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_proto: vlan encapsulation protocol
391 * @vlan_tci: vlan tag control information
392 * @inner_transport_header: Inner transport layer header (encapsulation)
393 * @inner_network_header: Network layer header (encapsulation)
394 * @inner_mac_header: Link layer header (encapsulation)
395 * @transport_header: Transport layer header
396 * @network_header: Network layer header
397 * @mac_header: Link layer header
398 * @tail: Tail pointer
400 * @head: Head of buffer
401 * @data: Data head pointer
402 * @truesize: Buffer size
403 * @users: User count - see {datagram,tcp}.c
407 /* These two members must be first. */
408 struct sk_buff
*next
;
409 struct sk_buff
*prev
;
414 struct net_device
*dev
;
417 * This is the control buffer. It is free to use for every
418 * layer. Please put your private variables there. If you
419 * want to keep them across layers you have to do a skb_clone()
420 * first. This is owned by whoever has the skb queued ATM.
422 char cb
[48] __aligned(8);
424 unsigned long _skb_refdst
;
440 kmemcheck_bitfield_begin(flags1
);
451 kmemcheck_bitfield_end(flags1
);
454 void (*destructor
)(struct sk_buff
*skb
);
455 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
456 struct nf_conntrack
*nfct
;
458 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
459 struct sk_buff
*nfct_reasm
;
461 #ifdef CONFIG_BRIDGE_NETFILTER
462 struct nf_bridge_info
*nf_bridge
;
472 #ifdef CONFIG_NET_SCHED
473 __u16 tc_index
; /* traffic control index */
474 #ifdef CONFIG_NET_CLS_ACT
475 __u16 tc_verd
; /* traffic control verdict */
480 kmemcheck_bitfield_begin(flags2
);
481 #ifdef CONFIG_IPV6_NDISC_NODETYPE
482 __u8 ndisc_nodetype
:2;
487 __u8 wifi_acked_valid
:1;
491 /* Encapsulation protocol and NIC drivers should use
492 * this flag to indicate to each other if the skb contains
493 * encapsulated packet or not and maybe use the inner packet
496 __u8 encapsulation
:1;
497 /* 7/9 bit hole (depending on ndisc_nodetype presence) */
498 kmemcheck_bitfield_end(flags2
);
500 #ifdef CONFIG_NET_DMA
501 dma_cookie_t dma_cookie
;
503 #ifdef CONFIG_NETWORK_SECMARK
509 __u32 reserved_tailroom
;
512 sk_buff_data_t inner_transport_header
;
513 sk_buff_data_t inner_network_header
;
514 sk_buff_data_t inner_mac_header
;
515 sk_buff_data_t transport_header
;
516 sk_buff_data_t network_header
;
517 sk_buff_data_t mac_header
;
518 /* These elements must be at the end, see alloc_skb() for details. */
523 unsigned int truesize
;
529 * Handling routines are only of interest to the kernel
531 #include <linux/slab.h>
534 #define SKB_ALLOC_FCLONE 0x01
535 #define SKB_ALLOC_RX 0x02
537 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
538 static inline bool skb_pfmemalloc(const struct sk_buff
*skb
)
540 return unlikely(skb
->pfmemalloc
);
544 * skb might have a dst pointer attached, refcounted or not.
545 * _skb_refdst low order bit is set if refcount was _not_ taken
547 #define SKB_DST_NOREF 1UL
548 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
551 * skb_dst - returns skb dst_entry
554 * Returns skb dst_entry, regardless of reference taken or not.
556 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
558 /* If refdst was not refcounted, check we still are in a
559 * rcu_read_lock section
561 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
562 !rcu_read_lock_held() &&
563 !rcu_read_lock_bh_held());
564 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
568 * skb_dst_set - sets skb dst
572 * Sets skb dst, assuming a reference was taken on dst and should
573 * be released by skb_dst_drop()
575 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
577 skb
->_skb_refdst
= (unsigned long)dst
;
580 extern void __skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
,
584 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
588 * Sets skb dst, assuming a reference was not taken on dst.
589 * If dst entry is cached, we do not take reference and dst_release
590 * will be avoided by refdst_drop. If dst entry is not cached, we take
591 * reference, so that last dst_release can destroy the dst immediately.
593 static inline void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
)
595 __skb_dst_set_noref(skb
, dst
, false);
599 * skb_dst_set_noref_force - sets skb dst, without taking reference
603 * Sets skb dst, assuming a reference was not taken on dst.
604 * No reference is taken and no dst_release will be called. While for
605 * cached dsts deferred reclaim is a basic feature, for entries that are
606 * not cached it is caller's job to guarantee that last dst_release for
607 * provided dst happens when nobody uses it, eg. after a RCU grace period.
609 static inline void skb_dst_set_noref_force(struct sk_buff
*skb
,
610 struct dst_entry
*dst
)
612 __skb_dst_set_noref(skb
, dst
, true);
616 * skb_dst_is_noref - Test if skb dst isn't refcounted
619 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
621 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
624 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
626 return (struct rtable
*)skb_dst(skb
);
629 extern void kfree_skb(struct sk_buff
*skb
);
630 extern void kfree_skb_list(struct sk_buff
*segs
);
631 extern void skb_tx_error(struct sk_buff
*skb
);
632 extern void consume_skb(struct sk_buff
*skb
);
633 extern void __kfree_skb(struct sk_buff
*skb
);
634 extern struct kmem_cache
*skbuff_head_cache
;
636 extern void kfree_skb_partial(struct sk_buff
*skb
, bool head_stolen
);
637 extern bool skb_try_coalesce(struct sk_buff
*to
, struct sk_buff
*from
,
638 bool *fragstolen
, int *delta_truesize
);
640 extern struct sk_buff
*__alloc_skb(unsigned int size
,
641 gfp_t priority
, int flags
, int node
);
642 extern struct sk_buff
*build_skb(void *data
, unsigned int frag_size
);
643 static inline struct sk_buff
*alloc_skb(unsigned int size
,
646 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
649 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
652 return __alloc_skb(size
, priority
, SKB_ALLOC_FCLONE
, NUMA_NO_NODE
);
655 extern struct sk_buff
*__alloc_skb_head(gfp_t priority
, int node
);
656 static inline struct sk_buff
*alloc_skb_head(gfp_t priority
)
658 return __alloc_skb_head(priority
, -1);
661 extern struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
662 extern int skb_copy_ubufs(struct sk_buff
*skb
, gfp_t gfp_mask
);
663 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
665 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
667 extern struct sk_buff
*__pskb_copy(struct sk_buff
*skb
,
668 int headroom
, gfp_t gfp_mask
);
670 extern int pskb_expand_head(struct sk_buff
*skb
,
671 int nhead
, int ntail
,
673 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
674 unsigned int headroom
);
675 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
676 int newheadroom
, int newtailroom
,
678 extern int skb_to_sgvec(struct sk_buff
*skb
,
679 struct scatterlist
*sg
, int offset
,
681 extern int skb_cow_data(struct sk_buff
*skb
, int tailbits
,
682 struct sk_buff
**trailer
);
683 extern int skb_pad(struct sk_buff
*skb
, int pad
);
684 #define dev_kfree_skb(a) consume_skb(a)
686 extern int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
687 int getfrag(void *from
, char *to
, int offset
,
688 int len
,int odd
, struct sk_buff
*skb
),
689 void *from
, int length
);
691 struct skb_seq_state
{
695 __u32 stepped_offset
;
696 struct sk_buff
*root_skb
;
697 struct sk_buff
*cur_skb
;
701 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
702 unsigned int from
, unsigned int to
,
703 struct skb_seq_state
*st
);
704 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
705 struct skb_seq_state
*st
);
706 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
708 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
709 unsigned int to
, struct ts_config
*config
,
710 struct ts_state
*state
);
712 extern void __skb_get_rxhash(struct sk_buff
*skb
);
713 static inline __u32
skb_get_rxhash(struct sk_buff
*skb
)
716 __skb_get_rxhash(skb
);
721 #ifdef NET_SKBUFF_DATA_USES_OFFSET
722 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
724 return skb
->head
+ skb
->end
;
727 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
732 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
737 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
739 return skb
->end
- skb
->head
;
744 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
746 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
748 return &skb_shinfo(skb
)->hwtstamps
;
752 * skb_queue_empty - check if a queue is empty
755 * Returns true if the queue is empty, false otherwise.
