
221 * care about modifications to the header part of skb->data.
222 */
223#define SKB_DATAREF_SHIFT 16
224#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
225
226
227enum {
228 SKB_FCLONE_UNAVAILABLE,
229 SKB_FCLONE_ORIG,
230 SKB_FCLONE_CLONE,
231};
232
233enum {
234 SKB_GSO_TCPV4 = 1 << 0,
235 SKB_GSO_UDP = 1 << 1,
236
237 /* This indicates the skb is from an untrusted source. */
238 SKB_GSO_DODGY = 1 << 2,
239
240 /* This indicates the tcp segment has CWR set. */
241 SKB_GSO_TCP_ECN = 1 << 3,
242
243 SKB_GSO_TCPV6 = 1 << 4,
244
245 SKB_GSO_FCOE = 1 << 5,
246};
247
248#if BITS_PER_LONG > 32
249#define NET_SKBUFF_DATA_USES_OFFSET 1
250#endif
251
252#ifdef NET_SKBUFF_DATA_USES_OFFSET
253typedef unsigned int sk_buff_data_t;
254#else
255typedef unsigned char *sk_buff_data_t;
256#endif
257
258/**
259 * struct sk_buff - socket buffer
260 * @next: Next buffer in list
261 * @prev: Previous buffer in list
262 * @sk: Socket we are owned by
263 * @tstamp: Time we arrived
264 * @dev: Device we arrived on/are leaving by
265 * @transport_header: Transport layer header
266 * @network_header: Network layer header
267 * @mac_header: Link layer header
268 * @_skb_dst: destination entry
269 * @sp: the security path, used for xfrm
270 * @cb: Control buffer. Free for use by every layer. Put private vars here
271 * @len: Length of actual data
272 * @data_len: Data length
273 * @mac_len: Length of link layer header
274 * @hdr_len: writable header length of cloned skb
275 * @csum: Checksum (must include start/offset pair)
276 * @csum_start: Offset from skb->head where checksumming should start
277 * @csum_offset: Offset from csum_start where checksum should be stored
278 * @local_df: allow local fragmentation
279 * @cloned: Head may be cloned (check refcnt to be sure)
280 * @nohdr: Payload reference only, must not modify header
281 * @pkt_type: Packet class
282 * @fclone: skbuff clone status
283 * @ip_summed: Driver fed us an IP checksum
284 * @priority: Packet queueing priority
285 * @users: User count - see {datagram,tcp}.c
286 * @protocol: Packet protocol from driver
287 * @truesize: Buffer size
288 * @head: Head of buffer
289 * @data: Data head pointer
290 * @tail: Tail pointer
291 * @end: End pointer
292 * @destructor: Destruct function
293 * @mark: Generic packet mark
294 * @nfct: Associated connection, if any
295 * @ipvs_property: skbuff is owned by ipvs
296 * @peeked: this packet has been seen already, so stats have been
297 * done for it, don't do them again
298 * @nf_trace: netfilter packet trace flag
299 * @nfctinfo: Relationship of this skb to the connection
300 * @nfct_reasm: netfilter conntrack re-assembly pointer
301 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
302 * @skb_iif: ifindex of device we arrived on
303 * @queue_mapping: Queue mapping for multiqueue devices
304 * @tc_index: Traffic control index
305 * @tc_verd: traffic control verdict
306 * @ndisc_nodetype: router type (from link layer)
307 * @dma_cookie: a cookie to one of several possible DMA operations
308 * done by skb DMA functions
309 * @secmark: security marking
310 * @vlan_tci: vlan tag control information
311 */
312
313struct sk_buff {
314 /* These two members must be first. */
315 struct sk_buff *next;
316 struct sk_buff *prev;
317
318 ktime_t tstamp;
319
320 struct sock *sk;
321 struct net_device *dev;
322
323 /*
324 * This is the control buffer. It is free to use for every
325 * layer. Please put your private variables there. If you
326 * want to keep them across layers you have to do a skb_clone()
327 * first. This is owned by whoever has the skb queued ATM.
328 */
329 char cb[48] __aligned(8);
330
331 unsigned long _skb_dst;
332#ifdef CONFIG_XFRM
333 struct sec_path *sp;
334#endif
335 unsigned int len,
336 data_len;
337 __u16 mac_len,
338 hdr_len;
339 union {
340 __wsum csum;
341 struct {
342 __u16 csum_start;
343 __u16 csum_offset;
344 };
345 };
346 __u32 priority;
347 kmemcheck_bitfield_begin(flags1);
348 __u8 local_df:1,
349 cloned:1,
350 ip_summed:2,
351 nohdr:1,
352 nfctinfo:3;
353 __u8 pkt_type:3,
354 fclone:2,
355 ipvs_property:1,
356 peeked:1,
357 nf_trace:1;
358 __be16 protocol:16;
359 kmemcheck_bitfield_end(flags1);
360
361 void (*destructor)(struct sk_buff *skb);
362#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
363 struct nf_conntrack *nfct;
364 struct sk_buff *nfct_reasm;
365#endif
366#ifdef CONFIG_BRIDGE_NETFILTER
367 struct nf_bridge_info *nf_bridge;
368#endif
369
370 int skb_iif;
371#ifdef CONFIG_NET_SCHED
372 __u16 tc_index; /* traffic control index */
373#ifdef CONFIG_NET_CLS_ACT
374 __u16 tc_verd; /* traffic control verdict */
375#endif
376#endif
377
378 kmemcheck_bitfield_begin(flags2);
379 __u16 queue_mapping:16;
380#ifdef CONFIG_IPV6_NDISC_NODETYPE
381 __u8 ndisc_nodetype:2;
382#endif
383 kmemcheck_bitfield_end(flags2);
384
385 /* 0/14 bit hole */
386
387#ifdef CONFIG_NET_DMA
388 dma_cookie_t dma_cookie;
389#endif
390#ifdef CONFIG_NETWORK_SECMARK
391 __u32 secmark;
392#endif
393 union {
394 __u32 mark;
395 __u32 dropcount;
396 };
397
398 __u16 vlan_tci;
399
400 sk_buff_data_t transport_header;
401 sk_buff_data_t network_header;
402 sk_buff_data_t mac_header;
403 /* These elements must be at the end, see alloc_skb() for details. */
404 sk_buff_data_t tail;
405 sk_buff_data_t end;
406 unsigned char *head,
407 *data;
408 unsigned int truesize;
409 atomic_t users;
410};
411
412#ifdef __KERNEL__
413/*
414 * Handling routines are only of interest to the kernel
415 */
416#include <linux/slab.h>
417
418#include <asm/system.h>
419
420static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
421{
422 return (struct dst_entry *)skb->_skb_dst;
423}
424
425static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
426{
427 skb->_skb_dst = (unsigned long)dst;
428}
429
430static inline struct rtable *skb_rtable(const struct sk_buff *skb)
431{
432 return (struct rtable *)skb_dst(skb);
433}
434
435extern void kfree_skb(struct sk_buff *skb);
436extern void consume_skb(struct sk_buff *skb);
437extern void __kfree_skb(struct sk_buff *skb);
438extern struct sk_buff *__alloc_skb(unsigned int size,
