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
{
132 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
141 #define HAVE_HW_TIME_STAMP
144 * struct skb_shared_hwtstamps - hardware time stamps
145 * @hwtstamp: hardware time stamp transformed into duration
146 * since arbitrary point in time
147 * @syststamp: hwtstamp transformed to system time base
149 * Software time stamps generated by ktime_get_real() are stored in
150 * skb->tstamp. The relation between the different kinds of time
151 * stamps is as follows:
153 * syststamp and tstamp can be compared against each other in
154 * arbitrary combinations. The accuracy of a
155 * syststamp/tstamp/"syststamp from other device" comparison is
156 * limited by the accuracy of the transformation into system time
157 * base. This depends on the device driver and its underlying
160 * hwtstamps can only be compared against other hwtstamps from
163 * This structure is attached to packets as part of the
164 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
166 struct skb_shared_hwtstamps
{
171 /* Definitions for tx_flags in struct skb_shared_info */
173 /* generate hardware time stamp */
174 SKBTX_HW_TSTAMP
= 1 << 0,
176 /* generate software time stamp */
177 SKBTX_SW_TSTAMP
= 1 << 1,
179 /* device driver is going to provide hardware time stamp */
180 SKBTX_IN_PROGRESS
= 1 << 2,
182 /* ensure the originating sk reference is available on driver level */
183 SKBTX_DRV_NEEDS_SK_REF
= 1 << 3,
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
;
197 struct sk_buff
*frag_list
;
198 struct skb_shared_hwtstamps hwtstamps
;
201 * Warning : all fields before dataref are cleared in __alloc_skb()
205 /* Intermediate layers must ensure that destructor_arg
206 * remains valid until skb destructor */
207 void * destructor_arg
;
208 /* must be last field, see pskb_expand_head() */
209 skb_frag_t frags
[MAX_SKB_FRAGS
];
212 /* We divide dataref into two halves. The higher 16 bits hold references
213 * to the payload part of skb->data. The lower 16 bits hold references to
214 * the entire skb->data. A clone of a headerless skb holds the length of
215 * the header in skb->hdr_len.
217 * All users must obey the rule that the skb->data reference count must be
218 * greater than or equal to the payload reference count.
220 * Holding a reference to the payload part means that the user does not
221 * care about modifications to the header part of skb->data.
223 #define SKB_DATAREF_SHIFT 16
224 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
228 SKB_FCLONE_UNAVAILABLE
,
234 SKB_GSO_TCPV4
= 1 << 0,
235 SKB_GSO_UDP
= 1 << 1,
237 /* This indicates the skb is from an untrusted source. */
238 SKB_GSO_DODGY
= 1 << 2,
240 /* This indicates the tcp segment has CWR set. */
241 SKB_GSO_TCP_ECN
= 1 << 3,
243 SKB_GSO_TCPV6
= 1 << 4,
245 SKB_GSO_FCOE
= 1 << 5,
248 #if BITS_PER_LONG > 32
249 #define NET_SKBUFF_DATA_USES_OFFSET 1
252 #ifdef NET_SKBUFF_DATA_USES_OFFSET
253 typedef unsigned int sk_buff_data_t
;
255 typedef unsigned char *sk_buff_data_t
;
258 #if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
259 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
260 #define NET_SKBUFF_NF_DEFRAG_NEEDED 1
264 * struct sk_buff - socket buffer
265 * @next: Next buffer in list
266 * @prev: Previous buffer in list
267 * @sk: Socket we are owned by
268 * @tstamp: Time we arrived
269 * @dev: Device we arrived on/are leaving by
270 * @transport_header: Transport layer header
271 * @network_header: Network layer header
272 * @mac_header: Link layer header
273 * @_skb_refdst: destination entry (with norefcount bit)
274 * @sp: the security path, used for xfrm
275 * @cb: Control buffer. Free for use by every layer. Put private vars here
276 * @len: Length of actual data
277 * @data_len: Data length
278 * @mac_len: Length of link layer header
279 * @hdr_len: writable header length of cloned skb
280 * @csum: Checksum (must include start/offset pair)
281 * @csum_start: Offset from skb->head where checksumming should start
282 * @csum_offset: Offset from csum_start where checksum should be stored
283 * @local_df: allow local fragmentation
284 * @cloned: Head may be cloned (check refcnt to be sure)
285 * @nohdr: Payload reference only, must not modify header
286 * @pkt_type: Packet class
287 * @fclone: skbuff clone status
288 * @ip_summed: Driver fed us an IP checksum
289 * @priority: Packet queueing priority
290 * @users: User count - see {datagram,tcp}.c
291 * @protocol: Packet protocol from driver
292 * @truesize: Buffer size
293 * @head: Head of buffer
294 * @data: Data head pointer
295 * @tail: Tail pointer
297 * @destructor: Destruct function
298 * @mark: Generic packet mark
299 * @nfct: Associated connection, if any
300 * @ipvs_property: skbuff is owned by ipvs
301 * @peeked: this packet has been seen already, so stats have been
302 * done for it, don't do them again
303 * @nf_trace: netfilter packet trace flag
304 * @nfctinfo: Relationship of this skb to the connection
305 * @nfct_reasm: netfilter conntrack re-assembly pointer
306 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
307 * @skb_iif: ifindex of device we arrived on
308 * @rxhash: the packet hash computed on receive
309 * @queue_mapping: Queue mapping for multiqueue devices
310 * @tc_index: Traffic control index
311 * @tc_verd: traffic control verdict
312 * @ndisc_nodetype: router type (from link layer)
313 * @dma_cookie: a cookie to one of several possible DMA operations
314 * done by skb DMA functions
315 * @secmark: security marking
316 * @vlan_tci: vlan tag control information
320 /* These two members must be first. */
321 struct sk_buff
*next
;
322 struct sk_buff
*prev
;
327 struct net_device
*dev
;
330 * This is the control buffer. It is free to use for every
331 * layer. Please put your private variables there. If you
332 * want to keep them across layers you have to do a skb_clone()
333 * first. This is owned by whoever has the skb queued ATM.