757 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
759 return list
->next
== (struct sk_buff
*)list
;
763 * skb_queue_is_last - check if skb is the last entry in the queue
767 * Returns true if @skb is the last buffer on the list.
769 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
770 const struct sk_buff
*skb
)
772 return skb
->next
== (struct sk_buff
*)list
;
776 * skb_queue_is_first - check if skb is the first entry in the queue
780 * Returns true if @skb is the first buffer on the list.
782 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
783 const struct sk_buff
*skb
)
785 return skb
->prev
== (struct sk_buff
*)list
;
789 * skb_queue_next - return the next packet in the queue
791 * @skb: current buffer
793 * Return the next packet in @list after @skb. It is only valid to
794 * call this if skb_queue_is_last() evaluates to false.
796 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
797 const struct sk_buff
*skb
)
799 /* This BUG_ON may seem severe, but if we just return then we
800 * are going to dereference garbage.
802 BUG_ON(skb_queue_is_last(list
, skb
));
807 * skb_queue_prev - return the prev packet in the queue
809 * @skb: current buffer
811 * Return the prev packet in @list before @skb. It is only valid to
812 * call this if skb_queue_is_first() evaluates to false.
814 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
815 const struct sk_buff
*skb
)
817 /* This BUG_ON may seem severe, but if we just return then we
818 * are going to dereference garbage.
820 BUG_ON(skb_queue_is_first(list
, skb
));
825 * skb_get - reference buffer
826 * @skb: buffer to reference
828 * Makes another reference to a socket buffer and returns a pointer
831 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
833 atomic_inc(&skb
->users
);
838 * If users == 1, we are the only owner and are can avoid redundant
843 * skb_cloned - is the buffer a clone
844 * @skb: buffer to check
846 * Returns true if the buffer was generated with skb_clone() and is
847 * one of multiple shared copies of the buffer. Cloned buffers are
848 * shared data so must not be written to under normal circumstances.
850 static inline int skb_cloned(const struct sk_buff
*skb
)
852 return skb
->cloned
&&
853 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
856 static inline int skb_unclone(struct sk_buff
*skb
, gfp_t pri
)
858 might_sleep_if(pri
& __GFP_WAIT
);
861 return pskb_expand_head(skb
, 0, 0, pri
);
867 * skb_header_cloned - is the header a clone
868 * @skb: buffer to check
870 * Returns true if modifying the header part of the buffer requires
871 * the data to be copied.
873 static inline int skb_header_cloned(const struct sk_buff
*skb
)
880 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
881 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
886 * skb_header_release - release reference to header
887 * @skb: buffer to operate on
889 * Drop a reference to the header part of the buffer. This is done
890 * by acquiring a payload reference. You must not read from the header
891 * part of skb->data after this.
893 static inline void skb_header_release(struct sk_buff
*skb
)
897 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
901 * skb_shared - is the buffer shared
902 * @skb: buffer to check
904 * Returns true if more than one person has a reference to this
907 static inline int skb_shared(const struct sk_buff
*skb
)
909 return atomic_read(&skb
->users
) != 1;
913 * skb_share_check - check if buffer is shared and if so clone it
914 * @skb: buffer to check
915 * @pri: priority for memory allocation
917 * If the buffer is shared the buffer is cloned and the old copy
918 * drops a reference. A new clone with a single reference is returned.
919 * If the buffer is not shared the original buffer is returned. When
920 * being called from interrupt status or with spinlocks held pri must
923 * NULL is returned on a memory allocation failure.
925 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
, gfp_t pri
)
927 might_sleep_if(pri
& __GFP_WAIT
);
928 if (skb_shared(skb
)) {
929 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
941 * Copy shared buffers into a new sk_buff. We effectively do COW on
942 * packets to handle cases where we have a local reader and forward
943 * and a couple of other messy ones. The normal one is tcpdumping
944 * a packet thats being forwarded.
948 * skb_unshare - make a copy of a shared buffer
949 * @skb: buffer to check
950 * @pri: priority for memory allocation
952 * If the socket buffer is a clone then this function creates a new
953 * copy of the data, drops a reference count on the old copy and returns
954 * the new copy with the reference count at 1. If the buffer is not a clone
955 * the original buffer is returned. When called with a spinlock held or
956 * from interrupt state @pri must be %GFP_ATOMIC
958 * %NULL is returned on a memory allocation failure.
960 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
963 might_sleep_if(pri
& __GFP_WAIT
);
964 if (skb_cloned(skb
)) {
965 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
966 kfree_skb(skb
); /* Free our shared copy */
973 * skb_peek - peek at the head of an &sk_buff_head
974 * @list_: list to peek at
976 * Peek an &sk_buff. Unlike most other operations you _MUST_
977 * be careful with this one. A peek leaves the buffer on the
978 * list and someone else may run off with it. You must hold
979 * the appropriate locks or have a private queue to do this.
981 * Returns %NULL for an empty list or a pointer to the head element.
982 * The reference count is not incremented and the reference is therefore
983 * volatile. Use with caution.
985 static inline struct sk_buff
*skb_peek(const struct sk_buff_head
*list_
)
987 struct sk_buff
*skb
= list_
->next
;
989 if (skb
== (struct sk_buff
*)list_
)
995 * skb_peek_next - peek skb following the given one from a queue
996 * @skb: skb to start from
997 * @list_: list to peek at
999 * Returns %NULL when the end of the list is met or a pointer to the
1000 * next element. The reference count is not incremented and the
1001 * reference is therefore volatile. Use with caution.
1003 static inline struct sk_buff
*skb_peek_next(struct sk_buff
*skb
,
1004 const struct sk_buff_head
*list_
)
1006 struct sk_buff
*next
= skb
->next
;
1008 if (next
== (struct sk_buff
*)list_
)
1014 * skb_peek_tail - peek at the tail of an &sk_buff_head
1015 * @list_: list to peek at
1017 * Peek an &sk_buff. Unlike most other operations you _MUST_
1018 * be careful with this one. A peek leaves the buffer on the
1019 * list and someone else may run off with it. You must hold
1020 * the appropriate locks or have a private queue to do this.
1022 * Returns %NULL for an empty list or a pointer to the tail element.
1023 * The reference count is not incremented and the reference is therefore
1024 * volatile. Use with caution.
1026 static inline struct sk_buff
*skb_peek_tail(const struct sk_buff_head
*list_
)
1028 struct sk_buff
*skb
= list_
->prev
;
1030 if (skb
== (struct sk_buff
*)list_
)
1037 * skb_queue_len - get queue length
1038 * @list_: list to measure
1040 * Return the length of an &sk_buff queue.