439 gfp_t priority, int fclone, int node);
440static inline struct sk_buff *alloc_skb(unsigned int size,
441 gfp_t priority)
442{
443 return __alloc_skb(size, priority, 0, -1);
444}
445
446static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
447 gfp_t priority)
448{
449 return __alloc_skb(size, priority, 1, -1);
450}
451
452extern int skb_recycle_check(struct sk_buff *skb, int skb_size);
453
454extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
455extern struct sk_buff *skb_clone(struct sk_buff *skb,
456 gfp_t priority);
457extern struct sk_buff *skb_copy(const struct sk_buff *skb,
458 gfp_t priority);
459extern struct sk_buff *pskb_copy(struct sk_buff *skb,
460 gfp_t gfp_mask);
461extern int pskb_expand_head(struct sk_buff *skb,
462 int nhead, int ntail,
463 gfp_t gfp_mask);
464extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
465 unsigned int headroom);
466extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
467 int newheadroom, int newtailroom,
468 gfp_t priority);
469extern int skb_to_sgvec(struct sk_buff *skb,
470 struct scatterlist *sg, int offset,
471 int len);
472extern int skb_cow_data(struct sk_buff *skb, int tailbits,
473 struct sk_buff **trailer);
474extern int skb_pad(struct sk_buff *skb, int pad);
475#define dev_kfree_skb(a) consume_skb(a)
476#define dev_consume_skb(a) kfree_skb_clean(a)
477extern void skb_over_panic(struct sk_buff *skb, int len,
478 void *here);
479extern void skb_under_panic(struct sk_buff *skb, int len,
480 void *here);
481
482extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
483 int getfrag(void *from, char *to, int offset,
484 int len,int odd, struct sk_buff *skb),
485 void *from, int length);
486
487struct skb_seq_state {
488 __u32 lower_offset;
489 __u32 upper_offset;
490 __u32 frag_idx;
491 __u32 stepped_offset;
492 struct sk_buff *root_skb;
493 struct sk_buff *cur_skb;
494 __u8 *frag_data;
495};
496
497extern void skb_prepare_seq_read(struct sk_buff *skb,
498 unsigned int from, unsigned int to,
499 struct skb_seq_state *st);
500extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
501 struct skb_seq_state *st);
502extern void skb_abort_seq_read(struct skb_seq_state *st);
503
504extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
505 unsigned int to, struct ts_config *config,
506 struct ts_state *state);
507
508#ifdef NET_SKBUFF_DATA_USES_OFFSET
509static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
510{
511 return skb->head + skb->end;
512}
513#else
514static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
515{
516 return skb->end;
517}
518#endif
519
520/* Internal */
521#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
522
523static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
524{
525 return &skb_shinfo(skb)->hwtstamps;
526}
527
528static inline union skb_shared_tx *skb_tx(struct sk_buff *skb)
529{
530 return &skb_shinfo(skb)->tx_flags;
531}
532
533/**
534 * skb_queue_empty - check if a queue is empty
535 * @list: queue head
536 *
537 * Returns true if the queue is empty, false otherwise.
538 */
539static inline int skb_queue_empty(const struct sk_buff_head *list)
540{
541 return list->next == (struct sk_buff *)list;
542}
543
544/**
545 * skb_queue_is_last - check if skb is the last entry in the queue
546 * @list: queue head
547 * @skb: buffer
548 *
549 * Returns true if @skb is the last buffer on the list.
550 */
551static inline bool skb_queue_is_last(const struct sk_buff_head *list,
552 const struct sk_buff *skb)
553{
554 return (skb->next == (struct sk_buff *) list);
555}
556
557/**
558 * skb_queue_is_first - check if skb is the first entry in the queue
559 * @list: queue head
560 * @skb: buffer
561 *
562 * Returns true if @skb is the first buffer on the list.
563 */
564static inline bool skb_queue_is_first(const struct sk_buff_head *list,
565 const struct sk_buff *skb)
566{
567 return (skb->prev == (struct sk_buff *) list);
568}
569
570/**
571 * skb_queue_next - return the next packet in the queue
572 * @list: queue head
573 * @skb: current buffer
574 *
575 * Return the next packet in @list after @skb. It is only valid to
576 * call this if skb_queue_is_last() evaluates to false.
577 */
578static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
579 const struct sk_buff *skb)
580{
581 /* This BUG_ON may seem severe, but if we just return then we
582 * are going to dereference garbage.
583 */
584 BUG_ON(skb_queue_is_last(list, skb));
585 return skb->next;
586}
587
588/**
589 * skb_queue_prev - return the prev packet in the queue
590 * @list: queue head
591 * @skb: current buffer
592 *
593 * Return the prev packet in @list before @skb. It is only valid to
594 * call this if skb_queue_is_first() evaluates to false.
595 */
596static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
597 const struct sk_buff *skb)
598{
599 /* This BUG_ON may seem severe, but if we just return then we
600 * are going to dereference garbage.
601 */
602 BUG_ON(skb_queue_is_first(list, skb));
603 return skb->prev;
604}
605
606/**
607 * skb_get - reference buffer
608 * @skb: buffer to reference
609 *
610 * Makes another reference to a socket buffer and returns a pointer
611 * to the buffer.
612 */
613static inline struct sk_buff *skb_get(struct sk_buff *skb)
614{
615 atomic_inc(&skb->users);
616 return skb;
617}
618
619/*
620 * If users == 1, we are the only owner and are can avoid redundant
621 * atomic change.
622 */
623
624/**
625 * skb_cloned - is the buffer a clone
626 * @skb: buffer to check
627 *
628 * Returns true if the buffer was generated with skb_clone() and is
629 * one of multiple shared copies of the buffer. Cloned buffers are
630 * shared data so must not be written to under normal circumstances.
631 */
632static inline int skb_cloned(const struct sk_buff *skb)
633{
634 return skb->cloned &&
635 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
636}
637
638/**
639 * skb_header_cloned - is the header a clone
640 * @skb: buffer to check
641 *
642 * Returns true if modifying the header part of the buffer requires
643 * the data to be copied.
