2 * Definitions for the 'struct sk_buff' memory handlers.
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
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.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/kernel.h>
18 #include <linux/kmemcheck.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/cache.h>
23 #include <asm/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
26 #include <linux/net.h>
27 #include <linux/textsearch.h>
28 #include <net/checksum.h>
29 #include <linux/rcupdate.h>
30 #include <linux/dmaengine.h>
31 #include <linux/hrtimer.h>
33 /* Don't change this without changing skb_csum_unnecessary! */
34 #define CHECKSUM_NONE 0
35 #define CHECKSUM_UNNECESSARY 1
36 #define CHECKSUM_COMPLETE 2
37 #define CHECKSUM_PARTIAL 3
39 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
40 ~(SMP_CACHE_BYTES - 1))
41 #define SKB_WITH_OVERHEAD(X) \
42 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
43 #define SKB_MAX_ORDER(X, ORDER) \
44 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
45 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
46 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
48 /* A. Checksumming of received packets by device.
50 * NONE: device failed to checksum this packet.
51 * skb->csum is undefined.
53 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
54 * skb->csum is undefined.
55 * It is bad option, but, unfortunately, many of vendors do this.
56 * Apparently with secret goal to sell you new device, when you
57 * will add new protocol to your host. F.e. IPv6. 8)
59 * COMPLETE: the most generic way. Device supplied checksum of _all_
60 * the packet as seen by netif_rx in skb->csum.
61 * NOTE: Even if device supports only some protocols, but
62 * is able to produce some skb->csum, it MUST use COMPLETE,
65 * PARTIAL: identical to the case for output below. This may occur
66 * on a packet received directly from another Linux OS, e.g.,
67 * a virtualised Linux kernel on the same host. The packet can
68 * be treated in the same way as UNNECESSARY except that on
69 * output (i.e., forwarding) the checksum must be filled in
70 * by the OS or the hardware.
72 * B. Checksumming on output.
74 * NONE: skb is checksummed by protocol or csum is not required.
76 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
77 * from skb->csum_start to the end and to record the checksum
78 * at skb->csum_start + skb->csum_offset.
80 * Device must show its capabilities in dev->features, set
81 * at device setup time.
82 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
84 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
85 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
86 * TCP/UDP over IPv4. Sigh. Vendors like this
87 * way by an unknown reason. Though, see comment above
88 * about CHECKSUM_UNNECESSARY. 8)
89 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
91 * Any questions? No questions, good. --ANK
96 struct pipe_inode_info
;
98 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
104 #ifdef CONFIG_BRIDGE_NETFILTER
105 struct nf_bridge_info
{
107 struct net_device
*physindev
;
108 struct net_device
*physoutdev
;
110 unsigned long data
[32 / sizeof(unsigned long)];
114 struct sk_buff_head
{
115 /* These two members must be first. */
116 struct sk_buff
*next
;
117 struct sk_buff
*prev
;
125 /* To allow 64K frame to be packed as single skb without frag_list */
126 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
128 typedef struct skb_frag_struct skb_frag_t
;
130 struct skb_frag_struct
{
136 #define HAVE_HW_TIME_STAMP
139 * struct skb_shared_hwtstamps - hardware time stamps
140 * @hwtstamp: hardware time stamp transformed into duration
141 * since arbitrary point in time
142 * @syststamp: hwtstamp transformed to system time base
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:
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
155 * hwtstamps can only be compared against other hwtstamps from
158 * This structure is attached to packets as part of the
159 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
161 struct skb_shared_hwtstamps
{
167 * struct skb_shared_tx - instructions for time stamping of outgoing packets
168 * @hardware: generate hardware time stamp
169 * @software: generate software time stamp
170 * @in_progress: device driver is going to provide
171 * hardware time stamp
172 * @flags: all shared_tx flags
174 * These flags are attached to packets as part of the
175 * &skb_shared_info. Use skb_tx() to get a pointer.
177 union skb_shared_tx
{
186 /* This data is invariant across clones and lives at
187 * the end of the header data, ie. at skb->end.
189 struct skb_shared_info
{
190 unsigned short nr_frags
;
191 unsigned short gso_size
;
192 /* Warning: this field is not always filled in (UFO)! */
193 unsigned short gso_segs
;
194 unsigned short gso_type
;
196 union skb_shared_tx tx_flags
;
197 struct sk_buff
*frag_list
;
198 struct skb_shared_hwtstamps hwtstamps
;
201 * Warning : all fields before dataref are cleared in __alloc_skb()
205 skb_frag_t frags
[MAX_SKB_FRAGS
];
206 /* Intermediate layers must ensure that destructor_arg
207 * remains valid until skb destructor */
208 void * destructor_arg
;
211 /* We divide dataref into two halves. The higher 16 bits hold references
212 * to the payload part of skb->data. The lower 16 bits hold references to
213 * the entire skb->data. A clone of a headerless skb holds the length of
214 * the header in skb->hdr_len.
216 * All users must obey the rule that the skb->data reference count must be
217 * greater than or equal to the payload reference count.
219 * Holding a reference to the payload part means that the user does not
220 * care about modifications to the header part of skb->data.
