Merge branch 'master' of /home/davem/src/GIT/linux-2.6/
[deliverable/linux.git] / include / linux / skbuff.h
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
93 struct net_device;
94 struct scatterlist;
95 struct pipe_inode_info;
96
97 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
98 struct nf_conntrack {
99 atomic_t use;
100 };
101 #endif
102
103 #ifdef CONFIG_BRIDGE_NETFILTER
104 struct 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
113 struct 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
122 struct 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
127 typedef struct skb_frag_struct skb_frag_t;
128
129 struct 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 * skb_shared_hwtstamps - hardware time stamps
139 *
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 */
161 struct skb_shared_hwtstamps {
162 ktime_t hwtstamp;
163 ktime_t syststamp;
164 };
165
166 /**
167 * skb_shared_tx - instructions for time stamping of outgoing packets
168 *
169 * @hardware: generate hardware time stamp
170 * @software: generate software time stamp
171 * @in_progress: device driver is going to provide
172 * hardware time stamp
173 *
174 * These flags are attached to packets as part of the
175 * &skb_shared_info. Use skb_tx() to get a pointer.
176 */
177 union 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 */
189 struct 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 #ifdef CONFIG_HAS_DMA
199 unsigned int num_dma_maps;
200 #endif
201 struct sk_buff *frag_list;
202 struct skb_shared_hwtstamps hwtstamps;
203 skb_frag_t frags[MAX_SKB_FRAGS];
204 #ifdef CONFIG_HAS_DMA
205 dma_addr_t dma_maps[MAX_SKB_FRAGS + 1];
206 #endif
207 };
208
209 /* We divide dataref into two halves. The higher 16 bits hold references
210 * to the payload part of skb->data. The lower 16 bits hold references to
211 * the entire skb->data. A clone of a headerless skb holds the length of
212 * the header in skb->hdr_len.
213 *
214 * All users must obey the rule that the skb->data reference count must be
215 * greater than or equal to the payload reference count.
216 *
217 * Holding a reference to the payload part means that the user does not
218 * care about modifications to the header part of skb->data.
219 */
220 #define SKB_DATAREF_SHIFT 16
221 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
222
223
224 enum {
225 SKB_FCLONE_UNAVAILABLE,
226 SKB_FCLONE_ORIG,
227 SKB_FCLONE_CLONE,
228 };
229
230 enum {
231 SKB_GSO_TCPV4 = 1 << 0,
232 SKB_GSO_UDP = 1 << 1,
233
234 /* This indicates the skb is from an untrusted source. */
235 SKB_GSO_DODGY = 1 << 2,
236
237 /* This indicates the tcp segment has CWR set. */
238 SKB_GSO_TCP_ECN = 1 << 3,
239
240 SKB_GSO_TCPV6 = 1 << 4,
241 };
242
243 #if BITS_PER_LONG > 32
244 #define NET_SKBUFF_DATA_USES_OFFSET 1
245 #endif
246
247 #ifdef NET_SKBUFF_DATA_USES_OFFSET
248 typedef unsigned int sk_buff_data_t;
249 #else
250 typedef unsigned char *sk_buff_data_t;
251 #endif
252
253 /**
254 * struct sk_buff - socket buffer
255 * @next: Next buffer in list
256 * @prev: Previous buffer in list
257 * @sk: Socket we are owned by
258 * @tstamp: Time we arrived
259 * @dev: Device we arrived on/are leaving by
260 * @transport_header: Transport layer header
261 * @network_header: Network layer header
262 * @mac_header: Link layer header
263 * @dst: destination entry
264 * @sp: the security path, used for xfrm
265 * @cb: Control buffer. Free for use by every layer. Put private vars here
266 * @len: Length of actual data
267 * @data_len: Data length
268 * @mac_len: Length of link layer header
269 * @hdr_len: writable header length of cloned skb
270 * @csum: Checksum (must include start/offset pair)
271 * @csum_start: Offset from skb->head where checksumming should start
272 * @csum_offset: Offset from csum_start where checksum should be stored
273 * @local_df: allow local fragmentation
274 * @cloned: Head may be cloned (check refcnt to be sure)
275 * @nohdr: Payload reference only, must not modify header
276 * @pkt_type: Packet class
277 * @fclone: skbuff clone status
278 * @ip_summed: Driver fed us an IP checksum
279 * @priority: Packet queueing priority
280 * @users: User count - see {datagram,tcp}.c
281 * @protocol: Packet protocol from driver
282 * @truesize: Buffer size
283 * @head: Head of buffer
284 * @data: Data head pointer
285 * @tail: Tail pointer
286 * @end: End pointer
287 * @destructor: Destruct function
288 * @mark: Generic packet mark
289 * @nfct: Associated connection, if any
290 * @ipvs_property: skbuff is owned by ipvs
291 * @peeked: this packet has been seen already, so stats have been
292 * done for it, don't do them again
293 * @nf_trace: netfilter packet trace flag
294 * @nfctinfo: Relationship of this skb to the connection
295 * @nfct_reasm: netfilter conntrack re-assembly pointer
296 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
297 * @iif: ifindex of device we arrived on
298 * @queue_mapping: Queue mapping for multiqueue devices
299 * @tc_index: Traffic control index
300 * @tc_verd: traffic control verdict
301 * @ndisc_nodetype: router type (from link layer)
302 * @do_not_encrypt: set to prevent encryption of this frame
303 * @requeue: set to indicate that the wireless core should attempt
304 * a software retry on this frame if we failed to
305 * receive an ACK for it
306 * @dma_cookie: a cookie to one of several possible DMA operations
307 * done by skb DMA functions
308 * @secmark: security marking
309 * @vlan_tci: vlan tag control information
310 */
311
312 struct sk_buff {
313 /* These two members must be first. */
314 struct sk_buff *next;
315 struct sk_buff *prev;
316
317 struct sock *sk;
318 ktime_t tstamp;
319 struct net_device *dev;
320
321 union {
322 struct dst_entry *dst;
323 struct rtable *rtable;
324 };
325 #ifdef CONFIG_XFRM
326 struct sec_path *sp;
327 #endif
328 /*
329 * This is the control buffer. It is free to use for every
330 * layer. Please put your private variables there. If you
331 * want to keep them across layers you have to do a skb_clone()
332 * first. This is owned by whoever has the skb queued ATM.