1042 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
1048 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1049 * @list: queue to initialize
1051 * This initializes only the list and queue length aspects of
1052 * an sk_buff_head object. This allows to initialize the list
1053 * aspects of an sk_buff_head without reinitializing things like
1054 * the spinlock. It can also be used for on-stack sk_buff_head
1055 * objects where the spinlock is known to not be used.
1057 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
1059 list
->prev
= list
->next
= (struct sk_buff
*)list
;
1064 * This function creates a split out lock class for each invocation;
1065 * this is needed for now since a whole lot of users of the skb-queue
1066 * infrastructure in drivers have different locking usage (in hardirq)
1067 * than the networking core (in softirq only). In the long run either the
1068 * network layer or drivers should need annotation to consolidate the
1069 * main types of usage into 3 classes.
1071 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
1073 spin_lock_init(&list
->lock
);
1074 __skb_queue_head_init(list
);
1077 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
1078 struct lock_class_key
*class)
1080 skb_queue_head_init(list
);
1081 lockdep_set_class(&list
->lock
, class);
1085 * Insert an sk_buff on a list.
1087 * The "__skb_xxxx()" functions are the non-atomic ones that
1088 * can only be called with interrupts disabled.
1090 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
1091 static inline void __skb_insert(struct sk_buff
*newsk
,
1092 struct sk_buff
*prev
, struct sk_buff
*next
,
1093 struct sk_buff_head
*list
)
1097 next
->prev
= prev
->next
= newsk
;
1101 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
1102 struct sk_buff
*prev
,
1103 struct sk_buff
*next
)
1105 struct sk_buff
*first
= list
->next
;
1106 struct sk_buff
*last
= list
->prev
;
1116 * skb_queue_splice - join two skb lists, this is designed for stacks
1117 * @list: the new list to add
1118 * @head: the place to add it in the first list
1120 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
1121 struct sk_buff_head
*head
)
1123 if (!skb_queue_empty(list
)) {
1124 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1125 head
->qlen
+= list
->qlen
;
1130 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1131 * @list: the new list to add
1132 * @head: the place to add it in the first list
1134 * The list at @list is reinitialised
1136 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
1137 struct sk_buff_head
*head
)
1139 if (!skb_queue_empty(list
)) {
1140 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1141 head
->qlen
+= list
->qlen
;
1142 __skb_queue_head_init(list
);
1147 * skb_queue_splice_tail - join two skb lists, each list being a queue
1148 * @list: the new list to add
1149 * @head: the place to add it in the first list
1151 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
1152 struct sk_buff_head
*head
)
1154 if (!skb_queue_empty(list
)) {
1155 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1156 head
->qlen
+= list
->qlen
;
1161 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1162 * @list: the new list to add
1163 * @head: the place to add it in the first list
1165 * Each of the lists is a queue.
1166 * The list at @list is reinitialised
1168 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
1169 struct sk_buff_head
*head
)
1171 if (!skb_queue_empty(list
)) {
1172 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1173 head
->qlen
+= list
->qlen
;
1174 __skb_queue_head_init(list
);
1179 * __skb_queue_after - queue a buffer at the list head
1180 * @list: list to use
1181 * @prev: place after this buffer
1182 * @newsk: buffer to queue
1184 * Queue a buffer int the middle of a list. This function takes no locks
1185 * and you must therefore hold required locks before calling it.
1187 * A buffer cannot be placed on two lists at the same time.
1189 static inline void __skb_queue_after(struct sk_buff_head
*list
,
1190 struct sk_buff
*prev
,
1191 struct sk_buff
*newsk
)
1193 __skb_insert(newsk
, prev
, prev
->next
, list
);
1196 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
1197 struct sk_buff_head
*list
);
1199 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1200 struct sk_buff
*next
,
1201 struct sk_buff
*newsk
)
1203 __skb_insert(newsk
, next
->prev
, next
, list
);
1207 * __skb_queue_head - queue a buffer at the list head
1208 * @list: list to use
1209 * @newsk: buffer to queue
1211 * Queue a buffer at the start of a list. This function takes no locks
1212 * and you must therefore hold required locks before calling it.
1214 * A buffer cannot be placed on two lists at the same time.
1216 extern void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1217 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1218 struct sk_buff
*newsk
)
1220 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1224 * __skb_queue_tail - queue a buffer at the list tail
1225 * @list: list to use
1226 * @newsk: buffer to queue
1228 * Queue a buffer at the end of a list. This function takes no locks
1229 * and you must therefore hold required locks before calling it.
1231 * A buffer cannot be placed on two lists at the same time.
1233 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1234 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1235 struct sk_buff
*newsk
)
1237 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1241 * remove sk_buff from list. _Must_ be called atomically, and with
1244 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1245 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1247 struct sk_buff
*next
, *prev
;
1252 skb
->next
= skb
->prev
= NULL
;
1258 * __skb_dequeue - remove from the head of the queue
1259 * @list: list to dequeue from
1261 * Remove the head of the list. This function does not take any locks
1262 * so must be used with appropriate locks held only. The head item is
1263 * returned or %NULL if the list is empty.
1265 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1266 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1268 struct sk_buff
*skb
= skb_peek(list
);
1270 __skb_unlink(skb
, list
);
1275 * __skb_dequeue_tail - remove from the tail of the queue
1276 * @list: list to dequeue from
1278 * Remove the tail of the list. This function does not take any locks
1279 * so must be used with appropriate locks held only. The tail item is
1280 * returned or %NULL if the list is empty.
1282 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1283 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1285 struct sk_buff
*skb
= skb_peek_tail(list
);
1287 __skb_unlink(skb
, list
);
1292 static inline bool skb_is_nonlinear(const struct sk_buff
*skb
)
1294 return skb
->data_len
;
1297 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1299 return skb
->len
- skb
->data_len
;
1302 static inline int skb_pagelen(const struct sk_buff
*skb
)
1306 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1307 len
+= skb_frag_size(&skb_shinfo(skb
)->frags
[i
]);
1308 return len
+ skb_headlen(skb
);
1312 * __skb_fill_page_desc - initialise a paged fragment in an skb
1313 * @skb: buffer containing fragment to be initialised
1314 * @i: paged fragment index to initialise
1315 * @page: the page to use for this fragment
1316 * @off: the offset to the data with @page
1317 * @size: the length of the data
1319 * Initialises the @i'th fragment of @skb to point to &size bytes at
1320 * offset @off within @page.
1322 * Does not take any additional reference on the fragment.
1324 static inline void __skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1325 struct page
*page
, int off
, int size
)
1327 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1330 * Propagate page->pfmemalloc to the skb if we can. The problem is
1331 * that not all callers have unique ownership of the page. If
1332 * pfmemalloc is set, we check the mapping as a mapping implies
1333 * page->index is set (index and pfmemalloc share space).
1334 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1335 * do not lose pfmemalloc information as the pages would not be
1336 * allocated using __GFP_MEMALLOC.
1338 frag
->page
.p
= page
;
1339 frag
->page_offset
= off
;
1340 skb_frag_size_set(frag
, size
);
1342 page
= compound_head(page
);
1343 if (page
->pfmemalloc
&& !page
->mapping
)
1344 skb
->pfmemalloc
= true;
1348 * skb_fill_page_desc - initialise a paged fragment in an skb
1349 * @skb: buffer containing fragment to be initialised
1350 * @i: paged fragment index to initialise
1351 * @page: the page to use for this fragment
1352 * @off: the offset to the data with @page
1353 * @size: the length of the data
1355 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1356 * @skb to point to &size bytes at offset @off within @page. In
1357 * addition updates @skb such that @i is the last fragment.
