]> bbs.cooldavid.org Git - net-next-2.6.git/blame_incremental - include/linux/skbuff.h
net: add _skb_dst opaque field
[net-next-2.6.git] / include / linux / skbuff.h
... / ...
CommitLineData
1/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
17#include <linux/kernel.h>
18#include <linux/compiler.h>
19#include <linux/time.h>
20#include <linux/cache.h>
21
22#include <asm/atomic.h>
23#include <asm/types.h>
24#include <linux/spinlock.h>
25#include <linux/net.h>
26#include <linux/textsearch.h>
27#include <net/checksum.h>
28#include <linux/rcupdate.h>
29#include <linux/dmaengine.h>
30#include <linux/hrtimer.h>
31
32/* Don't change this without changing skb_csum_unnecessary! */
33#define CHECKSUM_NONE 0
34#define CHECKSUM_UNNECESSARY 1
35#define CHECKSUM_COMPLETE 2
36#define CHECKSUM_PARTIAL 3
37
38#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
39 ~(SMP_CACHE_BYTES - 1))
40#define SKB_WITH_OVERHEAD(X) \
41 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
42#define SKB_MAX_ORDER(X, ORDER) \
43 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
44#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
45#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
46
47/* A. Checksumming of received packets by device.
48 *
49 * NONE: device failed to checksum this packet.
50 * skb->csum is undefined.
51 *
52 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
53 * skb->csum is undefined.
54 * It is bad option, but, unfortunately, many of vendors do this.
55 * Apparently with secret goal to sell you new device, when you
56 * will add new protocol to your host. F.e. IPv6. 8)
57 *
58 * COMPLETE: the most generic way. Device supplied checksum of _all_
59 * the packet as seen by netif_rx in skb->csum.
60 * NOTE: Even if device supports only some protocols, but
61 * is able to produce some skb->csum, it MUST use COMPLETE,
62 * not UNNECESSARY.
63 *
64 * PARTIAL: identical to the case for output below. This may occur
65 * on a packet received directly from another Linux OS, e.g.,
66 * a virtualised Linux kernel on the same host. The packet can
67 * be treated in the same way as UNNECESSARY except that on
68 * output (i.e., forwarding) the checksum must be filled in
69 * by the OS or the hardware.
70 *
71 * B. Checksumming on output.
72 *
73 * NONE: skb is checksummed by protocol or csum is not required.
74 *
75 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
76 * from skb->csum_start to the end and to record the checksum
77 * at skb->csum_start + skb->csum_offset.
78 *
79 * Device must show its capabilities in dev->features, set
80 * at device setup time.
81 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
82 * everything.
83 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
84 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
85 * TCP/UDP over IPv4. Sigh. Vendors like this
86 * way by an unknown reason. Though, see comment above
87 * about CHECKSUM_UNNECESSARY. 8)
88 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
89 *
90 * Any questions? No questions, good. --ANK
91 */
92
93struct net_device;
94struct scatterlist;
95struct pipe_inode_info;
96
97#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
98struct nf_conntrack {
99 atomic_t use;
100};
101#endif
102
103#ifdef CONFIG_BRIDGE_NETFILTER
104struct nf_bridge_info {
105 atomic_t use;
106 struct net_device *physindev;
107 struct net_device *physoutdev;
108 unsigned int mask;
109 unsigned long data[32 / sizeof(unsigned long)];
110};
111#endif
112
113struct sk_buff_head {
114 /* These two members must be first. */
115 struct sk_buff *next;
116 struct sk_buff *prev;
117
118 __u32 qlen;
119 spinlock_t lock;
120};
121
122struct sk_buff;
123
124/* To allow 64K frame to be packed as single skb without frag_list */
125#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
126
127typedef struct skb_frag_struct skb_frag_t;
128
129struct skb_frag_struct {
130 struct page *page;
131 __u32 page_offset;
132 __u32 size;
133};
134
135#define HAVE_HW_TIME_STAMP
136
137/**
138 * struct skb_shared_hwtstamps - hardware time stamps
139 * @hwtstamp: hardware time stamp transformed into duration
140 * since arbitrary point in time
141 * @syststamp: hwtstamp transformed to system time base
142 *
143 * Software time stamps generated by ktime_get_real() are stored in
144 * skb->tstamp. The relation between the different kinds of time
145 * stamps is as follows:
146 *
147 * syststamp and tstamp can be compared against each other in
148 * arbitrary combinations. The accuracy of a
149 * syststamp/tstamp/"syststamp from other device" comparison is
150 * limited by the accuracy of the transformation into system time
151 * base. This depends on the device driver and its underlying
152 * hardware.
153 *
154 * hwtstamps can only be compared against other hwtstamps from
155 * the same device.
156 *
157 * This structure is attached to packets as part of the
158 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
159 */
160struct skb_shared_hwtstamps {
161 ktime_t hwtstamp;
162 ktime_t syststamp;
163};
164
165/**
166 * struct skb_shared_tx - instructions for time stamping of outgoing packets
167 * @hardware: generate hardware time stamp
168 * @software: generate software time stamp
169 * @in_progress: device driver is going to provide
170 * hardware time stamp
171 * @flags: all shared_tx flags
172 *
173 * These flags are attached to packets as part of the
174 * &skb_shared_info. Use skb_tx() to get a pointer.
175 */
176union skb_shared_tx {
177 struct {
178 __u8 hardware:1,
179 software:1,
180 in_progress:1;
181 };
182 __u8 flags;
183};
184
185/* This data is invariant across clones and lives at
186 * the end of the header data, ie. at skb->end.
187 */
188struct skb_shared_info {
189 atomic_t dataref;
190 unsigned short nr_frags;
191 unsigned short gso_size;
192 /* Warning: this field is not always filled in (UFO)! */
193 unsigned short gso_segs;
194 unsigned short gso_type;
195 __be32 ip6_frag_id;
196 union skb_shared_tx tx_flags;
197#ifdef CONFIG_HAS_DMA
198 unsigned int num_dma_maps;
199#endif
200 struct sk_buff *frag_list;
201 struct skb_shared_hwtstamps hwtstamps;
202 skb_frag_t frags[MAX_SKB_FRAGS];
203#ifdef CONFIG_HAS_DMA
204 dma_addr_t dma_maps[MAX_SKB_FRAGS + 1];
205#endif
206 /* Intermediate layers must ensure that destructor_arg
207 * remains valid until skb destructor */
208 void * destructor_arg;
209};
210
211/* We divide dataref into two halves. The higher 16 bits hold references
212 * to the payload part of skb->data. The lower 16 bits hold references to
213 * the entire skb->data. A clone of a headerless skb holds the length of
214 * the header in skb->hdr_len.
215 *
216 * All users must obey the rule that the skb->data reference count must be
217 * greater than or equal to the payload reference count.
218 *
219 * Holding a reference to the payload part means that the user does not
220 * care about modifications to the header part of skb->data.