335 char cb
[48] __aligned(8);
337 unsigned long _skb_refdst
;
353 kmemcheck_bitfield_begin(flags1
);
364 kmemcheck_bitfield_end(flags1
);
367 void (*destructor
)(struct sk_buff
*skb
);
368 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
369 struct nf_conntrack
*nfct
;
371 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
372 struct sk_buff
*nfct_reasm
;
374 #ifdef CONFIG_BRIDGE_NETFILTER
375 struct nf_bridge_info
*nf_bridge
;
379 #ifdef CONFIG_NET_SCHED
380 __u16 tc_index
; /* traffic control index */
381 #ifdef CONFIG_NET_CLS_ACT
382 __u16 tc_verd
; /* traffic control verdict */
388 kmemcheck_bitfield_begin(flags2
);
389 __u16 queue_mapping
:16;
390 #ifdef CONFIG_IPV6_NDISC_NODETYPE
391 __u8 ndisc_nodetype
:2;
394 kmemcheck_bitfield_end(flags2
);
398 #ifdef CONFIG_NET_DMA
399 dma_cookie_t dma_cookie
;
401 #ifdef CONFIG_NETWORK_SECMARK
411 sk_buff_data_t transport_header
;
412 sk_buff_data_t network_header
;
413 sk_buff_data_t mac_header
;
414 /* These elements must be at the end, see alloc_skb() for details. */
419 unsigned int truesize
;
425 * Handling routines are only of interest to the kernel
427 #include <linux/slab.h>
429 #include <asm/system.h>
432 * skb might have a dst pointer attached, refcounted or not.
433 * _skb_refdst low order bit is set if refcount was _not_ taken
435 #define SKB_DST_NOREF 1UL
436 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
439 * skb_dst - returns skb dst_entry
442 * Returns skb dst_entry, regardless of reference taken or not.
444 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
446 /* If refdst was not refcounted, check we still are in a
447 * rcu_read_lock section
449 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
450 !rcu_read_lock_held() &&
451 !rcu_read_lock_bh_held());
452 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
456 * skb_dst_set - sets skb dst
460 * Sets skb dst, assuming a reference was taken on dst and should
461 * be released by skb_dst_drop()
463 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
465 skb
->_skb_refdst
= (unsigned long)dst
;
468 extern void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
);
471 * skb_dst_is_noref - Test if skb dst isnt refcounted
474 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
476 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
479 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
481 return (struct rtable
*)skb_dst(skb
);
484 extern void kfree_skb(struct sk_buff
*skb
);
485 extern void consume_skb(struct sk_buff
*skb
);
486 extern void __kfree_skb(struct sk_buff
*skb
);
487 extern struct sk_buff
*__alloc_skb(unsigned int size
,
488 gfp_t priority
, int fclone
, int node
);
489 static inline struct sk_buff
*alloc_skb(unsigned int size
,
492 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
495 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
498 return __alloc_skb(size
, priority
, 1, NUMA_NO_NODE
);
501 extern bool skb_recycle_check(struct sk_buff
*skb
, int skb_size
);
503 extern struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
504 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
506 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
508 extern struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
510 extern int pskb_expand_head(struct sk_buff
*skb
,
511 int nhead
, int ntail
,
513 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
514 unsigned int headroom
);
515 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
516 int newheadroom
, int newtailroom
,
518 extern int skb_to_sgvec(struct sk_buff
*skb
,
519 struct scatterlist
*sg
, int offset
,
521 extern int skb_cow_data(struct sk_buff
*skb
, int tailbits
,
522 struct sk_buff
**trailer
);
523 extern int skb_pad(struct sk_buff
*skb
, int pad
);
524 #define dev_kfree_skb(a) consume_skb(a)
526 extern int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
527 int getfrag(void *from
, char *to
, int offset
,
528 int len
,int odd
, struct sk_buff
*skb
),
529 void *from
, int length
);
531 struct skb_seq_state
{
535 __u32 stepped_offset
;
536 struct sk_buff
*root_skb
;
537 struct sk_buff
*cur_skb
;
541 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
542 unsigned int from
, unsigned int to
,
543 struct skb_seq_state
*st
);
544 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
545 struct skb_seq_state
*st
);
546 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
548 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
549 unsigned int to
, struct ts_config
*config
,
550 struct ts_state
*state
);
552 extern __u32
__skb_get_rxhash(struct sk_buff
*skb
);
553 static inline __u32
skb_get_rxhash(struct sk_buff
*skb
)
556 skb
->rxhash
= __skb_get_rxhash(skb
);
561 #ifdef NET_SKBUFF_DATA_USES_OFFSET
562 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
564 return skb
->head
+ skb
->end
;
567 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
574 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
576 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
578 return &skb_shinfo(skb
)->hwtstamps
;
582 * skb_queue_empty - check if a queue is empty
585 * Returns true if the queue is empty, false otherwise.
587 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
589 return list
->next
== (struct sk_buff
*)list
;
593 * skb_queue_is_last - check if skb is the last entry in the queue
597 * Returns true if @skb is the last buffer on the list.
599 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
600 const struct sk_buff
*skb
)
602 return skb
->next
== (struct sk_buff
*)list
;
606 * skb_queue_is_first - check if skb is the first entry in the queue
610 * Returns true if @skb is the first buffer on the list.
612 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
613 const struct sk_buff
*skb
)
615 return skb
->prev
== (struct sk_buff
*)list
;
619 * skb_queue_next - return the next packet in the queue
621 * @skb: current buffer
623 * Return the next packet in @list after @skb. It is only valid to
624 * call this if skb_queue_is_last() evaluates to false.