1359 * Does not take any additional reference on the fragment.
1361 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1362 struct page
*page
, int off
, int size
)
1364 __skb_fill_page_desc(skb
, i
, page
, off
, size
);
1365 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1368 extern void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
,
1369 int off
, int size
, unsigned int truesize
);
1371 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1372 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1373 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1375 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1376 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1378 return skb
->head
+ skb
->tail
;
1381 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1383 skb
->tail
= skb
->data
- skb
->head
;
1386 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1388 skb_reset_tail_pointer(skb
);
1389 skb
->tail
+= offset
;
1391 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1392 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1397 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1399 skb
->tail
= skb
->data
;
1402 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1404 skb
->tail
= skb
->data
+ offset
;
1407 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1410 * Add data to an sk_buff
1412 extern unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1413 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1415 unsigned char *tmp
= skb_tail_pointer(skb
);
1416 SKB_LINEAR_ASSERT(skb
);
1422 extern unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1423 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1430 extern unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1431 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1434 BUG_ON(skb
->len
< skb
->data_len
);
1435 return skb
->data
+= len
;
1438 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1440 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1443 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1445 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1447 if (len
> skb_headlen(skb
) &&
1448 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1451 return skb
->data
+= len
;
1454 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1456 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1459 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1461 if (likely(len
<= skb_headlen(skb
)))
1463 if (unlikely(len
> skb
->len
))
1465 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1469 * skb_headroom - bytes at buffer head
1470 * @skb: buffer to check
1472 * Return the number of bytes of free space at the head of an &sk_buff.
1474 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1476 return skb
->data
- skb
->head
;
1480 * skb_tailroom - bytes at buffer end
1481 * @skb: buffer to check
1483 * Return the number of bytes of free space at the tail of an sk_buff
1485 static inline int skb_tailroom(const struct sk_buff
*skb
)
1487 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1491 * skb_availroom - bytes at buffer end
1492 * @skb: buffer to check
1494 * Return the number of bytes of free space at the tail of an sk_buff
1495 * allocated by sk_stream_alloc()
1497 static inline int skb_availroom(const struct sk_buff
*skb
)
1499 if (skb_is_nonlinear(skb
))
1502 return skb
->end
- skb
->tail
- skb
->reserved_tailroom
;
1506 * skb_reserve - adjust headroom
1507 * @skb: buffer to alter
1508 * @len: bytes to move
1510 * Increase the headroom of an empty &sk_buff by reducing the tail
1511 * room. This is only allowed for an empty buffer.
1513 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
1519 static inline void skb_reset_inner_headers(struct sk_buff
*skb
)
1521 skb
->inner_mac_header
= skb
->mac_header
;
1522 skb
->inner_network_header
= skb
->network_header
;
1523 skb
->inner_transport_header
= skb
->transport_header
;
1526 static inline void skb_reset_mac_len(struct sk_buff
*skb
)
1528 skb
->mac_len
= skb
->network_header
- skb
->mac_header
;
1531 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1532 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1535 return skb
->head
+ skb
->inner_transport_header
;
1538 static inline void skb_reset_inner_transport_header(struct sk_buff
*skb
)
1540 skb
->inner_transport_header
= skb
->data
- skb
->head
;
1543 static inline void skb_set_inner_transport_header(struct sk_buff
*skb
,
1546 skb_reset_inner_transport_header(skb
);
1547 skb
->inner_transport_header
+= offset
;
1550 static inline unsigned char *skb_inner_network_header(const struct sk_buff
*skb
)
1552 return skb
->head
+ skb
->inner_network_header
;
1555 static inline void skb_reset_inner_network_header(struct sk_buff
*skb
)
1557 skb
->inner_network_header
= skb
->data
- skb
->head
;
1560 static inline void skb_set_inner_network_header(struct sk_buff
*skb
,
1563 skb_reset_inner_network_header(skb
);
1564 skb
->inner_network_header
+= offset
;
1567 static inline unsigned char *skb_inner_mac_header(const struct sk_buff
*skb
)
1569 return skb
->head
+ skb
->inner_mac_header
;
1572 static inline void skb_reset_inner_mac_header(struct sk_buff
*skb
)
1574 skb
->inner_mac_header
= skb
->data
- skb
->head
;
1577 static inline void skb_set_inner_mac_header(struct sk_buff
*skb
,
1580 skb_reset_inner_mac_header(skb
);
1581 skb
->inner_mac_header
+= offset
;
1583 static inline bool skb_transport_header_was_set(const struct sk_buff
*skb
)
1585 return skb
->transport_header
!= ~0U;
1588 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1590 return skb
->head
+ skb
->transport_header
;
1593 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1595 skb
->transport_header
= skb
->data
- skb
->head
;
1598 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1601 skb_reset_transport_header(skb
);
1602 skb
->transport_header
+= offset
;
1605 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1607 return skb
->head
+ skb
->network_header
;
1610 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1612 skb
->network_header
= skb
->data
- skb
->head
;
1615 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1617 skb_reset_network_header(skb
);
1618 skb
->network_header
+= offset
;
1621 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1623 return skb
->head
+ skb
->mac_header
;
1626 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1628 return skb
->mac_header
!= ~0U;
1631 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1633 skb
->mac_header
= skb
->data
- skb
->head
;
1636 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1638 skb_reset_mac_header(skb
);
1639 skb
->mac_header
+= offset
;
1642 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1643 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1646 return skb
->inner_transport_header
;
1649 static inline void skb_reset_inner_transport_header(struct sk_buff
*skb
)
1651 skb
->inner_transport_header
= skb
->data
;
1654 static inline void skb_set_inner_transport_header(struct sk_buff
*skb
,
1657 skb
->inner_transport_header
= skb
->data
+ offset
;
1660 static inline unsigned char *skb_inner_network_header(const struct sk_buff
*skb
)
1662 return skb
->inner_network_header
;
1665 static inline void skb_reset_inner_network_header(struct sk_buff
*skb
)
1667 skb
->inner_network_header
= skb
->data
;
1670 static inline void skb_set_inner_network_header(struct sk_buff
*skb
,
1673 skb
->inner_network_header
= skb
->data
+ offset
;
1676 static inline unsigned char *skb_inner_mac_header(const struct sk_buff
*skb
)
1678 return skb
->inner_mac_header
;
1681 static inline void skb_reset_inner_mac_header(struct sk_buff
*skb
)
1683 skb
->inner_mac_header
= skb
->data
;
1686 static inline void skb_set_inner_mac_header(struct sk_buff
*skb
,
1689 skb
->inner_mac_header
= skb
->data
+ offset
;
1691 static inline bool skb_transport_header_was_set(const struct sk_buff
*skb
)
1693 return skb
->transport_header
!= NULL
;
1696 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1698 return skb
->transport_header
;
1701 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1703 skb
->transport_header
= skb
->data
;
1706 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1709 skb
->transport_header
= skb
->data
+ offset
;
1712 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1714 return skb
->network_header
;
1717 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1719 skb
->network_header
= skb
->data
;
1722 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1724 skb
->network_header
= skb
->data
+ offset
;
1727 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1729 return skb
->mac_header
;