221 */
222#define SKB_DATAREF_SHIFT 16
223#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
224
225
226enum {
227 SKB_FCLONE_UNAVAILABLE,
228 SKB_FCLONE_ORIG,
229 SKB_FCLONE_CLONE,
230};
231
232enum {
233 SKB_GSO_TCPV4 = 1 << 0,
234 SKB_GSO_UDP = 1 << 1,
235
236 /* This indicates the skb is from an untrusted source. */
237 SKB_GSO_DODGY = 1 << 2,
238
239 /* This indicates the tcp segment has CWR set. */
240 SKB_GSO_TCP_ECN = 1 << 3,
241
242 SKB_GSO_TCPV6 = 1 << 4,
243
244 SKB_GSO_FCOE = 1 << 5,
245};
246
247#if BITS_PER_LONG > 32
248#define NET_SKBUFF_DATA_USES_OFFSET 1
249#endif
250
251#ifdef NET_SKBUFF_DATA_USES_OFFSET
252typedef unsigned int sk_buff_data_t;
253#else
254typedef unsigned char *sk_buff_data_t;
255#endif
256
257/**
258 * struct sk_buff - socket buffer
259 * @next: Next buffer in list
260 * @prev: Previous buffer in list
261 * @sk: Socket we are owned by
262 * @tstamp: Time we arrived
263 * @dev: Device we arrived on/are leaving by
264 * @transport_header: Transport layer header
265 * @network_header: Network layer header
266 * @mac_header: Link layer header
267 * @dst: destination entry
268 * @sp: the security path, used for xfrm
269 * @cb: Control buffer. Free for use by every layer. Put private vars here
270 * @len: Length of actual data
271 * @data_len: Data length
272 * @mac_len: Length of link layer header
273 * @hdr_len: writable header length of cloned skb
274 * @csum: Checksum (must include start/offset pair)
275 * @csum_start: Offset from skb->head where checksumming should start
276 * @csum_offset: Offset from csum_start where checksum should be stored
277 * @local_df: allow local fragmentation
278 * @cloned: Head may be cloned (check refcnt to be sure)
279 * @nohdr: Payload reference only, must not modify header
280 * @pkt_type: Packet class
281 * @fclone: skbuff clone status
282 * @ip_summed: Driver fed us an IP checksum
283 * @priority: Packet queueing priority
284 * @users: User count - see {datagram,tcp}.c
285 * @protocol: Packet protocol from driver
286 * @truesize: Buffer size
287 * @head: Head of buffer
288 * @data: Data head pointer
289 * @tail: Tail pointer
290 * @end: End pointer
291 * @destructor: Destruct function
292 * @mark: Generic packet mark
293 * @nfct: Associated connection, if any
294 * @ipvs_property: skbuff is owned by ipvs
295 * @peeked: this packet has been seen already, so stats have been
296 * done for it, don't do them again
297 * @nf_trace: netfilter packet trace flag
298 * @nfctinfo: Relationship of this skb to the connection
299 * @nfct_reasm: netfilter conntrack re-assembly pointer
300 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
301 * @iif: ifindex of device we arrived on
302 * @queue_mapping: Queue mapping for multiqueue devices
303 * @tc_index: Traffic control index
304 * @tc_verd: traffic control verdict
305 * @ndisc_nodetype: router type (from link layer)
306 * @do_not_encrypt: set to prevent encryption of this frame
307 * @requeue: set to indicate that the wireless core should attempt
308 * a software retry on this frame if we failed to
309 * receive an ACK for it
310 * @dma_cookie: a cookie to one of several possible DMA operations
311 * done by skb DMA functions
312 * @secmark: security marking
313 * @vlan_tci: vlan tag control information
314 */
315
316struct sk_buff {
317 /* These two members must be first. */
318 struct sk_buff *next;
319 struct sk_buff *prev;
320
321 struct sock *sk;
322 ktime_t tstamp;
323 struct net_device *dev;
324
325 union {
326 struct dst_entry *dst;
327 struct rtable *rtable;
328 unsigned long _skb_dst;
329 };
330#ifdef CONFIG_XFRM
331 struct sec_path *sp;
332#endif
333 /*
334 * This is the control buffer. It is free to use for every
335 * layer. Please put your private variables there. If you
336 * want to keep them across layers you have to do a skb_clone()
337 * first. This is owned by whoever has the skb queued ATM.
338 */
339 char cb[48];
340
341 unsigned int len,
342 data_len;
343 __u16 mac_len,
344 hdr_len;
345 union {
346 __wsum csum;
347 struct {
348 __u16 csum_start;
349 __u16 csum_offset;
350 };
351 };
352 __u32 priority;
353 __u8 local_df:1,
354 cloned:1,
355 ip_summed:2,
356 nohdr:1,
357 nfctinfo:3;
358 __u8 pkt_type:3,
359 fclone:2,
360 ipvs_property:1,
361 peeked:1,
362 nf_trace:1;
363 __be16 protocol;
364
365 void (*destructor)(struct sk_buff *skb);
366#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
367 struct nf_conntrack *nfct;
368 struct sk_buff *nfct_reasm;
369#endif
370#ifdef CONFIG_BRIDGE_NETFILTER
371 struct nf_bridge_info *nf_bridge;
372#endif
373
374 int iif;
375 __u16 queue_mapping;
376#ifdef CONFIG_NET_SCHED
377 __u16 tc_index; /* traffic control index */
378#ifdef CONFIG_NET_CLS_ACT
379 __u16 tc_verd; /* traffic control verdict */
380#endif
381#endif
382#ifdef CONFIG_IPV6_NDISC_NODETYPE
383 __u8 ndisc_nodetype:2;
384#endif
385#if defined(CONFIG_MAC80211) || defined(CONFIG_MAC80211_MODULE)
386 __u8 do_not_encrypt:1;
387 __u8 requeue:1;
388#endif
389 /* 0/13/14 bit hole */
390
391#ifdef CONFIG_NET_DMA
392 dma_cookie_t dma_cookie;
393#endif
394#ifdef CONFIG_NETWORK_SECMARK
395 __u32 secmark;
396#endif
397
398 __u32 mark;
399
400 __u16 vlan_tci;
401
402 sk_buff_data_t transport_header;
403 sk_buff_data_t network_header;
404 sk_buff_data_t mac_header;
405 /* These elements must be at the end, see alloc_skb() for details. */
406 sk_buff_data_t tail;
407 sk_buff_data_t end;
408 unsigned char *head,
409 *data;
410 unsigned int truesize;
411 atomic_t users;
412};
413
414#ifdef __KERNEL__
415/*
416 * Handling routines are only of interest to the kernel
417 */
418#include <linux/slab.h>
419
420#include <asm/system.h>
421
422#ifdef CONFIG_HAS_DMA
423#include <linux/dma-mapping.h>
424extern int skb_dma_map(struct device *dev, struct sk_buff *skb,
425 enum dma_data_direction dir);
426extern void skb_dma_unmap(struct device *dev, struct sk_buff *skb,
427 enum dma_data_direction dir);
428#endif
429
430extern void kfree_skb(struct sk_buff *skb);
431extern void consume_skb(struct sk_buff *skb);
432extern void __kfree_skb(struct sk_buff *skb);
433extern struct sk_buff *__alloc_skb(unsigned int size,
434 gfp_t priority, int fclone, int node);
435static inline struct sk_buff *alloc_skb(unsigned int size,
436 gfp_t priority)
437{
438 return __alloc_skb(size, priority, 0, -1);
439}
440
441static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
442 gfp_t priority)
443{
444 return __alloc_skb(size, priority, 1, -1);
445}
446
447extern int skb_recycle_check(struct sk_buff *skb, int skb_size);
448
449extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
450extern struct sk_buff *skb_clone(struct sk_buff *skb,
451 gfp_t priority);
452extern struct sk_buff *skb_copy(const struct sk_buff *skb,
453 gfp_t priority);
454extern struct sk_buff *pskb_copy(struct sk_buff *skb,
455 gfp_t gfp_mask);
456extern int pskb_expand_head(struct sk_buff *skb,
457 int nhead, int ntail,
458 gfp_t gfp_mask);
459extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
460 unsigned int headroom);
461extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
462 int newheadroom, int newtailroom,
463 gfp_t priority);
