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 skb_tx_error(struct sk_buff *skb);
631 extern void consume_skb(struct sk_buff *skb);
632 extern void __kfree_skb(struct sk_buff *skb);
633 extern struct kmem_cache *skbuff_head_cache;
635 extern void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
636 extern bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
637 bool *fragstolen, int *delta_truesize);
639 extern struct sk_buff *__alloc_skb(unsigned int size,
640 gfp_t priority, int flags, int node);
641 extern struct sk_buff *build_skb(void *data, unsigned int frag_size);
642 static inline struct sk_buff *alloc_skb(unsigned int size,
645 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
648 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
651 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
654 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
655 extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
656 extern struct sk_buff *skb_clone(struct sk_buff *skb,
658 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
660 extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
661 int headroom, gfp_t gfp_mask);
663 extern int pskb_expand_head(struct sk_buff *skb,
664 int nhead, int ntail,
666 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
667 unsigned int headroom);
668 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
669 int newheadroom, int newtailroom,
671 extern int skb_to_sgvec(struct sk_buff *skb,
672 struct scatterlist *sg, int offset,
674 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
675 struct sk_buff **trailer);
676 extern int skb_pad(struct sk_buff *skb, int pad);
677 #define dev_kfree_skb(a) consume_skb(a)
679 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
680 int getfrag(void *from, char *to, int offset,
681 int len,int odd, struct sk_buff *skb),
682 void *from, int length);
684 struct skb_seq_state {
688 __u32 stepped_offset;
689 struct sk_buff *root_skb;
690 struct sk_buff *cur_skb;
694 extern void skb_prepare_seq_read(struct sk_buff *skb,
695 unsigned int from, unsigned int to,
696 struct skb_seq_state *st);
697 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
698 struct skb_seq_state *st);
699 extern void skb_abort_seq_read(struct skb_seq_state *st);
701 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
702 unsigned int to, struct ts_config *config,
703 struct ts_state *state);
705 extern void __skb_get_rxhash(struct sk_buff *skb);
706 static inline __u32 skb_get_rxhash(struct sk_buff *skb)
709 __skb_get_rxhash(skb);
714 #ifdef NET_SKBUFF_DATA_USES_OFFSET
715 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
717 return skb->head + skb->end;
720 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
725 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
730 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
732 return skb->end - skb->head;
737 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
739 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
741 return &skb_shinfo(skb)->hwtstamps;
745 * skb_queue_empty - check if a queue is empty
748 * Returns true if the queue is empty, false otherwise.
750 static inline int skb_queue_empty(const struct sk_buff_head *list)
752 return list->next == (struct sk_buff *)list;
756 * skb_queue_is_last - check if skb is the last entry in the queue
760 * Returns true if @skb is the last buffer on the list.
762 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
763 const struct sk_buff *skb)
765 return skb->next == (struct sk_buff *)list;
769 * skb_queue_is_first - check if skb is the first entry in the queue
773 * Returns true if @skb is the first buffer on the list.
775 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
776 const struct sk_buff *skb)
778 return skb->prev == (struct sk_buff *)list;
782 * skb_queue_next - return the next packet in the queue
784 * @skb: current buffer
786 * Return the next packet in @list after @skb. It is only valid to
787 * call this if skb_queue_is_last() evaluates to false.
789 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
790 const struct sk_buff *skb)
792 /* This BUG_ON may seem severe, but if we just return then we
793 * are going to dereference garbage.
795 BUG_ON(skb_queue_is_last(list, skb));
800 * skb_queue_prev - return the prev packet in the queue
802 * @skb: current buffer
804 * Return the prev packet in @list before @skb. It is only valid to
805 * call this if skb_queue_is_first() evaluates to false.
807 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
808 const struct sk_buff *skb)
810 /* This BUG_ON may seem severe, but if we just return then we
811 * are going to dereference garbage.
813 BUG_ON(skb_queue_is_first(list, skb));
818 * skb_get - reference buffer
819 * @skb: buffer to reference
821 * Makes another reference to a socket buffer and returns a pointer
824 static inline struct sk_buff *skb_get(struct sk_buff *skb)
826 atomic_inc(&skb->users);
831 * If users == 1, we are the only owner and are can avoid redundant
836 * skb_cloned - is the buffer a clone
837 * @skb: buffer to check
839 * Returns true if the buffer was generated with skb_clone() and is
840 * one of multiple shared copies of the buffer. Cloned buffers are
841 * shared data so must not be written to under normal circumstances.
843 static inline int skb_cloned(const struct sk_buff *skb)
845 return skb->cloned &&
846 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
849 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
851 might_sleep_if(pri & __GFP_WAIT);
854 return pskb_expand_head(skb, 0, 0, pri);
860 * skb_header_cloned - is the header a clone
861 * @skb: buffer to check
863 * Returns true if modifying the header part of the buffer requires
864 * the data to be copied.
866 static inline int skb_header_cloned(const struct sk_buff *skb)
873 dataref = atomic_read(&skb_shinfo(skb)->dataref);
874 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
879 * skb_header_release - release reference to header
880 * @skb: buffer to operate on
882 * Drop a reference to the header part of the buffer. This is done
883 * by acquiring a payload reference. You must not read from the header
884 * part of skb->data after this.
886 static inline void skb_header_release(struct sk_buff *skb)
890 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
894 * skb_shared - is the buffer shared
895 * @skb: buffer to check
897 * Returns true if more than one person has a reference to this
900 static inline int skb_shared(const struct sk_buff *skb)
902 return atomic_read(&skb->users) != 1;
906 * skb_share_check - check if buffer is shared and if so clone it
907 * @skb: buffer to check
908 * @pri: priority for memory allocation
910 * If the buffer is shared the buffer is cloned and the old copy
911 * drops a reference. A new clone with a single reference is returned.
912 * If the buffer is not shared the original buffer is returned. When
913 * being called from interrupt status or with spinlocks held pri must
916 * NULL is returned on a memory allocation failure.