333 */
334 char cb[48];
335
336 unsigned int len,
337 data_len;
338 __u16 mac_len,
339 hdr_len;
340 union {
341 __wsum csum;
342 struct {
343 __u16 csum_start;
344 __u16 csum_offset;
345 };
346 };
347 __u32 priority;
348 __u8 local_df:1,
349 cloned:1,
350 ip_summed:2,
351 nohdr:1,
352 nfctinfo:3;
353 __u8 pkt_type:3,
354 fclone:2,
355 ipvs_property:1,
356 peeked:1,
357 nf_trace:1;
358 __be16 protocol;
359
360 void (*destructor)(struct sk_buff *skb);
361 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
362 struct nf_conntrack *nfct;
363 struct sk_buff *nfct_reasm;
364 #endif
365 #ifdef CONFIG_BRIDGE_NETFILTER
366 struct nf_bridge_info *nf_bridge;
367 #endif
368
369 int iif;
370 __u16 queue_mapping;
371 #ifdef CONFIG_NET_SCHED
372 __u16 tc_index; /* traffic control index */
373 #ifdef CONFIG_NET_CLS_ACT
374 __u16 tc_verd; /* traffic control verdict */
375 #endif
376 #endif
377 #ifdef CONFIG_IPV6_NDISC_NODETYPE
378 __u8 ndisc_nodetype:2;
379 #endif
380 #if defined(CONFIG_MAC80211) || defined(CONFIG_MAC80211_MODULE)
381 __u8 do_not_encrypt:1;
382 __u8 requeue:1;
383 #endif
384 /* 0/13/14 bit hole */
385
386 #ifdef CONFIG_NET_DMA
387 dma_cookie_t dma_cookie;
388 #endif
389 #ifdef CONFIG_NETWORK_SECMARK
390 __u32 secmark;
391 #endif
392
393 __u32 mark;
394
395 __u16 vlan_tci;
396
397 sk_buff_data_t transport_header;
398 sk_buff_data_t network_header;
399 sk_buff_data_t mac_header;
400 /* These elements must be at the end, see alloc_skb() for details. */
401 sk_buff_data_t tail;
402 sk_buff_data_t end;
403 unsigned char *head,
404 *data;
405 unsigned int truesize;
406 atomic_t users;
407 };
408
409 #ifdef __KERNEL__
410 /*
411 * Handling routines are only of interest to the kernel
412 */
413 #include <linux/slab.h>
414
415 #include <asm/system.h>
416
417 #ifdef CONFIG_HAS_DMA
418 #include <linux/dma-mapping.h>
419 extern int skb_dma_map(struct device *dev, struct sk_buff *skb,
420 enum dma_data_direction dir);
421 extern void skb_dma_unmap(struct device *dev, struct sk_buff *skb,
422 enum dma_data_direction dir);
423 #endif
424
425 extern void kfree_skb(struct sk_buff *skb);
426 extern void __kfree_skb(struct sk_buff *skb);
427 extern struct sk_buff *__alloc_skb(unsigned int size,
428 gfp_t priority, int fclone, int node);
429 static inline struct sk_buff *alloc_skb(unsigned int size,
430 gfp_t priority)
431 {
432 return __alloc_skb(size, priority, 0, -1);
433 }
434
435 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
436 gfp_t priority)
437 {
438 return __alloc_skb(size, priority, 1, -1);
439 }
440
441 extern int skb_recycle_check(struct sk_buff *skb, int skb_size);
442
443 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
444 extern struct sk_buff *skb_clone(struct sk_buff *skb,
445 gfp_t priority);
446 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
447 gfp_t priority);
448 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
449 gfp_t gfp_mask);
450 extern int pskb_expand_head(struct sk_buff *skb,
451 int nhead, int ntail,
452 gfp_t gfp_mask);
453 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
454 unsigned int headroom);
455 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
456 int newheadroom, int newtailroom,
457 gfp_t priority);
458 extern int skb_to_sgvec(struct sk_buff *skb,
459 struct scatterlist *sg, int offset,
460 int len);
461 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
462 struct sk_buff **trailer);
463 extern int skb_pad(struct sk_buff *skb, int pad);
464 #define dev_kfree_skb(a) kfree_skb(a)
465 extern void skb_over_panic(struct sk_buff *skb, int len,
466 void *here);
467 extern void skb_under_panic(struct sk_buff *skb, int len,
468 void *here);
469
470 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
471 int getfrag(void *from, char *to, int offset,
472 int len,int odd, struct sk_buff *skb),
473 void *from, int length);
474
475 struct skb_seq_state
476 {
477 __u32 lower_offset;
478 __u32 upper_offset;
479 __u32 frag_idx;
480 __u32 stepped_offset;
481 struct sk_buff *root_skb;
482 struct sk_buff *cur_skb;
483 __u8 *frag_data;
484 };
485
486 extern void skb_prepare_seq_read(struct sk_buff *skb,
487 unsigned int from, unsigned int to,
488 struct skb_seq_state *st);
489 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
490 struct skb_seq_state *st);
491 extern void skb_abort_seq_read(struct skb_seq_state *st);
492
493 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
494 unsigned int to, struct ts_config *config,
495 struct ts_state *state);
496
497 #ifdef NET_SKBUFF_DATA_USES_OFFSET
498 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
499 {
500 return skb->head + skb->end;
501 }
502 #else
503 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
504 {
505 return skb->end;
506 }
507 #endif
508
509 /* Internal */
510 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
511
512 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
513 {
514 return &skb_shinfo(skb)->hwtstamps;
515 }
516
517 static inline union skb_shared_tx *skb_tx(struct sk_buff *skb)
518 {
519 return &skb_shinfo(skb)->tx_flags;
520 }
521
522 /**
523 * skb_queue_empty - check if a queue is empty
524 * @list: queue head
525 *
526 * Returns true if the queue is empty, false otherwise.
527 */
528 static inline int skb_queue_empty(const struct sk_buff_head *list)
529 {
530 return list->next == (struct sk_buff *)list;
531 }
532
533 /**
534 * skb_queue_is_last - check if skb is the last entry in the queue
535 * @list: queue head
536 * @skb: buffer
537 *
538 * Returns true if @skb is the last buffer on the list.
539 */
540 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
541 const struct sk_buff *skb)
542 {
543 return (skb->next == (struct sk_buff *) list);
544 }
545
546 /**
547 * skb_queue_is_first - check if skb is the first entry in the queue
548 * @list: queue head
549 * @skb: buffer
550 *
551 * Returns true if @skb is the first buffer on the list.
552 */
553 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
554 const struct sk_buff *skb)
555 {
556 return (skb->prev == (struct sk_buff *) list);
557 }
558
559 /**
560 * skb_queue_next - return the next packet in the queue
561 * @list: queue head
562 * @skb: current buffer
563 *
564 * Return the next packet in @list after @skb. It is only valid to
565 * call this if skb_queue_is_last() evaluates to false.
566 */
567 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
568 const struct sk_buff *skb)
569 {
570 /* This BUG_ON may seem severe, but if we just return then we
571 * are going to dereference garbage.
