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/bug.h>
22 #include <linux/cache.h>
24 #include <linux/atomic.h>
25 #include <asm/types.h>
26 #include <linux/spinlock.h>
27 #include <linux/net.h>
28 #include <linux/textsearch.h>
29 #include <net/checksum.h>
30 #include <linux/rcupdate.h>
31 #include <linux/dmaengine.h>
32 #include <linux/hrtimer.h>
33 #include <linux/dma-mapping.h>
34 #include <linux/netdev_features.h>
35 #include <net/flow_keys.h>
37 /* Don't change this without changing skb_csum_unnecessary! */
38 #define CHECKSUM_NONE 0
39 #define CHECKSUM_UNNECESSARY 1
40 #define CHECKSUM_COMPLETE 2
41 #define CHECKSUM_PARTIAL 3
43 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
44 ~(SMP_CACHE_BYTES - 1))
45 #define SKB_WITH_OVERHEAD(X) \
46 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
47 #define SKB_MAX_ORDER(X, ORDER) \
48 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
49 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
50 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
52 /* return minimum truesize of one skb containing X bytes of data */
53 #define SKB_TRUESIZE(X) ((X) + \
54 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
55 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
57 /* A. Checksumming of received packets by device.
59 * NONE: device failed to checksum this packet.
60 * skb->csum is undefined.
62 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
63 * skb->csum is undefined.
64 * It is bad option, but, unfortunately, many of vendors do this.
65 * Apparently with secret goal to sell you new device, when you
66 * will add new protocol to your host. F.e. IPv6. 8)
68 * COMPLETE: the most generic way. Device supplied checksum of _all_
69 * the packet as seen by netif_rx in skb->csum.
70 * NOTE: Even if device supports only some protocols, but
71 * is able to produce some skb->csum, it MUST use COMPLETE,
74 * PARTIAL: identical to the case for output below. This may occur
75 * on a packet received directly from another Linux OS, e.g.,
76 * a virtualised Linux kernel on the same host. The packet can
77 * be treated in the same way as UNNECESSARY except that on
78 * output (i.e., forwarding) the checksum must be filled in
79 * by the OS or the hardware.
81 * B. Checksumming on output.
83 * NONE: skb is checksummed by protocol or csum is not required.
85 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
86 * from skb->csum_start to the end and to record the checksum
87 * at skb->csum_start + skb->csum_offset.
89 * Device must show its capabilities in dev->features, set
90 * at device setup time.
91 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
93 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
94 * TCP/UDP over IPv4. Sigh. Vendors like this
95 * way by an unknown reason. Though, see comment above
96 * about CHECKSUM_UNNECESSARY. 8)
97 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
99 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
100 * that do not want net to perform the checksum calculation should use
101 * this flag in their outgoing skbs.
102 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
103 * offload. Correspondingly, the FCoE protocol driver
104 * stack should use CHECKSUM_UNNECESSARY.
106 * Any questions? No questions, good. --ANK
111 struct pipe_inode_info
;
113 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
114 struct nf_conntrack
{
119 #ifdef CONFIG_BRIDGE_NETFILTER
120 struct nf_bridge_info
{
123 struct net_device
*physindev
;
124 struct net_device
*physoutdev
;
125 unsigned long data
[32 / sizeof(unsigned long)];
129 struct sk_buff_head
{
130 /* These two members must be first. */
131 struct sk_buff
*next
;
132 struct sk_buff
*prev
;
140 /* To allow 64K frame to be packed as single skb without frag_list we
141 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
142 * buffers which do not start on a page boundary.
144 * Since GRO uses frags we allocate at least 16 regardless of page
147 #if (65536/PAGE_SIZE + 1) < 16
148 #define MAX_SKB_FRAGS 16UL
150 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
153 typedef struct skb_frag_struct skb_frag_t
;
155 struct skb_frag_struct
{
159 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
168 static inline unsigned int skb_frag_size(const skb_frag_t
*frag
)
173 static inline void skb_frag_size_set(skb_frag_t
*frag
, unsigned int size
)
178 static inline void skb_frag_size_add(skb_frag_t
*frag
, int delta
)
183 static inline void skb_frag_size_sub(skb_frag_t
*frag
, int delta
)
188 #define HAVE_HW_TIME_STAMP
191 * struct skb_shared_hwtstamps - hardware time stamps
192 * @hwtstamp: hardware time stamp transformed into duration
193 * since arbitrary point in time
194 * @syststamp: hwtstamp transformed to system time base
196 * Software time stamps generated by ktime_get_real() are stored in
197 * skb->tstamp. The relation between the different kinds of time
198 * stamps is as follows:
200 * syststamp and tstamp can be compared against each other in
201 * arbitrary combinations. The accuracy of a
202 * syststamp/tstamp/"syststamp from other device" comparison is
203 * limited by the accuracy of the transformation into system time
204 * base. This depends on the device driver and its underlying
207 * hwtstamps can only be compared against other hwtstamps from
210 * This structure is attached to packets as part of the
211 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
213 struct skb_shared_hwtstamps
{
218 /* Definitions for tx_flags in struct skb_shared_info */
220 /* generate hardware time stamp */
221 SKBTX_HW_TSTAMP
= 1 << 0,
223 /* generate software time stamp */
224 SKBTX_SW_TSTAMP
= 1 << 1,
226 /* device driver is going to provide hardware time stamp */
227 SKBTX_IN_PROGRESS
= 1 << 2,
229 /* device driver supports TX zero-copy buffers */
230 SKBTX_DEV_ZEROCOPY
= 1 << 3,
232 /* generate wifi status information (where possible) */
233 SKBTX_WIFI_STATUS
= 1 << 4,
235 /* This indicates at least one fragment might be overwritten
236 * (as in vmsplice(), sendfile() ...)
237 * If we need to compute a TX checksum, we'll need to copy
238 * all frags to avoid possible bad checksum
240 SKBTX_SHARED_FRAG
= 1 << 5,
244 * The callback notifies userspace to release buffers when skb DMA is done in
245 * lower device, the skb last reference should be 0 when calling this.
246 * The zerocopy_success argument is true if zero copy transmit occurred,
247 * false on data copy or out of memory error caused by data copy attempt.
248 * The ctx field is used to track device context.
249 * The desc field is used to track userspace buffer index.
252 void (*callback
)(struct ubuf_info
*, bool zerocopy_success
);
257 /* This data is invariant across clones and lives at
258 * the end of the header data, ie. at skb->end.
260 struct skb_shared_info
{
261 unsigned char nr_frags
;
263 unsigned short gso_size
;
264 /* Warning: this field is not always filled in (UFO)! */
265 unsigned short gso_segs
;
266 unsigned short gso_type
;
267 struct sk_buff
*frag_list
;
268 struct skb_shared_hwtstamps hwtstamps
;
272 * Warning : all fields before dataref are cleared in __alloc_skb()
276 /* Intermediate layers must ensure that destructor_arg
277 * remains valid until skb destructor */
278 void * destructor_arg
;
280 /* must be last field, see pskb_expand_head() */
281 skb_frag_t frags
[MAX_SKB_FRAGS
];
284 /* We divide dataref into two halves. The higher 16 bits hold references
285 * to the payload part of skb->data. The lower 16 bits hold references to
286 * the entire skb->data. A clone of a headerless skb holds the length of
287 * the header in skb->hdr_len.
289 * All users must obey the rule that the skb->data reference count must be
290 * greater than or equal to the payload reference count.
292 * Holding a reference to the payload part means that the user does not
293 * care about modifications to the header part of skb->data.