1732 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1734 return skb
->mac_header
!= NULL
;
1737 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1739 skb
->mac_header
= skb
->data
;
1742 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1744 skb
->mac_header
= skb
->data
+ offset
;
1746 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1748 static inline void skb_probe_transport_header(struct sk_buff
*skb
,
1749 const int offset_hint
)
1751 struct flow_keys keys
;
1753 if (skb_transport_header_was_set(skb
))
1755 else if (skb_flow_dissect(skb
, &keys
))
1756 skb_set_transport_header(skb
, keys
.thoff
);
1758 skb_set_transport_header(skb
, offset_hint
);
1761 static inline void skb_mac_header_rebuild(struct sk_buff
*skb
)
1763 if (skb_mac_header_was_set(skb
)) {
1764 const unsigned char *old_mac
= skb_mac_header(skb
);
1766 skb_set_mac_header(skb
, -skb
->mac_len
);
1767 memmove(skb_mac_header(skb
), old_mac
, skb
->mac_len
);
1771 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
1773 return skb
->csum_start
- skb_headroom(skb
);
1776 static inline int skb_transport_offset(const struct sk_buff
*skb
)
1778 return skb_transport_header(skb
) - skb
->data
;
1781 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
1783 return skb
->transport_header
- skb
->network_header
;
1786 static inline u32
skb_inner_network_header_len(const struct sk_buff
*skb
)
1788 return skb
->inner_transport_header
- skb
->inner_network_header
;
1791 static inline int skb_network_offset(const struct sk_buff
*skb
)
1793 return skb_network_header(skb
) - skb
->data
;
1796 static inline int skb_inner_network_offset(const struct sk_buff
*skb
)
1798 return skb_inner_network_header(skb
) - skb
->data
;
1801 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
1803 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
1807 * CPUs often take a performance hit when accessing unaligned memory
1808 * locations. The actual performance hit varies, it can be small if the
1809 * hardware handles it or large if we have to take an exception and fix it
1812 * Since an ethernet header is 14 bytes network drivers often end up with
1813 * the IP header at an unaligned offset. The IP header can be aligned by
1814 * shifting the start of the packet by 2 bytes. Drivers should do this
1817 * skb_reserve(skb, NET_IP_ALIGN);
1819 * The downside to this alignment of the IP header is that the DMA is now
1820 * unaligned. On some architectures the cost of an unaligned DMA is high
1821 * and this cost outweighs the gains made by aligning the IP header.
1823 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1826 #ifndef NET_IP_ALIGN
1827 #define NET_IP_ALIGN 2
1831 * The networking layer reserves some headroom in skb data (via
1832 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1833 * the header has to grow. In the default case, if the header has to grow
1834 * 32 bytes or less we avoid the reallocation.
1836 * Unfortunately this headroom changes the DMA alignment of the resulting
1837 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1838 * on some architectures. An architecture can override this value,
1839 * perhaps setting it to a cacheline in size (since that will maintain
1840 * cacheline alignment of the DMA). It must be a power of 2.
1842 * Various parts of the networking layer expect at least 32 bytes of
1843 * headroom, you should not reduce this.
1845 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1846 * to reduce average number of cache lines per packet.
1847 * get_rps_cpus() for example only access one 64 bytes aligned block :
1848 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1851 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1854 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
1856 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
1858 if (unlikely(skb_is_nonlinear(skb
))) {
1863 skb_set_tail_pointer(skb
, len
);
1866 extern void skb_trim(struct sk_buff
*skb
, unsigned int len
);
1868 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1871 return ___pskb_trim(skb
, len
);
1872 __skb_trim(skb
, len
);
1876 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1878 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
1882 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1883 * @skb: buffer to alter
1886 * This is identical to pskb_trim except that the caller knows that
1887 * the skb is not cloned so we should never get an error due to out-
1890 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
1892 int err
= pskb_trim(skb
, len
);
1897 * skb_orphan - orphan a buffer
1898 * @skb: buffer to orphan
1900 * If a buffer currently has an owner then we call the owner's
1901 * destructor function and make the @skb unowned. The buffer continues
1902 * to exist but is no longer charged to its former owner.
1904 static inline void skb_orphan(struct sk_buff
*skb
)
1906 if (skb
->destructor
)
1907 skb
->destructor(skb
);
1908 skb
->destructor
= NULL
;
1913 * skb_orphan_frags - orphan the frags contained in a buffer
1914 * @skb: buffer to orphan frags from
1915 * @gfp_mask: allocation mask for replacement pages
1917 * For each frag in the SKB which needs a destructor (i.e. has an
1918 * owner) create a copy of that frag and release the original
1919 * page by calling the destructor.
1921 static inline int skb_orphan_frags(struct sk_buff
*skb
, gfp_t gfp_mask
)
1923 if (likely(!(skb_shinfo(skb
)->tx_flags
& SKBTX_DEV_ZEROCOPY
)))
1925 return skb_copy_ubufs(skb
, gfp_mask
);
1929 * __skb_queue_purge - empty a list
1930 * @list: list to empty
1932 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1933 * the list and one reference dropped. This function does not take the
1934 * list lock and the caller must hold the relevant locks to use it.
1936 extern void skb_queue_purge(struct sk_buff_head
*list
);
1937 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
1939 struct sk_buff
*skb
;
1940 while ((skb
= __skb_dequeue(list
)) != NULL
)
1944 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1945 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1946 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1948 extern void *netdev_alloc_frag(unsigned int fragsz
);
1950 extern struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
1951 unsigned int length
,
1955 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1956 * @dev: network device to receive on
1957 * @length: length to allocate
1959 * Allocate a new &sk_buff and assign it a usage count of one. The
1960 * buffer has unspecified headroom built in. Users should allocate
1961 * the headroom they think they need without accounting for the
1962 * built in space. The built in space is used for optimisations.
1964 * %NULL is returned if there is no free memory. Although this function
1965 * allocates memory it can be called from an interrupt.
1967 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
1968 unsigned int length
)
1970 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
1973 /* legacy helper around __netdev_alloc_skb() */
1974 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
1977 return __netdev_alloc_skb(NULL
, length
, gfp_mask
);
1980 /* legacy helper around netdev_alloc_skb() */
1981 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
1983 return netdev_alloc_skb(NULL
, length
);
1987 static inline struct sk_buff
*__netdev_alloc_skb_ip_align(struct net_device
*dev
,
1988 unsigned int length
, gfp_t gfp
)
1990 struct sk_buff
*skb
= __netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
, gfp
);
1992 if (NET_IP_ALIGN
&& skb
)
1993 skb_reserve(skb
, NET_IP_ALIGN
);
1997 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
1998 unsigned int length
)
2000 return __netdev_alloc_skb_ip_align(dev
, length
, GFP_ATOMIC
);
2004 * __skb_alloc_page - allocate pages for ps-rx on a skb and preserve pfmemalloc data
2005 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2006 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2007 * @order: size of the allocation
2009 * Allocate a new page.
2011 * %NULL is returned if there is no free memory.
2013 static inline struct page
*__skb_alloc_pages(gfp_t gfp_mask
,
2014 struct sk_buff
*skb
,
2019 gfp_mask
|= __GFP_COLD
;
2021 if (!(gfp_mask
& __GFP_NOMEMALLOC
))
2022 gfp_mask
|= __GFP_MEMALLOC
;
2024 page
= alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, order
);
2025 if (skb
&& page
&& page
->pfmemalloc
)
2026 skb
->pfmemalloc
= true;
2032 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2033 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2034 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2036 * Allocate a new page.