222 #define SKB_DATAREF_SHIFT 16
223 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
227 SKB_FCLONE_UNAVAILABLE
,
233 SKB_GSO_TCPV4
= 1 << 0,
234 SKB_GSO_UDP
= 1 << 1,
236 /* This indicates the skb is from an untrusted source. */
237 SKB_GSO_DODGY
= 1 << 2,
239 /* This indicates the tcp segment has CWR set. */
240 SKB_GSO_TCP_ECN
= 1 << 3,
242 SKB_GSO_TCPV6
= 1 << 4,
244 SKB_GSO_FCOE
= 1 << 5,
247 #if BITS_PER_LONG > 32
248 #define NET_SKBUFF_DATA_USES_OFFSET 1
251 #ifdef NET_SKBUFF_DATA_USES_OFFSET
252 typedef unsigned int sk_buff_data_t
;
254 typedef unsigned char *sk_buff_data_t
;
258 * struct sk_buff - socket buffer
259 * @next: Next buffer in list
260 * @prev: Previous buffer in list
261 * @sk: Socket we are owned by
262 * @tstamp: Time we arrived
263 * @dev: Device we arrived on/are leaving by
264 * @transport_header: Transport layer header
265 * @network_header: Network layer header
266 * @mac_header: Link layer header
267 * @_skb_refdst: destination entry (with norefcount bit)
268 * @sp: the security path, used for xfrm
269 * @cb: Control buffer. Free for use by every layer. Put private vars here
270 * @len: Length of actual data
271 * @data_len: Data length
272 * @mac_len: Length of link layer header
273 * @hdr_len: writable header length of cloned skb
274 * @csum: Checksum (must include start/offset pair)
275 * @csum_start: Offset from skb->head where checksumming should start
276 * @csum_offset: Offset from csum_start where checksum should be stored
277 * @local_df: allow local fragmentation
278 * @cloned: Head may be cloned (check refcnt to be sure)
279 * @nohdr: Payload reference only, must not modify header
280 * @pkt_type: Packet class
281 * @fclone: skbuff clone status
282 * @ip_summed: Driver fed us an IP checksum
283 * @priority: Packet queueing priority
284 * @users: User count - see {datagram,tcp}.c
285 * @protocol: Packet protocol from driver
286 * @truesize: Buffer size
287 * @head: Head of buffer
288 * @data: Data head pointer
289 * @tail: Tail pointer
291 * @destructor: Destruct function
292 * @mark: Generic packet mark
293 * @nfct: Associated connection, if any
294 * @ipvs_property: skbuff is owned by ipvs
295 * @peeked: this packet has been seen already, so stats have been
296 * done for it, don't do them again
297 * @nf_trace: netfilter packet trace flag
298 * @nfctinfo: Relationship of this skb to the connection
299 * @nfct_reasm: netfilter conntrack re-assembly pointer
300 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
301 * @skb_iif: ifindex of device we arrived on
302 * @rxhash: the packet hash computed on receive
303 * @queue_mapping: Queue mapping for multiqueue devices
304 * @tc_index: Traffic control index
305 * @tc_verd: traffic control verdict
306 * @ndisc_nodetype: router type (from link layer)
307 * @dma_cookie: a cookie to one of several possible DMA operations
308 * done by skb DMA functions
309 * @secmark: security marking
310 * @vlan_tci: vlan tag control information
314 /* These two members must be first. */
315 struct sk_buff
*next
;
316 struct sk_buff
*prev
;
321 struct net_device
*dev
;
324 * This is the control buffer. It is free to use for every
325 * layer. Please put your private variables there. If you
326 * want to keep them across layers you have to do a skb_clone()
327 * first. This is owned by whoever has the skb queued ATM.
329 char cb
[48] __aligned(8);
331 unsigned long _skb_refdst
;
347 kmemcheck_bitfield_begin(flags1
);
358 kmemcheck_bitfield_end(flags1
);
361 void (*destructor
)(struct sk_buff
*skb
);
362 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
363 struct nf_conntrack
*nfct
;
364 struct sk_buff
*nfct_reasm
;
366 #ifdef CONFIG_BRIDGE_NETFILTER
367 struct nf_bridge_info
*nf_bridge
;
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 */
380 kmemcheck_bitfield_begin(flags2
);
381 __u16 queue_mapping
:16;
382 #ifdef CONFIG_IPV6_NDISC_NODETYPE
383 __u8 ndisc_nodetype
:2;
385 kmemcheck_bitfield_end(flags2
);
389 #ifdef CONFIG_NET_DMA
390 dma_cookie_t dma_cookie
;
392 #ifdef CONFIG_NETWORK_SECMARK
402 sk_buff_data_t transport_header
;
403 sk_buff_data_t network_header
;
404 sk_buff_data_t mac_header
;
405 /* These elements must be at the end, see alloc_skb() for details. */
410 unsigned int truesize
;
416 * Handling routines are only of interest to the kernel
418 #include <linux/slab.h>
420 #include <asm/system.h>
423 * skb might have a dst pointer attached, refcounted or not.
424 * _skb_refdst low order bit is set if refcount was _not_ taken
426 #define SKB_DST_NOREF 1UL
427 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
430 * skb_dst - returns skb dst_entry
433 * Returns skb dst_entry, regardless of reference taken or not.
