xref: /linux-6.15/include/linux/skbuff.h (revision 54a8a222)
1 /*
2  *	Definitions for the 'struct sk_buff' memory handlers.
3  *
4  *	Authors:
5  *		Alan Cox, <[email protected]>
6  *		Florian La Roche, <[email protected]>
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/config.h>
18 #include <linux/kernel.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/cache.h>
22 
23 #include <asm/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
26 #include <linux/mm.h>
27 #include <linux/highmem.h>
28 #include <linux/poll.h>
29 #include <linux/net.h>
30 #include <linux/textsearch.h>
31 #include <net/checksum.h>
32 
33 #define HAVE_ALLOC_SKB		/* For the drivers to know */
34 #define HAVE_ALIGNABLE_SKB	/* Ditto 8)		   */
35 #define SLAB_SKB 		/* Slabified skbuffs 	   */
36 
37 #define CHECKSUM_NONE 0
38 #define CHECKSUM_HW 1
39 #define CHECKSUM_UNNECESSARY 2
40 
41 #define SKB_DATA_ALIGN(X)	(((X) + (SMP_CACHE_BYTES - 1)) & \
42 				 ~(SMP_CACHE_BYTES - 1))
43 #define SKB_MAX_ORDER(X, ORDER)	(((PAGE_SIZE << (ORDER)) - (X) - \
44 				  sizeof(struct skb_shared_info)) & \
45 				  ~(SMP_CACHE_BYTES - 1))
46 #define SKB_MAX_HEAD(X)		(SKB_MAX_ORDER((X), 0))
47 #define SKB_MAX_ALLOC		(SKB_MAX_ORDER(0, 2))
48 
49 /* A. Checksumming of received packets by device.
50  *
51  *	NONE: device failed to checksum this packet.
52  *		skb->csum is undefined.
53  *
54  *	UNNECESSARY: device parsed packet and wouldbe verified checksum.
55  *		skb->csum is undefined.
56  *	      It is bad option, but, unfortunately, many of vendors do this.
57  *	      Apparently with secret goal to sell you new device, when you
58  *	      will add new protocol to your host. F.e. IPv6. 8)
59  *
60  *	HW: the most generic way. Device supplied checksum of _all_
61  *	    the packet as seen by netif_rx in skb->csum.
62  *	    NOTE: Even if device supports only some protocols, but
63  *	    is able to produce some skb->csum, it MUST use HW,
64  *	    not UNNECESSARY.
65  *
66  * B. Checksumming on output.
67  *
68  *	NONE: skb is checksummed by protocol or csum is not required.
69  *
70  *	HW: device is required to csum packet as seen by hard_start_xmit
71  *	from skb->h.raw to the end and to record the checksum
72  *	at skb->h.raw+skb->csum.
73  *
74  *	Device must show its capabilities in dev->features, set
75  *	at device setup time.
76  *	NETIF_F_HW_CSUM	- it is clever device, it is able to checksum
77  *			  everything.
78  *	NETIF_F_NO_CSUM - loopback or reliable single hop media.
79  *	NETIF_F_IP_CSUM - device is dumb. It is able to csum only
80  *			  TCP/UDP over IPv4. Sigh. Vendors like this
81  *			  way by an unknown reason. Though, see comment above
82  *			  about CHECKSUM_UNNECESSARY. 8)
83  *
84  *	Any questions? No questions, good. 		--ANK
85  */
86 
87 struct net_device;
88 
89 #ifdef CONFIG_NETFILTER
90 struct nf_conntrack {
91 	atomic_t use;
92 	void (*destroy)(struct nf_conntrack *);
93 };
94 
95 #ifdef CONFIG_BRIDGE_NETFILTER
96 struct nf_bridge_info {
97 	atomic_t use;
98 	struct net_device *physindev;
99 	struct net_device *physoutdev;
100 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
101 	struct net_device *netoutdev;
102 #endif
103 	unsigned int mask;
104 	unsigned long data[32 / sizeof(unsigned long)];
105 };
106 #endif
107 
108 #endif
109 
110 struct sk_buff_head {
111 	/* These two members must be first. */
112 	struct sk_buff	*next;
113 	struct sk_buff	*prev;
114 
115 	__u32		qlen;
116 	spinlock_t	lock;
117 };
118 
119 struct sk_buff;
120 
121 /* To allow 64K frame to be packed as single skb without frag_list */
122 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
123 
124 typedef struct skb_frag_struct skb_frag_t;
125 
126 struct skb_frag_struct {
127 	struct page *page;
128 	__u16 page_offset;
129 	__u16 size;
130 };
131 
132 /* This data is invariant across clones and lives at
133  * the end of the header data, ie. at skb->end.
134  */
135 struct skb_shared_info {
136 	atomic_t	dataref;
137 	unsigned int	nr_frags;
138 	unsigned short	tso_size;
139 	unsigned short	tso_segs;
140 	struct sk_buff	*frag_list;
141 	skb_frag_t	frags[MAX_SKB_FRAGS];
142 };
143 
144 /* We divide dataref into two halves.  The higher 16 bits hold references
145  * to the payload part of skb->data.  The lower 16 bits hold references to
146  * the entire skb->data.  It is up to the users of the skb to agree on
147  * where the payload starts.
148  *
149  * All users must obey the rule that the skb->data reference count must be
150  * greater than or equal to the payload reference count.
