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 <linux/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. Since
126 * GRO uses frags we allocate at least 16 regardless of page size.
128 #if (65536/PAGE_SIZE + 2) < 16
129 #define MAX_SKB_FRAGS 16UL
131 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
134 typedef struct skb_frag_struct skb_frag_t
;
136 struct skb_frag_struct
{
138 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
147 #define HAVE_HW_TIME_STAMP
150 * struct skb_shared_hwtstamps - hardware time stamps
151 * @hwtstamp: hardware time stamp transformed into duration
152 * since arbitrary point in time
153 * @syststamp: hwtstamp transformed to system time base
155 * Software time stamps generated by ktime_get_real() are stored in
156 * skb->tstamp. The relation between the different kinds of time
157 * stamps is as follows:
159 * syststamp and tstamp can be compared against each other in
160 * arbitrary combinations. The accuracy of a
161 * syststamp/tstamp/"syststamp from other device" comparison is
162 * limited by the accuracy of the transformation into system time
163 * base. This depends on the device driver and its underlying
166 * hwtstamps can only be compared against other hwtstamps from
169 * This structure is attached to packets as part of the
170 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
172 struct skb_shared_hwtstamps
{
177 /* Definitions for tx_flags in struct skb_shared_info */
179 /* generate hardware time stamp */
180 SKBTX_HW_TSTAMP
= 1 << 0,
182 /* generate software time stamp */
183 SKBTX_SW_TSTAMP
= 1 << 1,
185 /* device driver is going to provide hardware time stamp */
186 SKBTX_IN_PROGRESS
= 1 << 2,
188 /* ensure the originating sk reference is available on driver level */
189 SKBTX_DRV_NEEDS_SK_REF
= 1 << 3,
191 /* device driver supports TX zero-copy buffers */
192 SKBTX_DEV_ZEROCOPY
= 1 << 4,
196 * The callback notifies userspace to release buffers when skb DMA is done in
197 * lower device, the skb last reference should be 0 when calling this.
198 * The desc is used to track userspace buffer index.
201 void (*callback
)(void *);
206 /* This data is invariant across clones and lives at
207 * the end of the header data, ie. at skb->end.
209 struct skb_shared_info
{
210 unsigned short nr_frags
;
211 unsigned short gso_size
;
212 /* Warning: this field is not always filled in (UFO)! */
213 unsigned short gso_segs
;
214 unsigned short gso_type
;
217 struct sk_buff
*frag_list
;
218 struct skb_shared_hwtstamps hwtstamps
;
221 * Warning : all fields before dataref are cleared in __alloc_skb()
225 /* Intermediate layers must ensure that destructor_arg
226 * remains valid until skb destructor */
227 void * destructor_arg
;
229 /* must be last field, see pskb_expand_head() */
230 skb_frag_t frags
[MAX_SKB_FRAGS
];
233 /* We divide dataref into two halves. The higher 16 bits hold references
234 * to the payload part of skb->data. The lower 16 bits hold references to
235 * the entire skb->data. A clone of a headerless skb holds the length of
236 * the header in skb->hdr_len.
238 * All users must obey the rule that the skb->data reference count must be
239 * greater than or equal to the payload reference count.
241 * Holding a reference to the payload part means that the user does not
242 * care about modifications to the header part of skb->data.
244 #define SKB_DATAREF_SHIFT 16
245 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
249 SKB_FCLONE_UNAVAILABLE
,
255 SKB_GSO_TCPV4
= 1 << 0,
256 SKB_GSO_UDP
= 1 << 1,
258 /* This indicates the skb is from an untrusted source. */
259 SKB_GSO_DODGY
= 1 << 2,
261 /* This indicates the tcp segment has CWR set. */
262 SKB_GSO_TCP_ECN
= 1 << 3,
264 SKB_GSO_TCPV6
= 1 << 4,
266 SKB_GSO_FCOE
= 1 << 5,
269 #if BITS_PER_LONG > 32
270 #define NET_SKBUFF_DATA_USES_OFFSET 1
273 #ifdef NET_SKBUFF_DATA_USES_OFFSET
274 typedef unsigned int sk_buff_data_t
;
276 typedef unsigned char *sk_buff_data_t
;
279 #if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
280 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
281 #define NET_SKBUFF_NF_DEFRAG_NEEDED 1
285 * struct sk_buff - socket buffer
286 * @next: Next buffer in list
287 * @prev: Previous buffer in list
288 * @tstamp: Time we arrived
289 * @sk: Socket we are owned by
290 * @dev: Device we arrived on/are leaving by
291 * @cb: Control buffer. Free for use by every layer. Put private vars here
292 * @_skb_refdst: destination entry (with norefcount bit)
293 * @sp: the security path, used for xfrm
294 * @len: Length of actual data
295 * @data_len: Data length
296 * @mac_len: Length of link layer header
297 * @hdr_len: writable header length of cloned skb
298 * @csum: Checksum (must include start/offset pair)
299 * @csum_start: Offset from skb->head where checksumming should start
300 * @csum_offset: Offset from csum_start where checksum should be stored
301 * @priority: Packet queueing priority
302 * @local_df: allow local fragmentation
303 * @cloned: Head may be cloned (check refcnt to be sure)
304 * @ip_summed: Driver fed us an IP checksum
305 * @nohdr: Payload reference only, must not modify header
306 * @nfctinfo: Relationship of this skb to the connection
307 * @pkt_type: Packet class
308 * @fclone: skbuff clone status
309 * @ipvs_property: skbuff is owned by ipvs
310 * @peeked: this packet has been seen already, so stats have been
311 * done for it, don't do them again
312 * @nf_trace: netfilter packet trace flag
313 * @protocol: Packet protocol from driver
314 * @destructor: Destruct function
315 * @nfct: Associated connection, if any
316 * @nfct_reasm: netfilter conntrack re-assembly pointer
317 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
318 * @skb_iif: ifindex of device we arrived on
319 * @tc_index: Traffic control index
320 * @tc_verd: traffic control verdict
321 * @rxhash: the packet hash computed on receive
322 * @queue_mapping: Queue mapping for multiqueue devices
323 * @ndisc_nodetype: router type (from link layer)
324 * @ooo_okay: allow the mapping of a socket to a queue to be changed
325 * @dma_cookie: a cookie to one of several possible DMA operations
326 * done by skb DMA functions
327 * @secmark: security marking
328 * @mark: Generic packet mark
329 * @dropcount: total number of sk_receive_queue overflows
330 * @vlan_tci: vlan tag control information
331 * @transport_header: Transport layer header
332 * @network_header: Network layer header
333 * @mac_header: Link layer header
334 * @tail: Tail pointer
336 * @head: Head of buffer
337 * @data: Data head pointer
338 * @truesize: Buffer size
339 * @users: User count - see {datagram,tcp}.c
343 /* These two members must be first. */
344 struct sk_buff
*next
;
345 struct sk_buff
*prev
;
350 struct net_device
*dev
;
353 * This is the control buffer. It is free to use for every
354 * layer. Please put your private variables there. If you
355 * want to keep them across layers you have to do a skb_clone()
356 * first. This is owned by whoever has the skb queued ATM.
