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