435 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
437 /* If refdst was not refcounted, check we still are in a
438 * rcu_read_lock section
440 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
441 !rcu_read_lock_held() &&
442 !rcu_read_lock_bh_held());
443 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
447 * skb_dst_set - sets skb dst
451 * Sets skb dst, assuming a reference was taken on dst and should
452 * be released by skb_dst_drop()
454 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
456 skb
->_skb_refdst
= (unsigned long)dst
;
460 * skb_dst_set_noref - sets skb dst, without a reference
464 * Sets skb dst, assuming a reference was not taken on dst
465 * skb_dst_drop() should not dst_release() this dst
467 static inline void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
)
469 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
470 skb
->_skb_refdst
= (unsigned long)dst
| SKB_DST_NOREF
;
474 * skb_dst_is_noref - Test if skb dst isnt refcounted
477 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
479 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
482 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
484 return (struct rtable
*)skb_dst(skb
);
487 extern void kfree_skb(struct sk_buff
*skb
);
488 extern void consume_skb(struct sk_buff
*skb
);
489 extern void __kfree_skb(struct sk_buff
*skb
);
490 extern struct sk_buff
*__alloc_skb(unsigned int size
,
491 gfp_t priority
, int fclone
, int node
);
492 static inline struct sk_buff
*alloc_skb(unsigned int size
,
495 return __alloc_skb(size
, priority
, 0, -1);
498 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
501 return __alloc_skb(size
, priority
, 1, -1);
504 extern int skb_recycle_check(struct sk_buff
*skb
, int skb_size
);
506 extern struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
507 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
509 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
511 extern struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
513 extern int pskb_expand_head(struct sk_buff
*skb
,
514 int nhead
, int ntail
,
516 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
517 unsigned int headroom
);
518 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
519 int newheadroom
, int newtailroom
,
521 extern int skb_to_sgvec(struct sk_buff
*skb
,
522 struct scatterlist
*sg
, int offset
,
524 extern int skb_cow_data(struct sk_buff
*skb
, int tailbits
,
525 struct sk_buff
**trailer
);
526 extern int skb_pad(struct sk_buff
*skb
, int pad
);
527 #define dev_kfree_skb(a) consume_skb(a)
529 extern int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
530 int getfrag(void *from
, char *to
, int offset
,
531 int len
,int odd
, struct sk_buff
*skb
),
532 void *from
, int length
);
534 struct skb_seq_state
{
538 __u32 stepped_offset
;
539 struct sk_buff
*root_skb
;
540 struct sk_buff
*cur_skb
;
544 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
545 unsigned int from
, unsigned int to
,
546 struct skb_seq_state
*st
);
547 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
548 struct skb_seq_state
*st
);
549 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
551 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
552 unsigned int to
, struct ts_config
*config
,
553 struct ts_state
*state
);
555 #ifdef NET_SKBUFF_DATA_USES_OFFSET
556 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
558 return skb
->head
+ skb
->end
;
561 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
568 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
570 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
572 return &skb_shinfo(skb
)->hwtstamps
;
575 static inline union skb_shared_tx
*skb_tx(struct sk_buff
*skb
)
577 return &skb_shinfo(skb
)->tx_flags
;
581 * skb_queue_empty - check if a queue is empty
584 * Returns true if the queue is empty, false otherwise.
586 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
588 return list
->next
== (struct sk_buff
*)list
;
592 * skb_queue_is_last - check if skb is the last entry in the queue
596 * Returns true if @skb is the last buffer on the list.
598 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
599 const struct sk_buff
*skb
)
601 return (skb
->next
== (struct sk_buff
*) list
);
605 * skb_queue_is_first - check if skb is the first entry in the queue
609 * Returns true if @skb is the first buffer on the list.
611 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
612 const struct sk_buff
*skb
)
614 return (skb
->prev
== (struct sk_buff
*) list
);
618 * skb_queue_next - return the next packet in the queue
620 * @skb: current buffer
622 * Return the next packet in @list after @skb. It is only valid to
623 * call this if skb_queue_is_last() evaluates to false.
625 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
626 const struct sk_buff
*skb
)
628 /* This BUG_ON may seem severe, but if we just return then we
629 * are going to dereference garbage.
631 BUG_ON(skb_queue_is_last(list
, skb
));
636 * skb_queue_prev - return the prev packet in the queue
638 * @skb: current buffer
640 * Return the prev packet in @list before @skb. It is only valid to
641 * call this if skb_queue_is_first() evaluates to false.
643 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
644 const struct sk_buff
*skb
)
646 /* This BUG_ON may seem severe, but if we just return then we
647 * are going to dereference garbage.
649 BUG_ON(skb_queue_is_first(list
, skb
));
654 * skb_get - reference buffer
655 * @skb: buffer to reference
657 * Makes another reference to a socket buffer and returns a pointer
660 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
662 atomic_inc(&skb
->users
);
667 * If users == 1, we are the only owner and are can avoid redundant
672 * skb_cloned - is the buffer a clone
673 * @skb: buffer to check
675 * Returns true if the buffer was generated with skb_clone() and is
676 * one of multiple shared copies of the buffer. Cloned buffers are
677 * shared data so must not be written to under normal circumstances.
679 static inline int skb_cloned(const struct sk_buff
*skb
)
681 return skb
->cloned
&&
682 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
686 * skb_header_cloned - is the header a clone
687 * @skb: buffer to check
689 * Returns true if modifying the header part of the buffer requires
690 * the data to be copied.
692 static inline int skb_header_cloned(const struct sk_buff
*skb
)
699 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
700 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
705 * skb_header_release - release reference to header
706 * @skb: buffer to operate on
708 * Drop a reference to the header part of the buffer. This is done
709 * by acquiring a payload reference. You must not read from the header
710 * part of skb->data after this.
712 static inline void skb_header_release(struct sk_buff
*skb
)
716 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
720 * skb_shared - is the buffer shared
721 * @skb: buffer to check
723 * Returns true if more than one person has a reference to this
726 static inline int skb_shared(const struct sk_buff
*skb
)
728 return atomic_read(&skb
->users
) != 1;
732 * skb_share_check - check if buffer is shared and if so clone it
733 * @skb: buffer to check
734 * @pri: priority for memory allocation
736 * If the buffer is shared the buffer is cloned and the old copy
737 * drops a reference. A new clone with a single reference is returned.
738 * If the buffer is not shared the original buffer is returned. When
739 * being called from interrupt status or with spinlocks held pri must
742 * NULL is returned on a memory allocation failure.
744 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
,
747 might_sleep_if(pri
& __GFP_WAIT
);
748 if (skb_shared(skb
)) {
749 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
757 * Copy shared buffers into a new sk_buff. We effectively do COW on
758 * packets to handle cases where we have a local reader and forward
759 * and a couple of other messy ones. The normal one is tcpdumping
760 * a packet thats being forwarded.
764 * skb_unshare - make a copy of a shared buffer
765 * @skb: buffer to check
766 * @pri: priority for memory allocation
768 * If the socket buffer is a clone then this function creates a new
769 * copy of the data, drops a reference count on the old copy and returns
770 * the new copy with the reference count at 1. If the buffer is not a clone
771 * the original buffer is returned. When called with a spinlock held or
772 * from interrupt state @pri must be %GFP_ATOMIC
774 * %NULL is returned on a memory allocation failure.
