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. 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,
192 /* This data is invariant across clones and lives at
193 * the end of the header data, ie. at skb->end.
195 struct skb_shared_info
{
196 unsigned short nr_frags
;
197 unsigned short gso_size
;
198 /* Warning: this field is not always filled in (UFO)! */
199 unsigned short gso_segs
;
200 unsigned short gso_type
;
203 struct sk_buff
*frag_list
;
204 struct skb_shared_hwtstamps hwtstamps
;
207 * Warning : all fields before dataref are cleared in __alloc_skb()
211 /* Intermediate layers must ensure that destructor_arg
212 * remains valid until skb destructor */
213 void * destructor_arg
;
214 /* must be last field, see pskb_expand_head() */
215 skb_frag_t frags
[MAX_SKB_FRAGS
];
218 /* We divide dataref into two halves. The higher 16 bits hold references
219 * to the payload part of skb->data. The lower 16 bits hold references to
220 * the entire skb->data. A clone of a headerless skb holds the length of
221 * the header in skb->hdr_len.
223 * All users must obey the rule that the skb->data reference count must be
224 * greater than or equal to the payload reference count.
226 * Holding a reference to the payload part means that the user does not
227 * care about modifications to the header part of skb->data.
229 #define SKB_DATAREF_SHIFT 16
230 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
234 SKB_FCLONE_UNAVAILABLE
,
240 SKB_GSO_TCPV4
= 1 << 0,
241 SKB_GSO_UDP
= 1 << 1,
243 /* This indicates the skb is from an untrusted source. */
244 SKB_GSO_DODGY
= 1 << 2,
246 /* This indicates the tcp segment has CWR set. */
247 SKB_GSO_TCP_ECN
= 1 << 3,
249 SKB_GSO_TCPV6
= 1 << 4,
251 SKB_GSO_FCOE
= 1 << 5,
254 #if BITS_PER_LONG > 32
255 #define NET_SKBUFF_DATA_USES_OFFSET 1
258 #ifdef NET_SKBUFF_DATA_USES_OFFSET
259 typedef unsigned int sk_buff_data_t
;
261 typedef unsigned char *sk_buff_data_t
;
264 #if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
265 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
266 #define NET_SKBUFF_NF_DEFRAG_NEEDED 1
270 * struct sk_buff - socket buffer
271 * @next: Next buffer in list
272 * @prev: Previous buffer in list
273 * @sk: Socket we are owned by
274 * @tstamp: Time we arrived
275 * @dev: Device we arrived on/are leaving by
276 * @transport_header: Transport layer header
277 * @network_header: Network layer header
278 * @mac_header: Link layer header
279 * @_skb_refdst: destination entry (with norefcount bit)
280 * @sp: the security path, used for xfrm
281 * @cb: Control buffer. Free for use by every layer. Put private vars here
282 * @len: Length of actual data
283 * @data_len: Data length
284 * @mac_len: Length of link layer header
285 * @hdr_len: writable header length of cloned skb
286 * @csum: Checksum (must include start/offset pair)
287 * @csum_start: Offset from skb->head where checksumming should start
288 * @csum_offset: Offset from csum_start where checksum should be stored
289 * @local_df: allow local fragmentation
290 * @cloned: Head may be cloned (check refcnt to be sure)
291 * @nohdr: Payload reference only, must not modify header
292 * @pkt_type: Packet class
293 * @fclone: skbuff clone status
294 * @ip_summed: Driver fed us an IP checksum
295 * @priority: Packet queueing priority
296 * @users: User count - see {datagram,tcp}.c
297 * @protocol: Packet protocol from driver
298 * @truesize: Buffer size
299 * @head: Head of buffer
300 * @data: Data head pointer
301 * @tail: Tail pointer
303 * @destructor: Destruct function
304 * @mark: Generic packet mark
305 * @nfct: Associated connection, if any
306 * @ipvs_property: skbuff is owned by ipvs
307 * @peeked: this packet has been seen already, so stats have been
308 * done for it, don't do them again
309 * @nf_trace: netfilter packet trace flag
310 * @nfctinfo: Relationship of this skb to the connection
311 * @nfct_reasm: netfilter conntrack re-assembly pointer
312 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
313 * @skb_iif: ifindex of device we arrived on
314 * @rxhash: the packet hash computed on receive
315 * @queue_mapping: Queue mapping for multiqueue devices
316 * @tc_index: Traffic control index
317 * @tc_verd: traffic control verdict
318 * @ndisc_nodetype: router type (from link layer)
319 * @dma_cookie: a cookie to one of several possible DMA operations
320 * done by skb DMA functions
321 * @secmark: security marking
322 * @vlan_tci: vlan tag control information
326 /* These two members must be first. */
327 struct sk_buff
*next
;
328 struct sk_buff
*prev
;
333 struct net_device
*dev
;
336 * This is the control buffer. It is free to use for every
337 * layer. Please put your private variables there. If you
338 * want to keep them across layers you have to do a skb_clone()
339 * first. This is owned by whoever has the skb queued ATM.