295 #define SKB_DATAREF_SHIFT 16
296 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
300 SKB_FCLONE_UNAVAILABLE
,
306 SKB_GSO_TCPV4
= 1 << 0,
307 SKB_GSO_UDP
= 1 << 1,
309 /* This indicates the skb is from an untrusted source. */
310 SKB_GSO_DODGY
= 1 << 2,
312 /* This indicates the tcp segment has CWR set. */
313 SKB_GSO_TCP_ECN
= 1 << 3,
315 SKB_GSO_TCPV6
= 1 << 4,
317 SKB_GSO_FCOE
= 1 << 5,
319 SKB_GSO_GRE
= 1 << 6,
321 SKB_GSO_UDP_TUNNEL
= 1 << 7,
324 #if BITS_PER_LONG > 32
325 #define NET_SKBUFF_DATA_USES_OFFSET 1
328 #ifdef NET_SKBUFF_DATA_USES_OFFSET
329 typedef unsigned int sk_buff_data_t
;
331 typedef unsigned char *sk_buff_data_t
;
334 #if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
335 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
336 #define NET_SKBUFF_NF_DEFRAG_NEEDED 1
340 * struct sk_buff - socket buffer
341 * @next: Next buffer in list
342 * @prev: Previous buffer in list
343 * @tstamp: Time we arrived
344 * @sk: Socket we are owned by
345 * @dev: Device we arrived on/are leaving by
346 * @cb: Control buffer. Free for use by every layer. Put private vars here
347 * @_skb_refdst: destination entry (with norefcount bit)
348 * @sp: the security path, used for xfrm
349 * @len: Length of actual data
350 * @data_len: Data length
351 * @mac_len: Length of link layer header
352 * @hdr_len: writable header length of cloned skb
353 * @csum: Checksum (must include start/offset pair)
354 * @csum_start: Offset from skb->head where checksumming should start
355 * @csum_offset: Offset from csum_start where checksum should be stored
356 * @priority: Packet queueing priority
357 * @local_df: allow local fragmentation
358 * @cloned: Head may be cloned (check refcnt to be sure)
359 * @ip_summed: Driver fed us an IP checksum
360 * @nohdr: Payload reference only, must not modify header
361 * @nfctinfo: Relationship of this skb to the connection
362 * @pkt_type: Packet class
363 * @fclone: skbuff clone status
364 * @ipvs_property: skbuff is owned by ipvs
365 * @peeked: this packet has been seen already, so stats have been
366 * done for it, don't do them again
367 * @nf_trace: netfilter packet trace flag
368 * @protocol: Packet protocol from driver
369 * @destructor: Destruct function
370 * @nfct: Associated connection, if any
371 * @nfct_reasm: netfilter conntrack re-assembly pointer
372 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
373 * @skb_iif: ifindex of device we arrived on
374 * @tc_index: Traffic control index
375 * @tc_verd: traffic control verdict
376 * @rxhash: the packet hash computed on receive
377 * @queue_mapping: Queue mapping for multiqueue devices
378 * @ndisc_nodetype: router type (from link layer)
379 * @ooo_okay: allow the mapping of a socket to a queue to be changed
380 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
382 * @wifi_acked_valid: wifi_acked was set
383 * @wifi_acked: whether frame was acked on wifi or not
384 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
385 * @dma_cookie: a cookie to one of several possible DMA operations
386 * done by skb DMA functions
387 * @secmark: security marking
388 * @mark: Generic packet mark
389 * @dropcount: total number of sk_receive_queue overflows
390 * @vlan_tci: vlan tag control information
391 * @inner_transport_header: Inner transport layer header (encapsulation)
392 * @inner_network_header: Network layer header (encapsulation)
393 * @inner_mac_header: Link layer header (encapsulation)
394 * @transport_header: Transport layer header
395 * @network_header: Network layer header
396 * @mac_header: Link layer header
397 * @tail: Tail pointer
399 * @head: Head of buffer
400 * @data: Data head pointer
401 * @truesize: Buffer size
402 * @users: User count - see {datagram,tcp}.c
406 /* These two members must be first. */
407 struct sk_buff
*next
;
408 struct sk_buff
*prev
;
413 struct net_device
*dev
;
416 * This is the control buffer. It is free to use for every
417 * layer. Please put your private variables there. If you
418 * want to keep them across layers you have to do a skb_clone()
419 * first. This is owned by whoever has the skb queued ATM.
421 char cb
[48] __aligned(8);
423 unsigned long _skb_refdst
;
439 kmemcheck_bitfield_begin(flags1
);
450 kmemcheck_bitfield_end(flags1
);
453 void (*destructor
)(struct sk_buff
*skb
);
454 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
455 struct nf_conntrack
*nfct
;
457 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
458 struct sk_buff
*nfct_reasm
;
460 #ifdef CONFIG_BRIDGE_NETFILTER
461 struct nf_bridge_info
*nf_bridge
;
470 #ifdef CONFIG_NET_SCHED
471 __u16 tc_index
; /* traffic control index */
472 #ifdef CONFIG_NET_CLS_ACT
473 __u16 tc_verd
; /* traffic control verdict */
478 kmemcheck_bitfield_begin(flags2
);
479 #ifdef CONFIG_IPV6_NDISC_NODETYPE
480 __u8 ndisc_nodetype
:2;
485 __u8 wifi_acked_valid
:1;
489 /* Encapsulation protocol and NIC drivers should use
490 * this flag to indicate to each other if the skb contains
491 * encapsulated packet or not and maybe use the inner packet
494 __u8 encapsulation
:1;
495 /* 7/9 bit hole (depending on ndisc_nodetype presence) */
496 kmemcheck_bitfield_end(flags2
);
498 #ifdef CONFIG_NET_DMA
499 dma_cookie_t dma_cookie
;
501 #ifdef CONFIG_NETWORK_SECMARK
507 __u32 reserved_tailroom
;
510 sk_buff_data_t inner_transport_header
;
511 sk_buff_data_t inner_network_header
;
512 sk_buff_data_t inner_mac_header
;
513 sk_buff_data_t transport_header
;
514 sk_buff_data_t network_header
;
515 sk_buff_data_t mac_header
;
516 /* These elements must be at the end, see alloc_skb() for details. */
521 unsigned int truesize
;
527 * Handling routines are only of interest to the kernel
529 #include <linux/slab.h>
532 #define SKB_ALLOC_FCLONE 0x01
533 #define SKB_ALLOC_RX 0x02
535 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
536 static inline bool skb_pfmemalloc(const struct sk_buff
*skb
)
538 return unlikely(skb
->pfmemalloc
);
542 * skb might have a dst pointer attached, refcounted or not.
543 * _skb_refdst low order bit is set if refcount was _not_ taken
545 #define SKB_DST_NOREF 1UL
546 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
549 * skb_dst - returns skb dst_entry
552 * Returns skb dst_entry, regardless of reference taken or not.
554 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
556 /* If refdst was not refcounted, check we still are in a
557 * rcu_read_lock section
559 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
560 !rcu_read_lock_held() &&
561 !rcu_read_lock_bh_held());
562 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
566 * skb_dst_set - sets skb dst
570 * Sets skb dst, assuming a reference was taken on dst and should
571 * be released by skb_dst_drop()
573 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
575 skb
->_skb_refdst
= (unsigned long)dst
;
578 extern void __skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
,
582 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
586 * Sets skb dst, assuming a reference was not taken on dst.
587 * If dst entry is cached, we do not take reference and dst_release
588 * will be avoided by refdst_drop. If dst entry is not cached, we take
589 * reference, so that last dst_release can destroy the dst immediately.
591 static inline void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
)
593 __skb_dst_set_noref(skb
, dst
, false);
597 * skb_dst_set_noref_force - sets skb dst, without taking reference
601 * Sets skb dst, assuming a reference was not taken on dst.
602 * No reference is taken and no dst_release will be called. While for
603 * cached dsts deferred reclaim is a basic feature, for entries that are
604 * not cached it is caller's job to guarantee that last dst_release for
605 * provided dst happens when nobody uses it, eg. after a RCU grace period.
607 static inline void skb_dst_set_noref_force(struct sk_buff
*skb
,
608 struct dst_entry
*dst
)
610 __skb_dst_set_noref(skb
, dst
, true);
614 * skb_dst_is_noref - Test if skb dst isn't refcounted
617 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
619 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
622 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
624 return (struct rtable
*)skb_dst(skb
);
627 extern void kfree_skb(struct sk_buff
*skb
);
628 extern void skb_tx_error(struct sk_buff
*skb
);
629 extern void consume_skb(struct sk_buff
*skb
);
630 extern void __kfree_skb(struct sk_buff
*skb
);
631 extern struct kmem_cache
*skbuff_head_cache
;
633 extern void kfree_skb_partial(struct sk_buff
*skb
, bool head_stolen
);
634 extern bool skb_try_coalesce(struct sk_buff
*to
, struct sk_buff
*from
,
635 bool *fragstolen
, int *delta_truesize
);
637 extern struct sk_buff
*__alloc_skb(unsigned int size
,
638 gfp_t priority
, int flags
, int node
);
639 extern struct sk_buff
*build_skb(void *data
, unsigned int frag_size
);
640 static inline struct sk_buff
*alloc_skb(unsigned int size
,
643 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
646 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
649 return __alloc_skb(size
, priority
, SKB_ALLOC_FCLONE
, NUMA_NO_NODE
);
652 extern struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
653 extern int skb_copy_ubufs(struct sk_buff
*skb
, gfp_t gfp_mask
);
654 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
656 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
658 extern struct sk_buff
*__pskb_copy(struct sk_buff
*skb
,
659 int headroom
, gfp_t gfp_mask
);
661 extern int pskb_expand_head(struct sk_buff
*skb
,
662 int nhead
, int ntail
,
664 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
665 unsigned int headroom
);
666 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
667 int newheadroom
, int newtailroom
,
669 extern int skb_to_sgvec(struct sk_buff
*skb
,
670 struct scatterlist
*sg
, int offset
,
672 extern int skb_cow_data(struct sk_buff
*skb
, int tailbits
,
673 struct sk_buff
**trailer
);
674 extern int skb_pad(struct sk_buff
*skb
, int pad
);
675 #define dev_kfree_skb(a) consume_skb(a)
677 extern int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
678 int getfrag(void *from
, char *to
, int offset
,
679 int len
,int odd
, struct sk_buff
*skb
),
680 void *from
, int length
);
682 struct skb_seq_state
{
686 __u32 stepped_offset
;
687 struct sk_buff
*root_skb
;
688 struct sk_buff
*cur_skb
;
692 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
693 unsigned int from
, unsigned int to
,
694 struct skb_seq_state
*st
);
695 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
696 struct skb_seq_state
*st
);
697 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
699 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
700 unsigned int to
, struct ts_config
*config
,
701 struct ts_state
*state
);
703 extern void __skb_get_rxhash(struct sk_buff
*skb
);
704 static inline __u32
skb_get_rxhash(struct sk_buff
*skb
)
707 __skb_get_rxhash(skb
);
712 #ifdef NET_SKBUFF_DATA_USES_OFFSET
713 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
715 return skb
->head
+ skb
->end
;
718 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
723 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
728 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
730 return skb
->end
- skb
->head
;
735 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
737 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
739 return &skb_shinfo(skb
)->hwtstamps
;
743 * skb_queue_empty - check if a queue is empty
746 * Returns true if the queue is empty, false otherwise.