918 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
920 might_sleep_if(pri & __GFP_WAIT);
921 if (skb_shared(skb)) {
922 struct sk_buff *nskb = skb_clone(skb, pri);
934 * Copy shared buffers into a new sk_buff. We effectively do COW on
935 * packets to handle cases where we have a local reader and forward
936 * and a couple of other messy ones. The normal one is tcpdumping
937 * a packet thats being forwarded.
941 * skb_unshare - make a copy of a shared buffer
942 * @skb: buffer to check
943 * @pri: priority for memory allocation
945 * If the socket buffer is a clone then this function creates a new
946 * copy of the data, drops a reference count on the old copy and returns
947 * the new copy with the reference count at 1. If the buffer is not a clone
948 * the original buffer is returned. When called with a spinlock held or
949 * from interrupt state @pri must be %GFP_ATOMIC
951 * %NULL is returned on a memory allocation failure.
953 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
956 might_sleep_if(pri & __GFP_WAIT);
957 if (skb_cloned(skb)) {
958 struct sk_buff *nskb = skb_copy(skb, pri);
959 kfree_skb(skb); /* Free our shared copy */
966 * skb_peek - peek at the head of an &sk_buff_head
967 * @list_: list to peek at
969 * Peek an &sk_buff. Unlike most other operations you _MUST_
970 * be careful with this one. A peek leaves the buffer on the
971 * list and someone else may run off with it. You must hold
972 * the appropriate locks or have a private queue to do this.
974 * Returns %NULL for an empty list or a pointer to the head element.
975 * The reference count is not incremented and the reference is therefore
976 * volatile. Use with caution.
978 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
980 struct sk_buff *skb = list_->next;
982 if (skb == (struct sk_buff *)list_)
988 * skb_peek_next - peek skb following the given one from a queue
989 * @skb: skb to start from
990 * @list_: list to peek at
992 * Returns %NULL when the end of the list is met or a pointer to the
993 * next element. The reference count is not incremented and the
994 * reference is therefore volatile. Use with caution.
996 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
997 const struct sk_buff_head *list_)
999 struct sk_buff *next = skb->next;
1001 if (next == (struct sk_buff *)list_)
1007 * skb_peek_tail - peek at the tail of an &sk_buff_head
1008 * @list_: list to peek at
1010 * Peek an &sk_buff. Unlike most other operations you _MUST_
1011 * be careful with this one. A peek leaves the buffer on the
1012 * list and someone else may run off with it. You must hold
1013 * the appropriate locks or have a private queue to do this.
1015 * Returns %NULL for an empty list or a pointer to the tail element.
1016 * The reference count is not incremented and the reference is therefore
1017 * volatile. Use with caution.
1019 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1021 struct sk_buff *skb = list_->prev;
1023 if (skb == (struct sk_buff *)list_)
1030 * skb_queue_len - get queue length
1031 * @list_: list to measure
1033 * Return the length of an &sk_buff queue.
1035 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1041 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1042 * @list: queue to initialize
1044 * This initializes only the list and queue length aspects of
1045 * an sk_buff_head object. This allows to initialize the list
1046 * aspects of an sk_buff_head without reinitializing things like
1047 * the spinlock. It can also be used for on-stack sk_buff_head
1048 * objects where the spinlock is known to not be used.
1050 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1052 list->prev = list->next = (struct sk_buff *)list;
1057 * This function creates a split out lock class for each invocation;
1058 * this is needed for now since a whole lot of users of the skb-queue
1059 * infrastructure in drivers have different locking usage (in hardirq)
1060 * than the networking core (in softirq only). In the long run either the
1061 * network layer or drivers should need annotation to consolidate the
1062 * main types of usage into 3 classes.
1064 static inline void skb_queue_head_init(struct sk_buff_head *list)
1066 spin_lock_init(&list->lock);
1067 __skb_queue_head_init(list);
1070 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1071 struct lock_class_key *class)
1073 skb_queue_head_init(list);
1074 lockdep_set_class(&list->lock, class);
1078 * Insert an sk_buff on a list.
1080 * The "__skb_xxxx()" functions are the non-atomic ones that
1081 * can only be called with interrupts disabled.
1083 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
1084 static inline void __skb_insert(struct sk_buff *newsk,
1085 struct sk_buff *prev, struct sk_buff *next,
1086 struct sk_buff_head *list)
1090 next->prev = prev->next = newsk;
1094 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1095 struct sk_buff *prev,
1096 struct sk_buff *next)
1098 struct sk_buff *first = list->next;
1099 struct sk_buff *last = list->prev;
1109 * skb_queue_splice - join two skb lists, this is designed for stacks
1110 * @list: the new list to add
1111 * @head: the place to add it in the first list
1113 static inline void skb_queue_splice(const struct sk_buff_head *list,
1114 struct sk_buff_head *head)
1116 if (!skb_queue_empty(list)) {
1117 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1118 head->qlen += list->qlen;
1123 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1124 * @list: the new list to add
1125 * @head: the place to add it in the first list
1127 * The list at @list is reinitialised
1129 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1130 struct sk_buff_head *head)
1132 if (!skb_queue_empty(list)) {
1133 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1134 head->qlen += list->qlen;
1135 __skb_queue_head_init(list);
1140 * skb_queue_splice_tail - join two skb lists, each list being a queue
1141 * @list: the new list to add
1142 * @head: the place to add it in the first list
1144 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1145 struct sk_buff_head *head)
1147 if (!skb_queue_empty(list)) {
1148 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1149 head->qlen += list->qlen;
1154 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1155 * @list: the new list to add
1156 * @head: the place to add it in the first list
1158 * Each of the lists is a queue.