644 */
645static inline int skb_header_cloned(const struct sk_buff *skb)
646{
647 int dataref;
648
649 if (!skb->cloned)
650 return 0;
651
652 dataref = atomic_read(&skb_shinfo(skb)->dataref);
653 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
654 return dataref != 1;
655}
656
657/**
658 * skb_header_release - release reference to header
659 * @skb: buffer to operate on
660 *
661 * Drop a reference to the header part of the buffer. This is done
662 * by acquiring a payload reference. You must not read from the header
663 * part of skb->data after this.
664 */
665static inline void skb_header_release(struct sk_buff *skb)
666{
667 BUG_ON(skb->nohdr);
668 skb->nohdr = 1;
669 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
670}
671
672/**
673 * skb_shared - is the buffer shared
674 * @skb: buffer to check
675 *
676 * Returns true if more than one person has a reference to this
677 * buffer.
678 */
679static inline int skb_shared(const struct sk_buff *skb)
680{
681 return atomic_read(&skb->users) != 1;
682}
683
684/**
685 * skb_share_check - check if buffer is shared and if so clone it
686 * @skb: buffer to check
687 * @pri: priority for memory allocation
688 *
689 * If the buffer is shared the buffer is cloned and the old copy
690 * drops a reference. A new clone with a single reference is returned.
691 * If the buffer is not shared the original buffer is returned. When
692 * being called from interrupt status or with spinlocks held pri must
693 * be GFP_ATOMIC.
694 *
695 * NULL is returned on a memory allocation failure.
696 */
697static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
698 gfp_t pri)
699{
700 might_sleep_if(pri & __GFP_WAIT);
701 if (skb_shared(skb)) {
702 struct sk_buff *nskb = skb_clone(skb, pri);
703 kfree_skb(skb);
704 skb = nskb;
705 }
706 return skb;
707}
708
709/*
710 * Copy shared buffers into a new sk_buff. We effectively do COW on
711 * packets to handle cases where we have a local reader and forward
712 * and a couple of other messy ones. The normal one is tcpdumping
713 * a packet thats being forwarded.
714 */
715
716/**
717 * skb_unshare - make a copy of a shared buffer
718 * @skb: buffer to check
719 * @pri: priority for memory allocation
720 *
721 * If the socket buffer is a clone then this function creates a new
722 * copy of the data, drops a reference count on the old copy and returns
723 * the new copy with the reference count at 1. If the buffer is not a clone
724 * the original buffer is returned. When called with a spinlock held or
725 * from interrupt state @pri must be %GFP_ATOMIC
726 *
727 * %NULL is returned on a memory allocation failure.
728 */
729static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
730 gfp_t pri)
731{
732 might_sleep_if(pri & __GFP_WAIT);
733 if (skb_cloned(skb)) {
734 struct sk_buff *nskb = skb_copy(skb, pri);
735 kfree_skb(skb); /* Free our shared copy */
736 skb = nskb;
737 }
738 return skb;
739}
740
741/**
742 * skb_peek - peek at the head of an &sk_buff_head
743 * @list_: list to peek at
744 *
745 * Peek an &sk_buff. Unlike most other operations you _MUST_
746 * be careful with this one. A peek leaves the buffer on the
747 * list and someone else may run off with it. You must hold
748 * the appropriate locks or have a private queue to do this.
749 *
750 * Returns %NULL for an empty list or a pointer to the head element.
751 * The reference count is not incremented and the reference is therefore
752 * volatile. Use with caution.
753 */
754static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
755{
756 struct sk_buff *list = ((struct sk_buff *)list_)->next;
757 if (list == (struct sk_buff *)list_)
758 list = NULL;
759 return list;
760}
761
762/**
763 * skb_peek_tail - peek at the tail of an &sk_buff_head
764 * @list_: list to peek at
765 *
766 * Peek an &sk_buff. Unlike most other operations you _MUST_
767 * be careful with this one. A peek leaves the buffer on the
768 * list and someone else may run off with it. You must hold
769 * the appropriate locks or have a private queue to do this.
770 *
771 * Returns %NULL for an empty list or a pointer to the tail element.
772 * The reference count is not incremented and the reference is therefore
773 * volatile. Use with caution.
774 */
775static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
776{
777 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
778 if (list == (struct sk_buff *)list_)
779 list = NULL;
780 return list;
781}
782
783/**
784 * skb_queue_len - get queue length
785 * @list_: list to measure
786 *
787 * Return the length of an &sk_buff queue.
788 */
789static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
790{
791 return list_->qlen;
792}
793
794/**
795 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
796 * @list: queue to initialize
797 *
798 * This initializes only the list and queue length aspects of
799 * an sk_buff_head object. This allows to initialize the list
800 * aspects of an sk_buff_head without reinitializing things like
801 * the spinlock. It can also be used for on-stack sk_buff_head
802 * objects where the spinlock is known to not be used.
803 */
804static inline void __skb_queue_head_init(struct sk_buff_head *list)
805{
806 list->prev = list->next = (struct sk_buff *)list;
807 list->qlen = 0;
808}
809
810/*
811 * This function creates a split out lock class for each invocation;
812 * this is needed for now since a whole lot of users of the skb-queue
813 * infrastructure in drivers have different locking usage (in hardirq)
814 * than the networking core (in softirq only). In the long run either the
815 * network layer or drivers should need annotation to consolidate the
816 * main types of usage into 3 classes.
817 */
818static inline void skb_queue_head_init(struct sk_buff_head *list)
819{
820 spin_lock_init(&list->lock);
821 __skb_queue_head_init(list);
822}
823
824static inline void skb_queue_head_init_class(struct sk_buff_head *list,
825 struct lock_class_key *class)
826{
827 skb_queue_head_init(list);
828 lockdep_set_class(&list->lock, class);
829}
830
831/*
832 * Insert an sk_buff on a list.
833 *
834 * The "__skb_xxxx()" functions are the non-atomic ones that
835 * can only be called with interrupts disabled.
