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