1159 * The list at @list is reinitialised
1161 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1162 struct sk_buff_head *head)
1164 if (!skb_queue_empty(list)) {
1165 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1166 head->qlen += list->qlen;
1167 __skb_queue_head_init(list);
1172 * __skb_queue_after - queue a buffer at the list head
1173 * @list: list to use
1174 * @prev: place after this buffer
1175 * @newsk: buffer to queue
1177 * Queue a buffer int the middle of a list. This function takes no locks
1178 * and you must therefore hold required locks before calling it.
1180 * A buffer cannot be placed on two lists at the same time.
1182 static inline void __skb_queue_after(struct sk_buff_head *list,
1183 struct sk_buff *prev,
1184 struct sk_buff *newsk)
1186 __skb_insert(newsk, prev, prev->next, list);
1189 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1190 struct sk_buff_head *list);
1192 static inline void __skb_queue_before(struct sk_buff_head *list,
1193 struct sk_buff *next,
1194 struct sk_buff *newsk)
1196 __skb_insert(newsk, next->prev, next, list);
1200 * __skb_queue_head - queue a buffer at the list head
1201 * @list: list to use
1202 * @newsk: buffer to queue
1204 * Queue a buffer at the start of a list. This function takes no locks
1205 * and you must therefore hold required locks before calling it.
1207 * A buffer cannot be placed on two lists at the same time.
1209 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1210 static inline void __skb_queue_head(struct sk_buff_head *list,
1211 struct sk_buff *newsk)
1213 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1217 * __skb_queue_tail - queue a buffer at the list tail
1218 * @list: list to use
1219 * @newsk: buffer to queue
1221 * Queue a buffer at the end of a list. This function takes no locks
1222 * and you must therefore hold required locks before calling it.
1224 * A buffer cannot be placed on two lists at the same time.
1226 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1227 static inline void __skb_queue_tail(struct sk_buff_head *list,
1228 struct sk_buff *newsk)
1230 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1234 * remove sk_buff from list. _Must_ be called atomically, and with
1237 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1238 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1240 struct sk_buff *next, *prev;
1245 skb->next = skb->prev = NULL;
1251 * __skb_dequeue - remove from the head of the queue
1252 * @list: list to dequeue from
1254 * Remove the head of the list. This function does not take any locks
1255 * so must be used with appropriate locks held only. The head item is
1256 * returned or %NULL if the list is empty.
1258 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1259 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1261 struct sk_buff *skb = skb_peek(list);
1263 __skb_unlink(skb, list);
1268 * __skb_dequeue_tail - remove from the tail of the queue
1269 * @list: list to dequeue from
1271 * Remove the tail of the list. This function does not take any locks
1272 * so must be used with appropriate locks held only. The tail item is
1273 * returned or %NULL if the list is empty.
1275 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1276 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1278 struct sk_buff *skb = skb_peek_tail(list);
1280 __skb_unlink(skb, list);
1285 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1287 return skb->data_len;
1290 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1292 return skb->len - skb->data_len;
1295 static inline int skb_pagelen(const struct sk_buff *skb)
1299 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1300 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1301 return len + skb_headlen(skb);
1305 * __skb_fill_page_desc - initialise a paged fragment in an skb
1306 * @skb: buffer containing fragment to be initialised
1307 * @i: paged fragment index to initialise
1308 * @page: the page to use for this fragment
1309 * @off: the offset to the data with @page
1310 * @size: the length of the data
1312 * Initialises the @i'th fragment of @skb to point to &size bytes at
1313 * offset @off within @page.
1315 * Does not take any additional reference on the fragment.
1317 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1318 struct page *page, int off, int size)
1320 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1323 * Propagate page->pfmemalloc to the skb if we can. The problem is
1324 * that not all callers have unique ownership of the page. If
1325 * pfmemalloc is set, we check the mapping as a mapping implies
1326 * page->index is set (index and pfmemalloc share space).
1327 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1328 * do not lose pfmemalloc information as the pages would not be
1329 * allocated using __GFP_MEMALLOC.
1331 frag->page.p = page;
1332 frag->page_offset = off;
1333 skb_frag_size_set(frag, size);
1335 page = compound_head(page);
1336 if (page->pfmemalloc && !page->mapping)
1337 skb->pfmemalloc = true;
1341 * skb_fill_page_desc - initialise a paged fragment in an skb
1342 * @skb: buffer containing fragment to be initialised
1343 * @i: paged fragment index to initialise
1344 * @page: the page to use for this fragment
1345 * @off: the offset to the data with @page
1346 * @size: the length of the data
1348 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1349 * @skb to point to &size bytes at offset @off within @page. In
1350 * addition updates @skb such that @i is the last fragment.
1352 * Does not take any additional reference on the fragment.
1354 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1355 struct page *page, int off, int size)
1357 __skb_fill_page_desc(skb, i, page, off, size);
1358 skb_shinfo(skb)->nr_frags = i + 1;
1361 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1362 int off, int size, unsigned int truesize);
1364 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1365 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1366 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1368 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1369 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1371 return skb->head + skb->tail;
1374 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1376 skb->tail = skb->data - skb->head;
1379 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1381 skb_reset_tail_pointer(skb);
1382 skb->tail += offset;
1384 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1385 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1390 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1392 skb->tail = skb->data;
1395 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1397 skb->tail = skb->data + offset;
1400 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1403 * Add data to an sk_buff
1405 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1406 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1408 unsigned char *tmp = skb_tail_pointer(skb);
1409 SKB_LINEAR_ASSERT(skb);
1415 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1416 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1423 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1424 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1427 BUG_ON(skb->len < skb->data_len);
1428 return skb->data += len;
1431 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1433 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1436 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1438 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1440 if (len > skb_headlen(skb) &&
1441 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1444 return skb->data += len;
1447 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1449 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1452 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1454 if (likely(len <= skb_headlen(skb)))
1456 if (unlikely(len > skb->len))
1458 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1462 * skb_headroom - bytes at buffer head
1463 * @skb: buffer to check
1465 * Return the number of bytes of free space at the head of an &sk_buff.