151  *
152  * Holding a reference to the payload part means that the user does not
153  * care about modifications to the header part of skb->data.
154  */
155 #define SKB_DATAREF_SHIFT 16
156 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
157 
158 struct skb_timeval {
159 	u32	off_sec;
160 	u32	off_usec;
161 };
162 
163 
164 enum {
165 	SKB_FCLONE_UNAVAILABLE,
166 	SKB_FCLONE_ORIG,
167 	SKB_FCLONE_CLONE,
168 };
169 
170 /**
171  *	struct sk_buff - socket buffer
172  *	@next: Next buffer in list
173  *	@prev: Previous buffer in list
174  *	@list: List we are on
175  *	@sk: Socket we are owned by
176  *	@tstamp: Time we arrived
177  *	@dev: Device we arrived on/are leaving by
178  *	@input_dev: Device we arrived on
179  *	@h: Transport layer header
180  *	@nh: Network layer header
181  *	@mac: Link layer header
182  *	@dst: destination entry
183  *	@sp: the security path, used for xfrm
184  *	@cb: Control buffer. Free for use by every layer. Put private vars here
185  *	@len: Length of actual data
186  *	@data_len: Data length
187  *	@mac_len: Length of link layer header
188  *	@csum: Checksum
189  *	@local_df: allow local fragmentation
190  *	@cloned: Head may be cloned (check refcnt to be sure)
191  *	@nohdr: Payload reference only, must not modify header
192  *	@pkt_type: Packet class
193  *	@ip_summed: Driver fed us an IP checksum
194  *	@priority: Packet queueing priority
195  *	@users: User count - see {datagram,tcp}.c
196  *	@protocol: Packet protocol from driver
197  *	@truesize: Buffer size
198  *	@head: Head of buffer
199  *	@data: Data head pointer
200  *	@tail: Tail pointer
201  *	@end: End pointer
202  *	@destructor: Destruct function
203  *	@nfmark: Can be used for communication between hooks
204  *	@nfct: Associated connection, if any
205  *	@nfctinfo: Relationship of this skb to the connection
206  *	@nf_bridge: Saved data about a bridged frame - see br_netfilter.c
207  *	@tc_index: Traffic control index
208  *	@tc_verd: traffic control verdict
209  */
210 
211 struct sk_buff {
212 	/* These two members must be first. */
213 	struct sk_buff		*next;
214 	struct sk_buff		*prev;
215 
216 	struct sock		*sk;
217 	struct skb_timeval	tstamp;
218 	struct net_device	*dev;
219 	struct net_device	*input_dev;
220 
221 	union {
222 		struct tcphdr	*th;
223 		struct udphdr	*uh;
224 		struct icmphdr	*icmph;
225 		struct igmphdr	*igmph;
226 		struct iphdr	*ipiph;
227 		struct ipv6hdr	*ipv6h;
228 		unsigned char	*raw;
229 	} h;
230 
231 	union {
232 		struct iphdr	*iph;
233 		struct ipv6hdr	*ipv6h;
234 		struct arphdr	*arph;
235 		unsigned char	*raw;
236 	} nh;
237 
238 	union {
239 	  	unsigned char 	*raw;
240 	} mac;
241 
242 	struct  dst_entry	*dst;
243 	struct	sec_path	*sp;
244 
245 	/*
246 	 * This is the control buffer. It is free to use for every
247 	 * layer. Please put your private variables there. If you
248 	 * want to keep them across layers you have to do a skb_clone()
249 	 * first. This is owned by whoever has the skb queued ATM.
250 	 */
251 	char			cb[40];
252 
253 	unsigned int		len,
254 				data_len,
255 				mac_len,
256 				csum;
257 	__u32			priority;
258 	__u8			local_df:1,
259 				cloned:1,
260 				ip_summed:2,
261 				nohdr:1,
262 				nfctinfo:3;
263 	__u8			pkt_type:3,
264 				fclone:2;
265 	__be16			protocol;
266 
267 	void			(*destructor)(struct sk_buff *skb);
268 #ifdef CONFIG_NETFILTER
269 	__u32			nfmark;
270 	struct nf_conntrack	*nfct;
271 #if defined(CONFIG_IP_VS) || defined(CONFIG_IP_VS_MODULE)
272 	__u8			ipvs_property:1;
273 #endif
274 #ifdef CONFIG_BRIDGE_NETFILTER
275 	struct nf_bridge_info	*nf_bridge;
276 #endif
277 #endif /* CONFIG_NETFILTER */
278 #ifdef CONFIG_NET_SCHED
279 	__u16			tc_index;	/* traffic control index */
280 #ifdef CONFIG_NET_CLS_ACT
281 	__u16			tc_verd;	/* traffic control verdict */
282 #endif
283 #endif
284 
285 
286 	/* These elements must be at the end, see alloc_skb() for details.  */
287 	unsigned int		truesize;
288 	atomic_t		users;
289 	unsigned char		*head,
290 				*data,
291 				*tail,
292 				*end;
293 };
294 
295 #ifdef __KERNEL__
296 /*
297  *	Handling routines are only of interest to the kernel
298  */
299 #include <linux/slab.h>
300 
301 #include <asm/system.h>
302 
303 extern void	       __kfree_skb(struct sk_buff *skb);
304 extern struct sk_buff *__alloc_skb(unsigned int size,
305 				   gfp_t priority, int fclone);
306 static inline struct sk_buff *alloc_skb(unsigned int size,
307 					gfp_t priority)
308 {
309 	return __alloc_skb(size, priority, 0);
310 }
311 
312 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
313 					       gfp_t priority)
314 {
315 	return __alloc_skb(size, priority, 1);
316 }
317 