572 */
573 BUG_ON(skb_queue_is_last(list, skb));
574 return skb->next;
575 }
576
577 /**
578 * skb_queue_prev - return the prev packet in the queue
579 * @list: queue head
580 * @skb: current buffer
581 *
582 * Return the prev packet in @list before @skb. It is only valid to
583 * call this if skb_queue_is_first() evaluates to false.
584 */
585 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
586 const struct sk_buff *skb)
587 {
588 /* This BUG_ON may seem severe, but if we just return then we
589 * are going to dereference garbage.
590 */
591 BUG_ON(skb_queue_is_first(list, skb));
592 return skb->prev;
593 }
594
595 /**
596 * skb_get - reference buffer
597 * @skb: buffer to reference
598 *
599 * Makes another reference to a socket buffer and returns a pointer
600 * to the buffer.
601 */
602 static inline struct sk_buff *skb_get(struct sk_buff *skb)
603 {
604 atomic_inc(&skb->users);
605 return skb;
606 }
607
608 /*
609 * If users == 1, we are the only owner and are can avoid redundant
610 * atomic change.
611 */
612
613 /**
614 * skb_cloned - is the buffer a clone
615 * @skb: buffer to check
616 *
617 * Returns true if the buffer was generated with skb_clone() and is
618 * one of multiple shared copies of the buffer. Cloned buffers are
619 * shared data so must not be written to under normal circumstances.
620 */
621 static inline int skb_cloned(const struct sk_buff *skb)
622 {
623 return skb->cloned &&
624 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
625 }
626
627 /**
628 * skb_header_cloned - is the header a clone
629 * @skb: buffer to check
630 *
631 * Returns true if modifying the header part of the buffer requires
632 * the data to be copied.
633 */
634 static inline int skb_header_cloned(const struct sk_buff *skb)
635 {
636 int dataref;
637
638 if (!skb->cloned)
639 return 0;
640
641 dataref = atomic_read(&skb_shinfo(skb)->dataref);
642 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
643 return dataref != 1;
644 }
645
646 /**
647 * skb_header_release - release reference to header
648 * @skb: buffer to operate on
649 *
650 * Drop a reference to the header part of the buffer. This is done
651 * by acquiring a payload reference. You must not read from the header
652 * part of skb->data after this.
653 */
654 static inline void skb_header_release(struct sk_buff *skb)
655 {
656 BUG_ON(skb->nohdr);
657 skb->nohdr = 1;
658 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
659 }
660
661 /**
662 * skb_shared - is the buffer shared
663 * @skb: buffer to check
664 *
665 * Returns true if more than one person has a reference to this
666 * buffer.
667 */
668 static inline int skb_shared(const struct sk_buff *skb)
669 {
670 return atomic_read(&skb->users) != 1;
671 }
672
673 /**
674 * skb_share_check - check if buffer is shared and if so clone it
675 * @skb: buffer to check
676 * @pri: priority for memory allocation
677 *
678 * If the buffer is shared the buffer is cloned and the old copy
679 * drops a reference. A new clone with a single reference is returned.
680 * If the buffer is not shared the original buffer is returned. When
681 * being called from interrupt status or with spinlocks held pri must
682 * be GFP_ATOMIC.
683 *
684 * NULL is returned on a memory allocation failure.
685 */
686 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
687 gfp_t pri)
688 {
689 might_sleep_if(pri & __GFP_WAIT);
690 if (skb_shared(skb)) {
691 struct sk_buff *nskb = skb_clone(skb, pri);
692 kfree_skb(skb);
693 skb = nskb;
694 }
695 return skb;
696 }
697
698 /*
699 * Copy shared buffers into a new sk_buff. We effectively do COW on
700 * packets to handle cases where we have a local reader and forward
701 * and a couple of other messy ones. The normal one is tcpdumping
702 * a packet thats being forwarded.
703 */
704
705 /**
706 * skb_unshare - make a copy of a shared buffer
707 * @skb: buffer to check
708 * @pri: priority for memory allocation
709 *
710 * If the socket buffer is a clone then this function creates a new
711 * copy of the data, drops a reference count on the old copy and returns
712 * the new copy with the reference count at 1. If the buffer is not a clone
713 * the original buffer is returned. When called with a spinlock held or
714 * from interrupt state @pri must be %GFP_ATOMIC
715 *
716 * %NULL is returned on a memory allocation failure.
717 */
718 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
719 gfp_t pri)
720 {
721 might_sleep_if(pri & __GFP_WAIT);
722 if (skb_cloned(skb)) {
723 struct sk_buff *nskb = skb_copy(skb, pri);
724 kfree_skb(skb); /* Free our shared copy */
725 skb = nskb;
726 }
727 return skb;
728 }
729
730 /**
731 * skb_peek
732 * @list_: list to peek at
733 *
734 * Peek an &sk_buff. Unlike most other operations you _MUST_
735 * be careful with this one. A peek leaves the buffer on the
736 * list and someone else may run off with it. You must hold
737 * the appropriate locks or have a private queue to do this.
738 *
739 * Returns %NULL for an empty list or a pointer to the head element.
740 * The reference count is not incremented and the reference is therefore
741 * volatile. Use with caution.
742 */
743 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
744 {
745 struct sk_buff *list = ((struct sk_buff *)list_)->next;
746 if (list == (struct sk_buff *)list_)
747 list = NULL;
748 return list;
749 }
750
751 /**
752 * skb_peek_tail
753 * @list_: list to peek at
754 *
755 * Peek an &sk_buff. Unlike most other operations you _MUST_
756 * be careful with this one. A peek leaves the buffer on the
757 * list and someone else may run off with it. You must hold
758 * the appropriate locks or have a private queue to do this.
759 *
760 * Returns %NULL for an empty list or a pointer to the tail element.
761 * The reference count is not incremented and the reference is therefore
762 * volatile. Use with caution.
763 */
764 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
765 {
766 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
767 if (list == (struct sk_buff *)list_)
768 list = NULL;
769 return list;
770 }
771
772 /**
773 * skb_queue_len - get queue length
774 * @list_: list to measure
775 *
776 * Return the length of an &sk_buff queue.
777 */
778 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
779 {
780 return list_->qlen;
781 }
782
783 /**
784 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
785 * @list: queue to initialize
786 *
787 * This initializes only the list and queue length aspects of
788 * an sk_buff_head object. This allows to initialize the list
789 * aspects of an sk_buff_head without reinitializing things like
790 * the spinlock. It can also be used for on-stack sk_buff_head
791 * objects where the spinlock is known to not be used.
792 */
793 static inline void __skb_queue_head_init(struct sk_buff_head *list)
794 {
795 list->prev = list->next = (struct sk_buff *)list;
796 list->qlen = 0;
797 }
798
799 /*
800 * This function creates a split out lock class for each invocation;
801 * this is needed for now since a whole lot of users of the skb-queue
802 * infrastructure in drivers have different locking usage (in hardirq)
803 * than the networking core (in softirq only). In the long run either the
804 * network layer or drivers should need annotation to consolidate the
805 * main types of usage into 3 classes.