341 char cb
[48] __aligned(8);
343 unsigned long _skb_refdst
;
359 kmemcheck_bitfield_begin(flags1
);
370 kmemcheck_bitfield_end(flags1
);
373 void (*destructor
)(struct sk_buff
*skb
);
374 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
375 struct nf_conntrack
*nfct
;
377 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
378 struct sk_buff
*nfct_reasm
;
380 #ifdef CONFIG_BRIDGE_NETFILTER
381 struct nf_bridge_info
*nf_bridge
;
385 #ifdef CONFIG_NET_SCHED
386 __u16 tc_index
; /* traffic control index */
387 #ifdef CONFIG_NET_CLS_ACT
388 __u16 tc_verd
; /* traffic control verdict */
395 kmemcheck_bitfield_begin(flags2
);
396 #ifdef CONFIG_IPV6_NDISC_NODETYPE
397 __u8 ndisc_nodetype
:2;
400 kmemcheck_bitfield_end(flags2
);
404 #ifdef CONFIG_NET_DMA
405 dma_cookie_t dma_cookie
;
407 #ifdef CONFIG_NETWORK_SECMARK
417 sk_buff_data_t transport_header
;
418 sk_buff_data_t network_header
;
419 sk_buff_data_t mac_header
;
420 /* These elements must be at the end, see alloc_skb() for details. */
425 unsigned int truesize
;
431 * Handling routines are only of interest to the kernel
433 #include <linux/slab.h>
435 #include <asm/system.h>
438 * skb might have a dst pointer attached, refcounted or not.
439 * _skb_refdst low order bit is set if refcount was _not_ taken
441 #define SKB_DST_NOREF 1UL
442 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
445 * skb_dst - returns skb dst_entry
448 * Returns skb dst_entry, regardless of reference taken or not.
450 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
452 /* If refdst was not refcounted, check we still are in a
453 * rcu_read_lock section
455 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
456 !rcu_read_lock_held() &&
457 !rcu_read_lock_bh_held());
458 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
462 * skb_dst_set - sets skb dst
466 * Sets skb dst, assuming a reference was taken on dst and should
467 * be released by skb_dst_drop()
469 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
471 skb
->_skb_refdst
= (unsigned long)dst
;
474 extern void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
);
477 * skb_dst_is_noref - Test if skb dst isn't refcounted
480 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
482 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
485 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
487 return (struct rtable
*)skb_dst(skb
);
490 extern void kfree_skb(struct sk_buff
*skb
);
491 extern void consume_skb(struct sk_buff
*skb
);
492 extern void __kfree_skb(struct sk_buff
*skb
);
493 extern struct sk_buff
*__alloc_skb(unsigned int size
,
494 gfp_t priority
, int fclone
, int node
);
495 static inline struct sk_buff
*alloc_skb(unsigned int size
,
498 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
501 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
504 return __alloc_skb(size
, priority
, 1, NUMA_NO_NODE
);
507 extern bool skb_recycle_check(struct sk_buff
*skb
, int skb_size
);
509 extern struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
510 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
512 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
514 extern struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
516 extern int pskb_expand_head(struct sk_buff
*skb
,
517 int nhead
, int ntail
,
519 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
520 unsigned int headroom
);
521 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
522 int newheadroom
, int newtailroom
,
524 extern int skb_to_sgvec(struct sk_buff
*skb
,
525 struct scatterlist
*sg
, int offset
,
527 extern int skb_cow_data(struct sk_buff
*skb
, int tailbits
,
528 struct sk_buff
**trailer
);
529 extern int skb_pad(struct sk_buff
*skb
, int pad
);
530 #define dev_kfree_skb(a) consume_skb(a)
532 extern int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
533 int getfrag(void *from
, char *to
, int offset
,
534 int len
,int odd
, struct sk_buff
*skb
),
535 void *from
, int length
);
537 struct skb_seq_state
{
541 __u32 stepped_offset
;
542 struct sk_buff
*root_skb
;
543 struct sk_buff
*cur_skb
;
547 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
548 unsigned int from
, unsigned int to
,
549 struct skb_seq_state
*st
);
550 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
551 struct skb_seq_state
*st
);
552 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
554 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
555 unsigned int to
, struct ts_config
*config
,
556 struct ts_state
*state
);
558 extern __u32
__skb_get_rxhash(struct sk_buff
*skb
);
559 static inline __u32
skb_get_rxhash(struct sk_buff
*skb
)
562 skb
->rxhash
= __skb_get_rxhash(skb
);
567 #ifdef NET_SKBUFF_DATA_USES_OFFSET
568 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
570 return skb
->head
+ skb
->end
;
573 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
580 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
582 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
584 return &skb_shinfo(skb
)->hwtstamps
;
588 * skb_queue_empty - check if a queue is empty
591 * Returns true if the queue is empty, false otherwise.
593 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
595 return list
->next
== (struct sk_buff
*)list
;
599 * skb_queue_is_last - check if skb is the last entry in the queue
603 * Returns true if @skb is the last buffer on the list.
605 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
606 const struct sk_buff
*skb
)
608 return skb
->next
== (struct sk_buff
*)list
;
612 * skb_queue_is_first - check if skb is the first entry in the queue
616 * Returns true if @skb is the first buffer on the list.
618 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
619 const struct sk_buff
*skb
)
621 return skb
->prev
== (struct sk_buff
*)list
;
625 * skb_queue_next - return the next packet in the queue
627 * @skb: current buffer
629 * Return the next packet in @list after @skb. It is only valid to
630 * call this if skb_queue_is_last() evaluates to false.
632 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
633 const struct sk_buff
*skb
)
635 /* This BUG_ON may seem severe, but if we just return then we
636 * are going to dereference garbage.
638 BUG_ON(skb_queue_is_last(list
, skb
));
643 * skb_queue_prev - return the prev packet in the queue
645 * @skb: current buffer
647 * Return the prev packet in @list before @skb. It is only valid to
648 * call this if skb_queue_is_first() evaluates to false.
650 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
651 const struct sk_buff
*skb
)
653 /* This BUG_ON may seem severe, but if we just return then we
654 * are going to dereference garbage.