1467 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1469 return skb->data - skb->head;
1473 * skb_tailroom - bytes at buffer end
1474 * @skb: buffer to check
1476 * Return the number of bytes of free space at the tail of an sk_buff
1478 static inline int skb_tailroom(const struct sk_buff *skb)
1480 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1484 * skb_availroom - bytes at buffer end
1485 * @skb: buffer to check
1487 * Return the number of bytes of free space at the tail of an sk_buff
1488 * allocated by sk_stream_alloc()
1490 static inline int skb_availroom(const struct sk_buff *skb)
1492 if (skb_is_nonlinear(skb))
1495 return skb->end - skb->tail - skb->reserved_tailroom;
1499 * skb_reserve - adjust headroom
1500 * @skb: buffer to alter
1501 * @len: bytes to move
1503 * Increase the headroom of an empty &sk_buff by reducing the tail
1504 * room. This is only allowed for an empty buffer.
1506 static inline void skb_reserve(struct sk_buff *skb, int len)
1512 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1514 skb->inner_mac_header = skb->mac_header;
1515 skb->inner_network_header = skb->network_header;
1516 skb->inner_transport_header = skb->transport_header;
1519 static inline void skb_reset_mac_len(struct sk_buff *skb)
1521 skb->mac_len = skb->network_header - skb->mac_header;
1524 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1525 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1528 return skb->head + skb->inner_transport_header;
1531 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1533 skb->inner_transport_header = skb->data - skb->head;
1536 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1539 skb_reset_inner_transport_header(skb);
1540 skb->inner_transport_header += offset;
1543 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1545 return skb->head + skb->inner_network_header;
1548 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1550 skb->inner_network_header = skb->data - skb->head;
1553 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1556 skb_reset_inner_network_header(skb);
1557 skb->inner_network_header += offset;
1560 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1562 return skb->head + skb->inner_mac_header;
1565 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1567 skb->inner_mac_header = skb->data - skb->head;
1570 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1573 skb_reset_inner_mac_header(skb);
1574 skb->inner_mac_header += offset;
1576 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1578 return skb->transport_header != ~0U;
1581 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1583 return skb->head + skb->transport_header;
1586 static inline void skb_reset_transport_header(struct sk_buff *skb)
1588 skb->transport_header = skb->data - skb->head;
1591 static inline void skb_set_transport_header(struct sk_buff *skb,
1594 skb_reset_transport_header(skb);
1595 skb->transport_header += offset;
1598 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1600 return skb->head + skb->network_header;
1603 static inline void skb_reset_network_header(struct sk_buff *skb)
1605 skb->network_header = skb->data - skb->head;
1608 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1610 skb_reset_network_header(skb);
1611 skb->network_header += offset;
1614 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1616 return skb->head + skb->mac_header;
1619 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1621 return skb->mac_header != ~0U;
1624 static inline void skb_reset_mac_header(struct sk_buff *skb)
1626 skb->mac_header = skb->data - skb->head;
1629 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1631 skb_reset_mac_header(skb);
1632 skb->mac_header += offset;
1635 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1636 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1639 return skb->inner_transport_header;
1642 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1644 skb->inner_transport_header = skb->data;
1647 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1650 skb->inner_transport_header = skb->data + offset;
1653 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1655 return skb->inner_network_header;
1658 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1660 skb->inner_network_header = skb->data;
1663 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1666 skb->inner_network_header = skb->data + offset;
1669 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1671 return skb->inner_mac_header;
1674 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1676 skb->inner_mac_header = skb->data;
1679 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1682 skb->inner_mac_header = skb->data + offset;
1684 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1686 return skb->transport_header != NULL;
1689 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1691 return skb->transport_header;
1694 static inline void skb_reset_transport_header(struct sk_buff *skb)
1696 skb->transport_header = skb->data;
1699 static inline void skb_set_transport_header(struct sk_buff *skb,
1702 skb->transport_header = skb->data + offset;
1705 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1707 return skb->network_header;
1710 static inline void skb_reset_network_header(struct sk_buff *skb)
1712 skb->network_header = skb->data;
1715 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1717 skb->network_header = skb->data + offset;
1720 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1722 return skb->mac_header;
1725 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1727 return skb->mac_header != NULL;
1730 static inline void skb_reset_mac_header(struct sk_buff *skb)
1732 skb->mac_header = skb->data;
1735 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1737 skb->mac_header = skb->data + offset;
1739 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1741 static inline void skb_probe_transport_header(struct sk_buff *skb,
1742 const int offset_hint)
1744 struct flow_keys keys;
1746 if (skb_transport_header_was_set(skb))
1748 else if (skb_flow_dissect(skb, &keys))
1749 skb_set_transport_header(skb, keys.thoff);
1751 skb_set_transport_header(skb, offset_hint);
1754 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1756 if (skb_mac_header_was_set(skb)) {
1757 const unsigned char *old_mac = skb_mac_header(skb);
1759 skb_set_mac_header(skb, -skb->mac_len);
1760 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1764 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1766 return skb->csum_start - skb_headroom(skb);
1769 static inline int skb_transport_offset(const struct sk_buff *skb)
1771 return skb_transport_header(skb) - skb->data;
1774 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1776 return skb->transport_header - skb->network_header;
1779 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1781 return skb->inner_transport_header - skb->inner_network_header;
1784 static inline int skb_network_offset(const struct sk_buff *skb)
1786 return skb_network_header(skb) - skb->data;
1789 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1791 return skb_inner_network_header(skb) - skb->data;
1794 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1796 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1800 * CPUs often take a performance hit when accessing unaligned memory
1801 * locations. The actual performance hit varies, it can be small if the
1802 * hardware handles it or large if we have to take an exception and fix it
1805 * Since an ethernet header is 14 bytes network drivers often end up with
1806 * the IP header at an unaligned offset. The IP header can be aligned by
1807 * shifting the start of the packet by 2 bytes. Drivers should do this
1810 * skb_reserve(skb, NET_IP_ALIGN);
1812 * The downside to this alignment of the IP header is that the DMA is now
1813 * unaligned. On some architectures the cost of an unaligned DMA is high
1814 * and this cost outweighs the gains made by aligning the IP header.