806 */
807 static inline void skb_queue_head_init(struct sk_buff_head *list)
808 {
809 spin_lock_init(&list->lock);
810 __skb_queue_head_init(list);
811 }
812
813 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
814 struct lock_class_key *class)
815 {
816 skb_queue_head_init(list);
817 lockdep_set_class(&list->lock, class);
818 }
819
820 /*
821 * Insert an sk_buff on a list.
822 *
823 * The "__skb_xxxx()" functions are the non-atomic ones that
824 * can only be called with interrupts disabled.
825 */
826 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
827 static inline void __skb_insert(struct sk_buff *newsk,
828 struct sk_buff *prev, struct sk_buff *next,
829 struct sk_buff_head *list)
830 {
831 newsk->next = next;
832 newsk->prev = prev;
833 next->prev = prev->next = newsk;
834 list->qlen++;
835 }
836
837 static inline void __skb_queue_splice(const struct sk_buff_head *list,
838 struct sk_buff *prev,
839 struct sk_buff *next)
840 {
841 struct sk_buff *first = list->next;
842 struct sk_buff *last = list->prev;
843
844 first->prev = prev;
845 prev->next = first;
846
847 last->next = next;
848 next->prev = last;
849 }
850
851 /**
852 * skb_queue_splice - join two skb lists, this is designed for stacks
853 * @list: the new list to add
854 * @head: the place to add it in the first list
855 */
856 static inline void skb_queue_splice(const struct sk_buff_head *list,
857 struct sk_buff_head *head)
858 {
859 if (!skb_queue_empty(list)) {
860 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
861 head->qlen += list->qlen;
862 }
863 }
864
865 /**
866 * skb_queue_splice - join two skb lists and reinitialise the emptied list
867 * @list: the new list to add
868 * @head: the place to add it in the first list
869 *
870 * The list at @list is reinitialised
871 */
872 static inline void skb_queue_splice_init(struct sk_buff_head *list,
873 struct sk_buff_head *head)
874 {
875 if (!skb_queue_empty(list)) {
876 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
877 head->qlen += list->qlen;
878 __skb_queue_head_init(list);
879 }
880 }
881
882 /**
883 * skb_queue_splice_tail - join two skb lists, each list being a queue
884 * @list: the new list to add
885 * @head: the place to add it in the first list
886 */
887 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
888 struct sk_buff_head *head)
889 {
890 if (!skb_queue_empty(list)) {
891 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
892 head->qlen += list->qlen;
893 }
894 }
895
896 /**
897 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
898 * @list: the new list to add
899 * @head: the place to add it in the first list
900 *
901 * Each of the lists is a queue.
902 * The list at @list is reinitialised
903 */
904 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
905 struct sk_buff_head *head)
906 {
907 if (!skb_queue_empty(list)) {
908 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
909 head->qlen += list->qlen;
910 __skb_queue_head_init(list);
911 }
912 }
913
914 /**
915 * __skb_queue_after - queue a buffer at the list head
916 * @list: list to use
917 * @prev: place after this buffer
918 * @newsk: buffer to queue
919 *
920 * Queue a buffer int the middle of a list. This function takes no locks
921 * and you must therefore hold required locks before calling it.
922 *
923 * A buffer cannot be placed on two lists at the same time.
924 */
925 static inline void __skb_queue_after(struct sk_buff_head *list,
926 struct sk_buff *prev,
927 struct sk_buff *newsk)
928 {
929 __skb_insert(newsk, prev, prev->next, list);
930 }
931
932 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
933 struct sk_buff_head *list);
934
935 static inline void __skb_queue_before(struct sk_buff_head *list,
936 struct sk_buff *next,
937 struct sk_buff *newsk)
938 {
939 __skb_insert(newsk, next->prev, next, list);
940 }
941
942 /**
943 * __skb_queue_head - queue a buffer at the list head
944 * @list: list to use
945 * @newsk: buffer to queue
946 *
947 * Queue a buffer at the start of a list. This function takes no locks
948 * and you must therefore hold required locks before calling it.
949 *
950 * A buffer cannot be placed on two lists at the same time.
951 */
952 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
953 static inline void __skb_queue_head(struct sk_buff_head *list,
954 struct sk_buff *newsk)
955 {
956 __skb_queue_after(list, (struct sk_buff *)list, newsk);
957 }
958
959 /**
960 * __skb_queue_tail - queue a buffer at the list tail
961 * @list: list to use
962 * @newsk: buffer to queue
963 *
964 * Queue a buffer at the end of a list. This function takes no locks
965 * and you must therefore hold required locks before calling it.
966 *
967 * A buffer cannot be placed on two lists at the same time.
968 */
969 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
970 static inline void __skb_queue_tail(struct sk_buff_head *list,
971 struct sk_buff *newsk)
972 {
973 __skb_queue_before(list, (struct sk_buff *)list, newsk);
974 }
975
976 /*
977 * remove sk_buff from list. _Must_ be called atomically, and with
978 * the list known..
979 */
980 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
981 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
982 {
983 struct sk_buff *next, *prev;
984
985 list->qlen--;
986 next = skb->next;
987 prev = skb->prev;
988 skb->next = skb->prev = NULL;
989 next->prev = prev;
990 prev->next = next;
991 }
992
993 /**
994 * __skb_dequeue - remove from the head of the queue
995 * @list: list to dequeue from
996 *
997 * Remove the head of the list. This function does not take any locks
998 * so must be used with appropriate locks held only. The head item is
999 * returned or %NULL if the list is empty.
1000 */
1001 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1002 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1003 {
1004 struct sk_buff *skb = skb_peek(list);
1005 if (skb)
1006 __skb_unlink(skb, list);
1007 return skb;
1008 }
1009
1010 /**
1011 * __skb_dequeue_tail - remove from the tail of the queue
1012 * @list: list to dequeue from
1013 *
1014 * Remove the tail of the list. This function does not take any locks
1015 * so must be used with appropriate locks held only. The tail item is
1016 * returned or %NULL if the list is empty.