656 BUG_ON(skb_queue_is_first(list
, skb
));
661 * skb_get - reference buffer
662 * @skb: buffer to reference
664 * Makes another reference to a socket buffer and returns a pointer
667 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
669 atomic_inc(&skb
->users
);
674 * If users == 1, we are the only owner and are can avoid redundant
679 * skb_cloned - is the buffer a clone
680 * @skb: buffer to check
682 * Returns true if the buffer was generated with skb_clone() and is
683 * one of multiple shared copies of the buffer. Cloned buffers are
684 * shared data so must not be written to under normal circumstances.
686 static inline int skb_cloned(const struct sk_buff
*skb
)
688 return skb
->cloned
&&
689 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
693 * skb_header_cloned - is the header a clone
694 * @skb: buffer to check
696 * Returns true if modifying the header part of the buffer requires
697 * the data to be copied.
699 static inline int skb_header_cloned(const struct sk_buff
*skb
)
706 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
707 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
712 * skb_header_release - release reference to header
713 * @skb: buffer to operate on
715 * Drop a reference to the header part of the buffer. This is done
716 * by acquiring a payload reference. You must not read from the header
717 * part of skb->data after this.
719 static inline void skb_header_release(struct sk_buff
*skb
)
723 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
727 * skb_shared - is the buffer shared
728 * @skb: buffer to check
730 * Returns true if more than one person has a reference to this
733 static inline int skb_shared(const struct sk_buff
*skb
)
735 return atomic_read(&skb
->users
) != 1;
739 * skb_share_check - check if buffer is shared and if so clone it
740 * @skb: buffer to check
741 * @pri: priority for memory allocation
743 * If the buffer is shared the buffer is cloned and the old copy
744 * drops a reference. A new clone with a single reference is returned.
745 * If the buffer is not shared the original buffer is returned. When
746 * being called from interrupt status or with spinlocks held pri must
749 * NULL is returned on a memory allocation failure.
751 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
,
754 might_sleep_if(pri
& __GFP_WAIT
);
755 if (skb_shared(skb
)) {
756 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
764 * Copy shared buffers into a new sk_buff. We effectively do COW on
765 * packets to handle cases where we have a local reader and forward
766 * and a couple of other messy ones. The normal one is tcpdumping
767 * a packet thats being forwarded.
771 * skb_unshare - make a copy of a shared buffer
772 * @skb: buffer to check
773 * @pri: priority for memory allocation
775 * If the socket buffer is a clone then this function creates a new
776 * copy of the data, drops a reference count on the old copy and returns
777 * the new copy with the reference count at 1. If the buffer is not a clone
778 * the original buffer is returned. When called with a spinlock held or
779 * from interrupt state @pri must be %GFP_ATOMIC
781 * %NULL is returned on a memory allocation failure.
783 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
786 might_sleep_if(pri
& __GFP_WAIT
);
787 if (skb_cloned(skb
)) {
788 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
789 kfree_skb(skb
); /* Free our shared copy */
796 * skb_peek - peek at the head of an &sk_buff_head
797 * @list_: list to peek at
799 * Peek an &sk_buff. Unlike most other operations you _MUST_
800 * be careful with this one. A peek leaves the buffer on the
801 * list and someone else may run off with it. You must hold
802 * the appropriate locks or have a private queue to do this.
804 * Returns %NULL for an empty list or a pointer to the head element.
805 * The reference count is not incremented and the reference is therefore
806 * volatile. Use with caution.
808 static inline struct sk_buff
*skb_peek(struct sk_buff_head
*list_
)
810 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->next
;
811 if (list
== (struct sk_buff
*)list_
)
817 * skb_peek_tail - peek at the tail of an &sk_buff_head
818 * @list_: list to peek at
820 * Peek an &sk_buff. Unlike most other operations you _MUST_
821 * be careful with this one. A peek leaves the buffer on the
822 * list and someone else may run off with it. You must hold
823 * the appropriate locks or have a private queue to do this.
825 * Returns %NULL for an empty list or a pointer to the tail element.
826 * The reference count is not incremented and the reference is therefore
827 * volatile. Use with caution.
829 static inline struct sk_buff
*skb_peek_tail(struct sk_buff_head
*list_
)
831 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->prev
;
832 if (list
== (struct sk_buff
*)list_
)
838 * skb_queue_len - get queue length
839 * @list_: list to measure
841 * Return the length of an &sk_buff queue.
843 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
849 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
850 * @list: queue to initialize
852 * This initializes only the list and queue length aspects of
853 * an sk_buff_head object. This allows to initialize the list
854 * aspects of an sk_buff_head without reinitializing things like
855 * the spinlock. It can also be used for on-stack sk_buff_head
856 * objects where the spinlock is known to not be used.
858 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
860 list
->prev
= list
->next
= (struct sk_buff
*)list
;
865 * This function creates a split out lock class for each invocation;
866 * this is needed for now since a whole lot of users of the skb-queue
867 * infrastructure in drivers have different locking usage (in hardirq)
868 * than the networking core (in softirq only). In the long run either the
869 * network layer or drivers should need annotation to consolidate the
870 * main types of usage into 3 classes.
872 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
874 spin_lock_init(&list
->lock
);
875 __skb_queue_head_init(list
);
878 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
879 struct lock_class_key
*class)
881 skb_queue_head_init(list
);
882 lockdep_set_class(&list
->lock
, class);
886 * Insert an sk_buff on a list.
888 * The "__skb_xxxx()" functions are the non-atomic ones that
889 * can only be called with interrupts disabled.