626 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
627 const struct sk_buff
*skb
)
629 /* This BUG_ON may seem severe, but if we just return then we
630 * are going to dereference garbage.
632 BUG_ON(skb_queue_is_last(list
, skb
));
637 * skb_queue_prev - return the prev packet in the queue
639 * @skb: current buffer
641 * Return the prev packet in @list before @skb. It is only valid to
642 * call this if skb_queue_is_first() evaluates to false.
644 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
645 const struct sk_buff
*skb
)
647 /* This BUG_ON may seem severe, but if we just return then we
648 * are going to dereference garbage.
650 BUG_ON(skb_queue_is_first(list
, skb
));
655 * skb_get - reference buffer
656 * @skb: buffer to reference
658 * Makes another reference to a socket buffer and returns a pointer
661 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
663 atomic_inc(&skb
->users
);
668 * If users == 1, we are the only owner and are can avoid redundant
673 * skb_cloned - is the buffer a clone
674 * @skb: buffer to check
676 * Returns true if the buffer was generated with skb_clone() and is
677 * one of multiple shared copies of the buffer. Cloned buffers are
678 * shared data so must not be written to under normal circumstances.
680 static inline int skb_cloned(const struct sk_buff
*skb
)
682 return skb
->cloned
&&
683 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
687 * skb_header_cloned - is the header a clone
688 * @skb: buffer to check
690 * Returns true if modifying the header part of the buffer requires
691 * the data to be copied.
693 static inline int skb_header_cloned(const struct sk_buff
*skb
)
700 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
701 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
706 * skb_header_release - release reference to header
707 * @skb: buffer to operate on
709 * Drop a reference to the header part of the buffer. This is done
710 * by acquiring a payload reference. You must not read from the header
711 * part of skb->data after this.
713 static inline void skb_header_release(struct sk_buff
*skb
)
717 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
721 * skb_shared - is the buffer shared
722 * @skb: buffer to check
724 * Returns true if more than one person has a reference to this
727 static inline int skb_shared(const struct sk_buff
*skb
)
729 return atomic_read(&skb
->users
) != 1;
733 * skb_share_check - check if buffer is shared and if so clone it
734 * @skb: buffer to check
735 * @pri: priority for memory allocation
737 * If the buffer is shared the buffer is cloned and the old copy
738 * drops a reference. A new clone with a single reference is returned.
739 * If the buffer is not shared the original buffer is returned. When
740 * being called from interrupt status or with spinlocks held pri must
743 * NULL is returned on a memory allocation failure.
745 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
,
748 might_sleep_if(pri
& __GFP_WAIT
);
749 if (skb_shared(skb
)) {
750 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
758 * Copy shared buffers into a new sk_buff. We effectively do COW on
759 * packets to handle cases where we have a local reader and forward
760 * and a couple of other messy ones. The normal one is tcpdumping
761 * a packet thats being forwarded.
765 * skb_unshare - make a copy of a shared buffer
766 * @skb: buffer to check
767 * @pri: priority for memory allocation
769 * If the socket buffer is a clone then this function creates a new
770 * copy of the data, drops a reference count on the old copy and returns
771 * the new copy with the reference count at 1. If the buffer is not a clone
772 * the original buffer is returned. When called with a spinlock held or
773 * from interrupt state @pri must be %GFP_ATOMIC
775 * %NULL is returned on a memory allocation failure.
777 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
780 might_sleep_if(pri
& __GFP_WAIT
);
781 if (skb_cloned(skb
)) {
782 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
783 kfree_skb(skb
); /* Free our shared copy */
790 * skb_peek - peek at the head of an &sk_buff_head
791 * @list_: list to peek at
793 * Peek an &sk_buff. Unlike most other operations you _MUST_
794 * be careful with this one. A peek leaves the buffer on the
795 * list and someone else may run off with it. You must hold
796 * the appropriate locks or have a private queue to do this.
798 * Returns %NULL for an empty list or a pointer to the head element.
799 * The reference count is not incremented and the reference is therefore
800 * volatile. Use with caution.
802 static inline struct sk_buff
*skb_peek(struct sk_buff_head
*list_
)
804 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->next
;
805 if (list
== (struct sk_buff
*)list_
)
811 * skb_peek_tail - peek at the tail of an &sk_buff_head
812 * @list_: list to peek at
814 * Peek an &sk_buff. Unlike most other operations you _MUST_
815 * be careful with this one. A peek leaves the buffer on the
816 * list and someone else may run off with it. You must hold
817 * the appropriate locks or have a private queue to do this.
819 * Returns %NULL for an empty list or a pointer to the tail element.
820 * The reference count is not incremented and the reference is therefore
821 * volatile. Use with caution.
823 static inline struct sk_buff
*skb_peek_tail(struct sk_buff_head
*list_
)
825 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->prev
;
826 if (list
== (struct sk_buff
*)list_
)
832 * skb_queue_len - get queue length
833 * @list_: list to measure
835 * Return the length of an &sk_buff queue.
837 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
843 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
844 * @list: queue to initialize
846 * This initializes only the list and queue length aspects of
847 * an sk_buff_head object. This allows to initialize the list
848 * aspects of an sk_buff_head without reinitializing things like
849 * the spinlock. It can also be used for on-stack sk_buff_head
850 * objects where the spinlock is known to not be used.
852 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
854 list
->prev
= list
->next
= (struct sk_buff
*)list
;
859 * This function creates a split out lock class for each invocation;
860 * this is needed for now since a whole lot of users of the skb-queue
861 * infrastructure in drivers have different locking usage (in hardirq)
862 * than the networking core (in softirq only). In the long run either the
863 * network layer or drivers should need annotation to consolidate the
864 * main types of usage into 3 classes.
866 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
868 spin_lock_init(&list
->lock
);
869 __skb_queue_head_init(list
);
872 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
873 struct lock_class_key
*class)
875 skb_queue_head_init(list
);
876 lockdep_set_class(&list
->lock
, class);
880 * Insert an sk_buff on a list.