358 char cb
[48] __aligned(8);
360 unsigned long _skb_refdst
;
376 kmemcheck_bitfield_begin(flags1
);
387 kmemcheck_bitfield_end(flags1
);
390 void (*destructor
)(struct sk_buff
*skb
);
391 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
392 struct nf_conntrack
*nfct
;
394 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
395 struct sk_buff
*nfct_reasm
;
397 #ifdef CONFIG_BRIDGE_NETFILTER
398 struct nf_bridge_info
*nf_bridge
;
402 #ifdef CONFIG_NET_SCHED
403 __u16 tc_index
; /* traffic control index */
404 #ifdef CONFIG_NET_CLS_ACT
405 __u16 tc_verd
; /* traffic control verdict */
412 kmemcheck_bitfield_begin(flags2
);
413 #ifdef CONFIG_IPV6_NDISC_NODETYPE
414 __u8 ndisc_nodetype
:2;
417 kmemcheck_bitfield_end(flags2
);
421 #ifdef CONFIG_NET_DMA
422 dma_cookie_t dma_cookie
;
424 #ifdef CONFIG_NETWORK_SECMARK
434 sk_buff_data_t transport_header
;
435 sk_buff_data_t network_header
;
436 sk_buff_data_t mac_header
;
437 /* These elements must be at the end, see alloc_skb() for details. */
442 unsigned int truesize
;
448 * Handling routines are only of interest to the kernel
450 #include <linux/slab.h>
452 #include <asm/system.h>
455 * skb might have a dst pointer attached, refcounted or not.
456 * _skb_refdst low order bit is set if refcount was _not_ taken
458 #define SKB_DST_NOREF 1UL
459 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
462 * skb_dst - returns skb dst_entry
465 * Returns skb dst_entry, regardless of reference taken or not.
467 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
469 /* If refdst was not refcounted, check we still are in a
470 * rcu_read_lock section
472 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
473 !rcu_read_lock_held() &&
474 !rcu_read_lock_bh_held());
475 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
479 * skb_dst_set - sets skb dst
483 * Sets skb dst, assuming a reference was taken on dst and should
484 * be released by skb_dst_drop()
486 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
488 skb
->_skb_refdst
= (unsigned long)dst
;
491 extern void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
);
494 * skb_dst_is_noref - Test if skb dst isn't refcounted
497 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
499 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
502 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
504 return (struct rtable
*)skb_dst(skb
);
507 extern void kfree_skb(struct sk_buff
*skb
);
508 extern void consume_skb(struct sk_buff
*skb
);
509 extern void __kfree_skb(struct sk_buff
*skb
);
510 extern struct sk_buff
*__alloc_skb(unsigned int size
,
511 gfp_t priority
, int fclone
, int node
);
512 static inline struct sk_buff
*alloc_skb(unsigned int size
,
515 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
518 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
521 return __alloc_skb(size
, priority
, 1, NUMA_NO_NODE
);
524 extern bool skb_recycle_check(struct sk_buff
*skb
, int skb_size
);
526 extern struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
527 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
529 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
531 extern struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
533 extern int pskb_expand_head(struct sk_buff
*skb
,
534 int nhead
, int ntail
,
536 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
537 unsigned int headroom
);
538 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
539 int newheadroom
, int newtailroom
,
541 extern int skb_to_sgvec(struct sk_buff
*skb
,
542 struct scatterlist
*sg
, int offset
,
544 extern int skb_cow_data(struct sk_buff
*skb
, int tailbits
,
545 struct sk_buff
**trailer
);
546 extern int skb_pad(struct sk_buff
*skb
, int pad
);
547 #define dev_kfree_skb(a) consume_skb(a)
549 extern int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
550 int getfrag(void *from
, char *to
, int offset
,
551 int len
,int odd
, struct sk_buff
*skb
),
552 void *from
, int length
);
554 struct skb_seq_state
{
558 __u32 stepped_offset
;
559 struct sk_buff
*root_skb
;
560 struct sk_buff
*cur_skb
;
564 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
565 unsigned int from
, unsigned int to
,
566 struct skb_seq_state
*st
);
567 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
568 struct skb_seq_state
*st
);
569 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
571 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
572 unsigned int to
, struct ts_config
*config
,
573 struct ts_state
*state
);
575 extern __u32
__skb_get_rxhash(struct sk_buff
*skb
);
576 static inline __u32
skb_get_rxhash(struct sk_buff
*skb
)
579 skb
->rxhash
= __skb_get_rxhash(skb
);
584 #ifdef NET_SKBUFF_DATA_USES_OFFSET
585 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
587 return skb
->head
+ skb
->end
;
590 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
597 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
599 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
601 return &skb_shinfo(skb
)->hwtstamps
;
605 * skb_queue_empty - check if a queue is empty
608 * Returns true if the queue is empty, false otherwise.
610 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
612 return list
->next
== (struct sk_buff
*)list
;
616 * skb_queue_is_last - check if skb is the last entry in the queue
620 * Returns true if @skb is the last buffer on the list.