891 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
892 static inline void __skb_insert(struct sk_buff
*newsk
,
893 struct sk_buff
*prev
, struct sk_buff
*next
,
894 struct sk_buff_head
*list
)
898 next
->prev
= prev
->next
= newsk
;
902 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
903 struct sk_buff
*prev
,
904 struct sk_buff
*next
)
906 struct sk_buff
*first
= list
->next
;
907 struct sk_buff
*last
= list
->prev
;
917 * skb_queue_splice - join two skb lists, this is designed for stacks
918 * @list: the new list to add
919 * @head: the place to add it in the first list
921 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
922 struct sk_buff_head
*head
)
924 if (!skb_queue_empty(list
)) {
925 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
926 head
->qlen
+= list
->qlen
;
931 * skb_queue_splice - join two skb lists and reinitialise the emptied list
932 * @list: the new list to add
933 * @head: the place to add it in the first list
935 * The list at @list is reinitialised
937 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
938 struct sk_buff_head
*head
)
940 if (!skb_queue_empty(list
)) {
941 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
942 head
->qlen
+= list
->qlen
;
943 __skb_queue_head_init(list
);
948 * skb_queue_splice_tail - join two skb lists, each list being a queue
949 * @list: the new list to add
950 * @head: the place to add it in the first list
952 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
953 struct sk_buff_head
*head
)
955 if (!skb_queue_empty(list
)) {
956 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
957 head
->qlen
+= list
->qlen
;
962 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
963 * @list: the new list to add
964 * @head: the place to add it in the first list
966 * Each of the lists is a queue.
967 * The list at @list is reinitialised
969 static inline void skb_queue_splice_tail_init(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
;
975 __skb_queue_head_init(list
);
980 * __skb_queue_after - queue a buffer at the list head
982 * @prev: place after this buffer
983 * @newsk: buffer to queue
985 * Queue a buffer int the middle of a list. This function takes no locks
986 * and you must therefore hold required locks before calling it.
988 * A buffer cannot be placed on two lists at the same time.
990 static inline void __skb_queue_after(struct sk_buff_head
*list
,
991 struct sk_buff
*prev
,
992 struct sk_buff
*newsk
)
994 __skb_insert(newsk
, prev
, prev
->next
, list
);
997 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
998 struct sk_buff_head
*list
);
1000 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1001 struct sk_buff
*next
,
1002 struct sk_buff
*newsk
)
1004 __skb_insert(newsk
, next
->prev
, next
, list
);
1008 * __skb_queue_head - queue a buffer at the list head
1009 * @list: list to use
1010 * @newsk: buffer to queue
1012 * Queue a buffer at the start of a list. This function takes no locks
1013 * and you must therefore hold required locks before calling it.
1015 * A buffer cannot be placed on two lists at the same time.
1017 extern void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1018 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1019 struct sk_buff
*newsk
)
1021 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1025 * __skb_queue_tail - queue a buffer at the list tail
1026 * @list: list to use
1027 * @newsk: buffer to queue
1029 * Queue a buffer at the end 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_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1035 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1036 struct sk_buff
*newsk
)
1038 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1042 * remove sk_buff from list. _Must_ be called atomically, and with
1045 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1046 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1048 struct sk_buff
*next
, *prev
;
1053 skb
->next
= skb
->prev
= NULL
;
1059 * __skb_dequeue - remove from the head of the queue
1060 * @list: list to dequeue from
1062 * Remove the head of the list. This function does not take any locks
1063 * so must be used with appropriate locks held only. The head item is
1064 * returned or %NULL if the list is empty.
1066 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1067 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1069 struct sk_buff
*skb
= skb_peek(list
);
1071 __skb_unlink(skb
, list
);
1076 * __skb_dequeue_tail - remove from the tail of the queue
1077 * @list: list to dequeue from
1079 * Remove the tail of the list. This function does not take any locks
1080 * so must be used with appropriate locks held only. The tail item is
1081 * returned or %NULL if the list is empty.
1083 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1084 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1086 struct sk_buff
*skb
= skb_peek_tail(list
);
1088 __skb_unlink(skb
, list
);
1093 static inline int skb_is_nonlinear(const struct sk_buff
*skb
)
1095 return skb
->data_len
;
1098 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1100 return skb
->len
- skb
->data_len
;
1103 static inline int skb_pagelen(const struct sk_buff
*skb
)
1107 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1108 len
+= skb_shinfo(skb
)->frags
[i
].size
;
1109 return len
+ skb_headlen(skb
);
1112 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1113 struct page
*page
, int off
, int size
)
1115 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1118 frag
->page_offset
= off
;
1120 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1123 extern void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
,
1126 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1127 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1128 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1130 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1131 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1133 return skb
->head
+ skb
->tail
;
1136 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1138 skb
->tail
= skb
->data
- skb
->head
;
1141 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1143 skb_reset_tail_pointer(skb
);
1144 skb
->tail
+= offset
;
1146 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1147 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1152 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1154 skb
->tail
= skb
->data
;
1157 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1159 skb
->tail
= skb
->data
+ offset
;
1162 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1165 * Add data to an sk_buff
1167 extern unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1168 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1170 unsigned char *tmp
= skb_tail_pointer(skb
);
1171 SKB_LINEAR_ASSERT(skb
);
1177 extern unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1178 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1185 extern unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1186 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1189 BUG_ON(skb
->len
< skb
->data_len
);
1190 return skb
->data
+= len
;
1193 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1195 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1198 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1200 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1202 if (len
> skb_headlen(skb
) &&
1203 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1206 return skb
->data
+= len
;
1209 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1211 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1214 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1216 if (likely(len
<= skb_headlen(skb
)))
1218 if (unlikely(len
> skb
->len
))
1220 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1224 * skb_headroom - bytes at buffer head
1225 * @skb: buffer to check
1227 * Return the number of bytes of free space at the head of an &sk_buff.