882 * The "__skb_xxxx()" functions are the non-atomic ones that
883 * can only be called with interrupts disabled.
885 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
886 static inline void __skb_insert(struct sk_buff
*newsk
,
887 struct sk_buff
*prev
, struct sk_buff
*next
,
888 struct sk_buff_head
*list
)
892 next
->prev
= prev
->next
= newsk
;
896 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
897 struct sk_buff
*prev
,
898 struct sk_buff
*next
)
900 struct sk_buff
*first
= list
->next
;
901 struct sk_buff
*last
= list
->prev
;
911 * skb_queue_splice - join two skb lists, this is designed for stacks
912 * @list: the new list to add
913 * @head: the place to add it in the first list
915 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
916 struct sk_buff_head
*head
)
918 if (!skb_queue_empty(list
)) {
919 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
920 head
->qlen
+= list
->qlen
;
925 * skb_queue_splice - join two skb lists and reinitialise the emptied list
926 * @list: the new list to add
927 * @head: the place to add it in the first list
929 * The list at @list is reinitialised
931 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
932 struct sk_buff_head
*head
)
934 if (!skb_queue_empty(list
)) {
935 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
936 head
->qlen
+= list
->qlen
;
937 __skb_queue_head_init(list
);
942 * skb_queue_splice_tail - join two skb lists, each list being a queue
943 * @list: the new list to add
944 * @head: the place to add it in the first list
946 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
947 struct sk_buff_head
*head
)
949 if (!skb_queue_empty(list
)) {
950 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
951 head
->qlen
+= list
->qlen
;
956 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
957 * @list: the new list to add
958 * @head: the place to add it in the first list
960 * Each of the lists is a queue.
961 * The list at @list is reinitialised
963 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
964 struct sk_buff_head
*head
)
966 if (!skb_queue_empty(list
)) {
967 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
968 head
->qlen
+= list
->qlen
;
969 __skb_queue_head_init(list
);
974 * __skb_queue_after - queue a buffer at the list head
976 * @prev: place after this buffer
977 * @newsk: buffer to queue
979 * Queue a buffer int the middle of a list. This function takes no locks
980 * and you must therefore hold required locks before calling it.
982 * A buffer cannot be placed on two lists at the same time.
984 static inline void __skb_queue_after(struct sk_buff_head
*list
,
985 struct sk_buff
*prev
,
986 struct sk_buff
*newsk
)
988 __skb_insert(newsk
, prev
, prev
->next
, list
);
991 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
992 struct sk_buff_head
*list
);
994 static inline void __skb_queue_before(struct sk_buff_head
*list
,
995 struct sk_buff
*next
,
996 struct sk_buff
*newsk
)
998 __skb_insert(newsk
, next
->prev
, next
, list
);
1002 * __skb_queue_head - queue a buffer at the list head
1003 * @list: list to use
1004 * @newsk: buffer to queue
1006 * Queue a buffer at the start of a list. This function takes no locks
1007 * and you must therefore hold required locks before calling it.
1009 * A buffer cannot be placed on two lists at the same time.
1011 extern void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1012 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1013 struct sk_buff
*newsk
)
1015 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1019 * __skb_queue_tail - queue a buffer at the list tail
1020 * @list: list to use
1021 * @newsk: buffer to queue
1023 * Queue a buffer at the end of a list. This function takes no locks
1024 * and you must therefore hold required locks before calling it.
1026 * A buffer cannot be placed on two lists at the same time.
1028 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1029 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1030 struct sk_buff
*newsk
)
1032 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1036 * remove sk_buff from list. _Must_ be called atomically, and with
1039 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1040 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1042 struct sk_buff
*next
, *prev
;
1047 skb
->next
= skb
->prev
= NULL
;
1053 * __skb_dequeue - remove from the head of the queue
1054 * @list: list to dequeue from
1056 * Remove the head of the list. This function does not take any locks
1057 * so must be used with appropriate locks held only. The head item is
1058 * returned or %NULL if the list is empty.
1060 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1061 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1063 struct sk_buff
*skb
= skb_peek(list
);
1065 __skb_unlink(skb
, list
);
1070 * __skb_dequeue_tail - remove from the tail of the queue
1071 * @list: list to dequeue from
1073 * Remove the tail of the list. This function does not take any locks
1074 * so must be used with appropriate locks held only. The tail item is
1075 * returned or %NULL if the list is empty.
1077 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1078 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1080 struct sk_buff
*skb
= skb_peek_tail(list
);
1082 __skb_unlink(skb
, list
);
1087 static inline int skb_is_nonlinear(const struct sk_buff
*skb
)
1089 return skb
->data_len
;
1092 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1094 return skb
->len
- skb
->data_len
;
1097 static inline int skb_pagelen(const struct sk_buff
*skb
)
1101 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1102 len
+= skb_shinfo(skb
)->frags
[i
].size
;
1103 return len
+ skb_headlen(skb
);
1106 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1107 struct page
*page
, int off
, int size
)
1109 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1112 frag
->page_offset
= off
;
1114 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1117 extern void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
,
1120 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1121 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1122 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1124 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1125 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1127 return skb
->head
+ skb
->tail
;
1130 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1132 skb
->tail
= skb
->data
- skb
->head
;
1135 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1137 skb_reset_tail_pointer(skb
);
1138 skb
->tail
+= offset
;
1140 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1141 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1146 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1148 skb
->tail
= skb
->data
;
1151 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1153 skb
->tail
= skb
->data
+ offset
;
1156 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1159 * Add data to an sk_buff
1161 extern unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1162 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1164 unsigned char *tmp
= skb_tail_pointer(skb
);
1165 SKB_LINEAR_ASSERT(skb
);
1171 extern unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1172 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1179 extern unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1180 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1183 BUG_ON(skb
->len
< skb
->data_len
);
1184 return skb
->data
+= len
;
1187 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1189 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1192 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1194 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1196 if (len
> skb_headlen(skb
) &&
1197 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1200 return skb
->data
+= len
;
1203 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1205 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1208 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1210 if (likely(len
<= skb_headlen(skb
)))
1212 if (unlikely(len
> skb
->len
))
1214 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1218 * skb_headroom - bytes at buffer head
1219 * @skb: buffer to check
1221 * Return the number of bytes of free space at the head of an &sk_buff.