2038 * %NULL is returned if there is no free memory.
2040 static inline struct page
*__skb_alloc_page(gfp_t gfp_mask
,
2041 struct sk_buff
*skb
)
2043 return __skb_alloc_pages(gfp_mask
, skb
, 0);
2047 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2048 * @page: The page that was allocated from skb_alloc_page
2049 * @skb: The skb that may need pfmemalloc set
2051 static inline void skb_propagate_pfmemalloc(struct page
*page
,
2052 struct sk_buff
*skb
)
2054 if (page
&& page
->pfmemalloc
)
2055 skb
->pfmemalloc
= true;
2059 * skb_frag_page - retrieve the page refered to by a paged fragment
2060 * @frag: the paged fragment
2062 * Returns the &struct page associated with @frag.
2064 static inline struct page
*skb_frag_page(const skb_frag_t
*frag
)
2066 return frag
->page
.p
;
2070 * __skb_frag_ref - take an addition reference on a paged fragment.
2071 * @frag: the paged fragment
2073 * Takes an additional reference on the paged fragment @frag.
2075 static inline void __skb_frag_ref(skb_frag_t
*frag
)
2077 get_page(skb_frag_page(frag
));
2081 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2083 * @f: the fragment offset.
2085 * Takes an additional reference on the @f'th paged fragment of @skb.
2087 static inline void skb_frag_ref(struct sk_buff
*skb
, int f
)
2089 __skb_frag_ref(&skb_shinfo(skb
)->frags
[f
]);
2093 * __skb_frag_unref - release a reference on a paged fragment.
2094 * @frag: the paged fragment
2096 * Releases a reference on the paged fragment @frag.
2098 static inline void __skb_frag_unref(skb_frag_t
*frag
)
2100 put_page(skb_frag_page(frag
));
2104 * skb_frag_unref - release a reference on a paged fragment of an skb.
2106 * @f: the fragment offset
2108 * Releases a reference on the @f'th paged fragment of @skb.
2110 static inline void skb_frag_unref(struct sk_buff
*skb
, int f
)
2112 __skb_frag_unref(&skb_shinfo(skb
)->frags
[f
]);
2116 * skb_frag_address - gets the address of the data contained in a paged fragment
2117 * @frag: the paged fragment buffer
2119 * Returns the address of the data within @frag. The page must already
2122 static inline void *skb_frag_address(const skb_frag_t
*frag
)
2124 return page_address(skb_frag_page(frag
)) + frag
->page_offset
;
2128 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2129 * @frag: the paged fragment buffer
2131 * Returns the address of the data within @frag. Checks that the page
2132 * is mapped and returns %NULL otherwise.
2134 static inline void *skb_frag_address_safe(const skb_frag_t
*frag
)
2136 void *ptr
= page_address(skb_frag_page(frag
));
2140 return ptr
+ frag
->page_offset
;
2144 * __skb_frag_set_page - sets the page contained in a paged fragment
2145 * @frag: the paged fragment
2146 * @page: the page to set
2148 * Sets the fragment @frag to contain @page.
2150 static inline void __skb_frag_set_page(skb_frag_t
*frag
, struct page
*page
)
2152 frag
->page
.p
= page
;
2156 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2158 * @f: the fragment offset
2159 * @page: the page to set
2161 * Sets the @f'th fragment of @skb to contain @page.
2163 static inline void skb_frag_set_page(struct sk_buff
*skb
, int f
,
2166 __skb_frag_set_page(&skb_shinfo(skb
)->frags
[f
], page
);
2170 * skb_frag_dma_map - maps a paged fragment via the DMA API
2171 * @dev: the device to map the fragment to
2172 * @frag: the paged fragment to map
2173 * @offset: the offset within the fragment (starting at the
2174 * fragment's own offset)
2175 * @size: the number of bytes to map
2176 * @dir: the direction of the mapping (%PCI_DMA_*)
2178 * Maps the page associated with @frag to @device.
2180 static inline dma_addr_t
skb_frag_dma_map(struct device
*dev
,
2181 const skb_frag_t
*frag
,
2182 size_t offset
, size_t size
,
2183 enum dma_data_direction dir
)
2185 return dma_map_page(dev
, skb_frag_page(frag
),
2186 frag
->page_offset
+ offset
, size
, dir
);
2189 static inline struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
2192 return __pskb_copy(skb
, skb_headroom(skb
), gfp_mask
);
2196 * skb_clone_writable - is the header of a clone writable
2197 * @skb: buffer to check
2198 * @len: length up to which to write
2200 * Returns true if modifying the header part of the cloned buffer
2201 * does not requires the data to be copied.
2203 static inline int skb_clone_writable(const struct sk_buff
*skb
, unsigned int len
)
2205 return !skb_header_cloned(skb
) &&
2206 skb_headroom(skb
) + len
<= skb
->hdr_len
;
2209 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
2214 if (headroom
> skb_headroom(skb
))
2215 delta
= headroom
- skb_headroom(skb
);
2217 if (delta
|| cloned
)
2218 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
2224 * skb_cow - copy header of skb when it is required
2225 * @skb: buffer to cow
2226 * @headroom: needed headroom
2228 * If the skb passed lacks sufficient headroom or its data part
2229 * is shared, data is reallocated. If reallocation fails, an error
2230 * is returned and original skb is not changed.
2232 * The result is skb with writable area skb->head...skb->tail
2233 * and at least @headroom of space at head.
2235 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
2237 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
2241 * skb_cow_head - skb_cow but only making the head writable
2242 * @skb: buffer to cow
2243 * @headroom: needed headroom
2245 * This function is identical to skb_cow except that we replace the
2246 * skb_cloned check by skb_header_cloned. It should be used when
2247 * you only need to push on some header and do not need to modify
2250 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
2252 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
2256 * skb_padto - pad an skbuff up to a minimal size
2257 * @skb: buffer to pad
2258 * @len: minimal length
2260 * Pads up a buffer to ensure the trailing bytes exist and are
2261 * blanked. If the buffer already contains sufficient data it
2262 * is untouched. Otherwise it is extended. Returns zero on
2263 * success. The skb is freed on error.
2266 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
2268 unsigned int size
= skb
->len
;
2269 if (likely(size
>= len
))
2271 return skb_pad(skb
, len
- size
);
2274 static inline int skb_add_data(struct sk_buff
*skb
,
2275 char __user
*from
, int copy
)
2277 const int off
= skb
->len
;
2279 if (skb
->ip_summed
== CHECKSUM_NONE
) {
2281 __wsum csum
= csum_and_copy_from_user(from
, skb_put(skb
, copy
),
2284 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
2287 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
2290 __skb_trim(skb
, off
);
2294 static inline bool skb_can_coalesce(struct sk_buff
*skb
, int i
,
2295 const struct page
*page
, int off
)
2298 const struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
2300 return page
== skb_frag_page(frag
) &&
2301 off
== frag
->page_offset
+ skb_frag_size(frag
);
2306 static inline int __skb_linearize(struct sk_buff
*skb
)
2308 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
2312 * skb_linearize - convert paged skb to linear one
2313 * @skb: buffer to linarize
2315 * If there is no free memory -ENOMEM is returned, otherwise zero
2316 * is returned and the old skb data released.