1017 */
1018 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1019 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1020 {
1021 struct sk_buff *skb = skb_peek_tail(list);
1022 if (skb)
1023 __skb_unlink(skb, list);
1024 return skb;
1025 }
1026
1027
1028 static inline int skb_is_nonlinear(const struct sk_buff *skb)
1029 {
1030 return skb->data_len;
1031 }
1032
1033 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1034 {
1035 return skb->len - skb->data_len;
1036 }
1037
1038 static inline int skb_pagelen(const struct sk_buff *skb)
1039 {
1040 int i, len = 0;
1041
1042 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1043 len += skb_shinfo(skb)->frags[i].size;
1044 return len + skb_headlen(skb);
1045 }
1046
1047 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1048 struct page *page, int off, int size)
1049 {
1050 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1051
1052 frag->page = page;
1053 frag->page_offset = off;
1054 frag->size = size;
1055 skb_shinfo(skb)->nr_frags = i + 1;
1056 }
1057
1058 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1059 int off, int size);
1060
1061 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1062 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
1063 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1064
1065 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1066 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1067 {
1068 return skb->head + skb->tail;
1069 }
1070
1071 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1072 {
1073 skb->tail = skb->data - skb->head;
1074 }
1075
1076 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1077 {
1078 skb_reset_tail_pointer(skb);
1079 skb->tail += offset;
1080 }
1081 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1082 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1083 {
1084 return skb->tail;
1085 }
1086
1087 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1088 {
1089 skb->tail = skb->data;
1090 }
1091
1092 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1093 {
1094 skb->tail = skb->data + offset;
1095 }
1096
1097 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1098
1099 /*
1100 * Add data to an sk_buff
1101 */
1102 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1103 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1104 {
1105 unsigned char *tmp = skb_tail_pointer(skb);
1106 SKB_LINEAR_ASSERT(skb);
1107 skb->tail += len;
1108 skb->len += len;
1109 return tmp;
1110 }
1111
1112 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1113 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1114 {
1115 skb->data -= len;
1116 skb->len += len;
1117 return skb->data;
1118 }
1119
1120 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1121 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1122 {
1123 skb->len -= len;
1124 BUG_ON(skb->len < skb->data_len);
1125 return skb->data += len;
1126 }
1127
1128 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1129
1130 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1131 {
1132 if (len > skb_headlen(skb) &&
1133 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1134 return NULL;
1135 skb->len -= len;
1136 return skb->data += len;
1137 }
1138
1139 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1140 {
1141 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1142 }
1143
1144 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1145 {
1146 if (likely(len <= skb_headlen(skb)))
1147 return 1;
1148 if (unlikely(len > skb->len))
1149 return 0;
1150 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1151 }
1152
1153 /**
1154 * skb_headroom - bytes at buffer head
1155 * @skb: buffer to check
1156 *
1157 * Return the number of bytes of free space at the head of an &sk_buff.
1158 */
1159 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1160 {
1161 return skb->data - skb->head;
1162 }
1163
1164 /**
1165 * skb_tailroom - bytes at buffer end
1166 * @skb: buffer to check
1167 *
1168 * Return the number of bytes of free space at the tail of an sk_buff
1169 */
1170 static inline int skb_tailroom(const struct sk_buff *skb)
1171 {
1172 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1173 }
1174
1175 /**
1176 * skb_reserve - adjust headroom
1177 * @skb: buffer to alter
1178 * @len: bytes to move
1179 *
1180 * Increase the headroom of an empty &sk_buff by reducing the tail
1181 * room. This is only allowed for an empty buffer.
1182 */
1183 static inline void skb_reserve(struct sk_buff *skb, int len)
1184 {
1185 skb->data += len;
1186 skb->tail += len;
1187 }
1188
1189 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1190 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1191 {
1192 return skb->head + skb->transport_header;
1193 }
1194
1195 static inline void skb_reset_transport_header(struct sk_buff *skb)
1196 {
1197 skb->transport_header = skb->data - skb->head;
1198 }
1199
1200 static inline void skb_set_transport_header(struct sk_buff *skb,
1201 const int offset)
1202 {
1203 skb_reset_transport_header(skb);
1204 skb->transport_header += offset;
1205 }
1206
1207 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1208 {
1209 return skb->head + skb->network_header;
1210 }
1211
1212 static inline void skb_reset_network_header(struct sk_buff *skb)
1213 {
1214 skb->network_header = skb->data - skb->head;
1215 }
1216
1217 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1218 {
1219 skb_reset_network_header(skb);
1220 skb->network_header += offset;
1221 }
1222
1223 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1224 {
1225 return skb->head + skb->mac_header;
1226 }
1227
1228 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1229 {
1230 return skb->mac_header != ~0U;
1231 }
1232
1233 static inline void skb_reset_mac_header(struct sk_buff *skb)
1234 {
1235 skb->mac_header = skb->data - skb->head;
1236 }
1237
1238 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1239 {
1240 skb_reset_mac_header(skb);
1241 skb->mac_header += offset;
1242 }
1243
1244 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1245
1246 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1247 {
1248 return skb->transport_header;
1249 }
1250
1251 static inline void skb_reset_transport_header(struct sk_buff *skb)
1252 {
1253 skb->transport_header = skb->data;
1254 }
1255
1256 static inline void skb_set_transport_header(struct sk_buff *skb,
1257 const int offset)
1258 {
1259 skb->transport_header = skb->data + offset;
1260 }
1261
1262 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1263 {
1264 return skb->network_header;
1265 }
1266
1267 static inline void skb_reset_network_header(struct sk_buff *skb)
1268 {
1269 skb->network_header = skb->data;
1270 }
1271
1272 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1273 {
1274 skb->network_header = skb->data + offset;
1275 }
1276
1277 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1278 {
1279 return skb->mac_header;
1280 }
1281
1282 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1283 {
1284 return skb->mac_header != NULL;
1285 }
1286
1287 static inline void skb_reset_mac_header(struct sk_buff *skb)
1288 {
1289 skb->mac_header = skb->data;
1290 }
1291
1292 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1293 {
1294 skb->mac_header = skb->data + offset;
1295 }
1296 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1297
1298 static inline int skb_transport_offset(const struct sk_buff *skb)
1299 {
1300 return skb_transport_header(skb) - skb->data;
1301 }
1302
1303 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1304 {
1305 return skb->transport_header - skb->network_header;
1306 }
1307
1308 static inline int skb_network_offset(const struct sk_buff *skb)
1309 {
1310 return skb_network_header(skb) - skb->data;
1311 }
1312
1313 /*
1314 * CPUs often take a performance hit when accessing unaligned memory
1315 * locations. The actual performance hit varies, it can be small if the
1316 * hardware handles it or large if we have to take an exception and fix it
1317 * in software.
1318 *
1319 * Since an ethernet header is 14 bytes network drivers often end up with
1320 * the IP header at an unaligned offset. The IP header can be aligned by
1321 * shifting the start of the packet by 2 bytes. Drivers should do this
1322 * with:
1323 *
1324 * skb_reserve(NET_IP_ALIGN);
1325 *
1326 * The downside to this alignment of the IP header is that the DMA is now
1327 * unaligned. On some architectures the cost of an unaligned DMA is high
1328 * and this cost outweighs the gains made by aligning the IP header.
1329 *
1330 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1331 * to be overridden.