1223 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1225 return skb
->data
- skb
->head
;
1229 * skb_tailroom - bytes at buffer end
1230 * @skb: buffer to check
1232 * Return the number of bytes of free space at the tail of an sk_buff
1234 static inline int skb_tailroom(const struct sk_buff
*skb
)
1236 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1240 * skb_reserve - adjust headroom
1241 * @skb: buffer to alter
1242 * @len: bytes to move
1244 * Increase the headroom of an empty &sk_buff by reducing the tail
1245 * room. This is only allowed for an empty buffer.
1247 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
1253 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1254 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1256 return skb
->head
+ skb
->transport_header
;
1259 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1261 skb
->transport_header
= skb
->data
- skb
->head
;
1264 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1267 skb_reset_transport_header(skb
);
1268 skb
->transport_header
+= offset
;
1271 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1273 return skb
->head
+ skb
->network_header
;
1276 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1278 skb
->network_header
= skb
->data
- skb
->head
;
1281 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1283 skb_reset_network_header(skb
);
1284 skb
->network_header
+= offset
;
1287 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1289 return skb
->head
+ skb
->mac_header
;
1292 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1294 return skb
->mac_header
!= ~0U;
1297 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1299 skb
->mac_header
= skb
->data
- skb
->head
;
1302 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1304 skb_reset_mac_header(skb
);
1305 skb
->mac_header
+= offset
;
1308 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1310 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1312 return skb
->transport_header
;
1315 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1317 skb
->transport_header
= skb
->data
;
1320 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1323 skb
->transport_header
= skb
->data
+ offset
;
1326 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1328 return skb
->network_header
;
1331 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1333 skb
->network_header
= skb
->data
;
1336 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1338 skb
->network_header
= skb
->data
+ offset
;
1341 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1343 return skb
->mac_header
;
1346 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1348 return skb
->mac_header
!= NULL
;
1351 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1353 skb
->mac_header
= skb
->data
;
1356 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1358 skb
->mac_header
= skb
->data
+ offset
;
1360 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1362 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
1364 return skb
->csum_start
- skb_headroom(skb
);
1367 static inline int skb_transport_offset(const struct sk_buff
*skb
)
1369 return skb_transport_header(skb
) - skb
->data
;
1372 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
1374 return skb
->transport_header
- skb
->network_header
;
1377 static inline int skb_network_offset(const struct sk_buff
*skb
)
1379 return skb_network_header(skb
) - skb
->data
;
1382 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
1384 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
1388 * CPUs often take a performance hit when accessing unaligned memory
1389 * locations. The actual performance hit varies, it can be small if the
1390 * hardware handles it or large if we have to take an exception and fix it
1393 * Since an ethernet header is 14 bytes network drivers often end up with
1394 * the IP header at an unaligned offset. The IP header can be aligned by
1395 * shifting the start of the packet by 2 bytes. Drivers should do this
1398 * skb_reserve(skb, NET_IP_ALIGN);
1400 * The downside to this alignment of the IP header is that the DMA is now
1401 * unaligned. On some architectures the cost of an unaligned DMA is high
1402 * and this cost outweighs the gains made by aligning the IP header.
1404 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1407 #ifndef NET_IP_ALIGN
1408 #define NET_IP_ALIGN 2
1412 * The networking layer reserves some headroom in skb data (via
1413 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1414 * the header has to grow. In the default case, if the header has to grow
1415 * 32 bytes or less we avoid the reallocation.
1417 * Unfortunately this headroom changes the DMA alignment of the resulting
1418 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1419 * on some architectures. An architecture can override this value,
1420 * perhaps setting it to a cacheline in size (since that will maintain
1421 * cacheline alignment of the DMA). It must be a power of 2.
1423 * Various parts of the networking layer expect at least 32 bytes of
1424 * headroom, you should not reduce this.
1426 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1427 * to reduce average number of cache lines per packet.
1428 * get_rps_cpus() for example only access one 64 bytes aligned block :
1429 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1432 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1435 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
1437 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
1439 if (unlikely(skb
->data_len
)) {
1444 skb_set_tail_pointer(skb
, len
);
1447 extern void skb_trim(struct sk_buff
*skb
, unsigned int len
);
1449 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1452 return ___pskb_trim(skb
, len
);
1453 __skb_trim(skb
, len
);
1457 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1459 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
1463 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1464 * @skb: buffer to alter
1467 * This is identical to pskb_trim except that the caller knows that
1468 * the skb is not cloned so we should never get an error due to out-
1471 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
1473 int err
= pskb_trim(skb
, len
);
1478 * skb_orphan - orphan a buffer
1479 * @skb: buffer to orphan
1481 * If a buffer currently has an owner then we call the owner's
1482 * destructor function and make the @skb unowned. The buffer continues
1483 * to exist but is no longer charged to its former owner.
1485 static inline void skb_orphan(struct sk_buff
*skb
)
1487 if (skb
->destructor
)
1488 skb
->destructor(skb
);
1489 skb
->destructor
= NULL
;
1494 * __skb_queue_purge - empty a list
1495 * @list: list to empty
1497 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1498 * the list and one reference dropped. This function does not take the
1499 * list lock and the caller must hold the relevant locks to use it.