464extern int skb_to_sgvec(struct sk_buff *skb,
465 struct scatterlist *sg, int offset,
466 int len);
467extern int skb_cow_data(struct sk_buff *skb, int tailbits,
468 struct sk_buff **trailer);
469extern int skb_pad(struct sk_buff *skb, int pad);
470#define dev_kfree_skb(a) consume_skb(a)
471#define dev_consume_skb(a) kfree_skb_clean(a)
472extern void skb_over_panic(struct sk_buff *skb, int len,
473 void *here);
474extern void skb_under_panic(struct sk_buff *skb, int len,
475 void *here);
476
477extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
478 int getfrag(void *from, char *to, int offset,
479 int len,int odd, struct sk_buff *skb),
480 void *from, int length);
481
482struct skb_seq_state
483{
484 __u32 lower_offset;
485 __u32 upper_offset;
486 __u32 frag_idx;
487 __u32 stepped_offset;
488 struct sk_buff *root_skb;
489 struct sk_buff *cur_skb;
490 __u8 *frag_data;
491};
492
493extern void skb_prepare_seq_read(struct sk_buff *skb,
494 unsigned int from, unsigned int to,
495 struct skb_seq_state *st);
496extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
497 struct skb_seq_state *st);
498extern void skb_abort_seq_read(struct skb_seq_state *st);
499
500extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
501 unsigned int to, struct ts_config *config,
502 struct ts_state *state);
503
504#ifdef NET_SKBUFF_DATA_USES_OFFSET
505static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
506{
507 return skb->head + skb->end;
508}
509#else
510static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
511{
512 return skb->end;
513}
514#endif
515
516/* Internal */
517#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
518
519static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
520{
521 return &skb_shinfo(skb)->hwtstamps;
522}
523
524static inline union skb_shared_tx *skb_tx(struct sk_buff *skb)
525{
526 return &skb_shinfo(skb)->tx_flags;
527}
528
529/**
530 * skb_queue_empty - check if a queue is empty
531 * @list: queue head
532 *
533 * Returns true if the queue is empty, false otherwise.
534 */
535static inline int skb_queue_empty(const struct sk_buff_head *list)
536{
537 return list->next == (struct sk_buff *)list;
538}
539
540/**
541 * skb_queue_is_last - check if skb is the last entry in the queue
542 * @list: queue head
543 * @skb: buffer
544 *
545 * Returns true if @skb is the last buffer on the list.
546 */
547static inline bool skb_queue_is_last(const struct sk_buff_head *list,
548 const struct sk_buff *skb)
549{
550 return (skb->next == (struct sk_buff *) list);
551}
552
553/**
554 * skb_queue_is_first - check if skb is the first entry in the queue
555 * @list: queue head
556 * @skb: buffer
557 *
558 * Returns true if @skb is the first buffer on the list.
559 */
560static inline bool skb_queue_is_first(const struct sk_buff_head *list,
561 const struct sk_buff *skb)
562{
563 return (skb->prev == (struct sk_buff *) list);
564}
565
566/**
567 * skb_queue_next - return the next packet in the queue
568 * @list: queue head
569 * @skb: current buffer
570 *
571 * Return the next packet in @list after @skb. It is only valid to
572 * call this if skb_queue_is_last() evaluates to false.
573 */
574static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
575 const struct sk_buff *skb)
576{
577 /* This BUG_ON may seem severe, but if we just return then we
578 * are going to dereference garbage.
579 */
580 BUG_ON(skb_queue_is_last(list, skb));
581 return skb->next;
582}
583
584/**
585 * skb_queue_prev - return the prev packet in the queue
586 * @list: queue head
587 * @skb: current buffer
588 *
589 * Return the prev packet in @list before @skb. It is only valid to
590 * call this if skb_queue_is_first() evaluates to false.
591 */
592static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
593 const struct sk_buff *skb)
594{
595 /* This BUG_ON may seem severe, but if we just return then we
596 * are going to dereference garbage.
597 */
598 BUG_ON(skb_queue_is_first(list, skb));
599 return skb->prev;
600}
601
602/**
603 * skb_get - reference buffer
604 * @skb: buffer to reference
605 *
606 * Makes another reference to a socket buffer and returns a pointer
607 * to the buffer.
608 */
609static inline struct sk_buff *skb_get(struct sk_buff *skb)
610{
611 atomic_inc(&skb->users);
612 return skb;
613}
614
615/*
616 * If users == 1, we are the only owner and are can avoid redundant
617 * atomic change.
618 */
619
620/**
621 * skb_cloned - is the buffer a clone
622 * @skb: buffer to check
623 *
624 * Returns true if the buffer was generated with skb_clone() and is
625 * one of multiple shared copies of the buffer. Cloned buffers are
626 * shared data so must not be written to under normal circumstances.
627 */
628static inline int skb_cloned(const struct sk_buff *skb)
629{
630 return skb->cloned &&
631 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
632}
633
634/**
635 * skb_header_cloned - is the header a clone
636 * @skb: buffer to check
637 *
638 * Returns true if modifying the header part of the buffer requires
639 * the data to be copied.
640 */
641static inline int skb_header_cloned(const struct sk_buff *skb)
642{
643 int dataref;
644
645 if (!skb->cloned)
646 return 0;
647
648 dataref = atomic_read(&skb_shinfo(skb)->dataref);
649 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
650 return dataref != 1;
651}
652
653/**
654 * skb_header_release - release reference to header
655 * @skb: buffer to operate on
656 *
657 * Drop a reference to the header part of the buffer. This is done
658 * by acquiring a payload reference. You must not read from the header
659 * part of skb->data after this.
660 */
661static inline void skb_header_release(struct sk_buff *skb)
662{
663 BUG_ON(skb->nohdr);
664 skb->nohdr = 1;
665 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
666}
667
668/**
669 * skb_shared - is the buffer shared
670 * @skb: buffer to check
671 *
672 * Returns true if more than one person has a reference to this
673 * buffer.
674 */
675static inline int skb_shared(const struct sk_buff *skb)
676{
677 return atomic_read(&skb->users) != 1;
678}
679
680/**
681 * skb_share_check - check if buffer is shared and if so clone it
682 * @skb: buffer to check
683 * @pri: priority for memory allocation
684 *
685 * If the buffer is shared the buffer is cloned and the old copy
686 * drops a reference. A new clone with a single reference is returned.
687 * If the buffer is not shared the original buffer is returned. When
688 * being called from interrupt status or with spinlocks held pri must
689 * be GFP_ATOMIC.
690 *
691 * NULL is returned on a memory allocation failure.
692 */
693static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
694 gfp_t pri)
695{
696 might_sleep_if(pri & __GFP_WAIT);
697 if (skb_shared(skb)) {
698 struct sk_buff *nskb = skb_clone(skb, pri);
699 kfree_skb(skb);
700 skb = nskb;
701 }
702 return skb;
703}
704
705/*
706 * Copy shared buffers into a new sk_buff. We effectively do COW on
707 * packets to handle cases where we have a local reader and forward
708 * and a couple of other messy ones. The normal one is tcpdumping
709 * a packet thats being forwarded.