836 */
837extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
838static inline void __skb_insert(struct sk_buff *newsk,
839 struct sk_buff *prev, struct sk_buff *next,
840 struct sk_buff_head *list)
841{
842 newsk->next = next;
843 newsk->prev = prev;
844 next->prev = prev->next = newsk;
845 list->qlen++;
846}
847
848static inline void __skb_queue_splice(const struct sk_buff_head *list,
849 struct sk_buff *prev,
850 struct sk_buff *next)
851{
852 struct sk_buff *first = list->next;
853 struct sk_buff *last = list->prev;
854
855 first->prev = prev;
856 prev->next = first;
857
858 last->next = next;
859 next->prev = last;
860}
861
862/**
863 * skb_queue_splice - join two skb lists, this is designed for stacks
864 * @list: the new list to add
865 * @head: the place to add it in the first list
866 */
867static inline void skb_queue_splice(const struct sk_buff_head *list,
868 struct sk_buff_head *head)
869{
870 if (!skb_queue_empty(list)) {
871 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
872 head->qlen += list->qlen;
873 }
874}
875
876/**
877 * skb_queue_splice - join two skb lists and reinitialise the emptied list
878 * @list: the new list to add
879 * @head: the place to add it in the first list
880 *
881 * The list at @list is reinitialised
882 */
883static inline void skb_queue_splice_init(struct sk_buff_head *list,
884 struct sk_buff_head *head)
885{
886 if (!skb_queue_empty(list)) {
887 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
888 head->qlen += list->qlen;
889 __skb_queue_head_init(list);
890 }
891}
892
893/**
894 * skb_queue_splice_tail - join two skb lists, each list being a queue
895 * @list: the new list to add
896 * @head: the place to add it in the first list
897 */
898static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
899 struct sk_buff_head *head)
900{
901 if (!skb_queue_empty(list)) {
902 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
903 head->qlen += list->qlen;
904 }
905}
906
907/**
908 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
909 * @list: the new list to add
910 * @head: the place to add it in the first list
911 *
912 * Each of the lists is a queue.
913 * The list at @list is reinitialised
914 */
915static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
916 struct sk_buff_head *head)
917{
918 if (!skb_queue_empty(list)) {
919 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
920 head->qlen += list->qlen;
921 __skb_queue_head_init(list);
922 }
923}
924
925/**
926 * __skb_queue_after - queue a buffer at the list head
927 * @list: list to use
928 * @prev: place after this buffer
929 * @newsk: buffer to queue
930 *
931 * Queue a buffer int the middle of a list. This function takes no locks
932 * and you must therefore hold required locks before calling it.
933 *
934 * A buffer cannot be placed on two lists at the same time.
935 */
936static inline void __skb_queue_after(struct sk_buff_head *list,
937 struct sk_buff *prev,
938 struct sk_buff *newsk)
939{
940 __skb_insert(newsk, prev, prev->next, list);
941}
942
943extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
944 struct sk_buff_head *list);
945
946static inline void __skb_queue_before(struct sk_buff_head *list,
947 struct sk_buff *next,
948 struct sk_buff *newsk)
949{
950 __skb_insert(newsk, next->prev, next, list);
951}
952
953/**
954 * __skb_queue_head - queue a buffer at the list head
955 * @list: list to use
956 * @newsk: buffer to queue
957 *
958 * Queue a buffer at the start of a list. This function takes no locks
959 * and you must therefore hold required locks before calling it.
960 *
961 * A buffer cannot be placed on two lists at the same time.
962 */
963extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
964static inline void __skb_queue_head(struct sk_buff_head *list,
965 struct sk_buff *newsk)
966{
967 __skb_queue_after(list, (struct sk_buff *)list, newsk);
968}
969
970/**
971 * __skb_queue_tail - queue a buffer at the list tail
972 * @list: list to use
973 * @newsk: buffer to queue
974 *
975 * Queue a buffer at the end of a list. This function takes no locks
976 * and you must therefore hold required locks before calling it.
977 *
978 * A buffer cannot be placed on two lists at the same time.
979 */
980extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
981static inline void __skb_queue_tail(struct sk_buff_head *list,
982 struct sk_buff *newsk)
983{
984 __skb_queue_before(list, (struct sk_buff *)list, newsk);
985}
986
987/*
988 * remove sk_buff from list. _Must_ be called atomically, and with
989 * the list known..
990 */
991extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
992static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
993{
994 struct sk_buff *next, *prev;
995
996 list->qlen--;
997 next = skb->next;
998 prev = skb->prev;
999 skb->next = skb->prev = NULL;
1000 next->prev = prev;
1001 prev->next = next;
1002}
1003
1004/**
1005 * __skb_dequeue - remove from the head of the queue
1006 * @list: list to dequeue from
1007 *
1008 * Remove the head of the list. This function does not take any locks
1009 * so must be used with appropriate locks held only. The head item is
1010 * returned or %NULL if the list is empty.
1011 */
1012extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1013static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1014{
1015 struct sk_buff *skb = skb_peek(list);
1016 if (skb)
1017 __skb_unlink(skb, list);
1018 return skb;
1019}
1020
1021/**
1022 * __skb_dequeue_tail - remove from the tail of the queue
1023 * @list: list to dequeue from
1024 *
1025 * Remove the tail of the list. This function does not take any locks
1026 * so must be used with appropriate locks held only. The tail item is
1027 * returned or %NULL if the list is empty.
1028 */
1029extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1030static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1031{
1032 struct sk_buff *skb = skb_peek_tail(list);
1033 if (skb)
1034 __skb_unlink(skb, list);
1035 return skb;
1036}
1037
1038
1039static inline int skb_is_nonlinear(const struct sk_buff *skb)
1040{
1041 return skb->data_len;
1042}
1043
1044static inline unsigned int skb_headlen(const struct sk_buff *skb)
1045{
1046 return skb->len - skb->data_len;
1047}
1048
1049static inline int skb_pagelen(const struct sk_buff *skb)
1050{
1051 int i, len = 0;
1052
1053 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1054 len += skb_shinfo(skb)->frags[i].size;
1055 return len + skb_headlen(skb);
1056}
1057
1058static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1059 struct page *page, int off, int size)
1060{
1061 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1062
1063 frag->page = page;
1064 frag->page_offset = off;
1065 frag->size = size;
1066 skb_shinfo(skb)->nr_frags = i + 1;
1067}
1068
1069extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1070 int off, int size);
1071
1072#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1073#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frags(skb))
1074#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1075
1076#ifdef NET_SKBUFF_DATA_USES_OFFSET
1077static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1078{
1079 return skb->head + skb->tail;
1080}
1081
1082static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1083{
1084 skb->tail = skb->data - skb->head;
1085}
1086
1087static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1088{
1089 skb_reset_tail_pointer(skb);
1090 skb->tail += offset;
1091}
1092#else /* NET_SKBUFF_DATA_USES_OFFSET */
1093static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1094{
1095 return skb->tail;
1096}
1097
1098static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1099{
1100 skb->tail = skb->data;
1101}
1102
1103static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1104{
1105 skb->tail = skb->data + offset;
1106}
1107
1108#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1109
1110/*
1111 * Add data to an sk_buff
1112 */
1113extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1114static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1115{
1116 unsigned char *tmp = skb_tail_pointer(skb);
1117 SKB_LINEAR_ASSERT(skb);
1118 skb->tail += len;
1119 skb->len += len;
1120 return tmp;
1121}
1122
1123extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1124static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1125{
1126 skb->data -= len;
1127 skb->len += len;
1128 return skb->data;
1129}
1130
1131extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1132static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1133{
1134 skb->len -= len;
1135 BUG_ON(skb->len < skb->data_len);
1136 return skb->data += len;
1137}
1138
1139extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1140
1141static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1142{
1143 if (len > skb_headlen(skb) &&
1144 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1145 return NULL;
1146 skb->len -= len;
1147 return skb->data += len;
1148}
1149
1150static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1151{
1152 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1153}
1154
1155static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1156{
1157 if (likely(len <= skb_headlen(skb)))
1158 return 1;
1159 if (unlikely(len > skb->len))
1160 return 0;
1161 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1162}
1163
1164/**
1165 * skb_headroom - bytes at buffer head
1166 * @skb: buffer to check
1167 *
1168 * Return the number of bytes of free space at the head of an &sk_buff.
