2 * Definitions for the 'struct sk_buff' memory handlers.
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>
36 /* Don't change this without changing skb_csum_unnecessary! */
37 #define CHECKSUM_NONE 0
38 #define CHECKSUM_UNNECESSARY 1
39 #define CHECKSUM_COMPLETE 2
40 #define CHECKSUM_PARTIAL 3
42 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
43 ~(SMP_CACHE_BYTES - 1))
44 #define SKB_WITH_OVERHEAD(X) \
45 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
46 #define SKB_MAX_ORDER(X, ORDER) \
47 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
48 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
49 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
51 /* return minimum truesize of one skb containing X bytes of data */
52 #define SKB_TRUESIZE(X) ((X) + \
53 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
54 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
56 /* A. Checksumming of received packets by device.
58 * NONE: device failed to checksum this packet.
59 * skb->csum is undefined.
61 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
62 * skb->csum is undefined.
63 * It is bad option, but, unfortunately, many of vendors do this.
64 * Apparently with secret goal to sell you new device, when you
65 * will add new protocol to your host. F.e. IPv6. 8)
67 * COMPLETE: the most generic way. Device supplied checksum of _all_
68 * the packet as seen by netif_rx in skb->csum.
69 * NOTE: Even if device supports only some protocols, but
70 * is able to produce some skb->csum, it MUST use COMPLETE,
73 * PARTIAL: identical to the case for output below. This may occur
74 * on a packet received directly from another Linux OS, e.g.,
75 * a virtualised Linux kernel on the same host. The packet can
76 * be treated in the same way as UNNECESSARY except that on
77 * output (i.e., forwarding) the checksum must be filled in
78 * by the OS or the hardware.
80 * B. Checksumming on output.
82 * NONE: skb is checksummed by protocol or csum is not required.
84 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
85 * from skb->csum_start to the end and to record the checksum
86 * at skb->csum_start + skb->csum_offset.
88 * Device must show its capabilities in dev->features, set
89 * at device setup time.
90 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
92 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
93 * TCP/UDP over IPv4. Sigh. Vendors like this
94 * way by an unknown reason. Though, see comment above
95 * about CHECKSUM_UNNECESSARY. 8)
96 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
98 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
99 * that do not want net to perform the checksum calculation should use
100 * this flag in their outgoing skbs.
101 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
102 * offload. Correspondingly, the FCoE protocol driver
103 * stack should use CHECKSUM_UNNECESSARY.
105 * Any questions? No questions, good. --ANK
110 struct pipe_inode_info;
112 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
113 struct nf_conntrack {
118 #ifdef CONFIG_BRIDGE_NETFILTER
119 struct nf_bridge_info {
122 struct net_device *physindev;
123 struct net_device *physoutdev;
124 unsigned long data[32 / sizeof(unsigned long)];
128 struct sk_buff_head {
129 /* These two members must be first. */
130 struct sk_buff *next;
131 struct sk_buff *prev;
139 /* To allow 64K frame to be packed as single skb without frag_list we
140 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
141 * buffers which do not start on a page boundary.
143 * Since GRO uses frags we allocate at least 16 regardless of page
146 #if (65536/PAGE_SIZE + 1) < 16
147 #define MAX_SKB_FRAGS 16UL
149 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
152 typedef struct skb_frag_struct skb_frag_t;
154 struct skb_frag_struct {
158 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
167 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
172 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
177 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
182 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
187 #define HAVE_HW_TIME_STAMP
190 * struct skb_shared_hwtstamps - hardware time stamps
191 * @hwtstamp: hardware time stamp transformed into duration
192 * since arbitrary point in time
193 * @syststamp: hwtstamp transformed to system time base
195 * Software time stamps generated by ktime_get_real() are stored in
196 * skb->tstamp. The relation between the different kinds of time
197 * stamps is as follows:
199 * syststamp and tstamp can be compared against each other in
200 * arbitrary combinations. The accuracy of a
201 * syststamp/tstamp/"syststamp from other device" comparison is
202 * limited by the accuracy of the transformation into system time
203 * base. This depends on the device driver and its underlying
206 * hwtstamps can only be compared against other hwtstamps from
209 * This structure is attached to packets as part of the
210 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
212 struct skb_shared_hwtstamps {
217 /* Definitions for tx_flags in struct skb_shared_info */
219 /* generate hardware time stamp */
220 SKBTX_HW_TSTAMP = 1 << 0,
222 /* generate software time stamp */
223 SKBTX_SW_TSTAMP = 1 << 1,
225 /* device driver is going to provide hardware time stamp */
226 SKBTX_IN_PROGRESS = 1 << 2,
228 /* device driver supports TX zero-copy buffers */
229 SKBTX_DEV_ZEROCOPY = 1 << 3,
231 /* generate wifi status information (where possible) */
232 SKBTX_WIFI_STATUS = 1 << 4,
234 /* This indicates at least one fragment might be overwritten
235 * (as in vmsplice(), sendfile() ...)