710 */
711
712/**
713 * skb_unshare - make a copy of a shared buffer
714 * @skb: buffer to check
715 * @pri: priority for memory allocation
716 *
717 * If the socket buffer is a clone then this function creates a new
718 * copy of the data, drops a reference count on the old copy and returns
719 * the new copy with the reference count at 1. If the buffer is not a clone
720 * the original buffer is returned. When called with a spinlock held or
721 * from interrupt state @pri must be %GFP_ATOMIC
722 *
723 * %NULL is returned on a memory allocation failure.
724 */
725static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
726 gfp_t pri)
727{
728 might_sleep_if(pri & __GFP_WAIT);
729 if (skb_cloned(skb)) {
730 struct sk_buff *nskb = skb_copy(skb, pri);
731 kfree_skb(skb); /* Free our shared copy */
732 skb = nskb;
733 }
734 return skb;
735}
736
737/**
738 * skb_peek
739 * @list_: list to peek at
740 *
741 * Peek an &sk_buff. Unlike most other operations you _MUST_
742 * be careful with this one. A peek leaves the buffer on the
743 * list and someone else may run off with it. You must hold
744 * the appropriate locks or have a private queue to do this.
745 *
746 * Returns %NULL for an empty list or a pointer to the head element.
747 * The reference count is not incremented and the reference is therefore
748 * volatile. Use with caution.
749 */
750static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
751{
752 struct sk_buff *list = ((struct sk_buff *)list_)->next;
753 if (list == (struct sk_buff *)list_)
754 list = NULL;
755 return list;
756}
757
758/**
759 * skb_peek_tail
760 * @list_: list to peek at
761 *
762 * Peek an &sk_buff. Unlike most other operations you _MUST_
763 * be careful with this one. A peek leaves the buffer on the
764 * list and someone else may run off with it. You must hold
765 * the appropriate locks or have a private queue to do this.
766 *
767 * Returns %NULL for an empty list or a pointer to the tail element.
768 * The reference count is not incremented and the reference is therefore
769 * volatile. Use with caution.
770 */
771static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
772{
773 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
774 if (list == (struct sk_buff *)list_)
775 list = NULL;
776 return list;
777}
778
779/**
780 * skb_queue_len - get queue length
781 * @list_: list to measure
782 *
783 * Return the length of an &sk_buff queue.
784 */
785static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
786{
787 return list_->qlen;
788}
789
790/**
791 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
792 * @list: queue to initialize
793 *
794 * This initializes only the list and queue length aspects of
795 * an sk_buff_head object. This allows to initialize the list
796 * aspects of an sk_buff_head without reinitializing things like
797 * the spinlock. It can also be used for on-stack sk_buff_head
798 * objects where the spinlock is known to not be used.
799 */
800static inline void __skb_queue_head_init(struct sk_buff_head *list)
801{
802 list->prev = list->next = (struct sk_buff *)list;
803 list->qlen = 0;
804}
805
806/*
807 * This function creates a split out lock class for each invocation;
808 * this is needed for now since a whole lot of users of the skb-queue
809 * infrastructure in drivers have different locking usage (in hardirq)
810 * than the networking core (in softirq only). In the long run either the
811 * network layer or drivers should need annotation to consolidate the
812 * main types of usage into 3 classes.
813 */
814static inline void skb_queue_head_init(struct sk_buff_head *list)
815{
816 spin_lock_init(&list->lock);
817 __skb_queue_head_init(list);
818}
819
820static inline void skb_queue_head_init_class(struct sk_buff_head *list,
821 struct lock_class_key *class)
822{
823 skb_queue_head_init(list);
824 lockdep_set_class(&list->lock, class);
825}
826
827/*
828 * Insert an sk_buff on a list.
829 *
830 * The "__skb_xxxx()" functions are the non-atomic ones that
831 * can only be called with interrupts disabled.
832 */
833extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
834static inline void __skb_insert(struct sk_buff *newsk,
835 struct sk_buff *prev, struct sk_buff *next,
836 struct sk_buff_head *list)
837{
838 newsk->next = next;
839 newsk->prev = prev;
840 next->prev = prev->next = newsk;
841 list->qlen++;
842}
843
844static inline void __skb_queue_splice(const struct sk_buff_head *list,
845 struct sk_buff *prev,
846 struct sk_buff *next)
847{
848 struct sk_buff *first = list->next;
849 struct sk_buff *last = list->prev;
850
851 first->prev = prev;
852 prev->next = first;
853
854 last->next = next;
855 next->prev = last;
856}
857
858/**
859 * skb_queue_splice - join two skb lists, this is designed for stacks
860 * @list: the new list to add
861 * @head: the place to add it in the first list
862 */
863static inline void skb_queue_splice(const struct sk_buff_head *list,
864 struct sk_buff_head *head)
865{
866 if (!skb_queue_empty(list)) {
867 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
868 head->qlen += list->qlen;
869 }
870}
871
872/**
873 * skb_queue_splice - join two skb lists and reinitialise the emptied list
874 * @list: the new list to add
875 * @head: the place to add it in the first list
876 *
877 * The list at @list is reinitialised
878 */
879static inline void skb_queue_splice_init(struct sk_buff_head *list,
880 struct sk_buff_head *head)
881{
882 if (!skb_queue_empty(list)) {
883 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
884 head->qlen += list->qlen;
885 __skb_queue_head_init(list);
886 }
887}
888
889/**
890 * skb_queue_splice_tail - join two skb lists, each list being a queue
891 * @list: the new list to add
892 * @head: the place to add it in the first list
893 */
894static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
895 struct sk_buff_head *head)
896{
897 if (!skb_queue_empty(list)) {
898 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
899 head->qlen += list->qlen;
900 }
901}
902
903/**
904 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
905 * @list: the new list to add
906 * @head: the place to add it in the first list
907 *
908 * Each of the lists is a queue.
909 * The list at @list is reinitialised
910 */
911static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
912 struct sk_buff_head *head)
913{
914 if (!skb_queue_empty(list)) {
915 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
916 head->qlen += list->qlen;
917 __skb_queue_head_init(list);
918 }
919}
920
921/**
922 * __skb_queue_after - queue a buffer at the list head
923 * @list: list to use
924 * @prev: place after this buffer
925 * @newsk: buffer to queue
926 *
927 * Queue a buffer int the middle of a list. This function takes no locks
928 * and you must therefore hold required locks before calling it.
929 *
930 * A buffer cannot be placed on two lists at the same time.
931 */
932static inline void __skb_queue_after(struct sk_buff_head *list,
933 struct sk_buff *prev,
934 struct sk_buff *newsk)
935{
936 __skb_insert(newsk, prev, prev->next, list);
937}
938
939extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
940 struct sk_buff_head *list);
941
942static inline void __skb_queue_before(struct sk_buff_head *list,
943 struct sk_buff *next,
944 struct sk_buff *newsk)
945{
946 __skb_insert(newsk, next->prev, next, list);
947}
948
949/**
950 * __skb_queue_head - queue a buffer at the list head
951 * @list: list to use
952 * @newsk: buffer to queue
953 *
954 * Queue a buffer at the start of a list. This function takes no locks
955 * and you must therefore hold required locks before calling it.
956 *
957 * A buffer cannot be placed on two lists at the same time.