1501 extern void skb_queue_purge(struct sk_buff_head
*list
);
1502 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
1504 struct sk_buff
*skb
;
1505 while ((skb
= __skb_dequeue(list
)) != NULL
)
1510 * __dev_alloc_skb - allocate an skbuff for receiving
1511 * @length: length to allocate
1512 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1514 * Allocate a new &sk_buff and assign it a usage count of one. The
1515 * buffer has unspecified headroom built in. Users should allocate
1516 * the headroom they think they need without accounting for the
1517 * built in space. The built in space is used for optimisations.
1519 * %NULL is returned if there is no free memory.
1521 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
1524 struct sk_buff
*skb
= alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
);
1526 skb_reserve(skb
, NET_SKB_PAD
);
1530 extern struct sk_buff
*dev_alloc_skb(unsigned int length
);
1532 extern struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
1533 unsigned int length
, gfp_t gfp_mask
);
1536 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1537 * @dev: network device to receive on
1538 * @length: length to allocate
1540 * Allocate a new &sk_buff and assign it a usage count of one. The
1541 * buffer has unspecified headroom built in. Users should allocate
1542 * the headroom they think they need without accounting for the
1543 * built in space. The built in space is used for optimisations.
1545 * %NULL is returned if there is no free memory. Although this function
1546 * allocates memory it can be called from an interrupt.
1548 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
1549 unsigned int length
)
1551 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
1554 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
1555 unsigned int length
)
1557 struct sk_buff
*skb
= netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
);
1559 if (NET_IP_ALIGN
&& skb
)
1560 skb_reserve(skb
, NET_IP_ALIGN
);
1565 * __netdev_alloc_page - allocate a page for ps-rx on a specific device
1566 * @dev: network device to receive on
1567 * @gfp_mask: alloc_pages_node mask
1569 * Allocate a new page. dev currently unused.
1571 * %NULL is returned if there is no free memory.
1573 static inline struct page
*__netdev_alloc_page(struct net_device
*dev
, gfp_t gfp_mask
)
1575 return alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, 0);
1579 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1580 * @dev: network device to receive on
1582 * Allocate a new page. dev currently unused.
1584 * %NULL is returned if there is no free memory.
1586 static inline struct page
*netdev_alloc_page(struct net_device
*dev
)
1588 return __netdev_alloc_page(dev
, GFP_ATOMIC
);
1591 static inline void netdev_free_page(struct net_device
*dev
, struct page
*page
)
1597 * skb_clone_writable - is the header of a clone writable
1598 * @skb: buffer to check
1599 * @len: length up to which to write
1601 * Returns true if modifying the header part of the cloned buffer
1602 * does not requires the data to be copied.
1604 static inline int skb_clone_writable(struct sk_buff
*skb
, unsigned int len
)
1606 return !skb_header_cloned(skb
) &&
1607 skb_headroom(skb
) + len
<= skb
->hdr_len
;
1610 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
1615 if (headroom
< NET_SKB_PAD
)
1616 headroom
= NET_SKB_PAD
;
1617 if (headroom
> skb_headroom(skb
))
1618 delta
= headroom
- skb_headroom(skb
);
1620 if (delta
|| cloned
)
1621 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
1627 * skb_cow - copy header of skb when it is required
1628 * @skb: buffer to cow
1629 * @headroom: needed headroom
1631 * If the skb passed lacks sufficient headroom or its data part
1632 * is shared, data is reallocated. If reallocation fails, an error
1633 * is returned and original skb is not changed.
1635 * The result is skb with writable area skb->head...skb->tail
1636 * and at least @headroom of space at head.
1638 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
1640 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
1644 * skb_cow_head - skb_cow but only making the head writable
1645 * @skb: buffer to cow
1646 * @headroom: needed headroom
1648 * This function is identical to skb_cow except that we replace the
1649 * skb_cloned check by skb_header_cloned. It should be used when
1650 * you only need to push on some header and do not need to modify
1653 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
1655 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
1659 * skb_padto - pad an skbuff up to a minimal size
1660 * @skb: buffer to pad
1661 * @len: minimal length
1663 * Pads up a buffer to ensure the trailing bytes exist and are
1664 * blanked. If the buffer already contains sufficient data it
1665 * is untouched. Otherwise it is extended. Returns zero on
1666 * success. The skb is freed on error.
1669 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
1671 unsigned int size
= skb
->len
;
1672 if (likely(size
>= len
))
1674 return skb_pad(skb
, len
- size
);
1677 static inline int skb_add_data(struct sk_buff
*skb
,
1678 char __user
*from
, int copy
)
1680 const int off
= skb
->len
;
1682 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1684 __wsum csum
= csum_and_copy_from_user(from
, skb_put(skb
, copy
),
1687 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
1690 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
1693 __skb_trim(skb
, off
);
1697 static inline int skb_can_coalesce(struct sk_buff
*skb
, int i
,
1698 struct page
*page
, int off
)
1701 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
1703 return page
== frag
->page
&&
1704 off
== frag
->page_offset
+ frag
->size
;
1709 static inline int __skb_linearize(struct sk_buff
*skb
)
1711 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
1715 * skb_linearize - convert paged skb to linear one
1716 * @skb: buffer to linarize
1718 * If there is no free memory -ENOMEM is returned, otherwise zero
1719 * is returned and the old skb data released.
1721 static inline int skb_linearize(struct sk_buff
*skb
)
1723 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
1727 * skb_linearize_cow - make sure skb is linear and writable
1728 * @skb: buffer to process
1730 * If there is no free memory -ENOMEM is returned, otherwise zero
1731 * is returned and the old skb data released.
1733 static inline int skb_linearize_cow(struct sk_buff
*skb
)
1735 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
1736 __skb_linearize(skb
) : 0;
1740 * skb_postpull_rcsum - update checksum for received skb after pull
1741 * @skb: buffer to update
1742 * @start: start of data before pull
1743 * @len: length of data pulled
1745 * After doing a pull on a received packet, you need to call this to
1746 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1747 * CHECKSUM_NONE so that it can be recomputed from scratch.