748 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
750 return list
->next
== (struct sk_buff
*)list
;
754 * skb_queue_is_last - check if skb is the last entry in the queue
758 * Returns true if @skb is the last buffer on the list.
760 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
761 const struct sk_buff
*skb
)
763 return skb
->next
== (struct sk_buff
*)list
;
767 * skb_queue_is_first - check if skb is the first entry in the queue
771 * Returns true if @skb is the first buffer on the list.
773 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
774 const struct sk_buff
*skb
)
776 return skb
->prev
== (struct sk_buff
*)list
;
780 * skb_queue_next - return the next packet in the queue
782 * @skb: current buffer
784 * Return the next packet in @list after @skb. It is only valid to
785 * call this if skb_queue_is_last() evaluates to false.
787 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
788 const struct sk_buff
*skb
)
790 /* This BUG_ON may seem severe, but if we just return then we
791 * are going to dereference garbage.
793 BUG_ON(skb_queue_is_last(list
, skb
));
798 * skb_queue_prev - return the prev packet in the queue
800 * @skb: current buffer
802 * Return the prev packet in @list before @skb. It is only valid to
803 * call this if skb_queue_is_first() evaluates to false.
805 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
806 const struct sk_buff
*skb
)
808 /* This BUG_ON may seem severe, but if we just return then we
809 * are going to dereference garbage.
811 BUG_ON(skb_queue_is_first(list
, skb
));
816 * skb_get - reference buffer
817 * @skb: buffer to reference
819 * Makes another reference to a socket buffer and returns a pointer
822 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
824 atomic_inc(&skb
->users
);
829 * If users == 1, we are the only owner and are can avoid redundant
834 * skb_cloned - is the buffer a clone
835 * @skb: buffer to check
837 * Returns true if the buffer was generated with skb_clone() and is
838 * one of multiple shared copies of the buffer. Cloned buffers are
839 * shared data so must not be written to under normal circumstances.
841 static inline int skb_cloned(const struct sk_buff
*skb
)
843 return skb
->cloned
&&
844 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
847 static inline int skb_unclone(struct sk_buff
*skb
, gfp_t pri
)
849 might_sleep_if(pri
& __GFP_WAIT
);
852 return pskb_expand_head(skb
, 0, 0, pri
);
858 * skb_header_cloned - is the header a clone
859 * @skb: buffer to check
861 * Returns true if modifying the header part of the buffer requires
862 * the data to be copied.
864 static inline int skb_header_cloned(const struct sk_buff
*skb
)
871 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
872 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
877 * skb_header_release - release reference to header
878 * @skb: buffer to operate on
880 * Drop a reference to the header part of the buffer. This is done
881 * by acquiring a payload reference. You must not read from the header
882 * part of skb->data after this.
884 static inline void skb_header_release(struct sk_buff
*skb
)
888 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
892 * skb_shared - is the buffer shared
893 * @skb: buffer to check
895 * Returns true if more than one person has a reference to this
898 static inline int skb_shared(const struct sk_buff
*skb
)
900 return atomic_read(&skb
->users
) != 1;
904 * skb_share_check - check if buffer is shared and if so clone it
905 * @skb: buffer to check
906 * @pri: priority for memory allocation
908 * If the buffer is shared the buffer is cloned and the old copy
909 * drops a reference. A new clone with a single reference is returned.
910 * If the buffer is not shared the original buffer is returned. When
911 * being called from interrupt status or with spinlocks held pri must
914 * NULL is returned on a memory allocation failure.
916 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
, gfp_t pri
)
918 might_sleep_if(pri
& __GFP_WAIT
);
919 if (skb_shared(skb
)) {
920 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
932 * Copy shared buffers into a new sk_buff. We effectively do COW on
933 * packets to handle cases where we have a local reader and forward
934 * and a couple of other messy ones. The normal one is tcpdumping
935 * a packet thats being forwarded.
939 * skb_unshare - make a copy of a shared buffer
940 * @skb: buffer to check
941 * @pri: priority for memory allocation
943 * If the socket buffer is a clone then this function creates a new
944 * copy of the data, drops a reference count on the old copy and returns
945 * the new copy with the reference count at 1. If the buffer is not a clone
946 * the original buffer is returned. When called with a spinlock held or
947 * from interrupt state @pri must be %GFP_ATOMIC
949 * %NULL is returned on a memory allocation failure.
951 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
954 might_sleep_if(pri
& __GFP_WAIT
);
955 if (skb_cloned(skb
)) {
956 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
957 kfree_skb(skb
); /* Free our shared copy */
964 * skb_peek - peek at the head of an &sk_buff_head
965 * @list_: list to peek at
967 * Peek an &sk_buff. Unlike most other operations you _MUST_
968 * be careful with this one. A peek leaves the buffer on the
969 * list and someone else may run off with it. You must hold
970 * the appropriate locks or have a private queue to do this.
972 * Returns %NULL for an empty list or a pointer to the head element.
973 * The reference count is not incremented and the reference is therefore
974 * volatile. Use with caution.
976 static inline struct sk_buff
*skb_peek(const struct sk_buff_head
*list_
)
978 struct sk_buff
*skb
= list_
->next
;
980 if (skb
== (struct sk_buff
*)list_
)
986 * skb_peek_next - peek skb following the given one from a queue
987 * @skb: skb to start from
988 * @list_: list to peek at
990 * Returns %NULL when the end of the list is met or a pointer to the
991 * next element. The reference count is not incremented and the
992 * reference is therefore volatile. Use with caution.
994 static inline struct sk_buff
*skb_peek_next(struct sk_buff
*skb
,
995 const struct sk_buff_head
*list_
)
997 struct sk_buff
*next
= skb
->next
;
999 if (next
== (struct sk_buff
*)list_
)
1005 * skb_peek_tail - peek at the tail of an &sk_buff_head
1006 * @list_: list to peek at
1008 * Peek an &sk_buff. Unlike most other operations you _MUST_
1009 * be careful with this one. A peek leaves the buffer on the
1010 * list and someone else may run off with it. You must hold
1011 * the appropriate locks or have a private queue to do this.
1013 * Returns %NULL for an empty list or a pointer to the tail element.
1014 * The reference count is not incremented and the reference is therefore
1015 * volatile. Use with caution.
1017 static inline struct sk_buff
*skb_peek_tail(const struct sk_buff_head
*list_
)
1019 struct sk_buff
*skb
= list_
->prev
;
1021 if (skb
== (struct sk_buff
*)list_
)
1028 * skb_queue_len - get queue length
1029 * @list_: list to measure
1031 * Return the length of an &sk_buff queue.
1033 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
1039 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1040 * @list: queue to initialize
1042 * This initializes only the list and queue length aspects of
1043 * an sk_buff_head object. This allows to initialize the list
1044 * aspects of an sk_buff_head without reinitializing things like
1045 * the spinlock. It can also be used for on-stack sk_buff_head
1046 * objects where the spinlock is known to not be used.