958 */
959extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
960static inline void __skb_queue_head(struct sk_buff_head *list,
961 struct sk_buff *newsk)
962{
963 __skb_queue_after(list, (struct sk_buff *)list, newsk);
964}
965
966/**
967 * __skb_queue_tail - queue a buffer at the list tail
968 * @list: list to use
969 * @newsk: buffer to queue
970 *
971 * Queue a buffer at the end of a list. This function takes no locks
972 * and you must therefore hold required locks before calling it.
973 *
974 * A buffer cannot be placed on two lists at the same time.
975 */
976extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
977static inline void __skb_queue_tail(struct sk_buff_head *list,
978 struct sk_buff *newsk)
979{
980 __skb_queue_before(list, (struct sk_buff *)list, newsk);
981}
982
983/*
984 * remove sk_buff from list. _Must_ be called atomically, and with
985 * the list known..
986 */
987extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
988static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
989{
990 struct sk_buff *next, *prev;
991
992 list->qlen--;
993 next = skb->next;
994 prev = skb->prev;
995 skb->next = skb->prev = NULL;
996 next->prev = prev;
997 prev->next = next;
998}
999
1000/**
1001 * __skb_dequeue - remove from the head of the queue
1002 * @list: list to dequeue from
1003 *
1004 * Remove the head of the list. This function does not take any locks
1005 * so must be used with appropriate locks held only. The head item is
1006 * returned or %NULL if the list is empty.
1007 */
1008extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1009static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1010{
1011 struct sk_buff *skb = skb_peek(list);
1012 if (skb)
1013 __skb_unlink(skb, list);
1014 return skb;
1015}
1016
1017/**
1018 * __skb_dequeue_tail - remove from the tail of the queue
1019 * @list: list to dequeue from
1020 *
1021 * Remove the tail of the list. This function does not take any locks
1022 * so must be used with appropriate locks held only. The tail item is
1023 * returned or %NULL if the list is empty.
1024 */
1025extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1026static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1027{
1028 struct sk_buff *skb = skb_peek_tail(list);
1029 if (skb)
1030 __skb_unlink(skb, list);
1031 return skb;
1032}
1033
1034
1035static inline int skb_is_nonlinear(const struct sk_buff *skb)
1036{
1037 return skb->data_len;
1038}
1039
1040static inline unsigned int skb_headlen(const struct sk_buff *skb)
1041{
1042 return skb->len - skb->data_len;
1043}
1044
1045static inline int skb_pagelen(const struct sk_buff *skb)
1046{
1047 int i, len = 0;
1048
1049 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1050 len += skb_shinfo(skb)->frags[i].size;
1051 return len + skb_headlen(skb);
1052}
1053
1054static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1055 struct page *page, int off, int size)
1056{
1057 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1058
1059 frag->page = page;
1060 frag->page_offset = off;
1061 frag->size = size;
1062 skb_shinfo(skb)->nr_frags = i + 1;
1063}
1064
1065extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1066 int off, int size);
1067
1068#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1069#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
1070#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1071
1072#ifdef NET_SKBUFF_DATA_USES_OFFSET
1073static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1074{
1075 return skb->head + skb->tail;
1076}
1077
1078static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1079{
1080 skb->tail = skb->data - skb->head;
1081}
1082
1083static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1084{
1085 skb_reset_tail_pointer(skb);
1086 skb->tail += offset;
1087}
1088#else /* NET_SKBUFF_DATA_USES_OFFSET */
1089static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1090{
1091 return skb->tail;
1092}
1093
1094static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1095{
1096 skb->tail = skb->data;
1097}
1098
1099static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1100{
1101 skb->tail = skb->data + offset;
1102}
1103
1104#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1105
1106/*
1107 * Add data to an sk_buff
1108 */
1109extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1110static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1111{
1112 unsigned char *tmp = skb_tail_pointer(skb);
1113 SKB_LINEAR_ASSERT(skb);
1114 skb->tail += len;
1115 skb->len += len;
1116 return tmp;
1117}
1118
1119extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1120static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1121{
1122 skb->data -= len;
1123 skb->len += len;
1124 return skb->data;
1125}
1126
1127extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1128static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1129{
1130 skb->len -= len;
1131 BUG_ON(skb->len < skb->data_len);
1132 return skb->data += len;
1133}
1134
1135extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1136
1137static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1138{
1139 if (len > skb_headlen(skb) &&
1140 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1141 return NULL;
1142 skb->len -= len;
1143 return skb->data += len;
1144}
1145
1146static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1147{
1148 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1149}
1150
1151static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1152{
1153 if (likely(len <= skb_headlen(skb)))
1154 return 1;
1155 if (unlikely(len > skb->len))
1156 return 0;
1157 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1158}
1159
1160/**
1161 * skb_headroom - bytes at buffer head
1162 * @skb: buffer to check
1163 *
1164 * Return the number of bytes of free space at the head of an &sk_buff.
1165 */
1166static inline unsigned int skb_headroom(const struct sk_buff *skb)
1167{
1168 return skb->data - skb->head;
1169}
1170
1171/**
1172 * skb_tailroom - bytes at buffer end
1173 * @skb: buffer to check
1174 *
1175 * Return the number of bytes of free space at the tail of an sk_buff
1176 */
1177static inline int skb_tailroom(const struct sk_buff *skb)
1178{
1179 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1180}
1181
1182/**
1183 * skb_reserve - adjust headroom
1184 * @skb: buffer to alter
1185 * @len: bytes to move
1186 *
1187 * Increase the headroom of an empty &sk_buff by reducing the tail
1188 * room. This is only allowed for an empty buffer.
1189 */
1190static inline void skb_reserve(struct sk_buff *skb, int len)
1191{
1192 skb->data += len;
1193 skb->tail += len;
1194}
1195
1196#ifdef NET_SKBUFF_DATA_USES_OFFSET
1197static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1198{
1199 return skb->head + skb->transport_header;
1200}
1201
1202static inline void skb_reset_transport_header(struct sk_buff *skb)
1203{
1204 skb->transport_header = skb->data - skb->head;
1205}
1206
1207static inline void skb_set_transport_header(struct sk_buff *skb,
1208 const int offset)
1209{
1210 skb_reset_transport_header(skb);
1211 skb->transport_header += offset;
1212}
1213
1214static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1215{
1216 return skb->head + skb->network_header;
1217}
1218
1219static inline void skb_reset_network_header(struct sk_buff *skb)
1220{
1221 skb->network_header = skb->data - skb->head;
1222}
1223
1224static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1225{
1226 skb_reset_network_header(skb);
1227 skb->network_header += offset;
1228}
1229
1230static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1231{
1232 return skb->head + skb->mac_header;
1233}
1234
1235static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1236{
1237 return skb->mac_header != ~0U;
1238}
1239
1240static inline void skb_reset_mac_header(struct sk_buff *skb)
1241{
1242 skb->mac_header = skb->data - skb->head;
1243}
1244
1245static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1246{
1247 skb_reset_mac_header(skb);
1248 skb->mac_header += offset;
1249}
1250
1251#else /* NET_SKBUFF_DATA_USES_OFFSET */
1252
1253static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1254{
1255 return skb->transport_header;
1256}
1257
1258static inline void skb_reset_transport_header(struct sk_buff *skb)
1259{
1260 skb->transport_header = skb->data;
1261}
1262
1263static inline void skb_set_transport_header(struct sk_buff *skb,
1264 const int offset)
1265{
1266 skb->transport_header = skb->data + offset;
1267}
1268
1269static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1270{
1271 return skb->network_header;
1272}
1273
1274static inline void skb_reset_network_header(struct sk_buff *skb)
1275{
1276 skb->network_header = skb->data;
1277}
1278
1279static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1280{
1281 skb->network_header = skb->data + offset;
1282}
1283
1284static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1285{
1286 return skb->mac_header;
1287}
1288
1289static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1290{
1291 return skb->mac_header != NULL;
1292}
1293
1294static inline void skb_reset_mac_header(struct sk_buff *skb)
1295{
1296 skb->mac_header = skb->data;
1297}
1298
1299static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1300{
1301 skb->mac_header = skb->data + offset;
1302}
1303#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1304
1305static inline int skb_transport_offset(const struct sk_buff *skb)
1306{
1307 return skb_transport_header(skb) - skb->data;
1308}
1309
1310static inline u32 skb_network_header_len(const struct sk_buff *skb)
1311{
1312 return skb->transport_header - skb->network_header;
1313}
1314
1315static inline int skb_network_offset(const struct sk_buff *skb)
1316{
1317 return skb_network_header(skb) - skb->data;
1318}
1319
1320/*
1321 * CPUs often take a performance hit when accessing unaligned memory
1322 * locations. The actual performance hit varies, it can be small if the
1323 * hardware handles it or large if we have to take an exception and fix it
1324 * in software.