1332 */
1333 #ifndef NET_IP_ALIGN
1334 #define NET_IP_ALIGN 2
1335 #endif
1336
1337 /*
1338 * The networking layer reserves some headroom in skb data (via
1339 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1340 * the header has to grow. In the default case, if the header has to grow
1341 * 32 bytes or less we avoid the reallocation.
1342 *
1343 * Unfortunately this headroom changes the DMA alignment of the resulting
1344 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1345 * on some architectures. An architecture can override this value,
1346 * perhaps setting it to a cacheline in size (since that will maintain
1347 * cacheline alignment of the DMA). It must be a power of 2.
1348 *
1349 * Various parts of the networking layer expect at least 32 bytes of
1350 * headroom, you should not reduce this.
1351 */
1352 #ifndef NET_SKB_PAD
1353 #define NET_SKB_PAD 32
1354 #endif
1355
1356 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1357
1358 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1359 {
1360 if (unlikely(skb->data_len)) {
1361 WARN_ON(1);
1362 return;
1363 }
1364 skb->len = len;
1365 skb_set_tail_pointer(skb, len);
1366 }
1367
1368 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1369
1370 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1371 {
1372 if (skb->data_len)
1373 return ___pskb_trim(skb, len);
1374 __skb_trim(skb, len);
1375 return 0;
1376 }
1377
1378 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1379 {
1380 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1381 }
1382
1383 /**
1384 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1385 * @skb: buffer to alter
1386 * @len: new length
1387 *
1388 * This is identical to pskb_trim except that the caller knows that
1389 * the skb is not cloned so we should never get an error due to out-
1390 * of-memory.
1391 */
1392 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1393 {
1394 int err = pskb_trim(skb, len);
1395 BUG_ON(err);
1396 }
1397
1398 /**
1399 * skb_orphan - orphan a buffer
1400 * @skb: buffer to orphan
1401 *
1402 * If a buffer currently has an owner then we call the owner's
1403 * destructor function and make the @skb unowned. The buffer continues
1404 * to exist but is no longer charged to its former owner.
1405 */
1406 static inline void skb_orphan(struct sk_buff *skb)
1407 {
1408 if (skb->destructor)
1409 skb->destructor(skb);
1410 skb->destructor = NULL;
1411 skb->sk = NULL;
1412 }
1413
1414 /**
1415 * __skb_queue_purge - empty a list
1416 * @list: list to empty
1417 *
1418 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1419 * the list and one reference dropped. This function does not take the
1420 * list lock and the caller must hold the relevant locks to use it.
1421 */
1422 extern void skb_queue_purge(struct sk_buff_head *list);
1423 static inline void __skb_queue_purge(struct sk_buff_head *list)
1424 {
1425 struct sk_buff *skb;
1426 while ((skb = __skb_dequeue(list)) != NULL)
1427 kfree_skb(skb);
1428 }
1429
1430 /**
1431 * __dev_alloc_skb - allocate an skbuff for receiving
1432 * @length: length to allocate
1433 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1434 *
1435 * Allocate a new &sk_buff and assign it a usage count of one. The
1436 * buffer has unspecified headroom built in. Users should allocate
1437 * the headroom they think they need without accounting for the
1438 * built in space. The built in space is used for optimisations.
1439 *
1440 * %NULL is returned if there is no free memory.
1441 */
1442 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1443 gfp_t gfp_mask)
1444 {
1445 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1446 if (likely(skb))
1447 skb_reserve(skb, NET_SKB_PAD);
1448 return skb;
1449 }
1450
1451 extern struct sk_buff *dev_alloc_skb(unsigned int length);
1452
1453 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1454 unsigned int length, gfp_t gfp_mask);
1455
1456 /**
1457 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1458 * @dev: network device to receive on
1459 * @length: length to allocate
1460 *
1461 * Allocate a new &sk_buff and assign it a usage count of one. The
1462 * buffer has unspecified headroom built in. Users should allocate
1463 * the headroom they think they need without accounting for the
1464 * built in space. The built in space is used for optimisations.
1465 *
1466 * %NULL is returned if there is no free memory. Although this function
1467 * allocates memory it can be called from an interrupt.
1468 */
1469 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1470 unsigned int length)
1471 {
1472 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1473 }
1474
1475 extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1476
1477 /**
1478 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1479 * @dev: network device to receive on
1480 *
1481 * Allocate a new page node local to the specified device.
1482 *
1483 * %NULL is returned if there is no free memory.
1484 */
1485 static inline struct page *netdev_alloc_page(struct net_device *dev)
1486 {
1487 return __netdev_alloc_page(dev, GFP_ATOMIC);
1488 }
1489
1490 static inline void netdev_free_page(struct net_device *dev, struct page *page)
1491 {
1492 __free_page(page);
1493 }
1494
1495 /**
1496 * skb_clone_writable - is the header of a clone writable
1497 * @skb: buffer to check
1498 * @len: length up to which to write
1499 *
1500 * Returns true if modifying the header part of the cloned buffer
1501 * does not requires the data to be copied.
1502 */
1503 static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1504 {
1505 return !skb_header_cloned(skb) &&
1506 skb_headroom(skb) + len <= skb->hdr_len;
1507 }
1508
1509 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1510 int cloned)
1511 {
1512 int delta = 0;
1513
1514 if (headroom < NET_SKB_PAD)
1515 headroom = NET_SKB_PAD;
1516 if (headroom > skb_headroom(skb))
1517 delta = headroom - skb_headroom(skb);
1518
1519 if (delta || cloned)
1520 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1521 GFP_ATOMIC);
1522 return 0;
1523 }
1524
1525 /**
1526 * skb_cow - copy header of skb when it is required
1527 * @skb: buffer to cow
1528 * @headroom: needed headroom
1529 *
1530 * If the skb passed lacks sufficient headroom or its data part
1531 * is shared, data is reallocated. If reallocation fails, an error
1532 * is returned and original skb is not changed.
1533 *
1534 * The result is skb with writable area skb->head...skb->tail
1535 * and at least @headroom of space at head.
1536 */
1537 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1538 {
1539 return __skb_cow(skb, headroom, skb_cloned(skb));
1540 }
1541
1542 /**
1543 * skb_cow_head - skb_cow but only making the head writable
1544 * @skb: buffer to cow
1545 * @headroom: needed headroom
1546 *
1547 * This function is identical to skb_cow except that we replace the
1548 * skb_cloned check by skb_header_cloned. It should be used when
1549 * you only need to push on some header and do not need to modify
1550 * the data.
1551 */
1552 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1553 {
1554 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1555 }
1556
1557 /**
1558 * skb_padto - pad an skbuff up to a minimal size
1559 * @skb: buffer to pad
1560 * @len: minimal length
1561 *
1562 * Pads up a buffer to ensure the trailing bytes exist and are
1563 * blanked. If the buffer already contains sufficient data it
1564 * is untouched. Otherwise it is extended. Returns zero on
1565 * success. The skb is freed on error.