236 * If we need to compute a TX checksum, we'll need to copy
237 * all frags to avoid possible bad checksum
239 SKBTX_SHARED_FRAG = 1 << 5,
243 * The callback notifies userspace to release buffers when skb DMA is done in
244 * lower device, the skb last reference should be 0 when calling this.
245 * The zerocopy_success argument is true if zero copy transmit occurred,
246 * false on data copy or out of memory error caused by data copy attempt.
247 * The ctx field is used to track device context.
248 * The desc field is used to track userspace buffer index.
251 void (*callback)(struct ubuf_info *, bool zerocopy_success);
256 /* This data is invariant across clones and lives at
257 * the end of the header data, ie. at skb->end.
259 struct skb_shared_info {
260 unsigned char nr_frags;
262 unsigned short gso_size;
263 /* Warning: this field is not always filled in (UFO)! */
264 unsigned short gso_segs;
265 unsigned short gso_type;
266 struct sk_buff *frag_list;
267 struct skb_shared_hwtstamps hwtstamps;
271 * Warning : all fields before dataref are cleared in __alloc_skb()
275 /* Intermediate layers must ensure that destructor_arg
276 * remains valid until skb destructor */
277 void * destructor_arg;
279 /* must be last field, see pskb_expand_head() */
280 skb_frag_t frags[MAX_SKB_FRAGS];
283 /* We divide dataref into two halves. The higher 16 bits hold references
284 * to the payload part of skb->data. The lower 16 bits hold references to
285 * the entire skb->data. A clone of a headerless skb holds the length of
286 * the header in skb->hdr_len.
288 * All users must obey the rule that the skb->data reference count must be
289 * greater than or equal to the payload reference count.
291 * Holding a reference to the payload part means that the user does not
292 * care about modifications to the header part of skb->data.
294 #define SKB_DATAREF_SHIFT 16
295 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
299 SKB_FCLONE_UNAVAILABLE,
305 SKB_GSO_TCPV4 = 1 << 0,
306 SKB_GSO_UDP = 1 << 1,
308 /* This indicates the skb is from an untrusted source. */
309 SKB_GSO_DODGY = 1 << 2,
311 /* This indicates the tcp segment has CWR set. */
312 SKB_GSO_TCP_ECN = 1 << 3,
314 SKB_GSO_TCPV6 = 1 << 4,
316 SKB_GSO_FCOE = 1 << 5,
318 SKB_GSO_GRE = 1 << 6,
321 #if BITS_PER_LONG > 32
322 #define NET_SKBUFF_DATA_USES_OFFSET 1
325 #ifdef NET_SKBUFF_DATA_USES_OFFSET
326 typedef unsigned int sk_buff_data_t;
328 typedef unsigned char *sk_buff_data_t;
331 #if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
332 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
333 #define NET_SKBUFF_NF_DEFRAG_NEEDED 1
337 * struct sk_buff - socket buffer
338 * @next: Next buffer in list
339 * @prev: Previous buffer in list
340 * @tstamp: Time we arrived
341 * @sk: Socket we are owned by
342 * @dev: Device we arrived on/are leaving by
343 * @cb: Control buffer. Free for use by every layer. Put private vars here
344 * @_skb_refdst: destination entry (with norefcount bit)
345 * @sp: the security path, used for xfrm
346 * @len: Length of actual data
347 * @data_len: Data length
348 * @mac_len: Length of link layer header
349 * @hdr_len: writable header length of cloned skb
350 * @csum: Checksum (must include start/offset pair)
351 * @csum_start: Offset from skb->head where checksumming should start
352 * @csum_offset: Offset from csum_start where checksum should be stored
353 * @priority: Packet queueing priority