622 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
623 const struct sk_buff
*skb
)
625 return skb
->next
== (struct sk_buff
*)list
;
629 * skb_queue_is_first - check if skb is the first entry in the queue
633 * Returns true if @skb is the first buffer on the list.
635 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
636 const struct sk_buff
*skb
)
638 return skb
->prev
== (struct sk_buff
*)list
;
642 * skb_queue_next - return the next packet in the queue
644 * @skb: current buffer
646 * Return the next packet in @list after @skb. It is only valid to
647 * call this if skb_queue_is_last() evaluates to false.
649 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
650 const struct sk_buff
*skb
)
652 /* This BUG_ON may seem severe, but if we just return then we
653 * are going to dereference garbage.
655 BUG_ON(skb_queue_is_last(list
, skb
));
660 * skb_queue_prev - return the prev packet in the queue
662 * @skb: current buffer
664 * Return the prev packet in @list before @skb. It is only valid to
665 * call this if skb_queue_is_first() evaluates to false.
667 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
668 const struct sk_buff
*skb
)
670 /* This BUG_ON may seem severe, but if we just return then we
671 * are going to dereference garbage.
673 BUG_ON(skb_queue_is_first(list
, skb
));
678 * skb_get - reference buffer
679 * @skb: buffer to reference
681 * Makes another reference to a socket buffer and returns a pointer
684 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
686 atomic_inc(&skb
->users
);
691 * If users == 1, we are the only owner and are can avoid redundant
696 * skb_cloned - is the buffer a clone
697 * @skb: buffer to check
699 * Returns true if the buffer was generated with skb_clone() and is
700 * one of multiple shared copies of the buffer. Cloned buffers are
701 * shared data so must not be written to under normal circumstances.
703 static inline int skb_cloned(const struct sk_buff
*skb
)
705 return skb
->cloned
&&
706 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
710 * skb_header_cloned - is the header a clone
711 * @skb: buffer to check
713 * Returns true if modifying the header part of the buffer requires
714 * the data to be copied.
716 static inline int skb_header_cloned(const struct sk_buff
*skb
)
723 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
724 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
729 * skb_header_release - release reference to header
730 * @skb: buffer to operate on
732 * Drop a reference to the header part of the buffer. This is done
733 * by acquiring a payload reference. You must not read from the header
734 * part of skb->data after this.
736 static inline void skb_header_release(struct sk_buff
*skb
)
740 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
744 * skb_shared - is the buffer shared
745 * @skb: buffer to check
747 * Returns true if more than one person has a reference to this
750 static inline int skb_shared(const struct sk_buff
*skb
)
752 return atomic_read(&skb
->users
) != 1;
756 * skb_share_check - check if buffer is shared and if so clone it
757 * @skb: buffer to check
758 * @pri: priority for memory allocation
760 * If the buffer is shared the buffer is cloned and the old copy
761 * drops a reference. A new clone with a single reference is returned.
762 * If the buffer is not shared the original buffer is returned. When
763 * being called from interrupt status or with spinlocks held pri must
766 * NULL is returned on a memory allocation failure.
768 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
,
771 might_sleep_if(pri
& __GFP_WAIT
);
772 if (skb_shared(skb
)) {
773 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
781 * Copy shared buffers into a new sk_buff. We effectively do COW on
782 * packets to handle cases where we have a local reader and forward
783 * and a couple of other messy ones. The normal one is tcpdumping
784 * a packet thats being forwarded.
788 * skb_unshare - make a copy of a shared buffer
789 * @skb: buffer to check
790 * @pri: priority for memory allocation
792 * If the socket buffer is a clone then this function creates a new
793 * copy of the data, drops a reference count on the old copy and returns
794 * the new copy with the reference count at 1. If the buffer is not a clone
795 * the original buffer is returned. When called with a spinlock held or
796 * from interrupt state @pri must be %GFP_ATOMIC
798 * %NULL is returned on a memory allocation failure.
800 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
803 might_sleep_if(pri
& __GFP_WAIT
);
804 if (skb_cloned(skb
)) {
805 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
806 kfree_skb(skb
); /* Free our shared copy */
813 * skb_peek - peek at the head of an &sk_buff_head
814 * @list_: list to peek at
816 * Peek an &sk_buff. Unlike most other operations you _MUST_
817 * be careful with this one. A peek leaves the buffer on the
818 * list and someone else may run off with it. You must hold
819 * the appropriate locks or have a private queue to do this.
821 * Returns %NULL for an empty list or a pointer to the head element.
822 * The reference count is not incremented and the reference is therefore
823 * volatile. Use with caution.
825 static inline struct sk_buff
*skb_peek(struct sk_buff_head
*list_
)
827 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->next
;
828 if (list
== (struct sk_buff
*)list_
)
834 * skb_peek_tail - peek at the tail of an &sk_buff_head
835 * @list_: list to peek at
837 * Peek an &sk_buff. Unlike most other operations you _MUST_
838 * be careful with this one. A peek leaves the buffer on the
839 * list and someone else may run off with it. You must hold
840 * the appropriate locks or have a private queue to do this.
842 * Returns %NULL for an empty list or a pointer to the tail element.
843 * The reference count is not incremented and the reference is therefore
844 * volatile. Use with caution.
846 static inline struct sk_buff
*skb_peek_tail(struct sk_buff_head
*list_
)
848 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->prev
;
849 if (list
== (struct sk_buff
*)list_
)
855 * skb_queue_len - get queue length
856 * @list_: list to measure
858 * Return the length of an &sk_buff queue.
860 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
866 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
867 * @list: queue to initialize
869 * This initializes only the list and queue length aspects of
870 * an sk_buff_head object. This allows to initialize the list
871 * aspects of an sk_buff_head without reinitializing things like
872 * the spinlock. It can also be used for on-stack sk_buff_head
873 * objects where the spinlock is known to not be used.