776 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
779 might_sleep_if(pri
& __GFP_WAIT
);
780 if (skb_cloned(skb
)) {
781 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
782 kfree_skb(skb
); /* Free our shared copy */
789 * skb_peek - peek at the head of an &sk_buff_head
790 * @list_: list to peek at
792 * Peek an &sk_buff. Unlike most other operations you _MUST_
793 * be careful with this one. A peek leaves the buffer on the
794 * list and someone else may run off with it. You must hold
795 * the appropriate locks or have a private queue to do this.
797 * Returns %NULL for an empty list or a pointer to the head element.
798 * The reference count is not incremented and the reference is therefore
799 * volatile. Use with caution.
801 static inline struct sk_buff
*skb_peek(struct sk_buff_head
*list_
)
803 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->next
;
804 if (list
== (struct sk_buff
*)list_
)
810 * skb_peek_tail - peek at the tail of an &sk_buff_head
811 * @list_: list to peek at
813 * Peek an &sk_buff. Unlike most other operations you _MUST_
814 * be careful with this one. A peek leaves the buffer on the
815 * list and someone else may run off with it. You must hold
816 * the appropriate locks or have a private queue to do this.
818 * Returns %NULL for an empty list or a pointer to the tail element.
819 * The reference count is not incremented and the reference is therefore
820 * volatile. Use with caution.
822 static inline struct sk_buff
*skb_peek_tail(struct sk_buff_head
*list_
)
824 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->prev
;
825 if (list
== (struct sk_buff
*)list_
)
831 * skb_queue_len - get queue length
832 * @list_: list to measure
834 * Return the length of an &sk_buff queue.
836 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
842 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
843 * @list: queue to initialize
845 * This initializes only the list and queue length aspects of
846 * an sk_buff_head object. This allows to initialize the list
847 * aspects of an sk_buff_head without reinitializing things like
848 * the spinlock. It can also be used for on-stack sk_buff_head
849 * objects where the spinlock is known to not be used.
851 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
853 list
->prev
= list
->next
= (struct sk_buff
*)list
;
858 * This function creates a split out lock class for each invocation;
859 * this is needed for now since a whole lot of users of the skb-queue
860 * infrastructure in drivers have different locking usage (in hardirq)
861 * than the networking core (in softirq only). In the long run either the
862 * network layer or drivers should need annotation to consolidate the
863 * main types of usage into 3 classes.
865 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
867 spin_lock_init(&list
->lock
);
868 __skb_queue_head_init(list
);
871 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
872 struct lock_class_key
*class)
874 skb_queue_head_init(list
);
875 lockdep_set_class(&list
->lock
, class);
879 * Insert an sk_buff on a list.
881 * The "__skb_xxxx()" functions are the non-atomic ones that
882 * can only be called with interrupts disabled.
884 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
885 static inline void __skb_insert(struct sk_buff
*newsk
,
886 struct sk_buff
*prev
, struct sk_buff
*next
,
887 struct sk_buff_head
*list
)
891 next
->prev
= prev
->next
= newsk
;
895 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
896 struct sk_buff
*prev
,
897 struct sk_buff
*next
)
899 struct sk_buff
*first
= list
->next
;
900 struct sk_buff
*last
= list
->prev
;
910 * skb_queue_splice - join two skb lists, this is designed for stacks
911 * @list: the new list to add
912 * @head: the place to add it in the first list
914 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
915 struct sk_buff_head
*head
)
917 if (!skb_queue_empty(list
)) {
918 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
919 head
->qlen
+= list
->qlen
;
924 * skb_queue_splice - join two skb lists and reinitialise the emptied list
925 * @list: the new list to add
926 * @head: the place to add it in the first list
928 * The list at @list is reinitialised
930 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
931 struct sk_buff_head
*head
)
933 if (!skb_queue_empty(list
)) {
934 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
935 head
->qlen
+= list
->qlen
;
936 __skb_queue_head_init(list
);
941 * skb_queue_splice_tail - join two skb lists, each list being a queue
942 * @list: the new list to add
943 * @head: the place to add it in the first list
945 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
946 struct sk_buff_head
*head
)
948 if (!skb_queue_empty(list
)) {
949 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
950 head
->qlen
+= list
->qlen
;
955 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
956 * @list: the new list to add
957 * @head: the place to add it in the first list
959 * Each of the lists is a queue.
960 * The list at @list is reinitialised
962 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
963 struct sk_buff_head
*head
)
965 if (!skb_queue_empty(list
)) {
966 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
967 head
->qlen
+= list
->qlen
;
968 __skb_queue_head_init(list
);
973 * __skb_queue_after - queue a buffer at the list head
975 * @prev: place after this buffer
976 * @newsk: buffer to queue
978 * Queue a buffer int the middle of a list. This function takes no locks
979 * and you must therefore hold required locks before calling it.
981 * A buffer cannot be placed on two lists at the same time.
983 static inline void __skb_queue_after(struct sk_buff_head
*list
,
984 struct sk_buff
*prev
,
985 struct sk_buff
*newsk
)
987 __skb_insert(newsk
, prev
, prev
->next
, list
);
990 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
991 struct sk_buff_head
*list
);
993 static inline void __skb_queue_before(struct sk_buff_head
*list
,
994 struct sk_buff
*next
,
995 struct sk_buff
*newsk
)
997 __skb_insert(newsk
, next
->prev
, next
, list
);
1001 * __skb_queue_head - queue a buffer at the list head
1002 * @list: list to use
1003 * @newsk: buffer to queue
1005 * Queue a buffer at the start of a list. This function takes no locks
1006 * and you must therefore hold required locks before calling it.
1008 * A buffer cannot be placed on two lists at the same time.
1010 extern void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1011 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1012 struct sk_buff
*newsk
)
1014 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1018 * __skb_queue_tail - queue a buffer at the list tail
1019 * @list: list to use
1020 * @newsk: buffer to queue
1022 * Queue a buffer at the end of a list. This function takes no locks
1023 * and you must therefore hold required locks before calling it.
1025 * A buffer cannot be placed on two lists at the same time.