318 extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
319 					    unsigned int size,
320 					    gfp_t priority);
321 extern void	       kfree_skbmem(struct sk_buff *skb);
322 extern struct sk_buff *skb_clone(struct sk_buff *skb,
323 				 gfp_t priority);
324 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
325 				gfp_t priority);
326 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
327 				 gfp_t gfp_mask);
328 extern int	       pskb_expand_head(struct sk_buff *skb,
329 					int nhead, int ntail,
330 					gfp_t gfp_mask);
331 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
332 					    unsigned int headroom);
333 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
334 				       int newheadroom, int newtailroom,
335 				       gfp_t priority);
336 extern struct sk_buff *		skb_pad(struct sk_buff *skb, int pad);
337 #define dev_kfree_skb(a)	kfree_skb(a)
338 extern void	      skb_over_panic(struct sk_buff *skb, int len,
339 				     void *here);
340 extern void	      skb_under_panic(struct sk_buff *skb, int len,
341 				      void *here);
342 
343 struct skb_seq_state
344 {
345 	__u32		lower_offset;
346 	__u32		upper_offset;
347 	__u32		frag_idx;
348 	__u32		stepped_offset;
349 	struct sk_buff	*root_skb;
350 	struct sk_buff	*cur_skb;
351 	__u8		*frag_data;
352 };
353 
354 extern void	      skb_prepare_seq_read(struct sk_buff *skb,
355 					   unsigned int from, unsigned int to,
356 					   struct skb_seq_state *st);
357 extern unsigned int   skb_seq_read(unsigned int consumed, const u8 **data,
358 				   struct skb_seq_state *st);
359 extern void	      skb_abort_seq_read(struct skb_seq_state *st);
360 
361 extern unsigned int   skb_find_text(struct sk_buff *skb, unsigned int from,
362 				    unsigned int to, struct ts_config *config,
363 				    struct ts_state *state);
364 
365 /* Internal */
366 #define skb_shinfo(SKB)		((struct skb_shared_info *)((SKB)->end))
367 
368 /**
369  *	skb_queue_empty - check if a queue is empty
370  *	@list: queue head
371  *
372  *	Returns true if the queue is empty, false otherwise.
373  */
374 static inline int skb_queue_empty(const struct sk_buff_head *list)
375 {
376 	return list->next == (struct sk_buff *)list;
377 }
378 
379 /**
380  *	skb_get - reference buffer
381  *	@skb: buffer to reference
382  *
383  *	Makes another reference to a socket buffer and returns a pointer
384  *	to the buffer.
385  */
386 static inline struct sk_buff *skb_get(struct sk_buff *skb)
387 {
388 	atomic_inc(&skb->users);
389 	return skb;
390 }
391 
392 /*
393  * If users == 1, we are the only owner and are can avoid redundant
394  * atomic change.
395  */
396 
397 /**
398  *	kfree_skb - free an sk_buff
399  *	@skb: buffer to free
400  *
401  *	Drop a reference to the buffer and free it if the usage count has
402  *	hit zero.
403  */
404 static inline void kfree_skb(struct sk_buff *skb)
405 {
406 	if (likely(atomic_read(&skb->users) == 1))
407 		smp_rmb();
408 	else if (likely(!atomic_dec_and_test(&skb->users)))
409 		return;
410 	__kfree_skb(skb);
411 }
412 
413 /**
414  *	skb_cloned - is the buffer a clone
415  *	@skb: buffer to check
416  *
417  *	Returns true if the buffer was generated with skb_clone() and is
418  *	one of multiple shared copies of the buffer. Cloned buffers are
419  *	shared data so must not be written to under normal circumstances.
420  */
421 static inline int skb_cloned(const struct sk_buff *skb)
422 {
423 	return skb->cloned &&
424 	       (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
425 }
426 
427 /**
428  *	skb_header_cloned - is the header a clone
429  *	@skb: buffer to check
430  *
431  *	Returns true if modifying the header part of the buffer requires
432  *	the data to be copied.
433  */
434 static inline int skb_header_cloned(const struct sk_buff *skb)
435 {
436 	int dataref;
437 
438 	if (!skb->cloned)
439 		return 0;
440 
441 	dataref = atomic_read(&skb_shinfo(skb)->dataref);
442 	dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
443 	return dataref != 1;
444 }
445 
446 /**
447  *	skb_header_release - release reference to header
448  *	@skb: buffer to operate on
449  *
450  *	Drop a reference to the header part of the buffer.  This is done
451  *	by acquiring a payload reference.  You must not read from the header
452  *	part of skb->data after this.
453  */
454 static inline void skb_header_release(struct sk_buff *skb)
455 {
456 	BUG_ON(skb->nohdr);
457 	skb->nohdr = 1;
458 	atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
459 }
460 
461 /**
462  *	skb_shared - is the buffer shared
463  *	@skb: buffer to check
464  *
465  *	Returns true if more than one person has a reference to this
466  *	buffer.