875 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
877 list
->prev
= list
->next
= (struct sk_buff
*)list
;
882 * This function creates a split out lock class for each invocation;
883 * this is needed for now since a whole lot of users of the skb-queue
884 * infrastructure in drivers have different locking usage (in hardirq)
885 * than the networking core (in softirq only). In the long run either the
886 * network layer or drivers should need annotation to consolidate the
887 * main types of usage into 3 classes.
889 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
891 spin_lock_init(&list
->lock
);
892 __skb_queue_head_init(list
);
895 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
896 struct lock_class_key
*class)
898 skb_queue_head_init(list
);
899 lockdep_set_class(&list
->lock
, class);
903 * Insert an sk_buff on a list.
905 * The "__skb_xxxx()" functions are the non-atomic ones that
906 * can only be called with interrupts disabled.
908 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
909 static inline void __skb_insert(struct sk_buff
*newsk
,
910 struct sk_buff
*prev
, struct sk_buff
*next
,
911 struct sk_buff_head
*list
)
915 next
->prev
= prev
->next
= newsk
;
919 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
920 struct sk_buff
*prev
,
921 struct sk_buff
*next
)
923 struct sk_buff
*first
= list
->next
;
924 struct sk_buff
*last
= list
->prev
;
934 * skb_queue_splice - join two skb lists, this is designed for stacks
935 * @list: the new list to add
936 * @head: the place to add it in the first list
938 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
939 struct sk_buff_head
*head
)
941 if (!skb_queue_empty(list
)) {
942 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
943 head
->qlen
+= list
->qlen
;
948 * skb_queue_splice - join two skb lists and reinitialise the emptied list
949 * @list: the new list to add
950 * @head: the place to add it in the first list
952 * The list at @list is reinitialised
954 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
955 struct sk_buff_head
*head
)
957 if (!skb_queue_empty(list
)) {
958 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
959 head
->qlen
+= list
->qlen
;
960 __skb_queue_head_init(list
);
965 * skb_queue_splice_tail - join two skb lists, each list being a queue
966 * @list: the new list to add
967 * @head: the place to add it in the first list
969 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
970 struct sk_buff_head
*head
)
972 if (!skb_queue_empty(list
)) {
973 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
974 head
->qlen
+= list
->qlen
;
979 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
980 * @list: the new list to add
981 * @head: the place to add it in the first list
983 * Each of the lists is a queue.
984 * The list at @list is reinitialised
986 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
987 struct sk_buff_head
*head
)
989 if (!skb_queue_empty(list
)) {
990 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
991 head
->qlen
+= list
->qlen
;
992 __skb_queue_head_init(list
);
997 * __skb_queue_after - queue a buffer at the list head
999 * @prev: place after this buffer
1000 * @newsk: buffer to queue
1002 * Queue a buffer int the middle of a list. This function takes no locks
1003 * and you must therefore hold required locks before calling it.
1005 * A buffer cannot be placed on two lists at the same time.
1007 static inline void __skb_queue_after(struct sk_buff_head
*list
,
1008 struct sk_buff
*prev
,
1009 struct sk_buff
*newsk
)
1011 __skb_insert(newsk
, prev
, prev
->next
, list
);
1014 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
1015 struct sk_buff_head
*list
);
1017 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1018 struct sk_buff
*next
,
1019 struct sk_buff
*newsk
)
1021 __skb_insert(newsk
, next
->prev
, next
, list
);
1025 * __skb_queue_head - queue a buffer at the list head
1026 * @list: list to use
1027 * @newsk: buffer to queue
1029 * Queue a buffer at the start of a list. This function takes no locks
1030 * and you must therefore hold required locks before calling it.
1032 * A buffer cannot be placed on two lists at the same time.
1034 extern void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1035 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1036 struct sk_buff
*newsk
)
1038 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1042 * __skb_queue_tail - queue a buffer at the list tail
1043 * @list: list to use
1044 * @newsk: buffer to queue
1046 * Queue a buffer at the end of a list. This function takes no locks
1047 * and you must therefore hold required locks before calling it.
1049 * A buffer cannot be placed on two lists at the same time.
1051 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1052 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1053 struct sk_buff
*newsk
)
1055 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1059 * remove sk_buff from list. _Must_ be called atomically, and with
1062 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1063 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1065 struct sk_buff
*next
, *prev
;
1070 skb
->next
= skb
->prev
= NULL
;
1076 * __skb_dequeue - remove from the head of the queue
1077 * @list: list to dequeue from
1079 * Remove the head of the list. This function does not take any locks
1080 * so must be used with appropriate locks held only. The head item is
1081 * returned or %NULL if the list is empty.
1083 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1084 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1086 struct sk_buff
*skb
= skb_peek(list
);
1088 __skb_unlink(skb
, list
);
1093 * __skb_dequeue_tail - remove from the tail of the queue
1094 * @list: list to dequeue from
1096 * Remove the tail of the list. This function does not take any locks
1097 * so must be used with appropriate locks held only. The tail item is
1098 * returned or %NULL if the list is empty.
1100 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1101 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1103 struct sk_buff
*skb
= skb_peek_tail(list
);
1105 __skb_unlink(skb
, list
);
1110 static inline int skb_is_nonlinear(const struct sk_buff
*skb
)
1112 return skb
->data_len
;
1115 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1117 return skb
->len
- skb
->data_len
;
1120 static inline int skb_pagelen(const struct sk_buff
*skb
)
1124 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1125 len
+= skb_shinfo(skb
)->frags
[i
].size
;
1126 return len
+ skb_headlen(skb
);
1129 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1130 struct page
*page
, int off
, int size
)
1132 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1135 frag
->page_offset
= off
;
1137 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1140 extern void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
,
1143 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1144 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1145 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1147 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1148 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1150 return skb
->head
+ skb
->tail
;
1153 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1155 skb
->tail
= skb
->data
- skb
->head
;
1158 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1160 skb_reset_tail_pointer(skb
);
1161 skb
->tail
+= offset
;
1163 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1164 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1169 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1171 skb
->tail
= skb
->data
;
1174 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1176 skb
->tail
= skb
->data
+ offset
;
1179 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1182 * Add data to an sk_buff
1184 extern unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1185 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1187 unsigned char *tmp
= skb_tail_pointer(skb
);
1188 SKB_LINEAR_ASSERT(skb
);
1194 extern unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1195 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1202 extern unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1203 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1206 BUG_ON(skb
->len
< skb
->data_len
);
1207 return skb
->data
+= len
;
1210 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1212 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1215 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1217 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1219 if (len
> skb_headlen(skb
) &&
1220 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1223 return skb
->data
+= len
;
1226 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1228 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1231 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1233 if (likely(len
<= skb_headlen(skb
)))
1235 if (unlikely(len
> skb
->len
))
1237 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1241 * skb_headroom - bytes at buffer head
1242 * @skb: buffer to check
1244 * Return the number of bytes of free space at the head of an &sk_buff.