1229 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1231 return skb
->data
- skb
->head
;
1235 * skb_tailroom - bytes at buffer end
1236 * @skb: buffer to check
1238 * Return the number of bytes of free space at the tail of an sk_buff
1240 static inline int skb_tailroom(const struct sk_buff
*skb
)
1242 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1246 * skb_reserve - adjust headroom
1247 * @skb: buffer to alter
1248 * @len: bytes to move
1250 * Increase the headroom of an empty &sk_buff by reducing the tail
1251 * room. This is only allowed for an empty buffer.
1253 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
1259 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1260 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1262 return skb
->head
+ skb
->transport_header
;
1265 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1267 skb
->transport_header
= skb
->data
- skb
->head
;
1270 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1273 skb_reset_transport_header(skb
);
1274 skb
->transport_header
+= offset
;
1277 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1279 return skb
->head
+ skb
->network_header
;
1282 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1284 skb
->network_header
= skb
->data
- skb
->head
;
1287 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1289 skb_reset_network_header(skb
);
1290 skb
->network_header
+= offset
;
1293 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1295 return skb
->head
+ skb
->mac_header
;
1298 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1300 return skb
->mac_header
!= ~0U;
1303 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1305 skb
->mac_header
= skb
->data
- skb
->head
;
1308 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1310 skb_reset_mac_header(skb
);
1311 skb
->mac_header
+= offset
;
1314 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1316 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1318 return skb
->transport_header
;
1321 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1323 skb
->transport_header
= skb
->data
;
1326 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1329 skb
->transport_header
= skb
->data
+ offset
;
1332 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1334 return skb
->network_header
;
1337 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1339 skb
->network_header
= skb
->data
;
1342 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1344 skb
->network_header
= skb
->data
+ offset
;
1347 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1349 return skb
->mac_header
;
1352 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1354 return skb
->mac_header
!= NULL
;
1357 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1359 skb
->mac_header
= skb
->data
;
1362 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1364 skb
->mac_header
= skb
->data
+ offset
;
1366 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1368 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
1370 return skb
->csum_start
- skb_headroom(skb
);
1373 static inline int skb_transport_offset(const struct sk_buff
*skb
)
1375 return skb_transport_header(skb
) - skb
->data
;
1378 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
1380 return skb
->transport_header
- skb
->network_header
;
1383 static inline int skb_network_offset(const struct sk_buff
*skb
)
1385 return skb_network_header(skb
) - skb
->data
;
1388 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
1390 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
1394 * CPUs often take a performance hit when accessing unaligned memory
1395 * locations. The actual performance hit varies, it can be small if the
1396 * hardware handles it or large if we have to take an exception and fix it
1399 * Since an ethernet header is 14 bytes network drivers often end up with
1400 * the IP header at an unaligned offset. The IP header can be aligned by
1401 * shifting the start of the packet by 2 bytes. Drivers should do this
1404 * skb_reserve(skb, NET_IP_ALIGN);
1406 * The downside to this alignment of the IP header is that the DMA is now
1407 * unaligned. On some architectures the cost of an unaligned DMA is high
1408 * and this cost outweighs the gains made by aligning the IP header.
1410 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1413 #ifndef NET_IP_ALIGN
1414 #define NET_IP_ALIGN 2
1418 * The networking layer reserves some headroom in skb data (via
1419 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1420 * the header has to grow. In the default case, if the header has to grow
1421 * 32 bytes or less we avoid the reallocation.
1423 * Unfortunately this headroom changes the DMA alignment of the resulting
1424 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1425 * on some architectures. An architecture can override this value,
1426 * perhaps setting it to a cacheline in size (since that will maintain
1427 * cacheline alignment of the DMA). It must be a power of 2.
1429 * Various parts of the networking layer expect at least 32 bytes of
1430 * headroom, you should not reduce this.
1432 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1433 * to reduce average number of cache lines per packet.
1434 * get_rps_cpus() for example only access one 64 bytes aligned block :
1435 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1438 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1441 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
1443 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
1445 if (unlikely(skb_is_nonlinear(skb
))) {
1450 skb_set_tail_pointer(skb
, len
);
1453 extern void skb_trim(struct sk_buff
*skb
, unsigned int len
);
1455 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1458 return ___pskb_trim(skb
, len
);
1459 __skb_trim(skb
, len
);
1463 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1465 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
1469 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1470 * @skb: buffer to alter
1473 * This is identical to pskb_trim except that the caller knows that
1474 * the skb is not cloned so we should never get an error due to out-
1477 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
1479 int err
= pskb_trim(skb
, len
);
1484 * skb_orphan - orphan a buffer
1485 * @skb: buffer to orphan
1487 * If a buffer currently has an owner then we call the owner's
1488 * destructor function and make the @skb unowned. The buffer continues
1489 * to exist but is no longer charged to its former owner.
1491 static inline void skb_orphan(struct sk_buff
*skb
)
1493 if (skb
->destructor
)
1494 skb
->destructor(skb
);
1495 skb
->destructor
= NULL
;
1500 * __skb_queue_purge - empty a list
1501 * @list: list to empty
1503 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1504 * the list and one reference dropped. This function does not take the
1505 * list lock and the caller must hold the relevant locks to use it.
1507 extern void skb_queue_purge(struct sk_buff_head
*list
);
1508 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
1510 struct sk_buff
*skb
;
1511 while ((skb
= __skb_dequeue(list
)) != NULL
)
1516 * __dev_alloc_skb - allocate an skbuff for receiving
1517 * @length: length to allocate
1518 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1520 * Allocate a new &sk_buff and assign it a usage count of one. The
1521 * buffer has unspecified headroom built in. Users should allocate
1522 * the headroom they think they need without accounting for the
1523 * built in space. The built in space is used for optimisations.
1525 * %NULL is returned if there is no free memory.