1027 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1028 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1029 struct sk_buff
*newsk
)
1031 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1035 * remove sk_buff from list. _Must_ be called atomically, and with
1038 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1039 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1041 struct sk_buff
*next
, *prev
;
1046 skb
->next
= skb
->prev
= NULL
;
1052 * __skb_dequeue - remove from the head of the queue
1053 * @list: list to dequeue from
1055 * Remove the head of the list. This function does not take any locks
1056 * so must be used with appropriate locks held only. The head item is
1057 * returned or %NULL if the list is empty.
1059 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1060 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1062 struct sk_buff
*skb
= skb_peek(list
);
1064 __skb_unlink(skb
, list
);
1069 * __skb_dequeue_tail - remove from the tail of the queue
1070 * @list: list to dequeue from
1072 * Remove the tail of the list. This function does not take any locks
1073 * so must be used with appropriate locks held only. The tail item is
1074 * returned or %NULL if the list is empty.
1076 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1077 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1079 struct sk_buff
*skb
= skb_peek_tail(list
);
1081 __skb_unlink(skb
, list
);
1086 static inline int skb_is_nonlinear(const struct sk_buff
*skb
)
1088 return skb
->data_len
;
1091 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1093 return skb
->len
- skb
->data_len
;
1096 static inline int skb_pagelen(const struct sk_buff
*skb
)
1100 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1101 len
+= skb_shinfo(skb
)->frags
[i
].size
;
1102 return len
+ skb_headlen(skb
);
1105 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1106 struct page
*page
, int off
, int size
)
1108 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1111 frag
->page_offset
= off
;
1113 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1116 extern void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
,
1119 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1120 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frags(skb))
1121 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1123 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1124 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1126 return skb
->head
+ skb
->tail
;
1129 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1131 skb
->tail
= skb
->data
- skb
->head
;
1134 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1136 skb_reset_tail_pointer(skb
);
1137 skb
->tail
+= offset
;
1139 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1140 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1145 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1147 skb
->tail
= skb
->data
;
1150 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1152 skb
->tail
= skb
->data
+ offset
;
1155 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1158 * Add data to an sk_buff
1160 extern unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1161 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1163 unsigned char *tmp
= skb_tail_pointer(skb
);
1164 SKB_LINEAR_ASSERT(skb
);
1170 extern unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1171 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1178 extern unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1179 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1182 BUG_ON(skb
->len
< skb
->data_len
);
1183 return skb
->data
+= len
;
1186 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1188 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1191 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1193 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1195 if (len
> skb_headlen(skb
) &&
1196 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1199 return skb
->data
+= len
;
1202 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1204 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1207 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1209 if (likely(len
<= skb_headlen(skb
)))
1211 if (unlikely(len
> skb
->len
))
1213 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1217 * skb_headroom - bytes at buffer head
1218 * @skb: buffer to check
1220 * Return the number of bytes of free space at the head of an &sk_buff.
1222 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1224 return skb
->data
- skb
->head
;
1228 * skb_tailroom - bytes at buffer end
1229 * @skb: buffer to check
1231 * Return the number of bytes of free space at the tail of an sk_buff
1233 static inline int skb_tailroom(const struct sk_buff
*skb
)
1235 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1239 * skb_reserve - adjust headroom
1240 * @skb: buffer to alter
1241 * @len: bytes to move
1243 * Increase the headroom of an empty &sk_buff by reducing the tail
1244 * room. This is only allowed for an empty buffer.
1246 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
1252 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1253 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1255 return skb
->head
+ skb
->transport_header
;
1258 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1260 skb
->transport_header
= skb
->data
- skb
->head
;
1263 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1266 skb_reset_transport_header(skb
);
1267 skb
->transport_header
+= offset
;
1270 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1272 return skb
->head
+ skb
->network_header
;
1275 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1277 skb
->network_header
= skb
->data
- skb
->head
;
1280 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1282 skb_reset_network_header(skb
);
1283 skb
->network_header
+= offset
;
1286 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1288 return skb
->head
+ skb
->mac_header
;
1291 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1293 return skb
->mac_header
!= ~0U;
1296 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1298 skb
->mac_header
= skb
->data
- skb
->head
;
1301 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1303 skb_reset_mac_header(skb
);
1304 skb
->mac_header
+= offset
;
1307 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1309 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1311 return skb
->transport_header
;
1314 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1316 skb
->transport_header
= skb
->data
;
1319 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1322 skb
->transport_header
= skb
->data
+ offset
;
1325 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1327 return skb
->network_header
;
1330 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1332 skb
->network_header
= skb
->data
;
1335 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1337 skb
->network_header
= skb
->data
+ offset
;
1340 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1342 return skb
->mac_header
;
1345 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1347 return skb
->mac_header
!= NULL
;
1350 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1352 skb
->mac_header
= skb
->data
;
1355 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1357 skb
->mac_header
= skb
->data
+ offset
;
1359 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1361 static inline int skb_transport_offset(const struct sk_buff
*skb
)
1363 return skb_transport_header(skb
) - skb
->data
;
1366 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
1368 return skb
->transport_header
- skb
->network_header
;
1371 static inline int skb_network_offset(const struct sk_buff
*skb
)
1373 return skb_network_header(skb
) - skb
->data
;
1377 * CPUs often take a performance hit when accessing unaligned memory
1378 * locations. The actual performance hit varies, it can be small if the
1379 * hardware handles it or large if we have to take an exception and fix it
1382 * Since an ethernet header is 14 bytes network drivers often end up with
1383 * the IP header at an unaligned offset. The IP header can be aligned by
1384 * shifting the start of the packet by 2 bytes. Drivers should do this
1387 * skb_reserve(skb, NET_IP_ALIGN);
1389 * The downside to this alignment of the IP header is that the DMA is now
1390 * unaligned. On some architectures the cost of an unaligned DMA is high
1391 * and this cost outweighs the gains made by aligning the IP header.