1246 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1248 return skb
->data
- skb
->head
;
1252 * skb_tailroom - bytes at buffer end
1253 * @skb: buffer to check
1255 * Return the number of bytes of free space at the tail of an sk_buff
1257 static inline int skb_tailroom(const struct sk_buff
*skb
)
1259 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1263 * skb_reserve - adjust headroom
1264 * @skb: buffer to alter
1265 * @len: bytes to move
1267 * Increase the headroom of an empty &sk_buff by reducing the tail
1268 * room. This is only allowed for an empty buffer.
1270 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
1276 static inline void skb_reset_mac_len(struct sk_buff
*skb
)
1278 skb
->mac_len
= skb
->network_header
- skb
->mac_header
;
1281 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1282 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1284 return skb
->head
+ skb
->transport_header
;
1287 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1289 skb
->transport_header
= skb
->data
- skb
->head
;
1292 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1295 skb_reset_transport_header(skb
);
1296 skb
->transport_header
+= offset
;
1299 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1301 return skb
->head
+ skb
->network_header
;
1304 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1306 skb
->network_header
= skb
->data
- skb
->head
;
1309 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1311 skb_reset_network_header(skb
);
1312 skb
->network_header
+= offset
;
1315 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1317 return skb
->head
+ skb
->mac_header
;
1320 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1322 return skb
->mac_header
!= ~0U;
1325 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1327 skb
->mac_header
= skb
->data
- skb
->head
;
1330 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1332 skb_reset_mac_header(skb
);
1333 skb
->mac_header
+= offset
;
1336 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1338 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1340 return skb
->transport_header
;
1343 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1345 skb
->transport_header
= skb
->data
;
1348 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1351 skb
->transport_header
= skb
->data
+ offset
;
1354 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1356 return skb
->network_header
;
1359 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1361 skb
->network_header
= skb
->data
;
1364 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1366 skb
->network_header
= skb
->data
+ offset
;
1369 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1371 return skb
->mac_header
;
1374 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1376 return skb
->mac_header
!= NULL
;
1379 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1381 skb
->mac_header
= skb
->data
;
1384 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1386 skb
->mac_header
= skb
->data
+ offset
;
1388 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1390 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
1392 return skb
->csum_start
- skb_headroom(skb
);
1395 static inline int skb_transport_offset(const struct sk_buff
*skb
)
1397 return skb_transport_header(skb
) - skb
->data
;
1400 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
1402 return skb
->transport_header
- skb
->network_header
;
1405 static inline int skb_network_offset(const struct sk_buff
*skb
)
1407 return skb_network_header(skb
) - skb
->data
;
1410 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
1412 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
1416 * CPUs often take a performance hit when accessing unaligned memory
1417 * locations. The actual performance hit varies, it can be small if the
1418 * hardware handles it or large if we have to take an exception and fix it
1421 * Since an ethernet header is 14 bytes network drivers often end up with
1422 * the IP header at an unaligned offset. The IP header can be aligned by
1423 * shifting the start of the packet by 2 bytes. Drivers should do this
1426 * skb_reserve(skb, NET_IP_ALIGN);
1428 * The downside to this alignment of the IP header is that the DMA is now
1429 * unaligned. On some architectures the cost of an unaligned DMA is high
1430 * and this cost outweighs the gains made by aligning the IP header.
1432 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1435 #ifndef NET_IP_ALIGN
1436 #define NET_IP_ALIGN 2
1440 * The networking layer reserves some headroom in skb data (via
1441 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1442 * the header has to grow. In the default case, if the header has to grow
1443 * 32 bytes or less we avoid the reallocation.
1445 * Unfortunately this headroom changes the DMA alignment of the resulting
1446 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1447 * on some architectures. An architecture can override this value,
1448 * perhaps setting it to a cacheline in size (since that will maintain
1449 * cacheline alignment of the DMA). It must be a power of 2.
1451 * Various parts of the networking layer expect at least 32 bytes of
1452 * headroom, you should not reduce this.
1454 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1455 * to reduce average number of cache lines per packet.
1456 * get_rps_cpus() for example only access one 64 bytes aligned block :
1457 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1460 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1463 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
1465 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
1467 if (unlikely(skb_is_nonlinear(skb
))) {
1472 skb_set_tail_pointer(skb
, len
);
1475 extern void skb_trim(struct sk_buff
*skb
, unsigned int len
);
1477 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1480 return ___pskb_trim(skb
, len
);
1481 __skb_trim(skb
, len
);
1485 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1487 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
1491 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1492 * @skb: buffer to alter
1495 * This is identical to pskb_trim except that the caller knows that
1496 * the skb is not cloned so we should never get an error due to out-
1499 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
1501 int err
= pskb_trim(skb
, len
);
1506 * skb_orphan - orphan a buffer
1507 * @skb: buffer to orphan
1509 * If a buffer currently has an owner then we call the owner's
1510 * destructor function and make the @skb unowned. The buffer continues
1511 * to exist but is no longer charged to its former owner.
1513 static inline void skb_orphan(struct sk_buff
*skb
)
1515 if (skb
->destructor
)
1516 skb
->destructor(skb
);
1517 skb
->destructor
= NULL
;
1522 * __skb_queue_purge - empty a list
1523 * @list: list to empty
1525 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1526 * the list and one reference dropped. This function does not take the
1527 * list lock and the caller must hold the relevant locks to use it.