467  */
468 static inline int skb_shared(const struct sk_buff *skb)
469 {
470 	return atomic_read(&skb->users) != 1;
471 }
472 
473 /**
474  *	skb_share_check - check if buffer is shared and if so clone it
475  *	@skb: buffer to check
476  *	@pri: priority for memory allocation
477  *
478  *	If the buffer is shared the buffer is cloned and the old copy
479  *	drops a reference. A new clone with a single reference is returned.
480  *	If the buffer is not shared the original buffer is returned. When
481  *	being called from interrupt status or with spinlocks held pri must
482  *	be GFP_ATOMIC.
483  *
484  *	NULL is returned on a memory allocation failure.
485  */
486 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
487 					      gfp_t pri)
488 {
489 	might_sleep_if(pri & __GFP_WAIT);
490 	if (skb_shared(skb)) {
491 		struct sk_buff *nskb = skb_clone(skb, pri);
492 		kfree_skb(skb);
493 		skb = nskb;
494 	}
495 	return skb;
496 }
497 
498 /*
499  *	Copy shared buffers into a new sk_buff. We effectively do COW on
500  *	packets to handle cases where we have a local reader and forward
501  *	and a couple of other messy ones. The normal one is tcpdumping
502  *	a packet thats being forwarded.
503  */
504 
505 /**
506  *	skb_unshare - make a copy of a shared buffer
507  *	@skb: buffer to check
508  *	@pri: priority for memory allocation
509  *
510  *	If the socket buffer is a clone then this function creates a new
511  *	copy of the data, drops a reference count on the old copy and returns
512  *	the new copy with the reference count at 1. If the buffer is not a clone
513  *	the original buffer is returned. When called with a spinlock held or
514  *	from interrupt state @pri must be %GFP_ATOMIC
515  *
516  *	%NULL is returned on a memory allocation failure.
517  */
518 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
519 					  gfp_t pri)
520 {
521 	might_sleep_if(pri & __GFP_WAIT);
522 	if (skb_cloned(skb)) {
523 		struct sk_buff *nskb = skb_copy(skb, pri);
524 		kfree_skb(skb);	/* Free our shared copy */
525 		skb = nskb;
526 	}
527 	return skb;
528 }
529 
530 /**
531  *	skb_peek
532  *	@list_: list to peek at
533  *
534  *	Peek an &sk_buff. Unlike most other operations you _MUST_
535  *	be careful with this one. A peek leaves the buffer on the
536  *	list and someone else may run off with it. You must hold
537  *	the appropriate locks or have a private queue to do this.
538  *
539  *	Returns %NULL for an empty list or a pointer to the head element.
540  *	The reference count is not incremented and the reference is therefore
541  *	volatile. Use with caution.
542  */
543 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
544 {
545 	struct sk_buff *list = ((struct sk_buff *)list_)->next;
546 	if (list == (struct sk_buff *)list_)
547 		list = NULL;
548 	return list;
549 }
550 
551 /**
552  *	skb_peek_tail
553  *	@list_: list to peek at
554  *
555  *	Peek an &sk_buff. Unlike most other operations you _MUST_
556  *	be careful with this one. A peek leaves the buffer on the
557  *	list and someone else may run off with it. You must hold
558  *	the appropriate locks or have a private queue to do this.
559  *
560  *	Returns %NULL for an empty list or a pointer to the tail element.
561  *	The reference count is not incremented and the reference is therefore
562  *	volatile. Use with caution.
563  */
564 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
565 {
566 	struct sk_buff *list = ((struct sk_buff *)list_)->prev;
567 	if (list == (struct sk_buff *)list_)
568 		list = NULL;
569 	return list;
570 }
571 
572 /**
573  *	skb_queue_len	- get queue length
574  *	@list_: list to measure
575  *
576  *	Return the length of an &sk_buff queue.
577  */
578 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
579 {
580 	return list_->qlen;
581 }
582 
583 static inline void skb_queue_head_init(struct sk_buff_head *list)
584 {
585 	spin_lock_init(&list->lock);
586 	list->prev = list->next = (struct sk_buff *)list;
587 	list->qlen = 0;
588 }
589 
590 /*
591  *	Insert an sk_buff at the start of a list.
592  *
593  *	The "__skb_xxxx()" functions are the non-atomic ones that
594  *	can only be called with interrupts disabled.
595  */
596 
597 /**
598  *	__skb_queue_head - queue a buffer at the list head
599  *	@list: list to use
600  *	@newsk: buffer to queue
601  *
602  *	Queue a buffer at the start of a list. This function takes no locks
603  *	and you must therefore hold required locks before calling it.
604  *
605  *	A buffer cannot be placed on two lists at the same time.
606  */
607 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
608 static inline void __skb_queue_head(struct sk_buff_head *list,
609 				    struct sk_buff *newsk)
610 {
611 	struct sk_buff *prev, *next;
612 
613 	list->qlen++;
614 	prev = (struct sk_buff *)list;
615 	next = prev->next;
616 	newsk->next = next;
617 	newsk->prev = prev;
618 	next->prev  = prev->next = newsk;
619 }
620 
621 /**
622  *	__skb_queue_tail - queue a buffer at the list tail
623  *	@list: list to use
624  *	@newsk: buffer to queue
625  *
626  *	Queue a buffer at the end of a list. This function takes no locks
627  *	and you must therefore hold required locks before calling it.
628  *
629  *	A buffer cannot be placed on two lists at the same time.