1527 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
1530 struct sk_buff
*skb
= alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
);
1532 skb_reserve(skb
, NET_SKB_PAD
);
1536 extern struct sk_buff
*dev_alloc_skb(unsigned int length
);
1538 extern struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
1539 unsigned int length
, gfp_t gfp_mask
);
1542 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1543 * @dev: network device to receive on
1544 * @length: length to allocate
1546 * Allocate a new &sk_buff and assign it a usage count of one. The
1547 * buffer has unspecified headroom built in. Users should allocate
1548 * the headroom they think they need without accounting for the
1549 * built in space. The built in space is used for optimisations.
1551 * %NULL is returned if there is no free memory. Although this function
1552 * allocates memory it can be called from an interrupt.
1554 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
1555 unsigned int length
)
1557 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
1560 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
1561 unsigned int length
)
1563 struct sk_buff
*skb
= netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
);
1565 if (NET_IP_ALIGN
&& skb
)
1566 skb_reserve(skb
, NET_IP_ALIGN
);
1571 * __netdev_alloc_page - allocate a page for ps-rx on a specific device
1572 * @dev: network device to receive on
1573 * @gfp_mask: alloc_pages_node mask
1575 * Allocate a new page. dev currently unused.
1577 * %NULL is returned if there is no free memory.
1579 static inline struct page
*__netdev_alloc_page(struct net_device
*dev
, gfp_t gfp_mask
)
1581 return alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, 0);
1585 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1586 * @dev: network device to receive on
1588 * Allocate a new page. dev currently unused.
1590 * %NULL is returned if there is no free memory.
1592 static inline struct page
*netdev_alloc_page(struct net_device
*dev
)
1594 return __netdev_alloc_page(dev
, GFP_ATOMIC
);
1597 static inline void netdev_free_page(struct net_device
*dev
, struct page
*page
)
1603 * skb_clone_writable - is the header of a clone writable
1604 * @skb: buffer to check
1605 * @len: length up to which to write
1607 * Returns true if modifying the header part of the cloned buffer
1608 * does not requires the data to be copied.
1610 static inline int skb_clone_writable(struct sk_buff
*skb
, unsigned int len
)
1612 return !skb_header_cloned(skb
) &&
1613 skb_headroom(skb
) + len
<= skb
->hdr_len
;
1616 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
1621 if (headroom
< NET_SKB_PAD
)
1622 headroom
= NET_SKB_PAD
;
1623 if (headroom
> skb_headroom(skb
))
1624 delta
= headroom
- skb_headroom(skb
);
1626 if (delta
|| cloned
)
1627 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
1633 * skb_cow - copy header of skb when it is required
1634 * @skb: buffer to cow
1635 * @headroom: needed headroom
1637 * If the skb passed lacks sufficient headroom or its data part
1638 * is shared, data is reallocated. If reallocation fails, an error
1639 * is returned and original skb is not changed.
1641 * The result is skb with writable area skb->head...skb->tail
1642 * and at least @headroom of space at head.
1644 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
1646 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
1650 * skb_cow_head - skb_cow but only making the head writable
1651 * @skb: buffer to cow
1652 * @headroom: needed headroom
1654 * This function is identical to skb_cow except that we replace the
1655 * skb_cloned check by skb_header_cloned. It should be used when
1656 * you only need to push on some header and do not need to modify
1659 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
1661 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
1665 * skb_padto - pad an skbuff up to a minimal size
1666 * @skb: buffer to pad
1667 * @len: minimal length
1669 * Pads up a buffer to ensure the trailing bytes exist and are
1670 * blanked. If the buffer already contains sufficient data it
1671 * is untouched. Otherwise it is extended. Returns zero on
1672 * success. The skb is freed on error.
1675 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
1677 unsigned int size
= skb
->len
;
1678 if (likely(size
>= len
))
1680 return skb_pad(skb
, len
- size
);
1683 static inline int skb_add_data(struct sk_buff
*skb
,
1684 char __user
*from
, int copy
)
1686 const int off
= skb
->len
;
1688 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1690 __wsum csum
= csum_and_copy_from_user(from
, skb_put(skb
, copy
),
1693 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
1696 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
1699 __skb_trim(skb
, off
);
1703 static inline int skb_can_coalesce(struct sk_buff
*skb
, int i
,
1704 struct page
*page
, int off
)
1707 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
1709 return page
== frag
->page
&&
1710 off
== frag
->page_offset
+ frag
->size
;
1715 static inline int __skb_linearize(struct sk_buff
*skb
)
1717 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
1721 * skb_linearize - convert paged skb to linear one
1722 * @skb: buffer to linarize
1724 * If there is no free memory -ENOMEM is returned, otherwise zero
1725 * is returned and the old skb data released.
1727 static inline int skb_linearize(struct sk_buff
*skb
)
1729 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
1733 * skb_linearize_cow - make sure skb is linear and writable
1734 * @skb: buffer to process
1736 * If there is no free memory -ENOMEM is returned, otherwise zero
1737 * is returned and the old skb data released.
1739 static inline int skb_linearize_cow(struct sk_buff
*skb
)
1741 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
1742 __skb_linearize(skb
) : 0;
1746 * skb_postpull_rcsum - update checksum for received skb after pull
1747 * @skb: buffer to update
1748 * @start: start of data before pull
1749 * @len: length of data pulled
1751 * After doing a pull on a received packet, you need to call this to
1752 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1753 * CHECKSUM_NONE so that it can be recomputed from scratch.
1756 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
1757 const void *start
, unsigned int len
)
1759 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1760 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
1763 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
1766 * pskb_trim_rcsum - trim received skb and update checksum
1767 * @skb: buffer to trim
1770 * This is exactly the same as pskb_trim except that it ensures the
1771 * checksum of received packets are still valid after the operation.