1566 */
1567
1568 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1569 {
1570 unsigned int size = skb->len;
1571 if (likely(size >= len))
1572 return 0;
1573 return skb_pad(skb, len - size);
1574 }
1575
1576 static inline int skb_add_data(struct sk_buff *skb,
1577 char __user *from, int copy)
1578 {
1579 const int off = skb->len;
1580
1581 if (skb->ip_summed == CHECKSUM_NONE) {
1582 int err = 0;
1583 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1584 copy, 0, &err);
1585 if (!err) {
1586 skb->csum = csum_block_add(skb->csum, csum, off);
1587 return 0;
1588 }
1589 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1590 return 0;
1591
1592 __skb_trim(skb, off);
1593 return -EFAULT;
1594 }
1595
1596 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1597 struct page *page, int off)
1598 {
1599 if (i) {
1600 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1601
1602 return page == frag->page &&
1603 off == frag->page_offset + frag->size;
1604 }
1605 return 0;
1606 }
1607
1608 static inline int __skb_linearize(struct sk_buff *skb)
1609 {
1610 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1611 }
1612
1613 /**
1614 * skb_linearize - convert paged skb to linear one
1615 * @skb: buffer to linarize
1616 *
1617 * If there is no free memory -ENOMEM is returned, otherwise zero
1618 * is returned and the old skb data released.
1619 */
1620 static inline int skb_linearize(struct sk_buff *skb)
1621 {
1622 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1623 }
1624
1625 /**
1626 * skb_linearize_cow - make sure skb is linear and writable
1627 * @skb: buffer to process
1628 *
1629 * If there is no free memory -ENOMEM is returned, otherwise zero
1630 * is returned and the old skb data released.
1631 */
1632 static inline int skb_linearize_cow(struct sk_buff *skb)
1633 {
1634 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1635 __skb_linearize(skb) : 0;
1636 }
1637
1638 /**
1639 * skb_postpull_rcsum - update checksum for received skb after pull
1640 * @skb: buffer to update
1641 * @start: start of data before pull
1642 * @len: length of data pulled
1643 *
1644 * After doing a pull on a received packet, you need to call this to
1645 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1646 * CHECKSUM_NONE so that it can be recomputed from scratch.
1647 */
1648
1649 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1650 const void *start, unsigned int len)
1651 {
1652 if (skb->ip_summed == CHECKSUM_COMPLETE)
1653 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1654 }
1655
1656 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1657
1658 /**
1659 * pskb_trim_rcsum - trim received skb and update checksum
1660 * @skb: buffer to trim
1661 * @len: new length
1662 *
1663 * This is exactly the same as pskb_trim except that it ensures the
1664 * checksum of received packets are still valid after the operation.
1665 */
1666
1667 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1668 {
1669 if (likely(len >= skb->len))
1670 return 0;
1671 if (skb->ip_summed == CHECKSUM_COMPLETE)
1672 skb->ip_summed = CHECKSUM_NONE;
1673 return __pskb_trim(skb, len);
1674 }
1675
1676 #define skb_queue_walk(queue, skb) \
1677 for (skb = (queue)->next; \
1678 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1679 skb = skb->next)
1680
1681 #define skb_queue_walk_safe(queue, skb, tmp) \
1682 for (skb = (queue)->next, tmp = skb->next; \
1683 skb != (struct sk_buff *)(queue); \
1684 skb = tmp, tmp = skb->next)
1685
1686 #define skb_queue_walk_from(queue, skb) \
1687 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1688 skb = skb->next)
1689
1690 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1691 for (tmp = skb->next; \
1692 skb != (struct sk_buff *)(queue); \
1693 skb = tmp, tmp = skb->next)
1694
1695 #define skb_queue_reverse_walk(queue, skb) \
1696 for (skb = (queue)->prev; \
1697 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1698 skb = skb->prev)
1699
1700
1701 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1702 int *peeked, int *err);
1703 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1704 int noblock, int *err);
1705 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1706 struct poll_table_struct *wait);
1707 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1708 int offset, struct iovec *to,
1709 int size);
1710 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1711 int hlen,
1712 struct iovec *iov);
1713 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1714 int offset,
1715 struct iovec *from,
1716 int len);
1717 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1718 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1719 unsigned int flags);
1720 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1721 int len, __wsum csum);
1722 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1723 void *to, int len);
1724 extern int skb_store_bits(struct sk_buff *skb, int offset,
1725 const void *from, int len);
1726 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1727 int offset, u8 *to, int len,
1728 __wsum csum);
1729 extern int skb_splice_bits(struct sk_buff *skb,
1730 unsigned int offset,
1731 struct pipe_inode_info *pipe,
1732 unsigned int len,
1733 unsigned int flags);
1734 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1735 extern void skb_split(struct sk_buff *skb,
1736 struct sk_buff *skb1, const u32 len);
1737 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1738 int shiftlen);
1739
1740 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1741
1742 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1743 int len, void *buffer)
1744 {
1745 int hlen = skb_headlen(skb);
1746
1747 if (hlen - offset >= len)
1748 return skb->data + offset;
1749
1750 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1751 return NULL;
1752
1753 return buffer;
1754 }
1755
1756 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1757 void *to,
1758 const unsigned int len)
1759 {
1760 memcpy(to, skb->data, len);
1761 }
1762
1763 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1764 const int offset, void *to,
1765 const unsigned int len)
1766 {
1767 memcpy(to, skb->data + offset, len);
1768 }
1769
1770 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1771 const void *from,
1772 const unsigned int len)
1773 {
1774 memcpy(skb->data, from, len);
1775 }
1776
1777 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1778 const int offset,
1779 const void *from,
1780 const unsigned int len)
1781 {
1782 memcpy(skb->data + offset, from, len);
1783 }
1784
1785 extern void skb_init(void);
1786
1787 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1788 {
1789 return skb->tstamp;
1790 }
1791
1792 /**
1793 * skb_get_timestamp - get timestamp from a skb
1794 * @skb: skb to get stamp from
1795 * @stamp: pointer to struct timeval to store stamp in
1796 *
1797 * Timestamps are stored in the skb as offsets to a base timestamp.
1798 * This function converts the offset back to a struct timeval and stores
1799 * it in stamp.