630  */
631 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
632 static inline void __skb_queue_tail(struct sk_buff_head *list,
633 				   struct sk_buff *newsk)
634 {
635 	struct sk_buff *prev, *next;
636 
637 	list->qlen++;
638 	next = (struct sk_buff *)list;
639 	prev = next->prev;
640 	newsk->next = next;
641 	newsk->prev = prev;
642 	next->prev  = prev->next = newsk;
643 }
644 
645 
646 /**
647  *	__skb_dequeue - remove from the head of the queue
648  *	@list: list to dequeue from
649  *
650  *	Remove the head of the list. This function does not take any locks
651  *	so must be used with appropriate locks held only. The head item is
652  *	returned or %NULL if the list is empty.
653  */
654 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
655 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
656 {
657 	struct sk_buff *next, *prev, *result;
658 
659 	prev = (struct sk_buff *) list;
660 	next = prev->next;
661 	result = NULL;
662 	if (next != prev) {
663 		result	     = next;
664 		next	     = next->next;
665 		list->qlen--;
666 		next->prev   = prev;
667 		prev->next   = next;
668 		result->next = result->prev = NULL;
669 	}
670 	return result;
671 }
672 
673 
674 /*
675  *	Insert a packet on a list.
676  */
677 extern void        skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
678 static inline void __skb_insert(struct sk_buff *newsk,
679 				struct sk_buff *prev, struct sk_buff *next,
680 				struct sk_buff_head *list)
681 {
682 	newsk->next = next;
683 	newsk->prev = prev;
684 	next->prev  = prev->next = newsk;
685 	list->qlen++;
686 }
687 
688 /*
689  *	Place a packet after a given packet in a list.
690  */
691 extern void	   skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
692 static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list)
693 {
694 	__skb_insert(newsk, old, old->next, list);
695 }
696 
697 /*
698  * remove sk_buff from list. _Must_ be called atomically, and with
699  * the list known..
700  */
701 extern void	   skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
702 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
703 {
704 	struct sk_buff *next, *prev;
705 
706 	list->qlen--;
707 	next	   = skb->next;
708 	prev	   = skb->prev;
709 	skb->next  = skb->prev = NULL;
710 	next->prev = prev;
711 	prev->next = next;
712 }
713 
714 
715 /* XXX: more streamlined implementation */
716 
717 /**
718  *	__skb_dequeue_tail - remove from the tail of the queue
719  *	@list: list to dequeue from
720  *
721  *	Remove the tail of the list. This function does not take any locks
722  *	so must be used with appropriate locks held only. The tail item is
723  *	returned or %NULL if the list is empty.
724  */
725 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
726 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
727 {
728 	struct sk_buff *skb = skb_peek_tail(list);
729 	if (skb)
730 		__skb_unlink(skb, list);
731 	return skb;
732 }
733 
734 
735 static inline int skb_is_nonlinear(const struct sk_buff *skb)
736 {
737 	return skb->data_len;
738 }
739 
740 static inline unsigned int skb_headlen(const struct sk_buff *skb)
741 {
742 	return skb->len - skb->data_len;
743 }
744 
745 static inline int skb_pagelen(const struct sk_buff *skb)
746 {
747 	int i, len = 0;
748 
749 	for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
750 		len += skb_shinfo(skb)->frags[i].size;
751 	return len + skb_headlen(skb);
752 }
753 
754 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
755 				      struct page *page, int off, int size)
756 {
757 	skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
758 
759 	frag->page		  = page;
760 	frag->page_offset	  = off;
761 	frag->size		  = size;
762 	skb_shinfo(skb)->nr_frags = i + 1;
763 }
764 
765 #define SKB_PAGE_ASSERT(skb) 	BUG_ON(skb_shinfo(skb)->nr_frags)
766 #define SKB_FRAG_ASSERT(skb) 	BUG_ON(skb_shinfo(skb)->frag_list)
767 #define SKB_LINEAR_ASSERT(skb)  BUG_ON(skb_is_nonlinear(skb))
768 
769 /*
770  *	Add data to an sk_buff
771  */
772 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
773 {
774 	unsigned char *tmp = skb->tail;
775 	SKB_LINEAR_ASSERT(skb);
776 	skb->tail += len;
777 	skb->len  += len;
778 	return tmp;
779 }
780 
781 /**
782  *	skb_put - add data to a buffer
783  *	@skb: buffer to use
784  *	@len: amount of data to add
785  *
786  *	This function extends the used data area of the buffer. If this would
787  *	exceed the total buffer size the kernel will panic. A pointer to the
788  *	first byte of the extra data is returned.
789  */
790 static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
791 {
792 	unsigned char *tmp = skb->tail;
793 	SKB_LINEAR_ASSERT(skb);
794 	skb->tail += len;
795 	skb->len  += len;
796 	if (unlikely(skb->tail>skb->end))
797 		skb_over_panic(skb, len, current_text_addr());
798 	return tmp;
799 }
800 
801 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
802 {
803 	skb->data -= len;
804 	skb->len  += len;
805 	return skb->data;
806 }
807 
808 /**
809  *	skb_push - add data to the start of a buffer
810  *	@skb: buffer to use
811  *	@len: amount of data to add
812  *
813  *	This function extends the used data area of the buffer at the buffer
814  *	start. If this would exceed the total buffer headroom the kernel will
815  *	panic. A pointer to the first byte of the extra data is returned.