1393 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1396 #ifndef NET_IP_ALIGN
1397 #define NET_IP_ALIGN 2
1401 * The networking layer reserves some headroom in skb data (via
1402 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1403 * the header has to grow. In the default case, if the header has to grow
1404 * 32 bytes or less we avoid the reallocation.
1406 * Unfortunately this headroom changes the DMA alignment of the resulting
1407 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1408 * on some architectures. An architecture can override this value,
1409 * perhaps setting it to a cacheline in size (since that will maintain
1410 * cacheline alignment of the DMA). It must be a power of 2.
1412 * Various parts of the networking layer expect at least 32 bytes of
1413 * headroom, you should not reduce this.
1414 * With RPS, we raised NET_SKB_PAD to 64 so that get_rps_cpus() fetches span
1415 * a 64 bytes aligned block to fit modern (>= 64 bytes) cache line sizes
1416 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1419 #define NET_SKB_PAD 64
1422 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
1424 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
1426 if (unlikely(skb
->data_len
)) {
1431 skb_set_tail_pointer(skb
, len
);
1434 extern void skb_trim(struct sk_buff
*skb
, unsigned int len
);
1436 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1439 return ___pskb_trim(skb
, len
);
1440 __skb_trim(skb
, len
);
1444 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1446 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
1450 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1451 * @skb: buffer to alter
1454 * This is identical to pskb_trim except that the caller knows that
1455 * the skb is not cloned so we should never get an error due to out-
1458 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
1460 int err
= pskb_trim(skb
, len
);
1465 * skb_orphan - orphan a buffer
1466 * @skb: buffer to orphan
1468 * If a buffer currently has an owner then we call the owner's
1469 * destructor function and make the @skb unowned. The buffer continues
1470 * to exist but is no longer charged to its former owner.
1472 static inline void skb_orphan(struct sk_buff
*skb
)
1474 if (skb
->destructor
)
1475 skb
->destructor(skb
);
1476 skb
->destructor
= NULL
;
1481 * __skb_queue_purge - empty a list
1482 * @list: list to empty
1484 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1485 * the list and one reference dropped. This function does not take the
1486 * list lock and the caller must hold the relevant locks to use it.
1488 extern void skb_queue_purge(struct sk_buff_head
*list
);
1489 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
1491 struct sk_buff
*skb
;
1492 while ((skb
= __skb_dequeue(list
)) != NULL
)
1497 * __dev_alloc_skb - allocate an skbuff for receiving
1498 * @length: length to allocate
1499 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1501 * Allocate a new &sk_buff and assign it a usage count of one. The
1502 * buffer has unspecified headroom built in. Users should allocate
1503 * the headroom they think they need without accounting for the
1504 * built in space. The built in space is used for optimisations.
1506 * %NULL is returned if there is no free memory.
1508 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
1511 struct sk_buff
*skb
= alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
);
1513 skb_reserve(skb
, NET_SKB_PAD
);
1517 extern struct sk_buff
*dev_alloc_skb(unsigned int length
);
1519 extern struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
1520 unsigned int length
, gfp_t gfp_mask
);
1523 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1524 * @dev: network device to receive on
1525 * @length: length to allocate
1527 * Allocate a new &sk_buff and assign it a usage count of one. The
1528 * buffer has unspecified headroom built in. Users should allocate
1529 * the headroom they think they need without accounting for the
1530 * built in space. The built in space is used for optimisations.
1532 * %NULL is returned if there is no free memory. Although this function
1533 * allocates memory it can be called from an interrupt.
1535 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
1536 unsigned int length
)
1538 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
1541 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
1542 unsigned int length
)
1544 struct sk_buff
*skb
= netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
);
1546 if (NET_IP_ALIGN
&& skb
)
1547 skb_reserve(skb
, NET_IP_ALIGN
);
1551 extern struct page
*__netdev_alloc_page(struct net_device
*dev
, gfp_t gfp_mask
);
1554 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1555 * @dev: network device to receive on
1557 * Allocate a new page node local to the specified device.
1559 * %NULL is returned if there is no free memory.
1561 static inline struct page
*netdev_alloc_page(struct net_device
*dev
)
1563 return __netdev_alloc_page(dev
, GFP_ATOMIC
);
1566 static inline void netdev_free_page(struct net_device
*dev
, struct page
*page
)
1572 * skb_clone_writable - is the header of a clone writable
1573 * @skb: buffer to check
1574 * @len: length up to which to write
1576 * Returns true if modifying the header part of the cloned buffer
1577 * does not requires the data to be copied.
1579 static inline int skb_clone_writable(struct sk_buff
*skb
, unsigned int len
)
1581 return !skb_header_cloned(skb
) &&
1582 skb_headroom(skb
) + len
<= skb
->hdr_len
;
1585 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
1590 if (headroom
< NET_SKB_PAD
)
1591 headroom
= NET_SKB_PAD
;
1592 if (headroom
> skb_headroom(skb
))
1593 delta
= headroom
- skb_headroom(skb
);
1595 if (delta
|| cloned
)
1596 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
1602 * skb_cow - copy header of skb when it is required
1603 * @skb: buffer to cow
1604 * @headroom: needed headroom
1606 * If the skb passed lacks sufficient headroom or its data part
1607 * is shared, data is reallocated. If reallocation fails, an error
1608 * is returned and original skb is not changed.
1610 * The result is skb with writable area skb->head...skb->tail
1611 * and at least @headroom of space at head.
1613 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
1615 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
1619 * skb_cow_head - skb_cow but only making the head writable
1620 * @skb: buffer to cow
1621 * @headroom: needed headroom
1623 * This function is identical to skb_cow except that we replace the
1624 * skb_cloned check by skb_header_cloned. It should be used when
1625 * you only need to push on some header and do not need to modify
1628 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
1630 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
1634 * skb_padto - pad an skbuff up to a minimal size
1635 * @skb: buffer to pad
1636 * @len: minimal length
1638 * Pads up a buffer to ensure the trailing bytes exist and are
1639 * blanked. If the buffer already contains sufficient data it
1640 * is untouched. Otherwise it is extended. Returns zero on
1641 * success. The skb is freed on error.