1529 extern void skb_queue_purge(struct sk_buff_head
*list
);
1530 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
1532 struct sk_buff
*skb
;
1533 while ((skb
= __skb_dequeue(list
)) != NULL
)
1538 * __dev_alloc_skb - allocate an skbuff for receiving
1539 * @length: length to allocate
1540 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1542 * Allocate a new &sk_buff and assign it a usage count of one. The
1543 * buffer has unspecified headroom built in. Users should allocate
1544 * the headroom they think they need without accounting for the
1545 * built in space. The built in space is used for optimisations.
1547 * %NULL is returned if there is no free memory.
1549 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
1552 struct sk_buff
*skb
= alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
);
1554 skb_reserve(skb
, NET_SKB_PAD
);
1558 extern struct sk_buff
*dev_alloc_skb(unsigned int length
);
1560 extern struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
1561 unsigned int length
, gfp_t gfp_mask
);
1564 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1565 * @dev: network device to receive on
1566 * @length: length to allocate
1568 * Allocate a new &sk_buff and assign it a usage count of one. The
1569 * buffer has unspecified headroom built in. Users should allocate
1570 * the headroom they think they need without accounting for the
1571 * built in space. The built in space is used for optimisations.
1573 * %NULL is returned if there is no free memory. Although this function
1574 * allocates memory it can be called from an interrupt.
1576 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
1577 unsigned int length
)
1579 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
1582 static inline struct sk_buff
*__netdev_alloc_skb_ip_align(struct net_device
*dev
,
1583 unsigned int length
, gfp_t gfp
)
1585 struct sk_buff
*skb
= __netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
, gfp
);
1587 if (NET_IP_ALIGN
&& skb
)
1588 skb_reserve(skb
, NET_IP_ALIGN
);
1592 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
1593 unsigned int length
)
1595 return __netdev_alloc_skb_ip_align(dev
, length
, GFP_ATOMIC
);
1599 * __netdev_alloc_page - allocate a page for ps-rx on a specific device
1600 * @dev: network device to receive on
1601 * @gfp_mask: alloc_pages_node mask
1603 * Allocate a new page. dev currently unused.
1605 * %NULL is returned if there is no free memory.
1607 static inline struct page
*__netdev_alloc_page(struct net_device
*dev
, gfp_t gfp_mask
)
1609 return alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, 0);
1613 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1614 * @dev: network device to receive on
1616 * Allocate a new page. dev currently unused.
1618 * %NULL is returned if there is no free memory.
1620 static inline struct page
*netdev_alloc_page(struct net_device
*dev
)
1622 return __netdev_alloc_page(dev
, GFP_ATOMIC
);
1625 static inline void netdev_free_page(struct net_device
*dev
, struct page
*page
)
1631 * skb_clone_writable - is the header of a clone writable
1632 * @skb: buffer to check
1633 * @len: length up to which to write
1635 * Returns true if modifying the header part of the cloned buffer
1636 * does not requires the data to be copied.
1638 static inline int skb_clone_writable(struct sk_buff
*skb
, unsigned int len
)
1640 return !skb_header_cloned(skb
) &&
1641 skb_headroom(skb
) + len
<= skb
->hdr_len
;
1644 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
1649 if (headroom
< NET_SKB_PAD
)
1650 headroom
= NET_SKB_PAD
;
1651 if (headroom
> skb_headroom(skb
))
1652 delta
= headroom
- skb_headroom(skb
);
1654 if (delta
|| cloned
)
1655 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
1661 * skb_cow - copy header of skb when it is required
1662 * @skb: buffer to cow
1663 * @headroom: needed headroom
1665 * If the skb passed lacks sufficient headroom or its data part
1666 * is shared, data is reallocated. If reallocation fails, an error
1667 * is returned and original skb is not changed.
1669 * The result is skb with writable area skb->head...skb->tail
1670 * and at least @headroom of space at head.
1672 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
1674 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
1678 * skb_cow_head - skb_cow but only making the head writable
1679 * @skb: buffer to cow
1680 * @headroom: needed headroom
1682 * This function is identical to skb_cow except that we replace the
1683 * skb_cloned check by skb_header_cloned. It should be used when
1684 * you only need to push on some header and do not need to modify
1687 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
1689 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
1693 * skb_padto - pad an skbuff up to a minimal size
1694 * @skb: buffer to pad
1695 * @len: minimal length
1697 * Pads up a buffer to ensure the trailing bytes exist and are
1698 * blanked. If the buffer already contains sufficient data it
1699 * is untouched. Otherwise it is extended. Returns zero on
1700 * success. The skb is freed on error.
1703 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
1705 unsigned int size
= skb
->len
;
1706 if (likely(size
>= len
))
1708 return skb_pad(skb
, len
- size
);
1711 static inline int skb_add_data(struct sk_buff
*skb
,
1712 char __user
*from
, int copy
)
1714 const int off
= skb
->len
;
1716 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1718 __wsum csum
= csum_and_copy_from_user(from
, skb_put(skb
, copy
),
1721 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
1724 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
1727 __skb_trim(skb
, off
);
1731 static inline int skb_can_coalesce(struct sk_buff
*skb
, int i
,
1732 struct page
*page
, int off
)
1735 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
1737 return page
== frag
->page
&&
1738 off
== frag
->page_offset
+ frag
->size
;
1743 static inline int __skb_linearize(struct sk_buff
*skb
)
1745 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
1749 * skb_linearize - convert paged skb to linear one
1750 * @skb: buffer to linarize
1752 * If there is no free memory -ENOMEM is returned, otherwise zero
1753 * is returned and the old skb data released.
1755 static inline int skb_linearize(struct sk_buff
*skb
)
1757 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
1761 * skb_linearize_cow - make sure skb is linear and writable
1762 * @skb: buffer to process
1764 * If there is no free memory -ENOMEM is returned, otherwise zero
1765 * is returned and the old skb data released.
