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