1644 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
1646 unsigned int size
= skb
->len
;
1647 if (likely(size
>= len
))
1649 return skb_pad(skb
, len
- size
);
1652 static inline int skb_add_data(struct sk_buff
*skb
,
1653 char __user
*from
, int copy
)
1655 const int off
= skb
->len
;
1657 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1659 __wsum csum
= csum_and_copy_from_user(from
, skb_put(skb
, copy
),
1662 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
1665 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
1668 __skb_trim(skb
, off
);
1672 static inline int skb_can_coalesce(struct sk_buff
*skb
, int i
,
1673 struct page
*page
, int off
)
1676 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
1678 return page
== frag
->page
&&
1679 off
== frag
->page_offset
+ frag
->size
;
1684 static inline int __skb_linearize(struct sk_buff
*skb
)
1686 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
1690 * skb_linearize - convert paged skb to linear one
1691 * @skb: buffer to linarize
1693 * If there is no free memory -ENOMEM is returned, otherwise zero
1694 * is returned and the old skb data released.
1696 static inline int skb_linearize(struct sk_buff
*skb
)
1698 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
1702 * skb_linearize_cow - make sure skb is linear and writable
1703 * @skb: buffer to process
1705 * If there is no free memory -ENOMEM is returned, otherwise zero
1706 * is returned and the old skb data released.
1708 static inline int skb_linearize_cow(struct sk_buff
*skb
)
1710 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
1711 __skb_linearize(skb
) : 0;
1715 * skb_postpull_rcsum - update checksum for received skb after pull
1716 * @skb: buffer to update
1717 * @start: start of data before pull
1718 * @len: length of data pulled
1720 * After doing a pull on a received packet, you need to call this to
1721 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1722 * CHECKSUM_NONE so that it can be recomputed from scratch.
1725 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
1726 const void *start
, unsigned int len
)
1728 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1729 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
1732 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
1735 * pskb_trim_rcsum - trim received skb and update checksum
1736 * @skb: buffer to trim
1739 * This is exactly the same as pskb_trim except that it ensures the
1740 * checksum of received packets are still valid after the operation.
1743 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
1745 if (likely(len
>= skb
->len
))
1747 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1748 skb
->ip_summed
= CHECKSUM_NONE
;
1749 return __pskb_trim(skb
, len
);
1752 #define skb_queue_walk(queue, skb) \
1753 for (skb = (queue)->next; \
1754 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1757 #define skb_queue_walk_safe(queue, skb, tmp) \
1758 for (skb = (queue)->next, tmp = skb->next; \
1759 skb != (struct sk_buff *)(queue); \
1760 skb = tmp, tmp = skb->next)
1762 #define skb_queue_walk_from(queue, skb) \
1763 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1766 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1767 for (tmp = skb->next; \
1768 skb != (struct sk_buff *)(queue); \
1769 skb = tmp, tmp = skb->next)
1771 #define skb_queue_reverse_walk(queue, skb) \
1772 for (skb = (queue)->prev; \
1773 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1777 static inline bool skb_has_frags(const struct sk_buff
*skb
)
1779 return skb_shinfo(skb
)->frag_list
!= NULL
;
1782 static inline void skb_frag_list_init(struct sk_buff
*skb
)
1784 skb_shinfo(skb
)->frag_list
= NULL
;
1787 static inline void skb_frag_add_head(struct sk_buff
*skb
, struct sk_buff
*frag
)
1789 frag
->next
= skb_shinfo(skb
)->frag_list
;
1790 skb_shinfo(skb
)->frag_list
= frag
;
1793 #define skb_walk_frags(skb, iter) \
1794 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1796 extern struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1797 int *peeked
, int *err
);
1798 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1799 int noblock
, int *err
);
1800 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
1801 struct poll_table_struct
*wait
);
1802 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
1803 int offset
, struct iovec
*to
,
1805 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
,
1808 extern int skb_copy_datagram_from_iovec(struct sk_buff
*skb
,
1810 const struct iovec
*from
,
1813 extern int skb_copy_datagram_const_iovec(const struct sk_buff
*from
,
1815 const struct iovec
*to
,
1818 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
1819 extern void skb_free_datagram_locked(struct sock
*sk
,
1820 struct sk_buff
*skb
);
1821 extern int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
,
1822 unsigned int flags
);
1823 extern __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
,
1824 int len
, __wsum csum
);
1825 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
1827 extern int skb_store_bits(struct sk_buff
*skb
, int offset
,
1828 const void *from
, int len
);
1829 extern __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
,
1830 int offset
, u8
*to
, int len
,
1832 extern int skb_splice_bits(struct sk_buff
*skb
,
1833 unsigned int offset
,
1834 struct pipe_inode_info
*pipe
,
1836 unsigned int flags
);
1837 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
1838 extern void skb_split(struct sk_buff
*skb
,
1839 struct sk_buff
*skb1
, const u32 len
);
1840 extern int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
,
1843 extern struct sk_buff
*skb_segment(struct sk_buff
*skb
, int features
);
1845 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
1846 int len
, void *buffer
)
1848 int hlen
= skb_headlen(skb
);
1850 if (hlen
- offset
>= len
)
1851 return skb
->data
+ offset
;
1853 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
1859 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
1861 const unsigned int len
)
1863 memcpy(to
, skb
->data
, len
);
1866 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
1867 const int offset
, void *to
,
1868 const unsigned int len
)
1870 memcpy(to
, skb
->data
+ offset
, len
);
1873 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
1875 const unsigned int len
)
1877 memcpy(skb
->data
, from
, len
);
1880 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
1883 const unsigned int len
)
1885 memcpy(skb
->data
+ offset
, from
, len
);
1888 extern void skb_init(void);
1890 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
1896 * skb_get_timestamp - get timestamp from a skb
1897 * @skb: skb to get stamp from
1898 * @stamp: pointer to struct timeval to store stamp in
1900 * Timestamps are stored in the skb as offsets to a base timestamp.