1048 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
1050 list
->prev
= list
->next
= (struct sk_buff
*)list
;
1055 * This function creates a split out lock class for each invocation;
1056 * this is needed for now since a whole lot of users of the skb-queue
1057 * infrastructure in drivers have different locking usage (in hardirq)
1058 * than the networking core (in softirq only). In the long run either the
1059 * network layer or drivers should need annotation to consolidate the
1060 * main types of usage into 3 classes.
1062 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
1064 spin_lock_init(&list
->lock
);
1065 __skb_queue_head_init(list
);
1068 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
1069 struct lock_class_key
*class)
1071 skb_queue_head_init(list
);
1072 lockdep_set_class(&list
->lock
, class);
1076 * Insert an sk_buff on a list.
1078 * The "__skb_xxxx()" functions are the non-atomic ones that
1079 * can only be called with interrupts disabled.
1081 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
1082 static inline void __skb_insert(struct sk_buff
*newsk
,
1083 struct sk_buff
*prev
, struct sk_buff
*next
,
1084 struct sk_buff_head
*list
)
1088 next
->prev
= prev
->next
= newsk
;
1092 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
1093 struct sk_buff
*prev
,
1094 struct sk_buff
*next
)
1096 struct sk_buff
*first
= list
->next
;
1097 struct sk_buff
*last
= list
->prev
;
1107 * skb_queue_splice - join two skb lists, this is designed for stacks
1108 * @list: the new list to add
1109 * @head: the place to add it in the first list
1111 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
1112 struct sk_buff_head
*head
)
1114 if (!skb_queue_empty(list
)) {
1115 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1116 head
->qlen
+= list
->qlen
;
1121 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1122 * @list: the new list to add
1123 * @head: the place to add it in the first list
1125 * The list at @list is reinitialised
1127 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
1128 struct sk_buff_head
*head
)
1130 if (!skb_queue_empty(list
)) {
1131 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1132 head
->qlen
+= list
->qlen
;
1133 __skb_queue_head_init(list
);
1138 * skb_queue_splice_tail - join two skb lists, each list being a queue
1139 * @list: the new list to add
1140 * @head: the place to add it in the first list
1142 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
1143 struct sk_buff_head
*head
)
1145 if (!skb_queue_empty(list
)) {
1146 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1147 head
->qlen
+= list
->qlen
;
1152 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1153 * @list: the new list to add
1154 * @head: the place to add it in the first list
1156 * Each of the lists is a queue.
1157 * The list at @list is reinitialised
1159 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
1160 struct sk_buff_head
*head
)
1162 if (!skb_queue_empty(list
)) {
1163 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1164 head
->qlen
+= list
->qlen
;
1165 __skb_queue_head_init(list
);
1170 * __skb_queue_after - queue a buffer at the list head
1171 * @list: list to use
1172 * @prev: place after this buffer
1173 * @newsk: buffer to queue
1175 * Queue a buffer int the middle of a list. This function takes no locks
1176 * and you must therefore hold required locks before calling it.
1178 * A buffer cannot be placed on two lists at the same time.
1180 static inline void __skb_queue_after(struct sk_buff_head
*list
,
1181 struct sk_buff
*prev
,
1182 struct sk_buff
*newsk
)
1184 __skb_insert(newsk
, prev
, prev
->next
, list
);
1187 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
1188 struct sk_buff_head
*list
);
1190 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1191 struct sk_buff
*next
,
1192 struct sk_buff
*newsk
)
1194 __skb_insert(newsk
, next
->prev
, next
, list
);
1198 * __skb_queue_head - queue a buffer at the list head
1199 * @list: list to use
1200 * @newsk: buffer to queue
1202 * Queue a buffer at the start of a list. This function takes no locks
1203 * and you must therefore hold required locks before calling it.
1205 * A buffer cannot be placed on two lists at the same time.
1207 extern void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1208 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1209 struct sk_buff
*newsk
)
1211 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1215 * __skb_queue_tail - queue a buffer at the list tail
1216 * @list: list to use
1217 * @newsk: buffer to queue
1219 * Queue a buffer at the end of a list. This function takes no locks
1220 * and you must therefore hold required locks before calling it.
1222 * A buffer cannot be placed on two lists at the same time.
1224 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1225 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1226 struct sk_buff
*newsk
)
1228 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1232 * remove sk_buff from list. _Must_ be called atomically, and with
1235 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1236 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1238 struct sk_buff
*next
, *prev
;
1243 skb
->next
= skb
->prev
= NULL
;
1249 * __skb_dequeue - remove from the head of the queue
1250 * @list: list to dequeue from
1252 * Remove the head of the list. This function does not take any locks
1253 * so must be used with appropriate locks held only. The head item is
1254 * returned or %NULL if the list is empty.
1256 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1257 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1259 struct sk_buff
*skb
= skb_peek(list
);
1261 __skb_unlink(skb
, list
);
1266 * __skb_dequeue_tail - remove from the tail of the queue
1267 * @list: list to dequeue from
1269 * Remove the tail of the list. This function does not take any locks
1270 * so must be used with appropriate locks held only. The tail item is
1271 * returned or %NULL if the list is empty.
1273 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1274 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1276 struct sk_buff
*skb
= skb_peek_tail(list
);
1278 __skb_unlink(skb
, list
);
1283 static inline bool skb_is_nonlinear(const struct sk_buff
*skb
)
1285 return skb
->data_len
;
1288 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1290 return skb
->len
- skb
->data_len
;
1293 static inline int skb_pagelen(const struct sk_buff
*skb
)
1297 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1298 len
+= skb_frag_size(&skb_shinfo(skb
)->frags
[i
]);
1299 return len
+ skb_headlen(skb
);
1303 * __skb_fill_page_desc - initialise a paged fragment in an skb
1304 * @skb: buffer containing fragment to be initialised
1305 * @i: paged fragment index to initialise
1306 * @page: the page to use for this fragment
1307 * @off: the offset to the data with @page
1308 * @size: the length of the data
1310 * Initialises the @i'th fragment of @skb to point to &size bytes at
1311 * offset @off within @page.
1313 * Does not take any additional reference on the fragment.
1315 static inline void __skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1316 struct page
*page
, int off
, int size
)
1318 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1321 * Propagate page->pfmemalloc to the skb if we can. The problem is
1322 * that not all callers have unique ownership of the page. If
1323 * pfmemalloc is set, we check the mapping as a mapping implies
1324 * page->index is set (index and pfmemalloc share space).
1325 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1326 * do not lose pfmemalloc information as the pages would not be
1327 * allocated using __GFP_MEMALLOC.
1329 frag
->page
.p
= page
;
1330 frag
->page_offset
= off
;
1331 skb_frag_size_set(frag
, size
);
1333 page
= compound_head(page
);
1334 if (page
->pfmemalloc
&& !page
->mapping
)
1335 skb
->pfmemalloc
= true;
1339 * skb_fill_page_desc - initialise a paged fragment in an skb
1340 * @skb: buffer containing fragment to be initialised
1341 * @i: paged fragment index to initialise
1342 * @page: the page to use for this fragment
1343 * @off: the offset to the data with @page
1344 * @size: the length of the data
1346 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1347 * @skb to point to &size bytes at offset @off within @page. In
1348 * addition updates @skb such that @i is the last fragment.
1350 * Does not take any additional reference on the fragment.
1352 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1353 struct page
*page
, int off
, int size
)
1355 __skb_fill_page_desc(skb
, i
, page
, off
, size
);
1356 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1359 extern void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
,
1360 int off
, int size
, unsigned int truesize
);
1362 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1363 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1364 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1366 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1367 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1369 return skb
->head
+ skb
->tail
;
1372 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1374 skb
->tail
= skb
->data
- skb
->head
;
1377 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1379 skb_reset_tail_pointer(skb
);
1380 skb
->tail
+= offset
;
1382 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1383 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1388 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1390 skb
->tail
= skb
->data
;
1393 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1395 skb
->tail
= skb
->data
+ offset
;
1398 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1401 * Add data to an sk_buff
1403 extern unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1404 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1406 unsigned char *tmp
= skb_tail_pointer(skb
);
1407 SKB_LINEAR_ASSERT(skb
);
1413 extern unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1414 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1421 extern unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1422 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1425 BUG_ON(skb
->len
< skb
->data_len
);
1426 return skb
->data
+= len
;
1429 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1431 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1434 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1436 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1438 if (len
> skb_headlen(skb
) &&
1439 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1442 return skb
->data
+= len
;
1445 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1447 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1450 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1452 if (likely(len
<= skb_headlen(skb
)))
1454 if (unlikely(len
> skb
->len
))
1456 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1460 * skb_headroom - bytes at buffer head
1461 * @skb: buffer to check
1463 * Return the number of bytes of free space at the head of an &sk_buff.