1767 static inline int skb_linearize_cow(struct sk_buff
*skb
)
1769 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
1770 __skb_linearize(skb
) : 0;
1774 * skb_postpull_rcsum - update checksum for received skb after pull
1775 * @skb: buffer to update
1776 * @start: start of data before pull
1777 * @len: length of data pulled
1779 * After doing a pull on a received packet, you need to call this to
1780 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1781 * CHECKSUM_NONE so that it can be recomputed from scratch.
1784 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
1785 const void *start
, unsigned int len
)
1787 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1788 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
1791 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
1794 * pskb_trim_rcsum - trim received skb and update checksum
1795 * @skb: buffer to trim
1798 * This is exactly the same as pskb_trim except that it ensures the
1799 * checksum of received packets are still valid after the operation.
1802 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
1804 if (likely(len
>= skb
->len
))
1806 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1807 skb
->ip_summed
= CHECKSUM_NONE
;
1808 return __pskb_trim(skb
, len
);
1811 #define skb_queue_walk(queue, skb) \
1812 for (skb = (queue)->next; \
1813 skb != (struct sk_buff *)(queue); \
1816 #define skb_queue_walk_safe(queue, skb, tmp) \
1817 for (skb = (queue)->next, tmp = skb->next; \
1818 skb != (struct sk_buff *)(queue); \
1819 skb = tmp, tmp = skb->next)
1821 #define skb_queue_walk_from(queue, skb) \
1822 for (; skb != (struct sk_buff *)(queue); \
1825 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1826 for (tmp = skb->next; \
1827 skb != (struct sk_buff *)(queue); \
1828 skb = tmp, tmp = skb->next)
1830 #define skb_queue_reverse_walk(queue, skb) \
1831 for (skb = (queue)->prev; \
1832 skb != (struct sk_buff *)(queue); \
1835 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
1836 for (skb = (queue)->prev, tmp = skb->prev; \
1837 skb != (struct sk_buff *)(queue); \
1838 skb = tmp, tmp = skb->prev)
1840 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
1841 for (tmp = skb->prev; \
1842 skb != (struct sk_buff *)(queue); \
1843 skb = tmp, tmp = skb->prev)
1845 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
1847 return skb_shinfo(skb
)->frag_list
!= NULL
;
1850 static inline void skb_frag_list_init(struct sk_buff
*skb
)
1852 skb_shinfo(skb
)->frag_list
= NULL
;
1855 static inline void skb_frag_add_head(struct sk_buff
*skb
, struct sk_buff
*frag
)
1857 frag
->next
= skb_shinfo(skb
)->frag_list
;
1858 skb_shinfo(skb
)->frag_list
= frag
;
1861 #define skb_walk_frags(skb, iter) \
1862 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1864 extern struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1865 int *peeked
, int *err
);
1866 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1867 int noblock
, int *err
);
1868 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
1869 struct poll_table_struct
*wait
);
1870 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
1871 int offset
, struct iovec
*to
,
1873 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
,
1876 extern int skb_copy_datagram_from_iovec(struct sk_buff
*skb
,
1878 const struct iovec
*from
,
1881 extern int skb_copy_datagram_const_iovec(const struct sk_buff
*from
,
1883 const struct iovec
*to
,
1886 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
1887 extern void skb_free_datagram_locked(struct sock
*sk
,
1888 struct sk_buff
*skb
);
1889 extern int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
,
1890 unsigned int flags
);
1891 extern __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
,
1892 int len
, __wsum csum
);
1893 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
1895 extern int skb_store_bits(struct sk_buff
*skb
, int offset
,
1896 const void *from
, int len
);
1897 extern __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
,
1898 int offset
, u8
*to
, int len
,
1900 extern int skb_splice_bits(struct sk_buff
*skb
,
1901 unsigned int offset
,
1902 struct pipe_inode_info
*pipe
,
1904 unsigned int flags
);
1905 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
1906 extern void skb_split(struct sk_buff
*skb
,
1907 struct sk_buff
*skb1
, const u32 len
);
1908 extern int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
,
1911 extern struct sk_buff
*skb_segment(struct sk_buff
*skb
, u32 features
);
1913 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
1914 int len
, void *buffer
)
1916 int hlen
= skb_headlen(skb
);
1918 if (hlen
- offset
>= len
)
1919 return skb
->data
+ offset
;
1921 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
1927 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
1929 const unsigned int len
)
1931 memcpy(to
, skb
->data
, len
);
1934 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
1935 const int offset
, void *to
,
1936 const unsigned int len
)
1938 memcpy(to
, skb
->data
+ offset
, len
);
1941 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
1943 const unsigned int len
)
1945 memcpy(skb
->data
, from
, len
);
1948 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
1951 const unsigned int len
)
1953 memcpy(skb
->data
+ offset
, from
, len
);
1956 extern void skb_init(void);
1958 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
1964 * skb_get_timestamp - get timestamp from a skb
1965 * @skb: skb to get stamp from
1966 * @stamp: pointer to struct timeval to store stamp in
1968 * Timestamps are stored in the skb as offsets to a base timestamp.
1969 * This function converts the offset back to a struct timeval and stores
1972 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
1973 struct timeval
*stamp
)
1975 *stamp
= ktime_to_timeval(skb
->tstamp
);
1978 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
1979 struct timespec
*stamp
)
1981 *stamp
= ktime_to_timespec(skb
->tstamp
);
1984 static inline void __net_timestamp(struct sk_buff
*skb
)
1986 skb
->tstamp
= ktime_get_real();
1989 static inline ktime_t
net_timedelta(ktime_t t
)
1991 return ktime_sub(ktime_get_real(), t
);
1994 static inline ktime_t
net_invalid_timestamp(void)
1996 return ktime_set(0, 0);
1999 extern void skb_timestamping_init(void);
2001 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2003 extern void skb_clone_tx_timestamp(struct sk_buff
*skb
);
2004 extern bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
2006 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2008 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
2012 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
2017 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2020 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2022 * @skb: clone of the the original outgoing packet
2023 * @hwtstamps: hardware time stamps
2026 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
2027 struct skb_shared_hwtstamps
*hwtstamps
);
2030 * skb_tstamp_tx - queue clone of skb with send time stamps
2031 * @orig_skb: the original outgoing packet
2032 * @hwtstamps: hardware time stamps, may be NULL if not available
2034 * If the skb has a socket associated, then this function clones the
2035 * skb (thus sharing the actual data and optional structures), stores
2036 * the optional hardware time stamping information (if non NULL) or
2037 * generates a software time stamp (otherwise), then queues the clone
2038 * to the error queue of the socket. Errors are silently ignored.