1169 */
1170static inline unsigned int skb_headroom(const struct sk_buff *skb)
1171{
1172 return skb->data - skb->head;
1173}
1174
1175/**
1176 * skb_tailroom - bytes at buffer end
1177 * @skb: buffer to check
1178 *
1179 * Return the number of bytes of free space at the tail of an sk_buff
1180 */
1181static inline int skb_tailroom(const struct sk_buff *skb)
1182{
1183 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1184}
1185
1186/**
1187 * skb_reserve - adjust headroom
1188 * @skb: buffer to alter
1189 * @len: bytes to move
1190 *
1191 * Increase the headroom of an empty &sk_buff by reducing the tail
1192 * room. This is only allowed for an empty buffer.
1193 */
1194static inline void skb_reserve(struct sk_buff *skb, int len)
1195{
1196 skb->data += len;
1197 skb->tail += len;
1198}
1199
1200#ifdef NET_SKBUFF_DATA_USES_OFFSET
1201static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1202{
1203 return skb->head + skb->transport_header;
1204}
1205
1206static inline void skb_reset_transport_header(struct sk_buff *skb)
1207{
1208 skb->transport_header = skb->data - skb->head;
1209}
1210
1211static inline void skb_set_transport_header(struct sk_buff *skb,
1212 const int offset)
1213{
1214 skb_reset_transport_header(skb);
1215 skb->transport_header += offset;
1216}
1217
1218static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1219{
1220 return skb->head + skb->network_header;
1221}
1222
1223static inline void skb_reset_network_header(struct sk_buff *skb)
1224{
1225 skb->network_header = skb->data - skb->head;
1226}
1227
1228static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1229{
1230 skb_reset_network_header(skb);
1231 skb->network_header += offset;
1232}
1233
1234static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1235{
1236 return skb->head + skb->mac_header;
1237}
1238
1239static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1240{
1241 return skb->mac_header != ~0U;
1242}
1243
1244static inline void skb_reset_mac_header(struct sk_buff *skb)
1245{
1246 skb->mac_header = skb->data - skb->head;
1247}
1248
1249static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1250{
1251 skb_reset_mac_header(skb);
1252 skb->mac_header += offset;
1253}
1254
1255#else /* NET_SKBUFF_DATA_USES_OFFSET */
1256
1257static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1258{
1259 return skb->transport_header;
1260}
1261
1262static inline void skb_reset_transport_header(struct sk_buff *skb)
1263{
1264 skb->transport_header = skb->data;
1265}
1266
1267static inline void skb_set_transport_header(struct sk_buff *skb,
1268 const int offset)
1269{
1270 skb->transport_header = skb->data + offset;
1271}
1272
1273static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1274{
1275 return skb->network_header;
1276}
1277
1278static inline void skb_reset_network_header(struct sk_buff *skb)
1279{
1280 skb->network_header = skb->data;
1281}
1282
1283static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1284{
1285 skb->network_header = skb->data + offset;
1286}
1287
1288static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1289{
1290 return skb->mac_header;
1291}
1292
1293static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1294{
1295 return skb->mac_header != NULL;
1296}
1297
1298static inline void skb_reset_mac_header(struct sk_buff *skb)
1299{
1300 skb->mac_header = skb->data;
1301}
1302
1303static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1304{
1305 skb->mac_header = skb->data + offset;
1306}
1307#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1308
1309static inline int skb_transport_offset(const struct sk_buff *skb)
1310{
1311 return skb_transport_header(skb) - skb->data;
1312}
1313
1314static inline u32 skb_network_header_len(const struct sk_buff *skb)
1315{
1316 return skb->transport_header - skb->network_header;
1317}
1318
1319static inline int skb_network_offset(const struct sk_buff *skb)
1320{
1321 return skb_network_header(skb) - skb->data;
1322}
1323
1324/*
1325 * CPUs often take a performance hit when accessing unaligned memory
1326 * locations. The actual performance hit varies, it can be small if the
1327 * hardware handles it or large if we have to take an exception and fix it
1328 * in software.
1329 *
1330 * Since an ethernet header is 14 bytes network drivers often end up with
1331 * the IP header at an unaligned offset. The IP header can be aligned by
1332 * shifting the start of the packet by 2 bytes. Drivers should do this
1333 * with:
1334 *
1335 * skb_reserve(skb, NET_IP_ALIGN);
1336 *
1337 * The downside to this alignment of the IP header is that the DMA is now
1338 * unaligned. On some architectures the cost of an unaligned DMA is high
1339 * and this cost outweighs the gains made by aligning the IP header.
1340 *
1341 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1342 * to be overridden.
1343 */
1344#ifndef NET_IP_ALIGN
1345#define NET_IP_ALIGN 2
1346#endif
1347
1348/*
1349 * The networking layer reserves some headroom in skb data (via
1350 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1351 * the header has to grow. In the default case, if the header has to grow
1352 * 32 bytes or less we avoid the reallocation.
1353 *
1354 * Unfortunately this headroom changes the DMA alignment of the resulting
1355 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1356 * on some architectures. An architecture can override this value,
1357 * perhaps setting it to a cacheline in size (since that will maintain
1358 * cacheline alignment of the DMA). It must be a power of 2.
1359 *
1360 * Various parts of the networking layer expect at least 32 bytes of
1361 * headroom, you should not reduce this.
1362 */
1363#ifndef NET_SKB_PAD
1364#define NET_SKB_PAD 32
1365#endif
1366
1367extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1368
1369static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1370{
1371 if (unlikely(skb->data_len)) {
1372 WARN_ON(1);
1373 return;
1374 }
1375 skb->len = len;
1376 skb_set_tail_pointer(skb, len);
1377}
1378
1379extern void skb_trim(struct sk_buff *skb, unsigned int len);
1380
1381static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1382{
1383 if (skb->data_len)
1384 return ___pskb_trim(skb, len);
1385 __skb_trim(skb, len);
1386 return 0;
1387}
1388
1389static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1390{
1391 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1392}
1393
1394/**
1395 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1396 * @skb: buffer to alter
1397 * @len: new length
1398 *
1399 * This is identical to pskb_trim except that the caller knows that
1400 * the skb is not cloned so we should never get an error due to out-
1401 * of-memory.