2318 static inline int skb_linearize(struct sk_buff
*skb
)
2320 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
2324 * skb_has_shared_frag - can any frag be overwritten
2325 * @skb: buffer to test
2327 * Return true if the skb has at least one frag that might be modified
2328 * by an external entity (as in vmsplice()/sendfile())
2330 static inline bool skb_has_shared_frag(const struct sk_buff
*skb
)
2332 return skb_is_nonlinear(skb
) &&
2333 skb_shinfo(skb
)->tx_flags
& SKBTX_SHARED_FRAG
;
2337 * skb_linearize_cow - make sure skb is linear and writable
2338 * @skb: buffer to process
2340 * If there is no free memory -ENOMEM is returned, otherwise zero
2341 * is returned and the old skb data released.
2343 static inline int skb_linearize_cow(struct sk_buff
*skb
)
2345 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
2346 __skb_linearize(skb
) : 0;
2350 * skb_postpull_rcsum - update checksum for received skb after pull
2351 * @skb: buffer to update
2352 * @start: start of data before pull
2353 * @len: length of data pulled
2355 * After doing a pull on a received packet, you need to call this to
2356 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2357 * CHECKSUM_NONE so that it can be recomputed from scratch.
2360 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
2361 const void *start
, unsigned int len
)
2363 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2364 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
2367 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
2370 * pskb_trim_rcsum - trim received skb and update checksum
2371 * @skb: buffer to trim
2374 * This is exactly the same as pskb_trim except that it ensures the
2375 * checksum of received packets are still valid after the operation.
2378 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
2380 if (likely(len
>= skb
->len
))
2382 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2383 skb
->ip_summed
= CHECKSUM_NONE
;
2384 return __pskb_trim(skb
, len
);
2387 #define skb_queue_walk(queue, skb) \
2388 for (skb = (queue)->next; \
2389 skb != (struct sk_buff *)(queue); \
2392 #define skb_queue_walk_safe(queue, skb, tmp) \
2393 for (skb = (queue)->next, tmp = skb->next; \
2394 skb != (struct sk_buff *)(queue); \
2395 skb = tmp, tmp = skb->next)
2397 #define skb_queue_walk_from(queue, skb) \
2398 for (; skb != (struct sk_buff *)(queue); \
2401 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2402 for (tmp = skb->next; \
2403 skb != (struct sk_buff *)(queue); \
2404 skb = tmp, tmp = skb->next)
2406 #define skb_queue_reverse_walk(queue, skb) \
2407 for (skb = (queue)->prev; \
2408 skb != (struct sk_buff *)(queue); \
2411 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2412 for (skb = (queue)->prev, tmp = skb->prev; \
2413 skb != (struct sk_buff *)(queue); \
2414 skb = tmp, tmp = skb->prev)
2416 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2417 for (tmp = skb->prev; \
2418 skb != (struct sk_buff *)(queue); \
2419 skb = tmp, tmp = skb->prev)
2421 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
2423 return skb_shinfo(skb
)->frag_list
!= NULL
;
2426 static inline void skb_frag_list_init(struct sk_buff
*skb
)
2428 skb_shinfo(skb
)->frag_list
= NULL
;
2431 static inline void skb_frag_add_head(struct sk_buff
*skb
, struct sk_buff
*frag
)
2433 frag
->next
= skb_shinfo(skb
)->frag_list
;
2434 skb_shinfo(skb
)->frag_list
= frag
;
2437 #define skb_walk_frags(skb, iter) \
2438 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2440 extern struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
2441 int *peeked
, int *off
, int *err
);
2442 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
2443 int noblock
, int *err
);
2444 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
2445 struct poll_table_struct
*wait
);
2446 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
2447 int offset
, struct iovec
*to
,
2449 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
,
2452 extern int skb_copy_datagram_from_iovec(struct sk_buff
*skb
,
2454 const struct iovec
*from
,
2457 extern int skb_copy_datagram_const_iovec(const struct sk_buff
*from
,
2459 const struct iovec
*to
,
2462 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
2463 extern void skb_free_datagram_locked(struct sock
*sk
,
2464 struct sk_buff
*skb
);
2465 extern int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
,
2466 unsigned int flags
);
2467 extern __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
,
2468 int len
, __wsum csum
);
2469 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
2471 extern int skb_store_bits(struct sk_buff
*skb
, int offset
,
2472 const void *from
, int len
);
2473 extern __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
,
2474 int offset
, u8
*to
, int len
,
2476 extern int skb_splice_bits(struct sk_buff
*skb
,
2477 unsigned int offset
,
2478 struct pipe_inode_info
*pipe
,
2480 unsigned int flags
);
2481 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
2482 extern void skb_split(struct sk_buff
*skb
,
2483 struct sk_buff
*skb1
, const u32 len
);
2484 extern int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
,
2487 extern struct sk_buff
*skb_segment(struct sk_buff
*skb
,
2488 netdev_features_t features
);
2490 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2491 int len
, void *buffer
)
2493 int hlen
= skb_headlen(skb
);
2495 if (hlen
- offset
>= len
)
2496 return skb
->data
+ offset
;
2498 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
2504 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
2506 const unsigned int len
)
2508 memcpy(to
, skb
->data
, len
);
2511 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
2512 const int offset
, void *to
,
2513 const unsigned int len
)
2515 memcpy(to
, skb
->data
+ offset
, len
);
2518 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
2520 const unsigned int len
)
2522 memcpy(skb
->data
, from
, len
);
2525 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
2528 const unsigned int len
)
2530 memcpy(skb
->data
+ offset
, from
, len
);
2533 extern void skb_init(void);
2535 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
2541 * skb_get_timestamp - get timestamp from a skb
2542 * @skb: skb to get stamp from
2543 * @stamp: pointer to struct timeval to store stamp in
2545 * Timestamps are stored in the skb as offsets to a base timestamp.
2546 * This function converts the offset back to a struct timeval and stores
2549 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
2550 struct timeval
*stamp
)
2552 *stamp
= ktime_to_timeval(skb
->tstamp
);
2555 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
2556 struct timespec
*stamp
)
2558 *stamp
= ktime_to_timespec(skb
->tstamp
);
2561 static inline void __net_timestamp(struct sk_buff
*skb
)
2563 skb
->tstamp
= ktime_get_real();
2566 static inline ktime_t
net_timedelta(ktime_t t
)
2568 return ktime_sub(ktime_get_real(), t
);
2571 static inline ktime_t
net_invalid_timestamp(void)
2573 return ktime_set(0, 0);
2576 extern void skb_timestamping_init(void);
2578 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2580 extern void skb_clone_tx_timestamp(struct sk_buff
*skb
);
2581 extern bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
2583 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2585 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
2589 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
2594 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2597 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2599 * PHY drivers may accept clones of transmitted packets for
2600 * timestamping via their phy_driver.txtstamp method. These drivers
2601 * must call this function to return the skb back to the stack, with
2602 * or without a timestamp.
2604 * @skb: clone of the the original outgoing packet
2605 * @hwtstamps: hardware time stamps, may be NULL if not available
2608 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
2609 struct skb_shared_hwtstamps
*hwtstamps
);
2612 * skb_tstamp_tx - queue clone of skb with send time stamps
2613 * @orig_skb: the original outgoing packet
2614 * @hwtstamps: hardware time stamps, may be NULL if not available
2616 * If the skb has a socket associated, then this function clones the
2617 * skb (thus sharing the actual data and optional structures), stores
2618 * the optional hardware time stamping information (if non NULL) or
2619 * generates a software time stamp (otherwise), then queues the clone
2620 * to the error queue of the socket. Errors are silently ignored.