816  */
817 static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
818 {
819 	skb->data -= len;
820 	skb->len  += len;
821 	if (unlikely(skb->data<skb->head))
822 		skb_under_panic(skb, len, current_text_addr());
823 	return skb->data;
824 }
825 
826 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
827 {
828 	skb->len -= len;
829 	BUG_ON(skb->len < skb->data_len);
830 	return skb->data += len;
831 }
832 
833 /**
834  *	skb_pull - remove data from the start of a buffer
835  *	@skb: buffer to use
836  *	@len: amount of data to remove
837  *
838  *	This function removes data from the start of a buffer, returning
839  *	the memory to the headroom. A pointer to the next data in the buffer
840  *	is returned. Once the data has been pulled future pushes will overwrite
841  *	the old data.
842  */
843 static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
844 {
845 	return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
846 }
847 
848 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
849 
850 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
851 {
852 	if (len > skb_headlen(skb) &&
853 	    !__pskb_pull_tail(skb, len-skb_headlen(skb)))
854 		return NULL;
855 	skb->len -= len;
856 	return skb->data += len;
857 }
858 
859 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
860 {
861 	return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
862 }
863 
864 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
865 {
866 	if (likely(len <= skb_headlen(skb)))
867 		return 1;
868 	if (unlikely(len > skb->len))
869 		return 0;
870 	return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
871 }
872 
873 /**
874  *	skb_headroom - bytes at buffer head
875  *	@skb: buffer to check
876  *
877  *	Return the number of bytes of free space at the head of an &sk_buff.
878  */
879 static inline int skb_headroom(const struct sk_buff *skb)
880 {
881 	return skb->data - skb->head;
882 }
883 
884 /**
885  *	skb_tailroom - bytes at buffer end
886  *	@skb: buffer to check
887  *
888  *	Return the number of bytes of free space at the tail of an sk_buff
889  */
890 static inline int skb_tailroom(const struct sk_buff *skb)
891 {
892 	return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
893 }
894 
895 /**
896  *	skb_reserve - adjust headroom
897  *	@skb: buffer to alter
898  *	@len: bytes to move
899  *
900  *	Increase the headroom of an empty &sk_buff by reducing the tail
901  *	room. This is only allowed for an empty buffer.
902  */
903 static inline void skb_reserve(struct sk_buff *skb, unsigned int len)
904 {
905 	skb->data += len;
906 	skb->tail += len;
907 }
908 
909 /*
910  * CPUs often take a performance hit when accessing unaligned memory
911  * locations. The actual performance hit varies, it can be small if the
912  * hardware handles it or large if we have to take an exception and fix it
913  * in software.
914  *
915  * Since an ethernet header is 14 bytes network drivers often end up with
916  * the IP header at an unaligned offset. The IP header can be aligned by
917  * shifting the start of the packet by 2 bytes. Drivers should do this
918  * with:
919  *
920  * skb_reserve(NET_IP_ALIGN);
921  *
922  * The downside to this alignment of the IP header is that the DMA is now
923  * unaligned. On some architectures the cost of an unaligned DMA is high
924  * and this cost outweighs the gains made by aligning the IP header.
925  *
926  * Since this trade off varies between architectures, we allow NET_IP_ALIGN
927  * to be overridden.
928  */
929 #ifndef NET_IP_ALIGN
930 #define NET_IP_ALIGN	2
931 #endif
932 
933 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
934 
935 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
936 {
937 	if (!skb->data_len) {
938 		skb->len  = len;
939 		skb->tail = skb->data + len;
940 	} else
941 		___pskb_trim(skb, len, 0);
942 }
943 
944 /**
945  *	skb_trim - remove end from a buffer
946  *	@skb: buffer to alter
947  *	@len: new length
948  *
949  *	Cut the length of a buffer down by removing data from the tail. If
950  *	the buffer is already under the length specified it is not modified.
951  */
952 static inline void skb_trim(struct sk_buff *skb, unsigned int len)
953 {
954 	if (skb->len > len)
955 		__skb_trim(skb, len);
956 }
957 
958 
959 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
960 {
961 	if (!skb->data_len) {
962 		skb->len  = len;
963 		skb->tail = skb->data+len;
964 		return 0;
965 	}
966 	return ___pskb_trim(skb, len, 1);
967 }
968 
969 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
970 {
971 	return (len < skb->len) ? __pskb_trim(skb, len) : 0;
972 }
973 
974 /**
975  *	skb_orphan - orphan a buffer
976  *	@skb: buffer to orphan
977  *
978  *	If a buffer currently has an owner then we call the owner's
979  *	destructor function and make the @skb unowned. The buffer continues
980  *	to exist but is no longer charged to its former owner.
981  */
982 static inline void skb_orphan(struct sk_buff *skb)
983 {
984 	if (skb->destructor)
985 		skb->destructor(skb);
986 	skb->destructor = NULL;
987 	skb->sk		= NULL;
988 }
989 
990 /**
991  *	__skb_queue_purge - empty a list
992  *	@list: list to empty
993  *
994  *	Delete all buffers on an &sk_buff list. Each buffer is removed from
995  *	the list and one reference dropped. This function does not take the
996  *	list lock and the caller must hold the relevant locks to use it.