2040 extern void skb_tstamp_tx(struct sk_buff
*orig_skb
,
2041 struct skb_shared_hwtstamps
*hwtstamps
);
2043 static inline void sw_tx_timestamp(struct sk_buff
*skb
)
2045 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
&&
2046 !(skb_shinfo(skb
)->tx_flags
& SKBTX_IN_PROGRESS
))
2047 skb_tstamp_tx(skb
, NULL
);
2051 * skb_tx_timestamp() - Driver hook for transmit timestamping
2053 * Ethernet MAC Drivers should call this function in their hard_xmit()
2054 * function immediately before giving the sk_buff to the MAC hardware.
2056 * @skb: A socket buffer.
2058 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
2060 skb_clone_tx_timestamp(skb
);
2061 sw_tx_timestamp(skb
);
2064 extern __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
2065 extern __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
2067 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
2069 return skb
->ip_summed
& CHECKSUM_UNNECESSARY
;
2073 * skb_checksum_complete - Calculate checksum of an entire packet
2074 * @skb: packet to process
2076 * This function calculates the checksum over the entire packet plus
2077 * the value of skb->csum. The latter can be used to supply the
2078 * checksum of a pseudo header as used by TCP/UDP. It returns the
2081 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2082 * this function can be used to verify that checksum on received
2083 * packets. In that case the function should return zero if the
2084 * checksum is correct. In particular, this function will return zero
2085 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2086 * hardware has already verified the correctness of the checksum.
2088 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
2090 return skb_csum_unnecessary(skb
) ?
2091 0 : __skb_checksum_complete(skb
);
2094 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2095 extern void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
2096 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
2098 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
2099 nf_conntrack_destroy(nfct
);
2101 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
2104 atomic_inc(&nfct
->use
);
2107 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2108 static inline void nf_conntrack_get_reasm(struct sk_buff
*skb
)
2111 atomic_inc(&skb
->users
);
2113 static inline void nf_conntrack_put_reasm(struct sk_buff
*skb
)
2119 #ifdef CONFIG_BRIDGE_NETFILTER
2120 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
2122 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
2125 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
2128 atomic_inc(&nf_bridge
->use
);
2130 #endif /* CONFIG_BRIDGE_NETFILTER */
2131 static inline void nf_reset(struct sk_buff
*skb
)
2133 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2134 nf_conntrack_put(skb
->nfct
);
2137 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2138 nf_conntrack_put_reasm(skb
->nfct_reasm
);
2139 skb
->nfct_reasm
= NULL
;
2141 #ifdef CONFIG_BRIDGE_NETFILTER
2142 nf_bridge_put(skb
->nf_bridge
);
2143 skb
->nf_bridge
= NULL
;
2147 /* Note: This doesn't put any conntrack and bridge info in dst. */
2148 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2150 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2151 dst
->nfct
= src
->nfct
;
2152 nf_conntrack_get(src
->nfct
);
2153 dst
->nfctinfo
= src
->nfctinfo
;
2155 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2156 dst
->nfct_reasm
= src
->nfct_reasm
;
2157 nf_conntrack_get_reasm(src
->nfct_reasm
);
2159 #ifdef CONFIG_BRIDGE_NETFILTER
2160 dst
->nf_bridge
= src
->nf_bridge
;
2161 nf_bridge_get(src
->nf_bridge
);
2165 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2167 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2168 nf_conntrack_put(dst
->nfct
);
2170 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2171 nf_conntrack_put_reasm(dst
->nfct_reasm
);
2173 #ifdef CONFIG_BRIDGE_NETFILTER
2174 nf_bridge_put(dst
->nf_bridge
);
2176 __nf_copy(dst
, src
);
2179 #ifdef CONFIG_NETWORK_SECMARK
2180 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2182 to
->secmark
= from
->secmark
;
2185 static inline void skb_init_secmark(struct sk_buff
*skb
)
2190 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2193 static inline void skb_init_secmark(struct sk_buff
*skb
)
2197 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
2199 skb
->queue_mapping
= queue_mapping
;
2202 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
2204 return skb
->queue_mapping
;
2207 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
2209 to
->queue_mapping
= from
->queue_mapping
;
2212 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
2214 skb
->queue_mapping
= rx_queue
+ 1;
2217 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
2219 return skb
->queue_mapping
- 1;
2222 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
2224 return skb
->queue_mapping
!= 0;
2227 extern u16
__skb_tx_hash(const struct net_device
*dev
,
2228 const struct sk_buff
*skb
,
2229 unsigned int num_tx_queues
);
2232 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2237 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2243 static inline int skb_is_gso(const struct sk_buff
*skb
)
2245 return skb_shinfo(skb
)->gso_size
;
2248 static inline int skb_is_gso_v6(const struct sk_buff
*skb
)
2250 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
2253 extern void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
2255 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
2257 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2258 * wanted then gso_type will be set. */
2259 struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
2260 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
2261 unlikely(shinfo
->gso_type
== 0)) {
2262 __skb_warn_lro_forwarding(skb
);
2268 static inline void skb_forward_csum(struct sk_buff
*skb
)
2270 /* Unfortunately we don't support this one. Any brave souls? */
2271 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2272 skb
->ip_summed
= CHECKSUM_NONE
;
2276 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2277 * @skb: skb to check
2279 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2280 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2281 * use this helper, to document places where we make this assertion.
2283 static inline void skb_checksum_none_assert(struct sk_buff
*skb
)
2286 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
2290 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
2292 #endif /* __KERNEL__ */
2293 #endif /* _LINUX_SKBUFF_H */