1465 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1467 return skb
->data
- skb
->head
;
1471 * skb_tailroom - bytes at buffer end
1472 * @skb: buffer to check
1474 * Return the number of bytes of free space at the tail of an sk_buff
1476 static inline int skb_tailroom(const struct sk_buff
*skb
)
1478 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1482 * skb_availroom - bytes at buffer end
1483 * @skb: buffer to check
1485 * Return the number of bytes of free space at the tail of an sk_buff
1486 * allocated by sk_stream_alloc()
1488 static inline int skb_availroom(const struct sk_buff
*skb
)
1490 if (skb_is_nonlinear(skb
))
1493 return skb
->end
- skb
->tail
- skb
->reserved_tailroom
;
1497 * skb_reserve - adjust headroom
1498 * @skb: buffer to alter
1499 * @len: bytes to move
1501 * Increase the headroom of an empty &sk_buff by reducing the tail
1502 * room. This is only allowed for an empty buffer.
1504 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
1510 static inline void skb_reset_inner_headers(struct sk_buff
*skb
)
1512 skb
->inner_mac_header
= skb
->mac_header
;
1513 skb
->inner_network_header
= skb
->network_header
;
1514 skb
->inner_transport_header
= skb
->transport_header
;
1517 static inline void skb_reset_mac_len(struct sk_buff
*skb
)
1519 skb
->mac_len
= skb
->network_header
- skb
->mac_header
;
1522 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1523 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1526 return skb
->head
+ skb
->inner_transport_header
;
1529 static inline void skb_reset_inner_transport_header(struct sk_buff
*skb
)
1531 skb
->inner_transport_header
= skb
->data
- skb
->head
;
1534 static inline void skb_set_inner_transport_header(struct sk_buff
*skb
,
1537 skb_reset_inner_transport_header(skb
);
1538 skb
->inner_transport_header
+= offset
;
1541 static inline unsigned char *skb_inner_network_header(const struct sk_buff
*skb
)
1543 return skb
->head
+ skb
->inner_network_header
;
1546 static inline void skb_reset_inner_network_header(struct sk_buff
*skb
)
1548 skb
->inner_network_header
= skb
->data
- skb
->head
;
1551 static inline void skb_set_inner_network_header(struct sk_buff
*skb
,
1554 skb_reset_inner_network_header(skb
);
1555 skb
->inner_network_header
+= offset
;
1558 static inline unsigned char *skb_inner_mac_header(const struct sk_buff
*skb
)
1560 return skb
->head
+ skb
->inner_mac_header
;
1563 static inline void skb_reset_inner_mac_header(struct sk_buff
*skb
)
1565 skb
->inner_mac_header
= skb
->data
- skb
->head
;
1568 static inline void skb_set_inner_mac_header(struct sk_buff
*skb
,
1571 skb_reset_inner_mac_header(skb
);
1572 skb
->inner_mac_header
+= offset
;
1574 static inline bool skb_transport_header_was_set(const struct sk_buff
*skb
)
1576 return skb
->transport_header
!= ~0U;
1579 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1581 return skb
->head
+ skb
->transport_header
;
1584 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1586 skb
->transport_header
= skb
->data
- skb
->head
;
1589 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1592 skb_reset_transport_header(skb
);
1593 skb
->transport_header
+= offset
;
1596 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1598 return skb
->head
+ skb
->network_header
;
1601 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1603 skb
->network_header
= skb
->data
- skb
->head
;
1606 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1608 skb_reset_network_header(skb
);
1609 skb
->network_header
+= offset
;
1612 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1614 return skb
->head
+ skb
->mac_header
;
1617 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1619 return skb
->mac_header
!= ~0U;
1622 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1624 skb
->mac_header
= skb
->data
- skb
->head
;
1627 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1629 skb_reset_mac_header(skb
);
1630 skb
->mac_header
+= offset
;
1633 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1634 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1637 return skb
->inner_transport_header
;
1640 static inline void skb_reset_inner_transport_header(struct sk_buff
*skb
)
1642 skb
->inner_transport_header
= skb
->data
;
1645 static inline void skb_set_inner_transport_header(struct sk_buff
*skb
,
1648 skb
->inner_transport_header
= skb
->data
+ offset
;
1651 static inline unsigned char *skb_inner_network_header(const struct sk_buff
*skb
)
1653 return skb
->inner_network_header
;
1656 static inline void skb_reset_inner_network_header(struct sk_buff
*skb
)
1658 skb
->inner_network_header
= skb
->data
;
1661 static inline void skb_set_inner_network_header(struct sk_buff
*skb
,
1664 skb
->inner_network_header
= skb
->data
+ offset
;
1667 static inline unsigned char *skb_inner_mac_header(const struct sk_buff
*skb
)
1669 return skb
->inner_mac_header
;
1672 static inline void skb_reset_inner_mac_header(struct sk_buff
*skb
)
1674 skb
->inner_mac_header
= skb
->data
;
1677 static inline void skb_set_inner_mac_header(struct sk_buff
*skb
,
1680 skb
->inner_mac_header
= skb
->data
+ offset
;
1682 static inline bool skb_transport_header_was_set(const struct sk_buff
*skb
)
1684 return skb
->transport_header
!= NULL
;
1687 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1689 return skb
->transport_header
;
1692 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1694 skb
->transport_header
= skb
->data
;
1697 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1700 skb
->transport_header
= skb
->data
+ offset
;
1703 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1705 return skb
->network_header
;
1708 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1710 skb
->network_header
= skb
->data
;
1713 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1715 skb
->network_header
= skb
->data
+ offset
;
1718 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1720 return skb
->mac_header
;
1723 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1725 return skb
->mac_header
!= NULL
;
1728 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1730 skb
->mac_header
= skb
->data
;
1733 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1735 skb
->mac_header
= skb
->data
+ offset
;
1737 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1739 static inline void skb_probe_transport_header(struct sk_buff
*skb
,
1740 const int offset_hint
)
1742 struct flow_keys keys
;
1744 if (skb_transport_header_was_set(skb
))
1746 else if (skb_flow_dissect(skb
, &keys
))
1747 skb_set_transport_header(skb
, keys
.thoff
);
1749 skb_set_transport_header(skb
, offset_hint
);
1752 static inline void skb_mac_header_rebuild(struct sk_buff
*skb
)
1754 if (skb_mac_header_was_set(skb
)) {
1755 const unsigned char *old_mac
= skb_mac_header(skb
);
1757 skb_set_mac_header(skb
, -skb
->mac_len
);
1758 memmove(skb_mac_header(skb
), old_mac
, skb
->mac_len
);
1762 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
1764 return skb
->csum_start
- skb_headroom(skb
);
1767 static inline int skb_transport_offset(const struct sk_buff
*skb
)
1769 return skb_transport_header(skb
) - skb
->data
;
1772 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
1774 return skb
->transport_header
- skb
->network_header
;
1777 static inline u32
skb_inner_network_header_len(const struct sk_buff
*skb
)
1779 return skb
->inner_transport_header
- skb
->inner_network_header
;
1782 static inline int skb_network_offset(const struct sk_buff
*skb
)
1784 return skb_network_header(skb
) - skb
->data
;
1787 static inline int skb_inner_network_offset(const struct sk_buff
*skb
)
1789 return skb_inner_network_header(skb
) - skb
->data
;
1792 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
1794 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
1798 * CPUs often take a performance hit when accessing unaligned memory
1799 * locations. The actual performance hit varies, it can be small if the
1800 * hardware handles it or large if we have to take an exception and fix it
1803 * Since an ethernet header is 14 bytes network drivers often end up with
1804 * the IP header at an unaligned offset. The IP header can be aligned by
1805 * shifting the start of the packet by 2 bytes. Drivers should do this
1808 * skb_reserve(skb, NET_IP_ALIGN);
1810 * The downside to this alignment of the IP header is that the DMA is now
1811 * unaligned. On some architectures the cost of an unaligned DMA is high
1812 * and this cost outweighs the gains made by aligning the IP header.
1814 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1817 #ifndef NET_IP_ALIGN
1818 #define NET_IP_ALIGN 2
1822 * The networking layer reserves some headroom in skb data (via
1823 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1824 * the header has to grow. In the default case, if the header has to grow
1825 * 32 bytes or less we avoid the reallocation.
1827 * Unfortunately this headroom changes the DMA alignment of the resulting
1828 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1829 * on some architectures. An architecture can override this value,
1830 * perhaps setting it to a cacheline in size (since that will maintain
1831 * cacheline alignment of the DMA). It must be a power of 2.
1833 * Various parts of the networking layer expect at least 32 bytes of
1834 * headroom, you should not reduce this.