1750 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
1751 const void *start
, unsigned int len
)
1753 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1754 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
1757 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
1760 * pskb_trim_rcsum - trim received skb and update checksum
1761 * @skb: buffer to trim
1764 * This is exactly the same as pskb_trim except that it ensures the
1765 * checksum of received packets are still valid after the operation.
1768 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
1770 if (likely(len
>= skb
->len
))
1772 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1773 skb
->ip_summed
= CHECKSUM_NONE
;
1774 return __pskb_trim(skb
, len
);
1777 #define skb_queue_walk(queue, skb) \
1778 for (skb = (queue)->next; \
1779 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1782 #define skb_queue_walk_safe(queue, skb, tmp) \
1783 for (skb = (queue)->next, tmp = skb->next; \
1784 skb != (struct sk_buff *)(queue); \
1785 skb = tmp, tmp = skb->next)
1787 #define skb_queue_walk_from(queue, skb) \
1788 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1791 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1792 for (tmp = skb->next; \
1793 skb != (struct sk_buff *)(queue); \
1794 skb = tmp, tmp = skb->next)
1796 #define skb_queue_reverse_walk(queue, skb) \
1797 for (skb = (queue)->prev; \
1798 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1801 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
1802 for (skb = (queue)->prev, tmp = skb->prev; \
1803 skb != (struct sk_buff *)(queue); \
1804 skb = tmp, tmp = skb->prev)
1806 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
1807 for (tmp = skb->prev; \
1808 skb != (struct sk_buff *)(queue); \
1809 skb = tmp, tmp = skb->prev)
1811 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
1813 return skb_shinfo(skb
)->frag_list
!= NULL
;
1816 static inline void skb_frag_list_init(struct sk_buff
*skb
)
1818 skb_shinfo(skb
)->frag_list
= NULL
;
1821 static inline void skb_frag_add_head(struct sk_buff
*skb
, struct sk_buff
*frag
)
1823 frag
->next
= skb_shinfo(skb
)->frag_list
;
1824 skb_shinfo(skb
)->frag_list
= frag
;
1827 #define skb_walk_frags(skb, iter) \
1828 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1830 extern struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1831 int *peeked
, int *err
);
1832 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1833 int noblock
, int *err
);
1834 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
1835 struct poll_table_struct
*wait
);
1836 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
1837 int offset
, struct iovec
*to
,
1839 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
,
1842 extern int skb_copy_datagram_from_iovec(struct sk_buff
*skb
,
1844 const struct iovec
*from
,
1847 extern int skb_copy_datagram_const_iovec(const struct sk_buff
*from
,
1849 const struct iovec
*to
,
1852 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
1853 extern void skb_free_datagram_locked(struct sock
*sk
,
1854 struct sk_buff
*skb
);
1855 extern int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
,
1856 unsigned int flags
);
1857 extern __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
,
1858 int len
, __wsum csum
);
1859 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
1861 extern int skb_store_bits(struct sk_buff
*skb
, int offset
,
1862 const void *from
, int len
);
1863 extern __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
,
1864 int offset
, u8
*to
, int len
,
1866 extern int skb_splice_bits(struct sk_buff
*skb
,
1867 unsigned int offset
,
1868 struct pipe_inode_info
*pipe
,
1870 unsigned int flags
);
1871 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
1872 extern void skb_split(struct sk_buff
*skb
,
1873 struct sk_buff
*skb1
, const u32 len
);
1874 extern int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
,
1877 extern struct sk_buff
*skb_segment(struct sk_buff
*skb
, u32 features
);
1879 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
1880 int len
, void *buffer
)
1882 int hlen
= skb_headlen(skb
);
1884 if (hlen
- offset
>= len
)
1885 return skb
->data
+ offset
;
1887 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
1893 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
1895 const unsigned int len
)
1897 memcpy(to
, skb
->data
, len
);
1900 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
1901 const int offset
, void *to
,
1902 const unsigned int len
)
1904 memcpy(to
, skb
->data
+ offset
, len
);
1907 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
1909 const unsigned int len
)
1911 memcpy(skb
->data
, from
, len
);
1914 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
1917 const unsigned int len
)
1919 memcpy(skb
->data
+ offset
, from
, len
);
1922 extern void skb_init(void);
1924 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
1930 * skb_get_timestamp - get timestamp from a skb
1931 * @skb: skb to get stamp from
1932 * @stamp: pointer to struct timeval to store stamp in
1934 * Timestamps are stored in the skb as offsets to a base timestamp.
1935 * This function converts the offset back to a struct timeval and stores
1938 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
1939 struct timeval
*stamp
)
1941 *stamp
= ktime_to_timeval(skb
->tstamp
);
1944 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
1945 struct timespec
*stamp
)
1947 *stamp
= ktime_to_timespec(skb
->tstamp
);
1950 static inline void __net_timestamp(struct sk_buff
*skb
)
1952 skb
->tstamp
= ktime_get_real();
1955 static inline ktime_t
net_timedelta(ktime_t t
)
1957 return ktime_sub(ktime_get_real(), t
);
1960 static inline ktime_t
net_invalid_timestamp(void)
1962 return ktime_set(0, 0);
1965 extern void skb_timestamping_init(void);
1967 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
1969 extern void skb_clone_tx_timestamp(struct sk_buff
*skb
);
1970 extern bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
1972 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
1974 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
1978 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
1983 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
1986 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
1988 * @skb: clone of the the original outgoing packet
1989 * @hwtstamps: hardware time stamps
1992 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
1993 struct skb_shared_hwtstamps
*hwtstamps
);
1996 * skb_tstamp_tx - queue clone of skb with send time stamps
1997 * @orig_skb: the original outgoing packet
1998 * @hwtstamps: hardware time stamps, may be NULL if not available
2000 * If the skb has a socket associated, then this function clones the
2001 * skb (thus sharing the actual data and optional structures), stores
2002 * the optional hardware time stamping information (if non NULL) or
2003 * generates a software time stamp (otherwise), then queues the clone
2004 * to the error queue of the socket. Errors are silently ignored.