1402 */
1403static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1404{
1405 int err = pskb_trim(skb, len);
1406 BUG_ON(err);
1407}
1408
1409/**
1410 * skb_orphan - orphan a buffer
1411 * @skb: buffer to orphan
1412 *
1413 * If a buffer currently has an owner then we call the owner's
1414 * destructor function and make the @skb unowned. The buffer continues
1415 * to exist but is no longer charged to its former owner.
1416 */
1417static inline void skb_orphan(struct sk_buff *skb)
1418{
1419 if (skb->destructor)
1420 skb->destructor(skb);
1421 skb->destructor = NULL;
1422 skb->sk = NULL;
1423}
1424
1425/**
1426 * __skb_queue_purge - empty a list
1427 * @list: list to empty
1428 *
1429 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1430 * the list and one reference dropped. This function does not take the
1431 * list lock and the caller must hold the relevant locks to use it.
1432 */
1433extern void skb_queue_purge(struct sk_buff_head *list);
1434static inline void __skb_queue_purge(struct sk_buff_head *list)
1435{
1436 struct sk_buff *skb;
1437 while ((skb = __skb_dequeue(list)) != NULL)
1438 kfree_skb(skb);
1439}
1440
1441/**
1442 * __dev_alloc_skb - allocate an skbuff for receiving
1443 * @length: length to allocate
1444 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1445 *
1446 * Allocate a new &sk_buff and assign it a usage count of one. The
1447 * buffer has unspecified headroom built in. Users should allocate
1448 * the headroom they think they need without accounting for the
1449 * built in space. The built in space is used for optimisations.
1450 *
1451 * %NULL is returned if there is no free memory.
1452 */
1453static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1454 gfp_t gfp_mask)
1455{
1456 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1457 if (likely(skb))
1458 skb_reserve(skb, NET_SKB_PAD);
1459 return skb;
1460}
1461
1462extern struct sk_buff *dev_alloc_skb(unsigned int length);
1463
1464extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1465 unsigned int length, gfp_t gfp_mask);
1466
1467/**
1468 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1469 * @dev: network device to receive on
1470 * @length: length to allocate
1471 *
1472 * Allocate a new &sk_buff and assign it a usage count of one. The
1473 * buffer has unspecified headroom built in. Users should allocate
1474 * the headroom they think they need without accounting for the
1475 * built in space. The built in space is used for optimisations.
1476 *
1477 * %NULL is returned if there is no free memory. Although this function
1478 * allocates memory it can be called from an interrupt.
1479 */
1480static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1481 unsigned int length)
1482{
1483 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1484}
1485
1486static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1487 unsigned int length)
1488{
1489 struct sk_buff *skb = netdev_alloc_skb(dev, length + NET_IP_ALIGN);
1490
1491 if (NET_IP_ALIGN && skb)
1492 skb_reserve(skb, NET_IP_ALIGN);
1493 return skb;
1494}
1495
1496extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1497
1498/**
1499 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1500 * @dev: network device to receive on
1501 *
1502 * Allocate a new page node local to the specified device.
1503 *
1504 * %NULL is returned if there is no free memory.
1505 */
1506static inline struct page *netdev_alloc_page(struct net_device *dev)
1507{
1508 return __netdev_alloc_page(dev, GFP_ATOMIC);
1509}
1510
1511static inline void netdev_free_page(struct net_device *dev, struct page *page)
1512{
1513 __free_page(page);
1514}
1515
1516/**
1517 * skb_clone_writable - is the header of a clone writable
1518 * @skb: buffer to check
1519 * @len: length up to which to write
1520 *
1521 * Returns true if modifying the header part of the cloned buffer
1522 * does not requires the data to be copied.
1523 */
1524static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1525{
1526 return !skb_header_cloned(skb) &&
1527 skb_headroom(skb) + len <= skb->hdr_len;
1528}
1529
1530static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1531 int cloned)
1532{
1533 int delta = 0;
1534
1535 if (headroom < NET_SKB_PAD)
1536 headroom = NET_SKB_PAD;
1537 if (headroom > skb_headroom(skb))
1538 delta = headroom - skb_headroom(skb);
1539
1540 if (delta || cloned)
1541 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1542 GFP_ATOMIC);
1543 return 0;
1544}
1545
1546/**
1547 * skb_cow - copy header of skb when it is required
1548 * @skb: buffer to cow
1549 * @headroom: needed headroom
1550 *
1551 * If the skb passed lacks sufficient headroom or its data part
1552 * is shared, data is reallocated. If reallocation fails, an error
1553 * is returned and original skb is not changed.
1554 *
1555 * The result is skb with writable area skb->head...skb->tail
1556 * and at least @headroom of space at head.
1557 */
1558static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1559{
1560 return __skb_cow(skb, headroom, skb_cloned(skb));
1561}
1562
1563/**
1564 * skb_cow_head - skb_cow but only making the head writable
1565 * @skb: buffer to cow
1566 * @headroom: needed headroom
1567 *
1568 * This function is identical to skb_cow except that we replace the
1569 * skb_cloned check by skb_header_cloned. It should be used when
1570 * you only need to push on some header and do not need to modify
1571 * the data.
1572 */
1573static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1574{
1575 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1576}
1577
1578/**
1579 * skb_padto - pad an skbuff up to a minimal size
1580 * @skb: buffer to pad
1581 * @len: minimal length
1582 *
1583 * Pads up a buffer to ensure the trailing bytes exist and are
1584 * blanked. If the buffer already contains sufficient data it
1585 * is untouched. Otherwise it is extended. Returns zero on
1586 * success. The skb is freed on error.
1587 */
1588
1589static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1590{
1591 unsigned int size = skb->len;
1592 if (likely(size >= len))
1593 return 0;
1594 return skb_pad(skb, len - size);
1595}
1596
1597static inline int skb_add_data(struct sk_buff *skb,
1598 char __user *from, int copy)
1599{
1600 const int off = skb->len;
1601
1602 if (skb->ip_summed == CHECKSUM_NONE) {
1603 int err = 0;
1604 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1605 copy, 0, &err);
1606 if (!err) {
1607 skb->csum = csum_block_add(skb->csum, csum, off);
1608 return 0;
1609 }
1610 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1611 return 0;
1612
1613 __skb_trim(skb, off);
1614 return -EFAULT;
1615}
1616
1617static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1618 struct page *page, int off)
1619{
1620 if (i) {
1621 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1622
1623 return page == frag->page &&
1624 off == frag->page_offset + frag->size;
1625 }
1626 return 0;
1627}
1628
1629static inline int __skb_linearize(struct sk_buff *skb)
1630{
1631 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1632}
1633
1634/**
1635 * skb_linearize - convert paged skb to linear one
1636 * @skb: buffer to linarize
1637 *
1638 * If there is no free memory -ENOMEM is returned, otherwise zero
1639 * is returned and the old skb data released.