1816 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1819 #ifndef NET_IP_ALIGN
1820 #define NET_IP_ALIGN 2
1824 * The networking layer reserves some headroom in skb data (via
1825 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1826 * the header has to grow. In the default case, if the header has to grow
1827 * 32 bytes or less we avoid the reallocation.
1829 * Unfortunately this headroom changes the DMA alignment of the resulting
1830 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1831 * on some architectures. An architecture can override this value,
1832 * perhaps setting it to a cacheline in size (since that will maintain
1833 * cacheline alignment of the DMA). It must be a power of 2.
1835 * Various parts of the networking layer expect at least 32 bytes of
1836 * headroom, you should not reduce this.
1838 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1839 * to reduce average number of cache lines per packet.
1840 * get_rps_cpus() for example only access one 64 bytes aligned block :
1841 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1844 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1847 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1849 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1851 if (unlikely(skb_is_nonlinear(skb))) {
1856 skb_set_tail_pointer(skb, len);
1859 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1861 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1864 return ___pskb_trim(skb, len);
1865 __skb_trim(skb, len);
1869 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1871 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1875 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1876 * @skb: buffer to alter
1879 * This is identical to pskb_trim except that the caller knows that
1880 * the skb is not cloned so we should never get an error due to out-
1883 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1885 int err = pskb_trim(skb, len);
1890 * skb_orphan - orphan a buffer
1891 * @skb: buffer to orphan
1893 * If a buffer currently has an owner then we call the owner's
1894 * destructor function and make the @skb unowned. The buffer continues
1895 * to exist but is no longer charged to its former owner.
1897 static inline void skb_orphan(struct sk_buff *skb)
1899 if (skb->destructor)
1900 skb->destructor(skb);
1901 skb->destructor = NULL;
1906 * skb_orphan_frags - orphan the frags contained in a buffer
1907 * @skb: buffer to orphan frags from
1908 * @gfp_mask: allocation mask for replacement pages
1910 * For each frag in the SKB which needs a destructor (i.e. has an
1911 * owner) create a copy of that frag and release the original
1912 * page by calling the destructor.
1914 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1916 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1918 return skb_copy_ubufs(skb, gfp_mask);
1922 * __skb_queue_purge - empty a list
1923 * @list: list to empty
1925 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1926 * the list and one reference dropped. This function does not take the
1927 * list lock and the caller must hold the relevant locks to use it.
1929 extern void skb_queue_purge(struct sk_buff_head *list);
1930 static inline void __skb_queue_purge(struct sk_buff_head *list)
1932 struct sk_buff *skb;
1933 while ((skb = __skb_dequeue(list)) != NULL)
1937 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1938 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1939 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1941 extern void *netdev_alloc_frag(unsigned int fragsz);
1943 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1944 unsigned int length,
1948 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1949 * @dev: network device to receive on
1950 * @length: length to allocate
1952 * Allocate a new &sk_buff and assign it a usage count of one. The
1953 * buffer has unspecified headroom built in. Users should allocate
1954 * the headroom they think they need without accounting for the
1955 * built in space. The built in space is used for optimisations.
1957 * %NULL is returned if there is no free memory. Although this function
1958 * allocates memory it can be called from an interrupt.
1960 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1961 unsigned int length)
1963 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1966 /* legacy helper around __netdev_alloc_skb() */
1967 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1970 return __netdev_alloc_skb(NULL, length, gfp_mask);
1973 /* legacy helper around netdev_alloc_skb() */
1974 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1976 return netdev_alloc_skb(NULL, length);
1980 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1981 unsigned int length, gfp_t gfp)
1983 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
1985 if (NET_IP_ALIGN && skb)
1986 skb_reserve(skb, NET_IP_ALIGN);
1990 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1991 unsigned int length)
1993 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1997 * __skb_alloc_page - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1998 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1999 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2000 * @order: size of the allocation
2002 * Allocate a new page.
2004 * %NULL is returned if there is no free memory.
2006 static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2007 struct sk_buff *skb,
2012 gfp_mask |= __GFP_COLD;
2014 if (!(gfp_mask & __GFP_NOMEMALLOC))
2015 gfp_mask |= __GFP_MEMALLOC;
2017 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2018 if (skb && page && page->pfmemalloc)
2019 skb->pfmemalloc = true;
2025 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2026 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2027 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2029 * Allocate a new page.
2031 * %NULL is returned if there is no free memory.
2033 static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2034 struct sk_buff *skb)
2036 return __skb_alloc_pages(gfp_mask, skb, 0);
2040 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2041 * @page: The page that was allocated from skb_alloc_page
2042 * @skb: The skb that may need pfmemalloc set
2044 static inline void skb_propagate_pfmemalloc(struct page *page,
2045 struct sk_buff *skb)
2047 if (page && page->pfmemalloc)
2048 skb->pfmemalloc = true;
2052 * skb_frag_page - retrieve the page refered to by a paged fragment
2053 * @frag: the paged fragment
2055 * Returns the &struct page associated with @frag.
2057 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2059 return frag->page.p;
2063 * __skb_frag_ref - take an addition reference on a paged fragment.
2064 * @frag: the paged fragment
2066 * Takes an additional reference on the paged fragment @frag.
2068 static inline void __skb_frag_ref(skb_frag_t *frag)
2070 get_page(skb_frag_page(frag));
2074 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2076 * @f: the fragment offset.
2078 * Takes an additional reference on the @f'th paged fragment of @skb.
2080 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2082 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2086 * __skb_frag_unref - release a reference on a paged fragment.