1836 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1837 * to reduce average number of cache lines per packet.
1838 * get_rps_cpus() for example only access one 64 bytes aligned block :
1839 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1842 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1845 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
1847 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
1849 if (unlikely(skb_is_nonlinear(skb
))) {
1854 skb_set_tail_pointer(skb
, len
);
1857 extern void skb_trim(struct sk_buff
*skb
, unsigned int len
);
1859 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1862 return ___pskb_trim(skb
, len
);
1863 __skb_trim(skb
, len
);
1867 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1869 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
1873 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1874 * @skb: buffer to alter
1877 * This is identical to pskb_trim except that the caller knows that
1878 * the skb is not cloned so we should never get an error due to out-
1881 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
1883 int err
= pskb_trim(skb
, len
);
1888 * skb_orphan - orphan a buffer
1889 * @skb: buffer to orphan
1891 * If a buffer currently has an owner then we call the owner's
1892 * destructor function and make the @skb unowned. The buffer continues
1893 * to exist but is no longer charged to its former owner.
1895 static inline void skb_orphan(struct sk_buff
*skb
)
1897 if (skb
->destructor
)
1898 skb
->destructor(skb
);
1899 skb
->destructor
= NULL
;
1904 * skb_orphan_frags - orphan the frags contained in a buffer
1905 * @skb: buffer to orphan frags from
1906 * @gfp_mask: allocation mask for replacement pages
1908 * For each frag in the SKB which needs a destructor (i.e. has an
1909 * owner) create a copy of that frag and release the original
1910 * page by calling the destructor.
1912 static inline int skb_orphan_frags(struct sk_buff
*skb
, gfp_t gfp_mask
)
1914 if (likely(!(skb_shinfo(skb
)->tx_flags
& SKBTX_DEV_ZEROCOPY
)))
1916 return skb_copy_ubufs(skb
, gfp_mask
);
1920 * __skb_queue_purge - empty a list
1921 * @list: list to empty
1923 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1924 * the list and one reference dropped. This function does not take the
1925 * list lock and the caller must hold the relevant locks to use it.
1927 extern void skb_queue_purge(struct sk_buff_head
*list
);
1928 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
1930 struct sk_buff
*skb
;
1931 while ((skb
= __skb_dequeue(list
)) != NULL
)
1935 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1936 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1937 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1939 extern void *netdev_alloc_frag(unsigned int fragsz
);
1941 extern struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
1942 unsigned int length
,
1946 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1947 * @dev: network device to receive on
1948 * @length: length to allocate
1950 * Allocate a new &sk_buff and assign it a usage count of one. The
1951 * buffer has unspecified headroom built in. Users should allocate
1952 * the headroom they think they need without accounting for the
1953 * built in space. The built in space is used for optimisations.
1955 * %NULL is returned if there is no free memory. Although this function
1956 * allocates memory it can be called from an interrupt.
1958 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
1959 unsigned int length
)
1961 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
1964 /* legacy helper around __netdev_alloc_skb() */
1965 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
1968 return __netdev_alloc_skb(NULL
, length
, gfp_mask
);
1971 /* legacy helper around netdev_alloc_skb() */
1972 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
1974 return netdev_alloc_skb(NULL
, length
);
1978 static inline struct sk_buff
*__netdev_alloc_skb_ip_align(struct net_device
*dev
,
1979 unsigned int length
, gfp_t gfp
)
1981 struct sk_buff
*skb
= __netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
, gfp
);
1983 if (NET_IP_ALIGN
&& skb
)
1984 skb_reserve(skb
, NET_IP_ALIGN
);
1988 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
1989 unsigned int length
)
1991 return __netdev_alloc_skb_ip_align(dev
, length
, GFP_ATOMIC
);
1995 * __skb_alloc_page - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1996 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1997 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1998 * @order: size of the allocation
2000 * Allocate a new page.
2002 * %NULL is returned if there is no free memory.
2004 static inline struct page
*__skb_alloc_pages(gfp_t gfp_mask
,
2005 struct sk_buff
*skb
,
2010 gfp_mask
|= __GFP_COLD
;
2012 if (!(gfp_mask
& __GFP_NOMEMALLOC
))
2013 gfp_mask
|= __GFP_MEMALLOC
;
2015 page
= alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, order
);
2016 if (skb
&& page
&& page
->pfmemalloc
)
2017 skb
->pfmemalloc
= true;
2023 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2024 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2025 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2027 * Allocate a new page.
2029 * %NULL is returned if there is no free memory.
2031 static inline struct page
*__skb_alloc_page(gfp_t gfp_mask
,
2032 struct sk_buff
*skb
)
2034 return __skb_alloc_pages(gfp_mask
, skb
, 0);
2038 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2039 * @page: The page that was allocated from skb_alloc_page
2040 * @skb: The skb that may need pfmemalloc set
2042 static inline void skb_propagate_pfmemalloc(struct page
*page
,
2043 struct sk_buff
*skb
)
2045 if (page
&& page
->pfmemalloc
)
2046 skb
->pfmemalloc
= true;
2050 * skb_frag_page - retrieve the page refered to by a paged fragment
2051 * @frag: the paged fragment
2053 * Returns the &struct page associated with @frag.
2055 static inline struct page
*skb_frag_page(const skb_frag_t
*frag
)
2057 return frag
->page
.p
;
2061 * __skb_frag_ref - take an addition reference on a paged fragment.
2062 * @frag: the paged fragment
2064 * Takes an additional reference on the paged fragment @frag.
2066 static inline void __skb_frag_ref(skb_frag_t
*frag
)
2068 get_page(skb_frag_page(frag
));
2072 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2074 * @f: the fragment offset.
2076 * Takes an additional reference on the @f'th paged fragment of @skb.
2078 static inline void skb_frag_ref(struct sk_buff
*skb
, int f
)
2080 __skb_frag_ref(&skb_shinfo(skb
)->frags
[f
]);
2084 * __skb_frag_unref - release a reference on a paged fragment.
2085 * @frag: the paged fragment
2087 * Releases a reference on the paged fragment @frag.
2089 static inline void __skb_frag_unref(skb_frag_t
*frag
)
2091 put_page(skb_frag_page(frag
));
2095 * skb_frag_unref - release a reference on a paged fragment of an skb.
2097 * @f: the fragment offset
2099 * Releases a reference on the @f'th paged fragment of @skb.
2101 static inline void skb_frag_unref(struct sk_buff
*skb
, int f
)
2103 __skb_frag_unref(&skb_shinfo(skb
)->frags
[f
]);
2107 * skb_frag_address - gets the address of the data contained in a paged fragment
2108 * @frag: the paged fragment buffer
2110 * Returns the address of the data within @frag. The page must already
2113 static inline void *skb_frag_address(const skb_frag_t
*frag
)
2115 return page_address(skb_frag_page(frag
)) + frag
->page_offset
;
2119 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2120 * @frag: the paged fragment buffer
2122 * Returns the address of the data within @frag. Checks that the page
2123 * is mapped and returns %NULL otherwise.
2125 static inline void *skb_frag_address_safe(const skb_frag_t
*frag
)
2127 void *ptr
= page_address(skb_frag_page(frag
));
2131 return ptr
+ frag
->page_offset
;
2135 * __skb_frag_set_page - sets the page contained in a paged fragment
2136 * @frag: the paged fragment
2137 * @page: the page to set
2139 * Sets the fragment @frag to contain @page.
2141 static inline void __skb_frag_set_page(skb_frag_t
*frag
, struct page
*page
)
2143 frag
->page
.p
= page
;
2147 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2149 * @f: the fragment offset
2150 * @page: the page to set
2152 * Sets the @f'th fragment of @skb to contain @page.
2154 static inline void skb_frag_set_page(struct sk_buff
*skb
, int f
,
2157 __skb_frag_set_page(&skb_shinfo(skb
)->frags
[f
], page
);
2161 * skb_frag_dma_map - maps a paged fragment via the DMA API
2162 * @dev: the device to map the fragment to
2163 * @frag: the paged fragment to map
2164 * @offset: the offset within the fragment (starting at the
2165 * fragment's own offset)
2166 * @size: the number of bytes to map
2167 * @dir: the direction of the mapping (%PCI_DMA_*)
2169 * Maps the page associated with @frag to @device.
2171 static inline dma_addr_t
skb_frag_dma_map(struct device
*dev
,
2172 const skb_frag_t
*frag
,
2173 size_t offset
, size_t size
,
2174 enum dma_data_direction dir
)
2176 return dma_map_page(dev
, skb_frag_page(frag
),
2177 frag
->page_offset
+ offset
, size
, dir
);
2180 static inline struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
2183 return __pskb_copy(skb
, skb_headroom(skb
), gfp_mask
);
2187 * skb_clone_writable - is the header of a clone writable
2188 * @skb: buffer to check
2189 * @len: length up to which to write
2191 * Returns true if modifying the header part of the cloned buffer
2192 * does not requires the data to be copied.