2622 extern void skb_tstamp_tx(struct sk_buff
*orig_skb
,
2623 struct skb_shared_hwtstamps
*hwtstamps
);
2625 static inline void sw_tx_timestamp(struct sk_buff
*skb
)
2627 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
&&
2628 !(skb_shinfo(skb
)->tx_flags
& SKBTX_IN_PROGRESS
))
2629 skb_tstamp_tx(skb
, NULL
);
2633 * skb_tx_timestamp() - Driver hook for transmit timestamping
2635 * Ethernet MAC Drivers should call this function in their hard_xmit()
2636 * function immediately before giving the sk_buff to the MAC hardware.
2638 * @skb: A socket buffer.
2640 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
2642 skb_clone_tx_timestamp(skb
);
2643 sw_tx_timestamp(skb
);
2647 * skb_complete_wifi_ack - deliver skb with wifi status
2649 * @skb: the original outgoing packet
2650 * @acked: ack status
2653 void skb_complete_wifi_ack(struct sk_buff
*skb
, bool acked
);
2655 extern __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
2656 extern __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
2658 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
2660 return skb
->ip_summed
& CHECKSUM_UNNECESSARY
;
2664 * skb_checksum_complete - Calculate checksum of an entire packet
2665 * @skb: packet to process
2667 * This function calculates the checksum over the entire packet plus
2668 * the value of skb->csum. The latter can be used to supply the
2669 * checksum of a pseudo header as used by TCP/UDP. It returns the
2672 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2673 * this function can be used to verify that checksum on received
2674 * packets. In that case the function should return zero if the
2675 * checksum is correct. In particular, this function will return zero
2676 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2677 * hardware has already verified the correctness of the checksum.
2679 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
2681 return skb_csum_unnecessary(skb
) ?
2682 0 : __skb_checksum_complete(skb
);
2685 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2686 extern void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
2687 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
2689 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
2690 nf_conntrack_destroy(nfct
);
2692 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
2695 atomic_inc(&nfct
->use
);
2698 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2699 static inline void nf_conntrack_get_reasm(struct sk_buff
*skb
)
2702 atomic_inc(&skb
->users
);
2704 static inline void nf_conntrack_put_reasm(struct sk_buff
*skb
)
2710 #ifdef CONFIG_BRIDGE_NETFILTER
2711 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
2713 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
2716 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
2719 atomic_inc(&nf_bridge
->use
);
2721 #endif /* CONFIG_BRIDGE_NETFILTER */
2722 static inline void nf_reset(struct sk_buff
*skb
)
2724 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2725 nf_conntrack_put(skb
->nfct
);
2728 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2729 nf_conntrack_put_reasm(skb
->nfct_reasm
);
2730 skb
->nfct_reasm
= NULL
;
2732 #ifdef CONFIG_BRIDGE_NETFILTER
2733 nf_bridge_put(skb
->nf_bridge
);
2734 skb
->nf_bridge
= NULL
;
2738 static inline void nf_reset_trace(struct sk_buff
*skb
)
2740 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
2745 /* Note: This doesn't put any conntrack and bridge info in dst. */
2746 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2748 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2749 dst
->nfct
= src
->nfct
;
2750 nf_conntrack_get(src
->nfct
);
2751 dst
->nfctinfo
= src
->nfctinfo
;
2753 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2754 dst
->nfct_reasm
= src
->nfct_reasm
;
2755 nf_conntrack_get_reasm(src
->nfct_reasm
);
2757 #ifdef CONFIG_BRIDGE_NETFILTER
2758 dst
->nf_bridge
= src
->nf_bridge
;
2759 nf_bridge_get(src
->nf_bridge
);
2763 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
2765 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2766 nf_conntrack_put(dst
->nfct
);
2768 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2769 nf_conntrack_put_reasm(dst
->nfct_reasm
);
2771 #ifdef CONFIG_BRIDGE_NETFILTER
2772 nf_bridge_put(dst
->nf_bridge
);
2774 __nf_copy(dst
, src
);
2777 #ifdef CONFIG_NETWORK_SECMARK
2778 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2780 to
->secmark
= from
->secmark
;
2783 static inline void skb_init_secmark(struct sk_buff
*skb
)
2788 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
2791 static inline void skb_init_secmark(struct sk_buff
*skb
)
2795 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
2797 skb
->queue_mapping
= queue_mapping
;
2800 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
2802 return skb
->queue_mapping
;
2805 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
2807 to
->queue_mapping
= from
->queue_mapping
;
2810 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
2812 skb
->queue_mapping
= rx_queue
+ 1;
2815 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
2817 return skb
->queue_mapping
- 1;
2820 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
2822 return skb
->queue_mapping
!= 0;
2825 extern u16
__skb_tx_hash(const struct net_device
*dev
,
2826 const struct sk_buff
*skb
,
2827 unsigned int num_tx_queues
);
2830 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2835 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
2841 /* Keeps track of mac header offset relative to skb->head.
2842 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2843 * For non-tunnel skb it points to skb_mac_header() and for
2844 * tunnel skb it points to outer mac header. */
2848 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2850 static inline int skb_tnl_header_len(const struct sk_buff
*inner_skb
)
2852 return (skb_mac_header(inner_skb
) - inner_skb
->head
) -
2853 SKB_GSO_CB(inner_skb
)->mac_offset
;
2856 static inline int gso_pskb_expand_head(struct sk_buff
*skb
, int extra
)
2858 int new_headroom
, headroom
;
2861 headroom
= skb_headroom(skb
);
2862 ret
= pskb_expand_head(skb
, extra
, 0, GFP_ATOMIC
);
2866 new_headroom
= skb_headroom(skb
);
2867 SKB_GSO_CB(skb
)->mac_offset
+= (new_headroom
- headroom
);
2871 static inline bool skb_is_gso(const struct sk_buff
*skb
)
2873 return skb_shinfo(skb
)->gso_size
;
2876 static inline bool skb_is_gso_v6(const struct sk_buff
*skb
)
2878 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
2881 extern void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
2883 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
2885 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2886 * wanted then gso_type will be set. */
2887 const struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
2889 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
2890 unlikely(shinfo
->gso_type
== 0)) {
2891 __skb_warn_lro_forwarding(skb
);
2897 static inline void skb_forward_csum(struct sk_buff
*skb
)
2899 /* Unfortunately we don't support this one. Any brave souls? */
2900 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2901 skb
->ip_summed
= CHECKSUM_NONE
;
2905 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2906 * @skb: skb to check
2908 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2909 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2910 * use this helper, to document places where we make this assertion.
2912 static inline void skb_checksum_none_assert(const struct sk_buff
*skb
)
2915 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
2919 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
2921 u32
__skb_get_poff(const struct sk_buff
*skb
);
2924 * skb_head_is_locked - Determine if the skb->head is locked down
2925 * @skb: skb to check
2927 * The head on skbs build around a head frag can be removed if they are
2928 * not cloned. This function returns true if the skb head is locked down
2929 * due to either being allocated via kmalloc, or by being a clone with
2930 * multiple references to the head.
2932 static inline bool skb_head_is_locked(const struct sk_buff
*skb
)
2934 return !skb
->head_frag
|| skb_cloned(skb
);
2936 #endif /* __KERNEL__ */
2937 #endif /* _LINUX_SKBUFF_H */