997  */
998 extern void skb_queue_purge(struct sk_buff_head *list);
999 static inline void __skb_queue_purge(struct sk_buff_head *list)
1000 {
1001 	struct sk_buff *skb;
1002 	while ((skb = __skb_dequeue(list)) != NULL)
1003 		kfree_skb(skb);
1004 }
1005 
1006 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
1007 /**
1008  *	__dev_alloc_skb - allocate an skbuff for sending
1009  *	@length: length to allocate
1010  *	@gfp_mask: get_free_pages mask, passed to alloc_skb
1011  *
1012  *	Allocate a new &sk_buff and assign it a usage count of one. The
1013  *	buffer has unspecified headroom built in. Users should allocate
1014  *	the headroom they think they need without accounting for the
1015  *	built in space. The built in space is used for optimisations.
1016  *
1017  *	%NULL is returned in there is no free memory.
1018  */
1019 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1020 					      gfp_t gfp_mask)
1021 {
1022 	struct sk_buff *skb = alloc_skb(length + 16, gfp_mask);
1023 	if (likely(skb))
1024 		skb_reserve(skb, 16);
1025 	return skb;
1026 }
1027 #else
1028 extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
1029 #endif
1030 
1031 /**
1032  *	dev_alloc_skb - allocate an skbuff for sending
1033  *	@length: length to allocate
1034  *
1035  *	Allocate a new &sk_buff and assign it a usage count of one. The
1036  *	buffer has unspecified headroom built in. Users should allocate
1037  *	the headroom they think they need without accounting for the
1038  *	built in space. The built in space is used for optimisations.
1039  *
1040  *	%NULL is returned in there is no free memory. Although this function
1041  *	allocates memory it can be called from an interrupt.
1042  */
1043 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1044 {
1045 	return __dev_alloc_skb(length, GFP_ATOMIC);
1046 }
1047 
1048 /**
1049  *	skb_cow - copy header of skb when it is required
1050  *	@skb: buffer to cow
1051  *	@headroom: needed headroom
1052  *
1053  *	If the skb passed lacks sufficient headroom or its data part
1054  *	is shared, data is reallocated. If reallocation fails, an error
1055  *	is returned and original skb is not changed.
1056  *
1057  *	The result is skb with writable area skb->head...skb->tail
1058  *	and at least @headroom of space at head.
1059  */
1060 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1061 {
1062 	int delta = (headroom > 16 ? headroom : 16) - skb_headroom(skb);
1063 
1064 	if (delta < 0)
1065 		delta = 0;
1066 
1067 	if (delta || skb_cloned(skb))
1068 		return pskb_expand_head(skb, (delta + 15) & ~15, 0, GFP_ATOMIC);
1069 	return 0;
1070 }
1071 
1072 /**
1073  *	skb_padto	- pad an skbuff up to a minimal size
1074  *	@skb: buffer to pad
1075  *	@len: minimal length
1076  *
1077  *	Pads up a buffer to ensure the trailing bytes exist and are
1078  *	blanked. If the buffer already contains sufficient data it
1079  *	is untouched. Returns the buffer, which may be a replacement
1080  *	for the original, or NULL for out of memory - in which case
1081  *	the original buffer is still freed.
1082  */
1083 
1084 static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
1085 {
1086 	unsigned int size = skb->len;
1087 	if (likely(size >= len))
1088 		return skb;
1089 	return skb_pad(skb, len-size);
1090 }
1091 
1092 static inline int skb_add_data(struct sk_buff *skb,
1093 			       char __user *from, int copy)
1094 {
1095 	const int off = skb->len;
1096 
1097 	if (skb->ip_summed == CHECKSUM_NONE) {
1098 		int err = 0;
1099 		unsigned int csum = csum_and_copy_from_user(from,
1100 							    skb_put(skb, copy),
1101 							    copy, 0, &err);
1102 		if (!err) {
1103 			skb->csum = csum_block_add(skb->csum, csum, off);
1104 			return 0;
1105 		}
1106 	} else if (!copy_from_user(skb_put(skb, copy), from, copy))
1107 		return 0;
1108 
1109 	__skb_trim(skb, off);
1110 	return -EFAULT;
1111 }
1112 
1113 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1114 				   struct page *page, int off)
1115 {
1116 	if (i) {
1117 		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1118 
1119 		return page == frag->page &&
1120 		       off == frag->page_offset + frag->size;
1121 	}
1122 	return 0;
1123 }
1124 
1125 /**
1126  *	skb_linearize - convert paged skb to linear one
1127  *	@skb: buffer to linarize
1128  *	@gfp: allocation mode
1129  *
1130  *	If there is no free memory -ENOMEM is returned, otherwise zero
1131  *	is returned and the old skb data released.
1132  */
1133 extern int __skb_linearize(struct sk_buff *skb, gfp_t gfp);
1134 static inline int skb_linearize(struct sk_buff *skb, gfp_t gfp)
1135 {
1136 	return __skb_linearize(skb, gfp);
1137 }
1138 
1139 /**
1140  *	skb_postpull_rcsum - update checksum for received skb after pull
1141  *	@skb: buffer to update
1142  *	@start: start of data before pull
1143  *	@len: length of data pulled
1144  *
1145  *	After doing a pull on a received packet, you need to call this to
1146  *	update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1147  *	so that it can be recomputed from scratch.
1148  */
1149 
1150 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1151 					 const void *start, int len)
1152 {
1153 	if (skb->ip_summed == CHECKSUM_HW)
1154 		skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1155 }
1156 
1157 /**
1158  *	pskb_trim_rcsum - trim received skb and update checksum
1159  *	@skb: buffer to trim
1160  *	@len: new length
1161  *
1162  *	This is exactly the same as pskb_trim except that it ensures the
1163  *	checksum of received packets are still valid after the operation.