2194 static inline int skb_clone_writable(const struct sk_buff
*skb
, unsigned int len
)
2196 return !skb_header_cloned(skb
) &&
2197 skb_headroom(skb
) + len
<= skb
->hdr_len
;
2200 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
2205 if (headroom
> skb_headroom(skb
))
2206 delta
= headroom
- skb_headroom(skb
);
2208 if (delta
|| cloned
)
2209 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
2215 * skb_cow - copy header of skb when it is required
2216 * @skb: buffer to cow
2217 * @headroom: needed headroom
2219 * If the skb passed lacks sufficient headroom or its data part
2220 * is shared, data is reallocated. If reallocation fails, an error
2221 * is returned and original skb is not changed.
2223 * The result is skb with writable area skb->head...skb->tail
2224 * and at least @headroom of space at head.
2226 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
2228 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
2232 * skb_cow_head - skb_cow but only making the head writable
2233 * @skb: buffer to cow
2234 * @headroom: needed headroom
2236 * This function is identical to skb_cow except that we replace the
2237 * skb_cloned check by skb_header_cloned. It should be used when
2238 * you only need to push on some header and do not need to modify
2241 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
2243 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
2247 * skb_padto - pad an skbuff up to a minimal size
2248 * @skb: buffer to pad
2249 * @len: minimal length
2251 * Pads up a buffer to ensure the trailing bytes exist and are
2252 * blanked. If the buffer already contains sufficient data it
2253 * is untouched. Otherwise it is extended. Returns zero on
2254 * success. The skb is freed on error.
2257 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
2259 unsigned int size
= skb
->len
;
2260 if (likely(size
>= len
))
2262 return skb_pad(skb
, len
- size
);
2265 static inline int skb_add_data(struct sk_buff
*skb
,
2266 char __user
*from
, int copy
)
2268 const int off
= skb
->len
;
2270 if (skb
->ip_summed
== CHECKSUM_NONE
) {
2272 __wsum csum
= csum_and_copy_from_user(from
, skb_put(skb
, copy
),
2275 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
2278 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
2281 __skb_trim(skb
, off
);
2285 static inline bool skb_can_coalesce(struct sk_buff
*skb
, int i
,
2286 const struct page
*page
, int off
)
2289 const struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
2291 return page
== skb_frag_page(frag
) &&
2292 off
== frag
->page_offset
+ skb_frag_size(frag
);
2297 static inline int __skb_linearize(struct sk_buff
*skb
)
2299 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
2303 * skb_linearize - convert paged skb to linear one
2304 * @skb: buffer to linarize
2306 * If there is no free memory -ENOMEM is returned, otherwise zero
2307 * is returned and the old skb data released.
2309 static inline int skb_linearize(struct sk_buff
*skb
)
2311 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
2315 * skb_has_shared_frag - can any frag be overwritten
2316 * @skb: buffer to test
2318 * Return true if the skb has at least one frag that might be modified
2319 * by an external entity (as in vmsplice()/sendfile())
2321 static inline bool skb_has_shared_frag(const struct sk_buff
*skb
)
2323 return skb_is_nonlinear(skb
) &&
2324 skb_shinfo(skb
)->tx_flags
& SKBTX_SHARED_FRAG
;
2328 * skb_linearize_cow - make sure skb is linear and writable
2329 * @skb: buffer to process
2331 * If there is no free memory -ENOMEM is returned, otherwise zero
2332 * is returned and the old skb data released.
2334 static inline int skb_linearize_cow(struct sk_buff
*skb
)
2336 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
2337 __skb_linearize(skb
) : 0;
2341 * skb_postpull_rcsum - update checksum for received skb after pull
2342 * @skb: buffer to update
2343 * @start: start of data before pull
2344 * @len: length of data pulled
2346 * After doing a pull on a received packet, you need to call this to
2347 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2348 * CHECKSUM_NONE so that it can be recomputed from scratch.
2351 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
2352 const void *start
, unsigned int len
)
2354 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2355 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
2358 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
2361 * pskb_trim_rcsum - trim received skb and update checksum
2362 * @skb: buffer to trim
2365 * This is exactly the same as pskb_trim except that it ensures the
2366 * checksum of received packets are still valid after the operation.
2369 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
2371 if (likely(len
>= skb
->len
))
2373 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2374 skb
->ip_summed
= CHECKSUM_NONE
;
2375 return __pskb_trim(skb
, len
);
2378 #define skb_queue_walk(queue, skb) \
2379 for (skb = (queue)->next; \
2380 skb != (struct sk_buff *)(queue); \
2383 #define skb_queue_walk_safe(queue, skb, tmp) \
2384 for (skb = (queue)->next, tmp = skb->next; \
2385 skb != (struct sk_buff *)(queue); \
2386 skb = tmp, tmp = skb->next)
2388 #define skb_queue_walk_from(queue, skb) \
2389 for (; skb != (struct sk_buff *)(queue); \
2392 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2393 for (tmp = skb->next; \
2394 skb != (struct sk_buff *)(queue); \
2395 skb = tmp, tmp = skb->next)
2397 #define skb_queue_reverse_walk(queue, skb) \
2398 for (skb = (queue)->prev; \
2399 skb != (struct sk_buff *)(queue); \
2402 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2403 for (skb = (queue)->prev, tmp = skb->prev; \
2404 skb != (struct sk_buff *)(queue); \
2405 skb = tmp, tmp = skb->prev)
2407 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2408 for (tmp = skb->prev; \
2409 skb != (struct sk_buff *)(queue); \
2410 skb = tmp, tmp = skb->prev)
2412 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
2414 return skb_shinfo(skb
)->frag_list
!= NULL
;
2417 static inline void skb_frag_list_init(struct sk_buff
*skb
)
2419 skb_shinfo(skb
)->frag_list
= NULL
;
2422 static inline void skb_frag_add_head(struct sk_buff
*skb
, struct sk_buff
*frag
)
2424 frag
->next
= skb_shinfo(skb
)->frag_list
;
2425 skb_shinfo(skb
)->frag_list
= frag
;
2428 #define skb_walk_frags(skb, iter) \
2429 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2431 extern struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
2432 int *peeked
, int *off
, int *err
);
2433 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
2434 int noblock
, int *err
);
2435 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
2436 struct poll_table_struct
*wait
);
2437 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
2438 int offset
, struct iovec
*to
,
2440 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
,
2443 extern int skb_copy_datagram_from_iovec(struct sk_buff
*skb
,
2445 const struct iovec
*from
,
2448 extern int skb_copy_datagram_const_iovec(const struct sk_buff
*from
,
2450 const struct iovec
*to
,
2453 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
2454 extern void skb_free_datagram_locked(struct sock
*sk
,
2455 struct sk_buff
*skb
);
2456 extern int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
,
2457 unsigned int flags
);
2458 extern __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
,
2459 int len
, __wsum csum
);
2460 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
2462 extern int skb_store_bits(struct sk_buff
*skb
, int offset
,
2463 const void *from
, int len
);
2464 extern __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
,
2465 int offset
, u8
*to
, int len
,
2467 extern int skb_splice_bits(struct sk_buff
*skb
,
2468 unsigned int offset
,
2469 struct pipe_inode_info
*pipe
,
2471 unsigned int flags
);
2472 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
2473 extern void skb_split(struct sk_buff
*skb
,
2474 struct sk_buff
*skb1
, const u32 len
);
2475 extern int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
,
2478 extern struct sk_buff
*skb_segment(struct sk_buff
*skb
,
2479 netdev_features_t features
);
2481 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2482 int len
, void *buffer
)
2484 int hlen
= skb_headlen(skb
);
2486 if (hlen
- offset
>= len
)
2487 return skb
->data
+ offset
;
2489 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
2495 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
2497 const unsigned int len
)
2499 memcpy(to
, skb
->data
, len
);
2502 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
2503 const int offset
, void *to
,
2504 const unsigned int len
)
2506 memcpy(to
, skb
->data
+ offset
, len
);
2509 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
2511 const unsigned int len
)
2513 memcpy(skb
->data
, from
, len
);
2516 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
2519 const unsigned int len
)
2521 memcpy(skb
->data
+ offset
, from
, len
);
2524 extern void skb_init(void);
2526 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
2532 * skb_get_timestamp - get timestamp from a skb
2533 * @skb: skb to get stamp from
2534 * @stamp: pointer to struct timeval to store stamp in
2536 * Timestamps are stored in the skb as offsets to a base timestamp.