1774 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
1776 if (likely(len
>= skb
->len
))
1778 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1779 skb
->ip_summed
= CHECKSUM_NONE
;
1780 return __pskb_trim(skb
, len
);
1783 #define skb_queue_walk(queue, skb) \
1784 for (skb = (queue)->next; \
1785 skb != (struct sk_buff *)(queue); \
1788 #define skb_queue_walk_safe(queue, skb, tmp) \
1789 for (skb = (queue)->next, tmp = skb->next; \
1790 skb != (struct sk_buff *)(queue); \
1791 skb = tmp, tmp = skb->next)
1793 #define skb_queue_walk_from(queue, skb) \
1794 for (; skb != (struct sk_buff *)(queue); \
1797 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1798 for (tmp = skb->next; \
1799 skb != (struct sk_buff *)(queue); \
1800 skb = tmp, tmp = skb->next)
1802 #define skb_queue_reverse_walk(queue, skb) \
1803 for (skb = (queue)->prev; \
1804 skb != (struct sk_buff *)(queue); \
1807 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
1808 for (skb = (queue)->prev, tmp = skb->prev; \
1809 skb != (struct sk_buff *)(queue); \
1810 skb = tmp, tmp = skb->prev)
1812 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
1813 for (tmp = skb->prev; \
1814 skb != (struct sk_buff *)(queue); \
1815 skb = tmp, tmp = skb->prev)
1817 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
1819 return skb_shinfo(skb
)->frag_list
!= NULL
;
1822 static inline void skb_frag_list_init(struct sk_buff
*skb
)
1824 skb_shinfo(skb
)->frag_list
= NULL
;
1827 static inline void skb_frag_add_head(struct sk_buff
*skb
, struct sk_buff
*frag
)
1829 frag
->next
= skb_shinfo(skb
)->frag_list
;
1830 skb_shinfo(skb
)->frag_list
= frag
;
1833 #define skb_walk_frags(skb, iter) \
1834 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1836 extern struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1837 int *peeked
, int *err
);
1838 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1839 int noblock
, int *err
);
1840 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
1841 struct poll_table_struct
*wait
);
1842 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
1843 int offset
, struct iovec
*to
,
1845 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
,
1848 extern int skb_copy_datagram_from_iovec(struct sk_buff
*skb
,
1850 const struct iovec
*from
,
1853 extern int skb_copy_datagram_const_iovec(const struct sk_buff
*from
,
1855 const struct iovec
*to
,
1858 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
1859 extern void skb_free_datagram_locked(struct sock
*sk
,
1860 struct sk_buff
*skb
);
1861 extern int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
,
1862 unsigned int flags
);
1863 extern __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
,
1864 int len
, __wsum csum
);
1865 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
1867 extern int skb_store_bits(struct sk_buff
*skb
, int offset
,
1868 const void *from
, int len
);
1869 extern __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
,
1870 int offset
, u8
*to
, int len
,
1872 extern int skb_splice_bits(struct sk_buff
*skb
,
1873 unsigned int offset
,
1874 struct pipe_inode_info
*pipe
,
1876 unsigned int flags
);
1877 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
1878 extern void skb_split(struct sk_buff
*skb
,
1879 struct sk_buff
*skb1
, const u32 len
);
1880 extern int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
,
1883 extern struct sk_buff
*skb_segment(struct sk_buff
*skb
, u32 features
);
1885 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
1886 int len
, void *buffer
)
1888 int hlen
= skb_headlen(skb
);
1890 if (hlen
- offset
>= len
)
1891 return skb
->data
+ offset
;
1893 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
1899 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
1901 const unsigned int len
)
1903 memcpy(to
, skb
->data
, len
);
1906 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
1907 const int offset
, void *to
,
1908 const unsigned int len
)
1910 memcpy(to
, skb
->data
+ offset
, len
);
1913 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
1915 const unsigned int len
)
1917 memcpy(skb
->data
, from
, len
);
1920 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
1923 const unsigned int len
)
1925 memcpy(skb
->data
+ offset
, from
, len
);
1928 extern void skb_init(void);
1930 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
1936 * skb_get_timestamp - get timestamp from a skb
1937 * @skb: skb to get stamp from
1938 * @stamp: pointer to struct timeval to store stamp in
1940 * Timestamps are stored in the skb as offsets to a base timestamp.
1941 * This function converts the offset back to a struct timeval and stores
1944 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
1945 struct timeval
*stamp
)
1947 *stamp
= ktime_to_timeval(skb
->tstamp
);
1950 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
1951 struct timespec
*stamp
)
1953 *stamp
= ktime_to_timespec(skb
->tstamp
);
1956 static inline void __net_timestamp(struct sk_buff
*skb
)
1958 skb
->tstamp
= ktime_get_real();
1961 static inline ktime_t
net_timedelta(ktime_t t
)
1963 return ktime_sub(ktime_get_real(), t
);
1966 static inline ktime_t
net_invalid_timestamp(void)
1968 return ktime_set(0, 0);
1971 extern void skb_timestamping_init(void);
1973 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
1975 extern void skb_clone_tx_timestamp(struct sk_buff
*skb
);
1976 extern bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
1978 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
1980 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
1984 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
1989 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
1992 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
1994 * @skb: clone of the the original outgoing packet
1995 * @hwtstamps: hardware time stamps
1998 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
1999 struct skb_shared_hwtstamps
*hwtstamps
);
2002 * skb_tstamp_tx - queue clone of skb with send time stamps
2003 * @orig_skb: the original outgoing packet
2004 * @hwtstamps: hardware time stamps, may be NULL if not available
2006 * If the skb has a socket associated, then this function clones the
2007 * skb (thus sharing the actual data and optional structures), stores
2008 * the optional hardware time stamping information (if non NULL) or
2009 * generates a software time stamp (otherwise), then queues the clone
2010 * to the error queue of the socket. Errors are silently ignored.