1325 *
1326 * Since an ethernet header is 14 bytes network drivers often end up with
1327 * the IP header at an unaligned offset. The IP header can be aligned by
1328 * shifting the start of the packet by 2 bytes. Drivers should do this
1329 * with:
1330 *
1331 * skb_reserve(NET_IP_ALIGN);
1332 *
1333 * The downside to this alignment of the IP header is that the DMA is now
1334 * unaligned. On some architectures the cost of an unaligned DMA is high
1335 * and this cost outweighs the gains made by aligning the IP header.
1336 *
1337 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1338 * to be overridden.
1339 */
1340#ifndef NET_IP_ALIGN
1341#define NET_IP_ALIGN 2
1342#endif
1343
1344/*
1345 * The networking layer reserves some headroom in skb data (via
1346 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1347 * the header has to grow. In the default case, if the header has to grow
1348 * 32 bytes or less we avoid the reallocation.
1349 *
1350 * Unfortunately this headroom changes the DMA alignment of the resulting
1351 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1352 * on some architectures. An architecture can override this value,
1353 * perhaps setting it to a cacheline in size (since that will maintain
1354 * cacheline alignment of the DMA). It must be a power of 2.
1355 *
1356 * Various parts of the networking layer expect at least 32 bytes of
1357 * headroom, you should not reduce this.
1358 */
1359#ifndef NET_SKB_PAD
1360#define NET_SKB_PAD 32
1361#endif
1362
1363extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1364
1365static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1366{
1367 if (unlikely(skb->data_len)) {
1368 WARN_ON(1);
1369 return;
1370 }
1371 skb->len = len;
1372 skb_set_tail_pointer(skb, len);
1373}
1374
1375extern void skb_trim(struct sk_buff *skb, unsigned int len);
1376
1377static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1378{
1379 if (skb->data_len)
1380 return ___pskb_trim(skb, len);
1381 __skb_trim(skb, len);
1382 return 0;
1383}
1384
1385static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1386{
1387 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1388}
1389
1390/**
1391 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1392 * @skb: buffer to alter
1393 * @len: new length
1394 *
1395 * This is identical to pskb_trim except that the caller knows that
1396 * the skb is not cloned so we should never get an error due to out-
1397 * of-memory.
1398 */
1399static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1400{
1401 int err = pskb_trim(skb, len);
1402 BUG_ON(err);
1403}
1404
1405/**
1406 * skb_orphan - orphan a buffer
1407 * @skb: buffer to orphan
1408 *
1409 * If a buffer currently has an owner then we call the owner's
1410 * destructor function and make the @skb unowned. The buffer continues
1411 * to exist but is no longer charged to its former owner.
1412 */
1413static inline void skb_orphan(struct sk_buff *skb)
1414{
1415 if (skb->destructor)
1416 skb->destructor(skb);
1417 skb->destructor = NULL;
1418 skb->sk = NULL;
1419}
1420
1421/**
1422 * __skb_queue_purge - empty a list
1423 * @list: list to empty
1424 *
1425 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1426 * the list and one reference dropped. This function does not take the
1427 * list lock and the caller must hold the relevant locks to use it.
1428 */
1429extern void skb_queue_purge(struct sk_buff_head *list);
1430static inline void __skb_queue_purge(struct sk_buff_head *list)
1431{
1432 struct sk_buff *skb;
1433 while ((skb = __skb_dequeue(list)) != NULL)
1434 kfree_skb(skb);
1435}
1436
1437/**
1438 * __dev_alloc_skb - allocate an skbuff for receiving
1439 * @length: length to allocate
1440 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1441 *
1442 * Allocate a new &sk_buff and assign it a usage count of one. The
1443 * buffer has unspecified headroom built in. Users should allocate
1444 * the headroom they think they need without accounting for the
1445 * built in space. The built in space is used for optimisations.
1446 *
1447 * %NULL is returned if there is no free memory.
1448 */
1449static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1450 gfp_t gfp_mask)
1451{
1452 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1453 if (likely(skb))
1454 skb_reserve(skb, NET_SKB_PAD);
1455 return skb;
1456}
1457
1458extern struct sk_buff *dev_alloc_skb(unsigned int length);
1459
1460extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1461 unsigned int length, gfp_t gfp_mask);
1462
1463/**
1464 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1465 * @dev: network device to receive on
1466 * @length: length to allocate
1467 *
1468 * Allocate a new &sk_buff and assign it a usage count of one. The
1469 * buffer has unspecified headroom built in. Users should allocate
1470 * the headroom they think they need without accounting for the
1471 * built in space. The built in space is used for optimisations.
1472 *
1473 * %NULL is returned if there is no free memory. Although this function
1474 * allocates memory it can be called from an interrupt.
1475 */
1476static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1477 unsigned int length)
1478{
1479 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1480}
1481
1482extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1483
1484/**
1485 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1486 * @dev: network device to receive on
1487 *
1488 * Allocate a new page node local to the specified device.
1489 *
1490 * %NULL is returned if there is no free memory.
1491 */
1492static inline struct page *netdev_alloc_page(struct net_device *dev)
1493{
1494 return __netdev_alloc_page(dev, GFP_ATOMIC);
1495}
1496
1497static inline void netdev_free_page(struct net_device *dev, struct page *page)
1498{
1499 __free_page(page);
1500}
1501
1502/**
1503 * skb_clone_writable - is the header of a clone writable
1504 * @skb: buffer to check
1505 * @len: length up to which to write
1506 *
1507 * Returns true if modifying the header part of the cloned buffer
1508 * does not requires the data to be copied.
1509 */
1510static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1511{
1512 return !skb_header_cloned(skb) &&
1513 skb_headroom(skb) + len <= skb->hdr_len;
1514}
1515
1516static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1517 int cloned)
1518{
1519 int delta = 0;
1520
1521 if (headroom < NET_SKB_PAD)
1522 headroom = NET_SKB_PAD;
1523 if (headroom > skb_headroom(skb))
1524 delta = headroom - skb_headroom(skb);
1525
1526 if (delta || cloned)
1527 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1528 GFP_ATOMIC);
1529 return 0;
1530}
1531
1532/**
1533 * skb_cow - copy header of skb when it is required
1534 * @skb: buffer to cow
1535 * @headroom: needed headroom
1536 *
1537 * If the skb passed lacks sufficient headroom or its data part
1538 * is shared, data is reallocated. If reallocation fails, an error
1539 * is returned and original skb is not changed.