1164  */
1165 
1166 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1167 {
1168 	if (likely(len >= skb->len))
1169 		return 0;
1170 	if (skb->ip_summed == CHECKSUM_HW)
1171 		skb->ip_summed = CHECKSUM_NONE;
1172 	return __pskb_trim(skb, len);
1173 }
1174 
1175 static inline void *kmap_skb_frag(const skb_frag_t *frag)
1176 {
1177 #ifdef CONFIG_HIGHMEM
1178 	BUG_ON(in_irq());
1179 
1180 	local_bh_disable();
1181 #endif
1182 	return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
1183 }
1184 
1185 static inline void kunmap_skb_frag(void *vaddr)
1186 {
1187 	kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
1188 #ifdef CONFIG_HIGHMEM
1189 	local_bh_enable();
1190 #endif
1191 }
1192 
1193 #define skb_queue_walk(queue, skb) \
1194 		for (skb = (queue)->next;					\
1195 		     prefetch(skb->next), (skb != (struct sk_buff *)(queue));	\
1196 		     skb = skb->next)
1197 
1198 
1199 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1200 					 int noblock, int *err);
1201 extern unsigned int    datagram_poll(struct file *file, struct socket *sock,
1202 				     struct poll_table_struct *wait);
1203 extern int	       skb_copy_datagram_iovec(const struct sk_buff *from,
1204 					       int offset, struct iovec *to,
1205 					       int size);
1206 extern int	       skb_copy_and_csum_datagram_iovec(const
1207 							struct sk_buff *skb,
1208 							int hlen,
1209 							struct iovec *iov);
1210 extern void	       skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1211 extern unsigned int    skb_checksum(const struct sk_buff *skb, int offset,
1212 				    int len, unsigned int csum);
1213 extern int	       skb_copy_bits(const struct sk_buff *skb, int offset,
1214 				     void *to, int len);
1215 extern int	       skb_store_bits(const struct sk_buff *skb, int offset,
1216 				      void *from, int len);
1217 extern unsigned int    skb_copy_and_csum_bits(const struct sk_buff *skb,
1218 					      int offset, u8 *to, int len,
1219 					      unsigned int csum);
1220 extern void	       skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1221 extern void	       skb_split(struct sk_buff *skb,
1222 				 struct sk_buff *skb1, const u32 len);
1223 
1224 extern void	       skb_release_data(struct sk_buff *skb);
1225 
1226 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1227 				       int len, void *buffer)
1228 {
1229 	int hlen = skb_headlen(skb);
1230 
1231 	if (hlen - offset >= len)
1232 		return skb->data + offset;
1233 
1234 	if (skb_copy_bits(skb, offset, buffer, len) < 0)
1235 		return NULL;
1236 
1237 	return buffer;
1238 }
1239 
1240 extern void skb_init(void);
1241 extern void skb_add_mtu(int mtu);
1242 
1243 /**
1244  *	skb_get_timestamp - get timestamp from a skb
1245  *	@skb: skb to get stamp from
1246  *	@stamp: pointer to struct timeval to store stamp in
1247  *
1248  *	Timestamps are stored in the skb as offsets to a base timestamp.
1249  *	This function converts the offset back to a struct timeval and stores
1250  *	it in stamp.
1251  */
1252 static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
1253 {
1254 	stamp->tv_sec  = skb->tstamp.off_sec;
1255 	stamp->tv_usec = skb->tstamp.off_usec;
1256 }
1257 
1258 /**
1259  * 	skb_set_timestamp - set timestamp of a skb
1260  *	@skb: skb to set stamp of
1261  *	@stamp: pointer to struct timeval to get stamp from
1262  *
1263  *	Timestamps are stored in the skb as offsets to a base timestamp.
1264  *	This function converts a struct timeval to an offset and stores
1265  *	it in the skb.
1266  */
1267 static inline void skb_set_timestamp(struct sk_buff *skb, const struct timeval *stamp)
1268 {
1269 	skb->tstamp.off_sec  = stamp->tv_sec;
1270 	skb->tstamp.off_usec = stamp->tv_usec;
1271 }
1272 
1273 extern void __net_timestamp(struct sk_buff *skb);
1274 
1275 #ifdef CONFIG_NETFILTER
1276 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1277 {
1278 	if (nfct && atomic_dec_and_test(&nfct->use))
1279 		nfct->destroy(nfct);
1280 }
1281 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1282 {
1283 	if (nfct)
1284 		atomic_inc(&nfct->use);
1285 }
1286 static inline void nf_reset(struct sk_buff *skb)
1287 {
1288 	nf_conntrack_put(skb->nfct);
1289 	skb->nfct = NULL;
1290 }
1291 
1292 #ifdef CONFIG_BRIDGE_NETFILTER
1293 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1294 {
1295 	if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1296 		kfree(nf_bridge);
1297 }
1298 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1299 {
1300 	if (nf_bridge)
1301 		atomic_inc(&nf_bridge->use);
1302 }
1303 #endif /* CONFIG_BRIDGE_NETFILTER */
1304 #else /* CONFIG_NETFILTER */
1305 static inline void nf_reset(struct sk_buff *skb) {}
1306 #endif /* CONFIG_NETFILTER */
1307 
1308 #endif	/* __KERNEL__ */
1309 #endif	/* _LINUX_SKBUFF_H */
1310