2087 * @frag: the paged fragment
2089 * Releases a reference on the paged fragment @frag.
2091 static inline void __skb_frag_unref(skb_frag_t *frag)
2093 put_page(skb_frag_page(frag));
2097 * skb_frag_unref - release a reference on a paged fragment of an skb.
2099 * @f: the fragment offset
2101 * Releases a reference on the @f'th paged fragment of @skb.
2103 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2105 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2109 * skb_frag_address - gets the address of the data contained in a paged fragment
2110 * @frag: the paged fragment buffer
2112 * Returns the address of the data within @frag. The page must already
2115 static inline void *skb_frag_address(const skb_frag_t *frag)
2117 return page_address(skb_frag_page(frag)) + frag->page_offset;
2121 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2122 * @frag: the paged fragment buffer
2124 * Returns the address of the data within @frag. Checks that the page
2125 * is mapped and returns %NULL otherwise.
2127 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2129 void *ptr = page_address(skb_frag_page(frag));
2133 return ptr + frag->page_offset;
2137 * __skb_frag_set_page - sets the page contained in a paged fragment
2138 * @frag: the paged fragment
2139 * @page: the page to set
2141 * Sets the fragment @frag to contain @page.
2143 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2145 frag->page.p = page;
2149 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2151 * @f: the fragment offset
2152 * @page: the page to set
2154 * Sets the @f'th fragment of @skb to contain @page.
2156 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2159 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2163 * skb_frag_dma_map - maps a paged fragment via the DMA API
2164 * @dev: the device to map the fragment to
2165 * @frag: the paged fragment to map
2166 * @offset: the offset within the fragment (starting at the
2167 * fragment's own offset)
2168 * @size: the number of bytes to map
2169 * @dir: the direction of the mapping (%PCI_DMA_*)
2171 * Maps the page associated with @frag to @device.
2173 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2174 const skb_frag_t *frag,
2175 size_t offset, size_t size,
2176 enum dma_data_direction dir)
2178 return dma_map_page(dev, skb_frag_page(frag),
2179 frag->page_offset + offset, size, dir);
2182 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2185 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2189 * skb_clone_writable - is the header of a clone writable
2190 * @skb: buffer to check
2191 * @len: length up to which to write
2193 * Returns true if modifying the header part of the cloned buffer
2194 * does not requires the data to be copied.
2196 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2198 return !skb_header_cloned(skb) &&
2199 skb_headroom(skb) + len <= skb->hdr_len;
2202 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2207 if (headroom > skb_headroom(skb))
2208 delta = headroom - skb_headroom(skb);
2210 if (delta || cloned)
2211 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2217 * skb_cow - copy header of skb when it is required
2218 * @skb: buffer to cow
2219 * @headroom: needed headroom
2221 * If the skb passed lacks sufficient headroom or its data part
2222 * is shared, data is reallocated. If reallocation fails, an error
2223 * is returned and original skb is not changed.
2225 * The result is skb with writable area skb->head...skb->tail
2226 * and at least @headroom of space at head.
2228 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2230 return __skb_cow(skb, headroom, skb_cloned(skb));
2234 * skb_cow_head - skb_cow but only making the head writable
2235 * @skb: buffer to cow
2236 * @headroom: needed headroom
2238 * This function is identical to skb_cow except that we replace the
2239 * skb_cloned check by skb_header_cloned. It should be used when
2240 * you only need to push on some header and do not need to modify
2243 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2245 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2249 * skb_padto - pad an skbuff up to a minimal size
2250 * @skb: buffer to pad
2251 * @len: minimal length
2253 * Pads up a buffer to ensure the trailing bytes exist and are
2254 * blanked. If the buffer already contains sufficient data it
2255 * is untouched. Otherwise it is extended. Returns zero on
2256 * success. The skb is freed on error.
2259 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2261 unsigned int size = skb->len;
2262 if (likely(size >= len))
2264 return skb_pad(skb, len - size);
2267 static inline int skb_add_data(struct sk_buff *skb,
2268 char __user *from, int copy)
2270 const int off = skb->len;
2272 if (skb->ip_summed == CHECKSUM_NONE) {
2274 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
2277 skb->csum = csum_block_add(skb->csum, csum, off);
2280 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2283 __skb_trim(skb, off);
2287 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2288 const struct page *page, int off)
2291 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2293 return page == skb_frag_page(frag) &&
2294 off == frag->page_offset + skb_frag_size(frag);
2299 static inline int __skb_linearize(struct sk_buff *skb)
2301 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2305 * skb_linearize - convert paged skb to linear one
2306 * @skb: buffer to linarize
2308 * If there is no free memory -ENOMEM is returned, otherwise zero
2309 * is returned and the old skb data released.
2311 static inline int skb_linearize(struct sk_buff *skb)
2313 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2317 * skb_has_shared_frag - can any frag be overwritten
2318 * @skb: buffer to test
2320 * Return true if the skb has at least one frag that might be modified
2321 * by an external entity (as in vmsplice()/sendfile())
2323 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2325 return skb_is_nonlinear(skb) &&
2326 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2330 * skb_linearize_cow - make sure skb is linear and writable
2331 * @skb: buffer to process
2333 * If there is no free memory -ENOMEM is returned, otherwise zero
2334 * is returned and the old skb data released.
2336 static inline int skb_linearize_cow(struct sk_buff *skb)
2338 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2339 __skb_linearize(skb) : 0;
2343 * skb_postpull_rcsum - update checksum for received skb after pull
2344 * @skb: buffer to update
2345 * @start: start of data before pull
2346 * @len: length of data pulled
2348 * After doing a pull on a received packet, you need to call this to
2349 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2350 * CHECKSUM_NONE so that it can be recomputed from scratch.
2353 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2354 const void *start, unsigned int len)
2356 if (skb->ip_summed == CHECKSUM_COMPLETE)
2357 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2360 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2363 * pskb_trim_rcsum - trim received skb and update checksum
2364 * @skb: buffer to trim
2367 * This is exactly the same as pskb_trim except that it ensures the
2368 * checksum of received packets are still valid after the operation.