1901 * This function converts the offset back to a struct timeval and stores
1904 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
1905 struct timeval
*stamp
)
1907 *stamp
= ktime_to_timeval(skb
->tstamp
);
1910 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
1911 struct timespec
*stamp
)
1913 *stamp
= ktime_to_timespec(skb
->tstamp
);
1916 static inline void __net_timestamp(struct sk_buff
*skb
)
1918 skb
->tstamp
= ktime_get_real();
1921 static inline ktime_t
net_timedelta(ktime_t t
)
1923 return ktime_sub(ktime_get_real(), t
);
1926 static inline ktime_t
net_invalid_timestamp(void)
1928 return ktime_set(0, 0);
1932 * skb_tstamp_tx - queue clone of skb with send time stamps
1933 * @orig_skb: the original outgoing packet
1934 * @hwtstamps: hardware time stamps, may be NULL if not available
1936 * If the skb has a socket associated, then this function clones the
1937 * skb (thus sharing the actual data and optional structures), stores
1938 * the optional hardware time stamping information (if non NULL) or
1939 * generates a software time stamp (otherwise), then queues the clone
1940 * to the error queue of the socket. Errors are silently ignored.
1942 extern void skb_tstamp_tx(struct sk_buff
*orig_skb
,
1943 struct skb_shared_hwtstamps
*hwtstamps
);
1945 extern __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
1946 extern __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
1948 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
1950 return skb
->ip_summed
& CHECKSUM_UNNECESSARY
;
1954 * skb_checksum_complete - Calculate checksum of an entire packet
1955 * @skb: packet to process
1957 * This function calculates the checksum over the entire packet plus
1958 * the value of skb->csum. The latter can be used to supply the
1959 * checksum of a pseudo header as used by TCP/UDP. It returns the
1962 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1963 * this function can be used to verify that checksum on received
1964 * packets. In that case the function should return zero if the
1965 * checksum is correct. In particular, this function will return zero
1966 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1967 * hardware has already verified the correctness of the checksum.
1969 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
1971 return skb_csum_unnecessary(skb
) ?
1972 0 : __skb_checksum_complete(skb
);
1975 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1976 extern void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
1977 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
1979 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
1980 nf_conntrack_destroy(nfct
);
1982 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
1985 atomic_inc(&nfct
->use
);
1987 static inline void nf_conntrack_get_reasm(struct sk_buff
*skb
)
1990 atomic_inc(&skb
->users
);
1992 static inline void nf_conntrack_put_reasm(struct sk_buff
*skb
)
1998 #ifdef CONFIG_BRIDGE_NETFILTER
1999 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
2001 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
2004 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
2007 atomic_inc(&nf_bridge
->use
);
2009 #endif /* CONFIG_BRIDGE_NETFILTER */
2010 static inline void nf_reset(struct sk_buff
*skb
)
2012 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2013 nf_conntrack_put(skb
->nfct
);
2015 nf_conntrack_put_reasm(skb
->nfct_reasm
);
2016 skb
->nfct_reasm
= NULL
;
2018 #ifdef CONFIG_BRIDGE_NETFILTER
2019 nf_bridge_put(skb
->nf_bridge
);
2020 skb
->nf_bridge
= NULL
;
2024 /* Note: This doesn't put any conntrack and bridge info in dst. */
2025 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2027 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2028 dst
->nfct
= src
->nfct
;
2029 nf_conntrack_get(src
->nfct
);
2030 dst
->nfctinfo
= src
->nfctinfo
;
2031 dst
->nfct_reasm
= src
->nfct_reasm
;
2032 nf_conntrack_get_reasm(src
->nfct_reasm
);
2034 #ifdef CONFIG_BRIDGE_NETFILTER
2035 dst
->nf_bridge
= src
->nf_bridge
;
2036 nf_bridge_get(src
->nf_bridge
);
2040 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2042 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2043 nf_conntrack_put(dst
->nfct
);
2044 nf_conntrack_put_reasm(dst
->nfct_reasm
);
2046 #ifdef CONFIG_BRIDGE_NETFILTER
2047 nf_bridge_put(dst
->nf_bridge
);
2049 __nf_copy(dst
, src
);
2052 #ifdef CONFIG_NETWORK_SECMARK
2053 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2055 to
->secmark
= from
->secmark
;
2058 static inline void skb_init_secmark(struct sk_buff
*skb
)
2063 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2066 static inline void skb_init_secmark(struct sk_buff
*skb
)
2070 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
2072 skb
->queue_mapping
= queue_mapping
;
2075 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
2077 return skb
->queue_mapping
;
2080 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
2082 to
->queue_mapping
= from
->queue_mapping
;
2085 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
2087 skb
->queue_mapping
= rx_queue
+ 1;
2090 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
2092 return skb
->queue_mapping
- 1;
2095 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
2097 return (skb
->queue_mapping
!= 0);
2100 extern u16
skb_tx_hash(const struct net_device
*dev
,
2101 const struct sk_buff
*skb
);
2104 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2109 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2115 static inline int skb_is_gso(const struct sk_buff
*skb
)
2117 return skb_shinfo(skb
)->gso_size
;
2120 static inline int skb_is_gso_v6(const struct sk_buff
*skb
)
2122 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
2125 extern void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
2127 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
2129 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2130 * wanted then gso_type will be set. */
2131 struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
2132 if (shinfo
->gso_size
!= 0 && unlikely(shinfo
->gso_type
== 0)) {
2133 __skb_warn_lro_forwarding(skb
);
2139 static inline void skb_forward_csum(struct sk_buff
*skb
)
2141 /* Unfortunately we don't support this one. Any brave souls? */
2142 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2143 skb
->ip_summed
= CHECKSUM_NONE
;
2146 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
2147 #endif /* __KERNEL__ */
2148 #endif /* _LINUX_SKBUFF_H */