2537 * This function converts the offset back to a struct timeval and stores
2540 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
2541 struct timeval
*stamp
)
2543 *stamp
= ktime_to_timeval(skb
->tstamp
);
2546 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
2547 struct timespec
*stamp
)
2549 *stamp
= ktime_to_timespec(skb
->tstamp
);
2552 static inline void __net_timestamp(struct sk_buff
*skb
)
2554 skb
->tstamp
= ktime_get_real();
2557 static inline ktime_t
net_timedelta(ktime_t t
)
2559 return ktime_sub(ktime_get_real(), t
);
2562 static inline ktime_t
net_invalid_timestamp(void)
2564 return ktime_set(0, 0);
2567 extern void skb_timestamping_init(void);
2569 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2571 extern void skb_clone_tx_timestamp(struct sk_buff
*skb
);
2572 extern bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
2574 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2576 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
2580 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
2585 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2588 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2590 * PHY drivers may accept clones of transmitted packets for
2591 * timestamping via their phy_driver.txtstamp method. These drivers
2592 * must call this function to return the skb back to the stack, with
2593 * or without a timestamp.
2595 * @skb: clone of the the original outgoing packet
2596 * @hwtstamps: hardware time stamps, may be NULL if not available
2599 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
2600 struct skb_shared_hwtstamps
*hwtstamps
);
2603 * skb_tstamp_tx - queue clone of skb with send time stamps
2604 * @orig_skb: the original outgoing packet
2605 * @hwtstamps: hardware time stamps, may be NULL if not available
2607 * If the skb has a socket associated, then this function clones the
2608 * skb (thus sharing the actual data and optional structures), stores
2609 * the optional hardware time stamping information (if non NULL) or
2610 * generates a software time stamp (otherwise), then queues the clone
2611 * to the error queue of the socket. Errors are silently ignored.
2613 extern void skb_tstamp_tx(struct sk_buff
*orig_skb
,
2614 struct skb_shared_hwtstamps
*hwtstamps
);
2616 static inline void sw_tx_timestamp(struct sk_buff
*skb
)
2618 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
&&
2619 !(skb_shinfo(skb
)->tx_flags
& SKBTX_IN_PROGRESS
))
2620 skb_tstamp_tx(skb
, NULL
);
2624 * skb_tx_timestamp() - Driver hook for transmit timestamping
2626 * Ethernet MAC Drivers should call this function in their hard_xmit()
2627 * function immediately before giving the sk_buff to the MAC hardware.
2629 * @skb: A socket buffer.
2631 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
2633 skb_clone_tx_timestamp(skb
);
2634 sw_tx_timestamp(skb
);
2638 * skb_complete_wifi_ack - deliver skb with wifi status
2640 * @skb: the original outgoing packet
2641 * @acked: ack status
2644 void skb_complete_wifi_ack(struct sk_buff
*skb
, bool acked
);
2646 extern __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
2647 extern __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
2649 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
2651 return skb
->ip_summed
& CHECKSUM_UNNECESSARY
;
2655 * skb_checksum_complete - Calculate checksum of an entire packet
2656 * @skb: packet to process
2658 * This function calculates the checksum over the entire packet plus
2659 * the value of skb->csum. The latter can be used to supply the
2660 * checksum of a pseudo header as used by TCP/UDP. It returns the
2663 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2664 * this function can be used to verify that checksum on received
2665 * packets. In that case the function should return zero if the
2666 * checksum is correct. In particular, this function will return zero
2667 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2668 * hardware has already verified the correctness of the checksum.
2670 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
2672 return skb_csum_unnecessary(skb
) ?
2673 0 : __skb_checksum_complete(skb
);
2676 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2677 extern void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
2678 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
2680 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
2681 nf_conntrack_destroy(nfct
);
2683 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
2686 atomic_inc(&nfct
->use
);
2689 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2690 static inline void nf_conntrack_get_reasm(struct sk_buff
*skb
)
2693 atomic_inc(&skb
->users
);
2695 static inline void nf_conntrack_put_reasm(struct sk_buff
*skb
)
2701 #ifdef CONFIG_BRIDGE_NETFILTER
2702 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
2704 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
2707 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
2710 atomic_inc(&nf_bridge
->use
);
2712 #endif /* CONFIG_BRIDGE_NETFILTER */
2713 static inline void nf_reset(struct sk_buff
*skb
)
2715 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2716 nf_conntrack_put(skb
->nfct
);
2719 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2720 nf_conntrack_put_reasm(skb
->nfct_reasm
);
2721 skb
->nfct_reasm
= NULL
;
2723 #ifdef CONFIG_BRIDGE_NETFILTER
2724 nf_bridge_put(skb
->nf_bridge
);
2725 skb
->nf_bridge
= NULL
;
2729 static inline void nf_reset_trace(struct sk_buff
*skb
)
2731 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
2736 /* Note: This doesn't put any conntrack and bridge info in dst. */
2737 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2739 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2740 dst
->nfct
= src
->nfct
;
2741 nf_conntrack_get(src
->nfct
);
2742 dst
->nfctinfo
= src
->nfctinfo
;
2744 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2745 dst
->nfct_reasm
= src
->nfct_reasm
;
2746 nf_conntrack_get_reasm(src
->nfct_reasm
);
2748 #ifdef CONFIG_BRIDGE_NETFILTER
2749 dst
->nf_bridge
= src
->nf_bridge
;
2750 nf_bridge_get(src
->nf_bridge
);
2754 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2756 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2757 nf_conntrack_put(dst
->nfct
);
2759 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2760 nf_conntrack_put_reasm(dst
->nfct_reasm
);
2762 #ifdef CONFIG_BRIDGE_NETFILTER
2763 nf_bridge_put(dst
->nf_bridge
);
2765 __nf_copy(dst
, src
);
2768 #ifdef CONFIG_NETWORK_SECMARK
2769 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2771 to
->secmark
= from
->secmark
;
2774 static inline void skb_init_secmark(struct sk_buff
*skb
)
2779 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2782 static inline void skb_init_secmark(struct sk_buff
*skb
)
2786 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
2788 skb
->queue_mapping
= queue_mapping
;
2791 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
2793 return skb
->queue_mapping
;
2796 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
2798 to
->queue_mapping
= from
->queue_mapping
;
2801 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
2803 skb
->queue_mapping
= rx_queue
+ 1;
2806 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
2808 return skb
->queue_mapping
- 1;
2811 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
2813 return skb
->queue_mapping
!= 0;
2816 extern u16
__skb_tx_hash(const struct net_device
*dev
,
2817 const struct sk_buff
*skb
,
2818 unsigned int num_tx_queues
);
2821 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2826 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2832 /* Keeps track of mac header offset relative to skb->head.
2833 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2834 * For non-tunnel skb it points to skb_mac_header() and for
2835 * tunnel skb it points to outer mac header. */
2839 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2841 static inline int skb_tnl_header_len(const struct sk_buff
*inner_skb
)
2843 return (skb_mac_header(inner_skb
) - inner_skb
->head
) -
2844 SKB_GSO_CB(inner_skb
)->mac_offset
;
2847 static inline bool skb_is_gso(const struct sk_buff
*skb
)
2849 return skb_shinfo(skb
)->gso_size
;
2852 static inline bool skb_is_gso_v6(const struct sk_buff
*skb
)
2854 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
2857 extern void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
2859 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
2861 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2862 * wanted then gso_type will be set. */
2863 const struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
2865 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
2866 unlikely(shinfo
->gso_type
== 0)) {
2867 __skb_warn_lro_forwarding(skb
);
2873 static inline void skb_forward_csum(struct sk_buff
*skb
)
2875 /* Unfortunately we don't support this one. Any brave souls? */
2876 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2877 skb
->ip_summed
= CHECKSUM_NONE
;
2881 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2882 * @skb: skb to check
2884 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2885 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2886 * use this helper, to document places where we make this assertion.
2888 static inline void skb_checksum_none_assert(const struct sk_buff
*skb
)
2891 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
2895 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
2897 u32
__skb_get_poff(const struct sk_buff
*skb
);
2900 * skb_head_is_locked - Determine if the skb->head is locked down
2901 * @skb: skb to check
2903 * The head on skbs build around a head frag can be removed if they are
2904 * not cloned. This function returns true if the skb head is locked down
2905 * due to either being allocated via kmalloc, or by being a clone with
2906 * multiple references to the head.
2908 static inline bool skb_head_is_locked(const struct sk_buff
*skb
)
2910 return !skb
->head_frag
|| skb_cloned(skb
);
2912 #endif /* __KERNEL__ */
2913 #endif /* _LINUX_SKBUFF_H */