2006 extern void skb_tstamp_tx(struct sk_buff
*orig_skb
,
2007 struct skb_shared_hwtstamps
*hwtstamps
);
2009 static inline void sw_tx_timestamp(struct sk_buff
*skb
)
2011 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
&&
2012 !(skb_shinfo(skb
)->tx_flags
& SKBTX_IN_PROGRESS
))
2013 skb_tstamp_tx(skb
, NULL
);
2017 * skb_tx_timestamp() - Driver hook for transmit timestamping
2019 * Ethernet MAC Drivers should call this function in their hard_xmit()
2020 * function as soon as possible after giving the sk_buff to the MAC
2021 * hardware, but before freeing the sk_buff.
2023 * @skb: A socket buffer.
2025 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
2027 skb_clone_tx_timestamp(skb
);
2028 sw_tx_timestamp(skb
);
2031 extern __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
2032 extern __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
2034 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
2036 return skb
->ip_summed
& CHECKSUM_UNNECESSARY
;
2040 * skb_checksum_complete - Calculate checksum of an entire packet
2041 * @skb: packet to process
2043 * This function calculates the checksum over the entire packet plus
2044 * the value of skb->csum. The latter can be used to supply the
2045 * checksum of a pseudo header as used by TCP/UDP. It returns the
2048 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2049 * this function can be used to verify that checksum on received
2050 * packets. In that case the function should return zero if the
2051 * checksum is correct. In particular, this function will return zero
2052 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2053 * hardware has already verified the correctness of the checksum.
2055 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
2057 return skb_csum_unnecessary(skb
) ?
2058 0 : __skb_checksum_complete(skb
);
2061 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2062 extern void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
2063 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
2065 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
2066 nf_conntrack_destroy(nfct
);
2068 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
2071 atomic_inc(&nfct
->use
);
2074 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2075 static inline void nf_conntrack_get_reasm(struct sk_buff
*skb
)
2078 atomic_inc(&skb
->users
);
2080 static inline void nf_conntrack_put_reasm(struct sk_buff
*skb
)
2086 #ifdef CONFIG_BRIDGE_NETFILTER
2087 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
2089 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
2092 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
2095 atomic_inc(&nf_bridge
->use
);
2097 #endif /* CONFIG_BRIDGE_NETFILTER */
2098 static inline void nf_reset(struct sk_buff
*skb
)
2100 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2101 nf_conntrack_put(skb
->nfct
);
2104 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2105 nf_conntrack_put_reasm(skb
->nfct_reasm
);
2106 skb
->nfct_reasm
= NULL
;
2108 #ifdef CONFIG_BRIDGE_NETFILTER
2109 nf_bridge_put(skb
->nf_bridge
);
2110 skb
->nf_bridge
= NULL
;
2114 /* Note: This doesn't put any conntrack and bridge info in dst. */
2115 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2117 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2118 dst
->nfct
= src
->nfct
;
2119 nf_conntrack_get(src
->nfct
);
2120 dst
->nfctinfo
= src
->nfctinfo
;
2122 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2123 dst
->nfct_reasm
= src
->nfct_reasm
;
2124 nf_conntrack_get_reasm(src
->nfct_reasm
);
2126 #ifdef CONFIG_BRIDGE_NETFILTER
2127 dst
->nf_bridge
= src
->nf_bridge
;
2128 nf_bridge_get(src
->nf_bridge
);
2132 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2134 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2135 nf_conntrack_put(dst
->nfct
);
2137 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2138 nf_conntrack_put_reasm(dst
->nfct_reasm
);
2140 #ifdef CONFIG_BRIDGE_NETFILTER
2141 nf_bridge_put(dst
->nf_bridge
);
2143 __nf_copy(dst
, src
);
2146 #ifdef CONFIG_NETWORK_SECMARK
2147 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2149 to
->secmark
= from
->secmark
;
2152 static inline void skb_init_secmark(struct sk_buff
*skb
)
2157 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2160 static inline void skb_init_secmark(struct sk_buff
*skb
)
2164 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
2166 skb
->queue_mapping
= queue_mapping
;
2169 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
2171 return skb
->queue_mapping
;
2174 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
2176 to
->queue_mapping
= from
->queue_mapping
;
2179 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
2181 skb
->queue_mapping
= rx_queue
+ 1;
2184 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
2186 return skb
->queue_mapping
- 1;
2189 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
2191 return skb
->queue_mapping
!= 0;
2194 extern u16
__skb_tx_hash(const struct net_device
*dev
,
2195 const struct sk_buff
*skb
,
2196 unsigned int num_tx_queues
);
2199 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2204 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2210 static inline int skb_is_gso(const struct sk_buff
*skb
)
2212 return skb_shinfo(skb
)->gso_size
;
2215 static inline int skb_is_gso_v6(const struct sk_buff
*skb
)
2217 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
2220 extern void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
2222 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
2224 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2225 * wanted then gso_type will be set. */
2226 struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
2227 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
2228 unlikely(shinfo
->gso_type
== 0)) {
2229 __skb_warn_lro_forwarding(skb
);
2235 static inline void skb_forward_csum(struct sk_buff
*skb
)
2237 /* Unfortunately we don't support this one. Any brave souls? */
2238 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2239 skb
->ip_summed
= CHECKSUM_NONE
;
2243 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2244 * @skb: skb to check
2246 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2247 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2248 * use this helper, to document places where we make this assertion.
2250 static inline void skb_checksum_none_assert(struct sk_buff
*skb
)
2253 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
2257 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
2258 #endif /* __KERNEL__ */
2259 #endif /* _LINUX_SKBUFF_H */