1540 *
1541 * The result is skb with writable area skb->head...skb->tail
1542 * and at least @headroom of space at head.
1543 */
1544static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1545{
1546 return __skb_cow(skb, headroom, skb_cloned(skb));
1547}
1548
1549/**
1550 * skb_cow_head - skb_cow but only making the head writable
1551 * @skb: buffer to cow
1552 * @headroom: needed headroom
1553 *
1554 * This function is identical to skb_cow except that we replace the
1555 * skb_cloned check by skb_header_cloned. It should be used when
1556 * you only need to push on some header and do not need to modify
1557 * the data.
1558 */
1559static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1560{
1561 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1562}
1563
1564/**
1565 * skb_padto - pad an skbuff up to a minimal size
1566 * @skb: buffer to pad
1567 * @len: minimal length
1568 *
1569 * Pads up a buffer to ensure the trailing bytes exist and are
1570 * blanked. If the buffer already contains sufficient data it
1571 * is untouched. Otherwise it is extended. Returns zero on
1572 * success. The skb is freed on error.
1573 */
1574
1575static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1576{
1577 unsigned int size = skb->len;
1578 if (likely(size >= len))
1579 return 0;
1580 return skb_pad(skb, len - size);
1581}
1582
1583static inline int skb_add_data(struct sk_buff *skb,
1584 char __user *from, int copy)
1585{
1586 const int off = skb->len;
1587
1588 if (skb->ip_summed == CHECKSUM_NONE) {
1589 int err = 0;
1590 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1591 copy, 0, &err);
1592 if (!err) {
1593 skb->csum = csum_block_add(skb->csum, csum, off);
1594 return 0;
1595 }
1596 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1597 return 0;
1598
1599 __skb_trim(skb, off);
1600 return -EFAULT;
1601}
1602
1603static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1604 struct page *page, int off)
1605{
1606 if (i) {
1607 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1608
1609 return page == frag->page &&
1610 off == frag->page_offset + frag->size;
1611 }
1612 return 0;
1613}
1614
1615static inline int __skb_linearize(struct sk_buff *skb)
1616{
1617 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1618}
1619
1620/**
1621 * skb_linearize - convert paged skb to linear one
1622 * @skb: buffer to linarize
1623 *
1624 * If there is no free memory -ENOMEM is returned, otherwise zero
1625 * is returned and the old skb data released.
1626 */
1627static inline int skb_linearize(struct sk_buff *skb)
1628{
1629 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1630}
1631
1632/**
1633 * skb_linearize_cow - make sure skb is linear and writable
1634 * @skb: buffer to process
1635 *
1636 * If there is no free memory -ENOMEM is returned, otherwise zero
1637 * is returned and the old skb data released.
1638 */
1639static inline int skb_linearize_cow(struct sk_buff *skb)
1640{
1641 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1642 __skb_linearize(skb) : 0;
1643}
1644
1645/**
1646 * skb_postpull_rcsum - update checksum for received skb after pull
1647 * @skb: buffer to update
1648 * @start: start of data before pull
1649 * @len: length of data pulled
1650 *
1651 * After doing a pull on a received packet, you need to call this to
1652 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1653 * CHECKSUM_NONE so that it can be recomputed from scratch.
1654 */
1655
1656static inline void skb_postpull_rcsum(struct sk_buff *skb,
1657 const void *start, unsigned int len)
1658{
1659 if (skb->ip_summed == CHECKSUM_COMPLETE)
1660 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1661}
1662
1663unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1664
1665/**
1666 * pskb_trim_rcsum - trim received skb and update checksum
1667 * @skb: buffer to trim
1668 * @len: new length
1669 *
1670 * This is exactly the same as pskb_trim except that it ensures the
1671 * checksum of received packets are still valid after the operation.
1672 */
1673
1674static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1675{
1676 if (likely(len >= skb->len))
1677 return 0;
1678 if (skb->ip_summed == CHECKSUM_COMPLETE)
1679 skb->ip_summed = CHECKSUM_NONE;
1680 return __pskb_trim(skb, len);
1681}
1682
1683#define skb_queue_walk(queue, skb) \
1684 for (skb = (queue)->next; \
1685 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1686 skb = skb->next)
1687
1688#define skb_queue_walk_safe(queue, skb, tmp) \
1689 for (skb = (queue)->next, tmp = skb->next; \
1690 skb != (struct sk_buff *)(queue); \
1691 skb = tmp, tmp = skb->next)
1692
1693#define skb_queue_walk_from(queue, skb) \
1694 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1695 skb = skb->next)
1696
1697#define skb_queue_walk_from_safe(queue, skb, tmp) \
1698 for (tmp = skb->next; \
1699 skb != (struct sk_buff *)(queue); \
1700 skb = tmp, tmp = skb->next)
1701
1702#define skb_queue_reverse_walk(queue, skb) \
1703 for (skb = (queue)->prev; \
1704 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1705 skb = skb->prev)
1706
1707
1708extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1709 int *peeked, int *err);
1710extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1711 int noblock, int *err);
1712extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1713 struct poll_table_struct *wait);
1714extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1715 int offset, struct iovec *to,
1716 int size);
1717extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1718 int hlen,
1719 struct iovec *iov);
1720extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1721 int offset,
1722 const struct iovec *from,
1723 int from_offset,
1724 int len);
1725extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1726 int offset,
1727 const struct iovec *to,
1728 int to_offset,
1729 int size);
1730extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1731extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1732 unsigned int flags);
1733extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1734 int len, __wsum csum);
1735extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1736 void *to, int len);
1737extern int skb_store_bits(struct sk_buff *skb, int offset,
1738 const void *from, int len);
1739extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1740 int offset, u8 *to, int len,
1741 __wsum csum);
1742extern int skb_splice_bits(struct sk_buff *skb,
1743 unsigned int offset,
1744 struct pipe_inode_info *pipe,
1745 unsigned int len,
1746 unsigned int flags);
1747extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1748extern void skb_split(struct sk_buff *skb,
1749 struct sk_buff *skb1, const u32 len);
1750extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1751 int shiftlen);
1752
1753extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1754
1755static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1756 int len, void *buffer)
1757{
1758 int hlen = skb_headlen(skb);
1759
1760 if (hlen - offset >= len)
1761 return skb->data + offset;
1762
1763 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1764 return NULL;
1765
1766 return buffer;
1767}
1768
1769static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1770 void *to,
1771 const unsigned int len)
1772{
1773 memcpy(to, skb->data, len);
1774}
1775
1776static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1777 const int offset, void *to,
1778 const unsigned int len)
1779{
1780 memcpy(to, skb->data + offset, len);
1781}
1782
1783static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1784 const void *from,
1785 const unsigned int len)
1786{
1787 memcpy(skb->data, from, len);
1788}
1789
1790static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1791 const int offset,
1792 const void *from,
1793 const unsigned int len)
1794{
1795 memcpy(skb->data + offset, from, len);
1796}
1797
1798extern void skb_init(void);
1799
1800static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1801{
1802 return skb->tstamp;
1803}
1804
1805/**
1806 * skb_get_timestamp - get timestamp from a skb
1807 * @skb: skb to get stamp from
1808 * @stamp: pointer to struct timeval to store stamp in
1809 *
1810 * Timestamps are stored in the skb as offsets to a base timestamp.