2371 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2373 if (likely(len >= skb->len))
2375 if (skb->ip_summed == CHECKSUM_COMPLETE)
2376 skb->ip_summed = CHECKSUM_NONE;
2377 return __pskb_trim(skb, len);
2380 #define skb_queue_walk(queue, skb) \
2381 for (skb = (queue)->next; \
2382 skb != (struct sk_buff *)(queue); \
2385 #define skb_queue_walk_safe(queue, skb, tmp) \
2386 for (skb = (queue)->next, tmp = skb->next; \
2387 skb != (struct sk_buff *)(queue); \
2388 skb = tmp, tmp = skb->next)
2390 #define skb_queue_walk_from(queue, skb) \
2391 for (; skb != (struct sk_buff *)(queue); \
2394 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2395 for (tmp = skb->next; \
2396 skb != (struct sk_buff *)(queue); \
2397 skb = tmp, tmp = skb->next)
2399 #define skb_queue_reverse_walk(queue, skb) \
2400 for (skb = (queue)->prev; \
2401 skb != (struct sk_buff *)(queue); \
2404 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2405 for (skb = (queue)->prev, tmp = skb->prev; \
2406 skb != (struct sk_buff *)(queue); \
2407 skb = tmp, tmp = skb->prev)
2409 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2410 for (tmp = skb->prev; \
2411 skb != (struct sk_buff *)(queue); \
2412 skb = tmp, tmp = skb->prev)
2414 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2416 return skb_shinfo(skb)->frag_list != NULL;
2419 static inline void skb_frag_list_init(struct sk_buff *skb)
2421 skb_shinfo(skb)->frag_list = NULL;
2424 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2426 frag->next = skb_shinfo(skb)->frag_list;
2427 skb_shinfo(skb)->frag_list = frag;
2430 #define skb_walk_frags(skb, iter) \
2431 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2433 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2434 int *peeked, int *off, int *err);
2435 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2436 int noblock, int *err);
2437 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2438 struct poll_table_struct *wait);
2439 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2440 int offset, struct iovec *to,
2442 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
2445 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2447 const struct iovec *from,
2450 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2452 const struct iovec *to,
2455 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2456 extern void skb_free_datagram_locked(struct sock *sk,
2457 struct sk_buff *skb);
2458 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
2459 unsigned int flags);
2460 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2461 int len, __wsum csum);
2462 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2464 extern int skb_store_bits(struct sk_buff *skb, int offset,
2465 const void *from, int len);
2466 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
2467 int offset, u8 *to, int len,
2469 extern int skb_splice_bits(struct sk_buff *skb,
2470 unsigned int offset,
2471 struct pipe_inode_info *pipe,
2473 unsigned int flags);
2474 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2475 extern void skb_split(struct sk_buff *skb,
2476 struct sk_buff *skb1, const u32 len);
2477 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2480 extern struct sk_buff *skb_segment(struct sk_buff *skb,
2481 netdev_features_t features);
2483 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2484 int len, void *buffer)
2486 int hlen = skb_headlen(skb);
2488 if (hlen - offset >= len)
2489 return skb->data + offset;
2491 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2497 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2499 const unsigned int len)
2501 memcpy(to, skb->data, len);
2504 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2505 const int offset, void *to,
2506 const unsigned int len)
2508 memcpy(to, skb->data + offset, len);
2511 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2513 const unsigned int len)
2515 memcpy(skb->data, from, len);
2518 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2521 const unsigned int len)
2523 memcpy(skb->data + offset, from, len);
2526 extern void skb_init(void);
2528 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2534 * skb_get_timestamp - get timestamp from a skb
2535 * @skb: skb to get stamp from
2536 * @stamp: pointer to struct timeval to store stamp in
2538 * Timestamps are stored in the skb as offsets to a base timestamp.
2539 * This function converts the offset back to a struct timeval and stores
2542 static inline void skb_get_timestamp(const struct sk_buff *skb,
2543 struct timeval *stamp)
2545 *stamp = ktime_to_timeval(skb->tstamp);
2548 static inline void skb_get_timestampns(const struct sk_buff *skb,
2549 struct timespec *stamp)
2551 *stamp = ktime_to_timespec(skb->tstamp);
2554 static inline void __net_timestamp(struct sk_buff *skb)
2556 skb->tstamp = ktime_get_real();
2559 static inline ktime_t net_timedelta(ktime_t t)
2561 return ktime_sub(ktime_get_real(), t);
2564 static inline ktime_t net_invalid_timestamp(void)
2566 return ktime_set(0, 0);
2569 extern void skb_timestamping_init(void);
2571 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2573 extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2574 extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2576 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2578 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2582 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2587 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2590 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2592 * PHY drivers may accept clones of transmitted packets for
2593 * timestamping via their phy_driver.txtstamp method. These drivers
2594 * must call this function to return the skb back to the stack, with
2595 * or without a timestamp.
2597 * @skb: clone of the the original outgoing packet
2598 * @hwtstamps: hardware time stamps, may be NULL if not available
2601 void skb_complete_tx_timestamp(struct sk_buff *skb,
2602 struct skb_shared_hwtstamps *hwtstamps);
2605 * skb_tstamp_tx - queue clone of skb with send time stamps
2606 * @orig_skb: the original outgoing packet
2607 * @hwtstamps: hardware time stamps, may be NULL if not available
2609 * If the skb has a socket associated, then this function clones the
2610 * skb (thus sharing the actual data and optional structures), stores
2611 * the optional hardware time stamping information (if non NULL) or
2612 * generates a software time stamp (otherwise), then queues the clone
2613 * to the error queue of the socket. Errors are silently ignored.