1640 */
1641static inline int skb_linearize(struct sk_buff *skb)
1642{
1643 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1644}
1645
1646/**
1647 * skb_linearize_cow - make sure skb is linear and writable
1648 * @skb: buffer to process
1649 *
1650 * If there is no free memory -ENOMEM is returned, otherwise zero
1651 * is returned and the old skb data released.
1652 */
1653static inline int skb_linearize_cow(struct sk_buff *skb)
1654{
1655 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1656 __skb_linearize(skb) : 0;
1657}
1658
1659/**
1660 * skb_postpull_rcsum - update checksum for received skb after pull
1661 * @skb: buffer to update
1662 * @start: start of data before pull
1663 * @len: length of data pulled
1664 *
1665 * After doing a pull on a received packet, you need to call this to
1666 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1667 * CHECKSUM_NONE so that it can be recomputed from scratch.
1668 */
1669
1670static inline void skb_postpull_rcsum(struct sk_buff *skb,
1671 const void *start, unsigned int len)
1672{
1673 if (skb->ip_summed == CHECKSUM_COMPLETE)
1674 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1675}
1676
1677unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1678
1679/**
1680 * pskb_trim_rcsum - trim received skb and update checksum
1681 * @skb: buffer to trim
1682 * @len: new length
1683 *
1684 * This is exactly the same as pskb_trim except that it ensures the
1685 * checksum of received packets are still valid after the operation.
1686 */
1687
1688static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1689{
1690 if (likely(len >= skb->len))
1691 return 0;
1692 if (skb->ip_summed == CHECKSUM_COMPLETE)
1693 skb->ip_summed = CHECKSUM_NONE;
1694 return __pskb_trim(skb, len);
1695}
1696
1697#define skb_queue_walk(queue, skb) \
1698 for (skb = (queue)->next; \
1699 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1700 skb = skb->next)
1701
1702#define skb_queue_walk_safe(queue, skb, tmp) \
1703 for (skb = (queue)->next, tmp = skb->next; \
1704 skb != (struct sk_buff *)(queue); \
1705 skb = tmp, tmp = skb->next)
1706
1707#define skb_queue_walk_from(queue, skb) \
1708 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1709 skb = skb->next)
1710
1711#define skb_queue_walk_from_safe(queue, skb, tmp) \
1712 for (tmp = skb->next; \
1713 skb != (struct sk_buff *)(queue); \
1714 skb = tmp, tmp = skb->next)
1715
1716#define skb_queue_reverse_walk(queue, skb) \
1717 for (skb = (queue)->prev; \
1718 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1719 skb = skb->prev)
1720
1721
1722static inline bool skb_has_frags(const struct sk_buff *skb)
1723{
1724 return skb_shinfo(skb)->frag_list != NULL;
1725}
1726
1727static inline void skb_frag_list_init(struct sk_buff *skb)
1728{
1729 skb_shinfo(skb)->frag_list = NULL;
1730}
1731
1732static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1733{
1734 frag->next = skb_shinfo(skb)->frag_list;
1735 skb_shinfo(skb)->frag_list = frag;
1736}
1737
1738#define skb_walk_frags(skb, iter) \
1739 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1740
1741extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1742 int *peeked, int *err);
1743extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1744 int noblock, int *err);
1745extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1746 struct poll_table_struct *wait);
1747extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1748 int offset, struct iovec *to,
1749 int size);
1750extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1751 int hlen,
1752 struct iovec *iov);
1753extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1754 int offset,
1755 const struct iovec *from,
1756 int from_offset,
1757 int len);
1758extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1759 int offset,
1760 const struct iovec *to,
1761 int to_offset,
1762 int size);
1763extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1764extern void skb_free_datagram_locked(struct sock *sk,
1765 struct sk_buff *skb);
1766extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1767 unsigned int flags);
1768extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1769 int len, __wsum csum);
1770extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1771 void *to, int len);
1772extern int skb_store_bits(struct sk_buff *skb, int offset,
1773 const void *from, int len);
1774extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1775 int offset, u8 *to, int len,
1776 __wsum csum);
1777extern int skb_splice_bits(struct sk_buff *skb,
1778 unsigned int offset,
1779 struct pipe_inode_info *pipe,
1780 unsigned int len,
1781 unsigned int flags);
1782extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1783extern void skb_split(struct sk_buff *skb,
1784 struct sk_buff *skb1, const u32 len);
1785extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1786 int shiftlen);
1787
1788extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1789
1790static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1791 int len, void *buffer)
1792{
1793 int hlen = skb_headlen(skb);
1794
1795 if (hlen - offset >= len)
1796 return skb->data + offset;
1797
1798 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1799 return NULL;
1800
1801 return buffer;
1802}
1803
1804static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1805 void *to,
1806 const unsigned int len)
1807{
1808 memcpy(to, skb->data, len);
1809}
1810
1811static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1812 const int offset, void *to,
1813 const unsigned int len)
1814{
1815 memcpy(to, skb->data + offset, len);
1816}
1817
1818static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1819 const void *from,
1820 const unsigned int len)
1821{
1822 memcpy(skb->data, from, len);
1823}
1824
1825static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1826 const int offset,
1827 const void *from,
1828 const unsigned int len)
1829{
1830 memcpy(skb->data + offset, from, len);
1831}
1832
1833extern void skb_init(void);
1834
1835static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1836{
1837 return skb->tstamp;
1838}
1839
1840/**
1841 * skb_get_timestamp - get timestamp from a skb
1842 * @skb: skb to get stamp from
1843 * @stamp: pointer to struct timeval to store stamp in
1844 *
1845 * Timestamps are stored in the skb as offsets to a base timestamp.
1846 * This function converts the offset back to a struct timeval and stores
1847 * it in stamp.