2012 extern void skb_tstamp_tx(struct sk_buff
*orig_skb
,
2013 struct skb_shared_hwtstamps
*hwtstamps
);
2015 static inline void sw_tx_timestamp(struct sk_buff
*skb
)
2017 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
&&
2018 !(skb_shinfo(skb
)->tx_flags
& SKBTX_IN_PROGRESS
))
2019 skb_tstamp_tx(skb
, NULL
);
2023 * skb_tx_timestamp() - Driver hook for transmit timestamping
2025 * Ethernet MAC Drivers should call this function in their hard_xmit()
2026 * function as soon as possible after giving the sk_buff to the MAC
2027 * hardware, but before freeing the sk_buff.
2029 * @skb: A socket buffer.
2031 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
2033 skb_clone_tx_timestamp(skb
);
2034 sw_tx_timestamp(skb
);
2037 extern __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
2038 extern __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
2040 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
2042 return skb
->ip_summed
& CHECKSUM_UNNECESSARY
;
2046 * skb_checksum_complete - Calculate checksum of an entire packet
2047 * @skb: packet to process
2049 * This function calculates the checksum over the entire packet plus
2050 * the value of skb->csum. The latter can be used to supply the
2051 * checksum of a pseudo header as used by TCP/UDP. It returns the
2054 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2055 * this function can be used to verify that checksum on received
2056 * packets. In that case the function should return zero if the
2057 * checksum is correct. In particular, this function will return zero
2058 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2059 * hardware has already verified the correctness of the checksum.
2061 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
2063 return skb_csum_unnecessary(skb
) ?
2064 0 : __skb_checksum_complete(skb
);
2067 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2068 extern void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
2069 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
2071 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
2072 nf_conntrack_destroy(nfct
);
2074 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
2077 atomic_inc(&nfct
->use
);
2080 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2081 static inline void nf_conntrack_get_reasm(struct sk_buff
*skb
)
2084 atomic_inc(&skb
->users
);
2086 static inline void nf_conntrack_put_reasm(struct sk_buff
*skb
)
2092 #ifdef CONFIG_BRIDGE_NETFILTER
2093 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
2095 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
2098 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
2101 atomic_inc(&nf_bridge
->use
);
2103 #endif /* CONFIG_BRIDGE_NETFILTER */
2104 static inline void nf_reset(struct sk_buff
*skb
)
2106 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2107 nf_conntrack_put(skb
->nfct
);
2110 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2111 nf_conntrack_put_reasm(skb
->nfct_reasm
);
2112 skb
->nfct_reasm
= NULL
;
2114 #ifdef CONFIG_BRIDGE_NETFILTER
2115 nf_bridge_put(skb
->nf_bridge
);
2116 skb
->nf_bridge
= NULL
;
2120 /* Note: This doesn't put any conntrack and bridge info in dst. */
2121 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2123 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2124 dst
->nfct
= src
->nfct
;
2125 nf_conntrack_get(src
->nfct
);
2126 dst
->nfctinfo
= src
->nfctinfo
;
2128 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2129 dst
->nfct_reasm
= src
->nfct_reasm
;
2130 nf_conntrack_get_reasm(src
->nfct_reasm
);
2132 #ifdef CONFIG_BRIDGE_NETFILTER
2133 dst
->nf_bridge
= src
->nf_bridge
;
2134 nf_bridge_get(src
->nf_bridge
);
2138 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2140 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2141 nf_conntrack_put(dst
->nfct
);
2143 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2144 nf_conntrack_put_reasm(dst
->nfct_reasm
);
2146 #ifdef CONFIG_BRIDGE_NETFILTER
2147 nf_bridge_put(dst
->nf_bridge
);
2149 __nf_copy(dst
, src
);
2152 #ifdef CONFIG_NETWORK_SECMARK
2153 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2155 to
->secmark
= from
->secmark
;
2158 static inline void skb_init_secmark(struct sk_buff
*skb
)
2163 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2166 static inline void skb_init_secmark(struct sk_buff
*skb
)
2170 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
2172 skb
->queue_mapping
= queue_mapping
;
2175 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
2177 return skb
->queue_mapping
;
2180 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
2182 to
->queue_mapping
= from
->queue_mapping
;
2185 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
2187 skb
->queue_mapping
= rx_queue
+ 1;
2190 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
2192 return skb
->queue_mapping
- 1;
2195 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
2197 return skb
->queue_mapping
!= 0;
2200 extern u16
__skb_tx_hash(const struct net_device
*dev
,
2201 const struct sk_buff
*skb
,
2202 unsigned int num_tx_queues
);
2205 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2210 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2216 static inline int skb_is_gso(const struct sk_buff
*skb
)
2218 return skb_shinfo(skb
)->gso_size
;
2221 static inline int skb_is_gso_v6(const struct sk_buff
*skb
)
2223 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
2226 extern void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
2228 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
2230 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2231 * wanted then gso_type will be set. */
2232 struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
2233 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
2234 unlikely(shinfo
->gso_type
== 0)) {
2235 __skb_warn_lro_forwarding(skb
);
2241 static inline void skb_forward_csum(struct sk_buff
*skb
)
2243 /* Unfortunately we don't support this one. Any brave souls? */
2244 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2245 skb
->ip_summed
= CHECKSUM_NONE
;
2249 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2250 * @skb: skb to check
2252 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2253 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2254 * use this helper, to document places where we make this assertion.
2256 static inline void skb_checksum_none_assert(struct sk_buff
*skb
)
2259 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
2263 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
2264 #endif /* __KERNEL__ */
2265 #endif /* _LINUX_SKBUFF_H */