1800 */
1801 static inline void skb_get_timestamp(const struct sk_buff *skb,
1802 struct timeval *stamp)
1803 {
1804 *stamp = ktime_to_timeval(skb->tstamp);
1805 }
1806
1807 static inline void skb_get_timestampns(const struct sk_buff *skb,
1808 struct timespec *stamp)
1809 {
1810 *stamp = ktime_to_timespec(skb->tstamp);
1811 }
1812
1813 static inline void __net_timestamp(struct sk_buff *skb)
1814 {
1815 skb->tstamp = ktime_get_real();
1816 }
1817
1818 static inline ktime_t net_timedelta(ktime_t t)
1819 {
1820 return ktime_sub(ktime_get_real(), t);
1821 }
1822
1823 static inline ktime_t net_invalid_timestamp(void)
1824 {
1825 return ktime_set(0, 0);
1826 }
1827
1828 /**
1829 * skb_tstamp_tx - queue clone of skb with send time stamps
1830 * @orig_skb: the original outgoing packet
1831 * @hwtstamps: hardware time stamps, may be NULL if not available
1832 *
1833 * If the skb has a socket associated, then this function clones the
1834 * skb (thus sharing the actual data and optional structures), stores
1835 * the optional hardware time stamping information (if non NULL) or
1836 * generates a software time stamp (otherwise), then queues the clone
1837 * to the error queue of the socket. Errors are silently ignored.
1838 */
1839 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
1840 struct skb_shared_hwtstamps *hwtstamps);
1841
1842 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
1843 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
1844
1845 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
1846 {
1847 return skb->ip_summed & CHECKSUM_UNNECESSARY;
1848 }
1849
1850 /**
1851 * skb_checksum_complete - Calculate checksum of an entire packet
1852 * @skb: packet to process
1853 *
1854 * This function calculates the checksum over the entire packet plus
1855 * the value of skb->csum. The latter can be used to supply the
1856 * checksum of a pseudo header as used by TCP/UDP. It returns the
1857 * checksum.
1858 *
1859 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1860 * this function can be used to verify that checksum on received
1861 * packets. In that case the function should return zero if the
1862 * checksum is correct. In particular, this function will return zero
1863 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1864 * hardware has already verified the correctness of the checksum.
1865 */
1866 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
1867 {
1868 return skb_csum_unnecessary(skb) ?
1869 0 : __skb_checksum_complete(skb);
1870 }
1871
1872 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1873 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
1874 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1875 {
1876 if (nfct && atomic_dec_and_test(&nfct->use))
1877 nf_conntrack_destroy(nfct);
1878 }
1879 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1880 {
1881 if (nfct)
1882 atomic_inc(&nfct->use);
1883 }
1884 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1885 {
1886 if (skb)
1887 atomic_inc(&skb->users);
1888 }
1889 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1890 {
1891 if (skb)
1892 kfree_skb(skb);
1893 }
1894 #endif
1895 #ifdef CONFIG_BRIDGE_NETFILTER
1896 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1897 {
1898 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1899 kfree(nf_bridge);
1900 }
1901 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1902 {
1903 if (nf_bridge)
1904 atomic_inc(&nf_bridge->use);
1905 }
1906 #endif /* CONFIG_BRIDGE_NETFILTER */
1907 static inline void nf_reset(struct sk_buff *skb)
1908 {
1909 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1910 nf_conntrack_put(skb->nfct);
1911 skb->nfct = NULL;
1912 nf_conntrack_put_reasm(skb->nfct_reasm);
1913 skb->nfct_reasm = NULL;
1914 #endif
1915 #ifdef CONFIG_BRIDGE_NETFILTER
1916 nf_bridge_put(skb->nf_bridge);
1917 skb->nf_bridge = NULL;
1918 #endif
1919 }
1920
1921 /* Note: This doesn't put any conntrack and bridge info in dst. */
1922 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1923 {
1924 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1925 dst->nfct = src->nfct;
1926 nf_conntrack_get(src->nfct);
1927 dst->nfctinfo = src->nfctinfo;
1928 dst->nfct_reasm = src->nfct_reasm;
1929 nf_conntrack_get_reasm(src->nfct_reasm);
1930 #endif
1931 #ifdef CONFIG_BRIDGE_NETFILTER
1932 dst->nf_bridge = src->nf_bridge;
1933 nf_bridge_get(src->nf_bridge);
1934 #endif
1935 }
1936
1937 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1938 {
1939 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1940 nf_conntrack_put(dst->nfct);
1941 nf_conntrack_put_reasm(dst->nfct_reasm);
1942 #endif
1943 #ifdef CONFIG_BRIDGE_NETFILTER
1944 nf_bridge_put(dst->nf_bridge);
1945 #endif
1946 __nf_copy(dst, src);
1947 }
1948
1949 #ifdef CONFIG_NETWORK_SECMARK
1950 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1951 {
1952 to->secmark = from->secmark;
1953 }
1954
1955 static inline void skb_init_secmark(struct sk_buff *skb)
1956 {
1957 skb->secmark = 0;
1958 }
1959 #else
1960 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1961 { }
1962
1963 static inline void skb_init_secmark(struct sk_buff *skb)
1964 { }
1965 #endif
1966
1967 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
1968 {
1969 skb->queue_mapping = queue_mapping;
1970 }
1971
1972 static inline u16 skb_get_queue_mapping(struct sk_buff *skb)
1973 {
1974 return skb->queue_mapping;
1975 }
1976
1977 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
1978 {
1979 to->queue_mapping = from->queue_mapping;
1980 }
1981
1982 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
1983 {
1984 skb->queue_mapping = rx_queue + 1;
1985 }
1986
1987 static inline u16 skb_get_rx_queue(struct sk_buff *skb)
1988 {
1989 return skb->queue_mapping - 1;
1990 }
1991
1992 static inline bool skb_rx_queue_recorded(struct sk_buff *skb)
1993 {
1994 return (skb->queue_mapping != 0);
1995 }
1996
1997 #ifdef CONFIG_XFRM
1998 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
1999 {
2000 return skb->sp;
2001 }
2002 #else
2003 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2004 {
2005 return NULL;
2006 }
2007 #endif
2008
2009 static inline int skb_is_gso(const struct sk_buff *skb)
2010 {
2011 return skb_shinfo(skb)->gso_size;
2012 }
2013
2014 static inline int skb_is_gso_v6(const struct sk_buff *skb)
2015 {
2016 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2017 }
2018
2019 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2020
2021 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2022 {
2023 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2024 * wanted then gso_type will be set. */
2025 struct skb_shared_info *shinfo = skb_shinfo(skb);
2026 if (shinfo->gso_size != 0 && unlikely(shinfo->gso_type == 0)) {
2027 __skb_warn_lro_forwarding(skb);
2028 return true;
2029 }
2030 return false;
2031 }
2032
2033 static inline void skb_forward_csum(struct sk_buff *skb)
2034 {
2035 /* Unfortunately we don't support this one. Any brave souls? */
2036 if (skb->ip_summed == CHECKSUM_COMPLETE)
2037 skb->ip_summed = CHECKSUM_NONE;
2038 }
2039
2040 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2041 #endif /* __KERNEL__ */
2042 #endif /* _LINUX_SKBUFF_H */
This page took 0.110031 seconds and 5 git commands to generate.