1848 */
1849static inline void skb_get_timestamp(const struct sk_buff *skb,
1850 struct timeval *stamp)
1851{
1852 *stamp = ktime_to_timeval(skb->tstamp);
1853}
1854
1855static inline void skb_get_timestampns(const struct sk_buff *skb,
1856 struct timespec *stamp)
1857{
1858 *stamp = ktime_to_timespec(skb->tstamp);
1859}
1860
1861static inline void __net_timestamp(struct sk_buff *skb)
1862{
1863 skb->tstamp = ktime_get_real();
1864}
1865
1866static inline ktime_t net_timedelta(ktime_t t)
1867{
1868 return ktime_sub(ktime_get_real(), t);
1869}
1870
1871static inline ktime_t net_invalid_timestamp(void)
1872{
1873 return ktime_set(0, 0);
1874}
1875
1876/**
1877 * skb_tstamp_tx - queue clone of skb with send time stamps
1878 * @orig_skb: the original outgoing packet
1879 * @hwtstamps: hardware time stamps, may be NULL if not available
1880 *
1881 * If the skb has a socket associated, then this function clones the
1882 * skb (thus sharing the actual data and optional structures), stores
1883 * the optional hardware time stamping information (if non NULL) or
1884 * generates a software time stamp (otherwise), then queues the clone
1885 * to the error queue of the socket. Errors are silently ignored.
1886 */
1887extern void skb_tstamp_tx(struct sk_buff *orig_skb,
1888 struct skb_shared_hwtstamps *hwtstamps);
1889
1890extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
1891extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
1892
1893static inline int skb_csum_unnecessary(const struct sk_buff *skb)
1894{
1895 return skb->ip_summed & CHECKSUM_UNNECESSARY;
1896}
1897
1898/**
1899 * skb_checksum_complete - Calculate checksum of an entire packet
1900 * @skb: packet to process
1901 *
1902 * This function calculates the checksum over the entire packet plus
1903 * the value of skb->csum. The latter can be used to supply the
1904 * checksum of a pseudo header as used by TCP/UDP. It returns the
1905 * checksum.
1906 *
1907 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1908 * this function can be used to verify that checksum on received
1909 * packets. In that case the function should return zero if the
1910 * checksum is correct. In particular, this function will return zero
1911 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1912 * hardware has already verified the correctness of the checksum.
1913 */
1914static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
1915{
1916 return skb_csum_unnecessary(skb) ?
1917 0 : __skb_checksum_complete(skb);
1918}
1919
1920#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1921extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
1922static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1923{
1924 if (nfct && atomic_dec_and_test(&nfct->use))
1925 nf_conntrack_destroy(nfct);
1926}
1927static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1928{
1929 if (nfct)
1930 atomic_inc(&nfct->use);
1931}
1932static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1933{
1934 if (skb)
1935 atomic_inc(&skb->users);
1936}
1937static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1938{
1939 if (skb)
1940 kfree_skb(skb);
1941}
1942#endif
1943#ifdef CONFIG_BRIDGE_NETFILTER
1944static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1945{
1946 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1947 kfree(nf_bridge);
1948}
1949static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1950{
1951 if (nf_bridge)
1952 atomic_inc(&nf_bridge->use);
1953}
1954#endif /* CONFIG_BRIDGE_NETFILTER */
1955static inline void nf_reset(struct sk_buff *skb)
1956{
1957#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1958 nf_conntrack_put(skb->nfct);
1959 skb->nfct = NULL;
1960 nf_conntrack_put_reasm(skb->nfct_reasm);
1961 skb->nfct_reasm = NULL;
1962#endif
1963#ifdef CONFIG_BRIDGE_NETFILTER
1964 nf_bridge_put(skb->nf_bridge);
1965 skb->nf_bridge = NULL;
1966#endif
1967}
1968
1969/* Note: This doesn't put any conntrack and bridge info in dst. */
1970static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1971{
1972#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1973 dst->nfct = src->nfct;
1974 nf_conntrack_get(src->nfct);
1975 dst->nfctinfo = src->nfctinfo;
1976 dst->nfct_reasm = src->nfct_reasm;
1977 nf_conntrack_get_reasm(src->nfct_reasm);
1978#endif
1979#ifdef CONFIG_BRIDGE_NETFILTER
1980 dst->nf_bridge = src->nf_bridge;
1981 nf_bridge_get(src->nf_bridge);
1982#endif
1983}
1984
1985static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1986{
1987#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1988 nf_conntrack_put(dst->nfct);
1989 nf_conntrack_put_reasm(dst->nfct_reasm);
1990#endif
1991#ifdef CONFIG_BRIDGE_NETFILTER
1992 nf_bridge_put(dst->nf_bridge);
1993#endif
1994 __nf_copy(dst, src);
1995}
1996
1997#ifdef CONFIG_NETWORK_SECMARK
1998static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1999{
2000 to->secmark = from->secmark;
2001}
2002
2003static inline void skb_init_secmark(struct sk_buff *skb)
2004{
2005 skb->secmark = 0;
2006}
2007#else
2008static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2009{ }
2010
2011static inline void skb_init_secmark(struct sk_buff *skb)
2012{ }
2013#endif
2014
2015static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2016{
2017 skb->queue_mapping = queue_mapping;
2018}
2019
2020static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2021{
2022 return skb->queue_mapping;
2023}
2024
2025static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2026{
2027 to->queue_mapping = from->queue_mapping;
2028}
2029
2030static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2031{
2032 skb->queue_mapping = rx_queue + 1;
2033}
2034
2035static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2036{
2037 return skb->queue_mapping - 1;
2038}
2039
2040static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2041{
2042 return (skb->queue_mapping != 0);
2043}
2044
2045extern u16 skb_tx_hash(const struct net_device *dev,
2046 const struct sk_buff *skb);
2047
2048#ifdef CONFIG_XFRM
2049static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2050{
2051 return skb->sp;
2052}
2053#else
2054static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2055{
2056 return NULL;
2057}
2058#endif
2059
2060static inline int skb_is_gso(const struct sk_buff *skb)
2061{
2062 return skb_shinfo(skb)->gso_size;
2063}
2064
2065static inline int skb_is_gso_v6(const struct sk_buff *skb)
2066{
2067 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2068}
2069
2070extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2071
2072static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2073{
2074 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2075 * wanted then gso_type will be set. */
2076 struct skb_shared_info *shinfo = skb_shinfo(skb);
2077 if (shinfo->gso_size != 0 && unlikely(shinfo->gso_type == 0)) {
2078 __skb_warn_lro_forwarding(skb);
2079 return true;
2080 }
2081 return false;
2082}
2083
2084static inline void skb_forward_csum(struct sk_buff *skb)
2085{
2086 /* Unfortunately we don't support this one. Any brave souls? */
2087 if (skb->ip_summed == CHECKSUM_COMPLETE)
2088 skb->ip_summed = CHECKSUM_NONE;
2089}
2090
2091bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2092#endif /* __KERNEL__ */
2093#endif /* _LINUX_SKBUFF_H */