1811 * This function converts the offset back to a struct timeval and stores
1812 * it in stamp.
1813 */
1814static inline void skb_get_timestamp(const struct sk_buff *skb,
1815 struct timeval *stamp)
1816{
1817 *stamp = ktime_to_timeval(skb->tstamp);
1818}
1819
1820static inline void skb_get_timestampns(const struct sk_buff *skb,
1821 struct timespec *stamp)
1822{
1823 *stamp = ktime_to_timespec(skb->tstamp);
1824}
1825
1826static inline void __net_timestamp(struct sk_buff *skb)
1827{
1828 skb->tstamp = ktime_get_real();
1829}
1830
1831static inline ktime_t net_timedelta(ktime_t t)
1832{
1833 return ktime_sub(ktime_get_real(), t);
1834}
1835
1836static inline ktime_t net_invalid_timestamp(void)
1837{
1838 return ktime_set(0, 0);
1839}
1840
1841/**
1842 * skb_tstamp_tx - queue clone of skb with send time stamps
1843 * @orig_skb: the original outgoing packet
1844 * @hwtstamps: hardware time stamps, may be NULL if not available
1845 *
1846 * If the skb has a socket associated, then this function clones the
1847 * skb (thus sharing the actual data and optional structures), stores
1848 * the optional hardware time stamping information (if non NULL) or
1849 * generates a software time stamp (otherwise), then queues the clone
1850 * to the error queue of the socket. Errors are silently ignored.
1851 */
1852extern void skb_tstamp_tx(struct sk_buff *orig_skb,
1853 struct skb_shared_hwtstamps *hwtstamps);
1854
1855extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
1856extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
1857
1858static inline int skb_csum_unnecessary(const struct sk_buff *skb)
1859{
1860 return skb->ip_summed & CHECKSUM_UNNECESSARY;
1861}
1862
1863/**
1864 * skb_checksum_complete - Calculate checksum of an entire packet
1865 * @skb: packet to process
1866 *
1867 * This function calculates the checksum over the entire packet plus
1868 * the value of skb->csum. The latter can be used to supply the
1869 * checksum of a pseudo header as used by TCP/UDP. It returns the
1870 * checksum.
1871 *
1872 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1873 * this function can be used to verify that checksum on received
1874 * packets. In that case the function should return zero if the
1875 * checksum is correct. In particular, this function will return zero
1876 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1877 * hardware has already verified the correctness of the checksum.
1878 */
1879static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
1880{
1881 return skb_csum_unnecessary(skb) ?
1882 0 : __skb_checksum_complete(skb);
1883}
1884
1885#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1886extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
1887static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1888{
1889 if (nfct && atomic_dec_and_test(&nfct->use))
1890 nf_conntrack_destroy(nfct);
1891}
1892static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1893{
1894 if (nfct)
1895 atomic_inc(&nfct->use);
1896}
1897static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1898{
1899 if (skb)
1900 atomic_inc(&skb->users);
1901}
1902static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1903{
1904 if (skb)
1905 kfree_skb(skb);
1906}
1907#endif
1908#ifdef CONFIG_BRIDGE_NETFILTER
1909static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1910{
1911 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1912 kfree(nf_bridge);
1913}
1914static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1915{
1916 if (nf_bridge)
1917 atomic_inc(&nf_bridge->use);
1918}
1919#endif /* CONFIG_BRIDGE_NETFILTER */
1920static inline void nf_reset(struct sk_buff *skb)
1921{
1922#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1923 nf_conntrack_put(skb->nfct);
1924 skb->nfct = NULL;
1925 nf_conntrack_put_reasm(skb->nfct_reasm);
1926 skb->nfct_reasm = NULL;
1927#endif
1928#ifdef CONFIG_BRIDGE_NETFILTER
1929 nf_bridge_put(skb->nf_bridge);
1930 skb->nf_bridge = NULL;
1931#endif
1932}
1933
1934/* Note: This doesn't put any conntrack and bridge info in dst. */
1935static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1936{
1937#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1938 dst->nfct = src->nfct;
1939 nf_conntrack_get(src->nfct);
1940 dst->nfctinfo = src->nfctinfo;
1941 dst->nfct_reasm = src->nfct_reasm;
1942 nf_conntrack_get_reasm(src->nfct_reasm);
1943#endif
1944#ifdef CONFIG_BRIDGE_NETFILTER
1945 dst->nf_bridge = src->nf_bridge;
1946 nf_bridge_get(src->nf_bridge);
1947#endif
1948}
1949
1950static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1951{
1952#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1953 nf_conntrack_put(dst->nfct);
1954 nf_conntrack_put_reasm(dst->nfct_reasm);
1955#endif
1956#ifdef CONFIG_BRIDGE_NETFILTER
1957 nf_bridge_put(dst->nf_bridge);
1958#endif
1959 __nf_copy(dst, src);
1960}
1961
1962#ifdef CONFIG_NETWORK_SECMARK
1963static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1964{
1965 to->secmark = from->secmark;
1966}
1967
1968static inline void skb_init_secmark(struct sk_buff *skb)
1969{
1970 skb->secmark = 0;
1971}
1972#else
1973static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1974{ }
1975
1976static inline void skb_init_secmark(struct sk_buff *skb)
1977{ }
1978#endif
1979
1980static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
1981{
1982 skb->queue_mapping = queue_mapping;
1983}
1984
1985static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
1986{
1987 return skb->queue_mapping;
1988}
1989
1990static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
1991{
1992 to->queue_mapping = from->queue_mapping;
1993}
1994
1995static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
1996{
1997 skb->queue_mapping = rx_queue + 1;
1998}
1999
2000static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2001{
2002 return skb->queue_mapping - 1;
2003}
2004
2005static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2006{
2007 return (skb->queue_mapping != 0);
2008}
2009
2010extern u16 skb_tx_hash(const struct net_device *dev,
2011 const struct sk_buff *skb);
2012
2013#ifdef CONFIG_XFRM
2014static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2015{
2016 return skb->sp;
2017}
2018#else
2019static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2020{
2021 return NULL;
2022}
2023#endif
2024
2025static inline int skb_is_gso(const struct sk_buff *skb)
2026{
2027 return skb_shinfo(skb)->gso_size;
2028}
2029
2030static inline int skb_is_gso_v6(const struct sk_buff *skb)
2031{
2032 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2033}
2034
2035extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2036
2037static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2038{
2039 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2040 * wanted then gso_type will be set. */
2041 struct skb_shared_info *shinfo = skb_shinfo(skb);
2042 if (shinfo->gso_size != 0 && unlikely(shinfo->gso_type == 0)) {
2043 __skb_warn_lro_forwarding(skb);
2044 return true;
2045 }
2046 return false;
2047}
2048
2049static inline void skb_forward_csum(struct sk_buff *skb)
2050{
2051 /* Unfortunately we don't support this one. Any brave souls? */
2052 if (skb->ip_summed == CHECKSUM_COMPLETE)
2053 skb->ip_summed = CHECKSUM_NONE;
2054}
2055
2056bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2057#endif /* __KERNEL__ */
2058#endif /* _LINUX_SKBUFF_H */