2615 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2616 struct skb_shared_hwtstamps *hwtstamps);
2618 static inline void sw_tx_timestamp(struct sk_buff *skb)
2620 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2621 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2622 skb_tstamp_tx(skb, NULL);
2626 * skb_tx_timestamp() - Driver hook for transmit timestamping
2628 * Ethernet MAC Drivers should call this function in their hard_xmit()
2629 * function immediately before giving the sk_buff to the MAC hardware.
2631 * @skb: A socket buffer.
2633 static inline void skb_tx_timestamp(struct sk_buff *skb)
2635 skb_clone_tx_timestamp(skb);
2636 sw_tx_timestamp(skb);
2640 * skb_complete_wifi_ack - deliver skb with wifi status
2642 * @skb: the original outgoing packet
2643 * @acked: ack status
2646 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2648 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2649 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
2651 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2653 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2657 * skb_checksum_complete - Calculate checksum of an entire packet
2658 * @skb: packet to process
2660 * This function calculates the checksum over the entire packet plus
2661 * the value of skb->csum. The latter can be used to supply the
2662 * checksum of a pseudo header as used by TCP/UDP. It returns the
2665 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2666 * this function can be used to verify that checksum on received
2667 * packets. In that case the function should return zero if the
2668 * checksum is correct. In particular, this function will return zero
2669 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2670 * hardware has already verified the correctness of the checksum.
2672 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2674 return skb_csum_unnecessary(skb) ?
2675 0 : __skb_checksum_complete(skb);
2678 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2679 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
2680 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2682 if (nfct && atomic_dec_and_test(&nfct->use))
2683 nf_conntrack_destroy(nfct);
2685 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2688 atomic_inc(&nfct->use);
2691 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2692 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2695 atomic_inc(&skb->users);
2697 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2703 #ifdef CONFIG_BRIDGE_NETFILTER
2704 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2706 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2709 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2712 atomic_inc(&nf_bridge->use);
2714 #endif /* CONFIG_BRIDGE_NETFILTER */
2715 static inline void nf_reset(struct sk_buff *skb)
2717 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2718 nf_conntrack_put(skb->nfct);
2721 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2722 nf_conntrack_put_reasm(skb->nfct_reasm);
2723 skb->nfct_reasm = NULL;
2725 #ifdef CONFIG_BRIDGE_NETFILTER
2726 nf_bridge_put(skb->nf_bridge);
2727 skb->nf_bridge = NULL;
2731 static inline void nf_reset_trace(struct sk_buff *skb)
2733 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
2738 /* Note: This doesn't put any conntrack and bridge info in dst. */
2739 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2741 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2742 dst->nfct = src->nfct;
2743 nf_conntrack_get(src->nfct);
2744 dst->nfctinfo = src->nfctinfo;
2746 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2747 dst->nfct_reasm = src->nfct_reasm;
2748 nf_conntrack_get_reasm(src->nfct_reasm);
2750 #ifdef CONFIG_BRIDGE_NETFILTER
2751 dst->nf_bridge = src->nf_bridge;
2752 nf_bridge_get(src->nf_bridge);
2756 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2758 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2759 nf_conntrack_put(dst->nfct);
2761 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2762 nf_conntrack_put_reasm(dst->nfct_reasm);
2764 #ifdef CONFIG_BRIDGE_NETFILTER
2765 nf_bridge_put(dst->nf_bridge);
2767 __nf_copy(dst, src);
2770 #ifdef CONFIG_NETWORK_SECMARK
2771 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2773 to->secmark = from->secmark;
2776 static inline void skb_init_secmark(struct sk_buff *skb)
2781 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2784 static inline void skb_init_secmark(struct sk_buff *skb)
2788 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2790 skb->queue_mapping = queue_mapping;
2793 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2795 return skb->queue_mapping;
2798 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2800 to->queue_mapping = from->queue_mapping;
2803 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2805 skb->queue_mapping = rx_queue + 1;
2808 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2810 return skb->queue_mapping - 1;
2813 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2815 return skb->queue_mapping != 0;
2818 extern u16 __skb_tx_hash(const struct net_device *dev,
2819 const struct sk_buff *skb,
2820 unsigned int num_tx_queues);
2823 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2828 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2834 /* Keeps track of mac header offset relative to skb->head.
2835 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2836 * For non-tunnel skb it points to skb_mac_header() and for
2837 * tunnel skb it points to outer mac header. */
2841 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2843 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2845 return (skb_mac_header(inner_skb) - inner_skb->head) -
2846 SKB_GSO_CB(inner_skb)->mac_offset;
2849 static inline bool skb_is_gso(const struct sk_buff *skb)
2851 return skb_shinfo(skb)->gso_size;
2854 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
2856 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2859 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2861 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2863 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2864 * wanted then gso_type will be set. */
2865 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2867 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2868 unlikely(shinfo->gso_type == 0)) {
2869 __skb_warn_lro_forwarding(skb);
2875 static inline void skb_forward_csum(struct sk_buff *skb)
2877 /* Unfortunately we don't support this one. Any brave souls? */
2878 if (skb->ip_summed == CHECKSUM_COMPLETE)
2879 skb->ip_summed = CHECKSUM_NONE;
2883 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2884 * @skb: skb to check
2886 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2887 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2888 * use this helper, to document places where we make this assertion.
2890 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
2893 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2897 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2899 u32 __skb_get_poff(const struct sk_buff *skb);
2902 * skb_head_is_locked - Determine if the skb->head is locked down
2903 * @skb: skb to check
2905 * The head on skbs build around a head frag can be removed if they are
2906 * not cloned. This function returns true if the skb head is locked down
2907 * due to either being allocated via kmalloc, or by being a clone with
2908 * multiple references to the head.
2910 static inline bool skb_head_is_locked(const struct sk_buff *skb)
2912 return !skb->head_frag || skb_cloned(skb);
2914 #endif /* __KERNEL__ */
2915 #endif /* _LINUX_SKBUFF_H */