354 * @local_df: allow local fragmentation
355 * @cloned: Head may be cloned (check refcnt to be sure)
356 * @ip_summed: Driver fed us an IP checksum
357 * @nohdr: Payload reference only, must not modify header
358 * @nfctinfo: Relationship of this skb to the connection
359 * @pkt_type: Packet class
360 * @fclone: skbuff clone status
361 * @ipvs_property: skbuff is owned by ipvs
362 * @peeked: this packet has been seen already, so stats have been
363 * done for it, don't do them again
364 * @nf_trace: netfilter packet trace flag
365 * @protocol: Packet protocol from driver
366 * @destructor: Destruct function
367 * @nfct: Associated connection, if any
368 * @nfct_reasm: netfilter conntrack re-assembly pointer
369 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
370 * @skb_iif: ifindex of device we arrived on
371 * @tc_index: Traffic control index
372 * @tc_verd: traffic control verdict
373 * @rxhash: the packet hash computed on receive
374 * @queue_mapping: Queue mapping for multiqueue devices
375 * @ndisc_nodetype: router type (from link layer)
376 * @ooo_okay: allow the mapping of a socket to a queue to be changed
377 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
379 * @wifi_acked_valid: wifi_acked was set
380 * @wifi_acked: whether frame was acked on wifi or not
381 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
382 * @dma_cookie: a cookie to one of several possible DMA operations
383 * done by skb DMA functions
384 * @secmark: security marking
385 * @mark: Generic packet mark
386 * @dropcount: total number of sk_receive_queue overflows
387 * @vlan_tci: vlan tag control information
388 * @inner_transport_header: Inner transport layer header (encapsulation)
389 * @inner_network_header: Network layer header (encapsulation)
390 * @transport_header: Transport layer header
391 * @network_header: Network layer header
392 * @mac_header: Link layer header
393 * @tail: Tail pointer
395 * @head: Head of buffer
396 * @data: Data head pointer
397 * @truesize: Buffer size
398 * @users: User count - see {datagram,tcp}.c
402 /* These two members must be first. */
403 struct sk_buff *next;
404 struct sk_buff *prev;
409 struct net_device *dev;
412 * This is the control buffer. It is free to use for every
413 * layer. Please put your private variables there. If you
414 * want to keep them across layers you have to do a skb_clone()
415 * first. This is owned by whoever has the skb queued ATM.
417 char cb[48] __aligned(8);
419 unsigned long _skb_refdst;
435 kmemcheck_bitfield_begin(flags1);
446 kmemcheck_bitfield_end(flags1);
449 void (*destructor)(struct sk_buff *skb);
450 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
451 struct nf_conntrack *nfct;
453 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
454 struct sk_buff *nfct_reasm;
456 #ifdef CONFIG_BRIDGE_NETFILTER
457 struct nf_bridge_info *nf_bridge;
466 #ifdef CONFIG_NET_SCHED
467 __u16 tc_index; /* traffic control index */
468 #ifdef CONFIG_NET_CLS_ACT
469 __u16 tc_verd; /* traffic control verdict */
474 kmemcheck_bitfield_begin(flags2);
475 #ifdef CONFIG_IPV6_NDISC_NODETYPE
476 __u8 ndisc_nodetype:2;
481 __u8 wifi_acked_valid:1;
485 /* Encapsulation protocol and NIC drivers should use
486 * this flag to indicate to each other if the skb contains
487 * encapsulated packet or not and maybe use the inner packet
490 __u8 encapsulation:1;
491 /* 7/9 bit hole (depending on ndisc_nodetype presence) */
492 kmemcheck_bitfield_end(flags2);
494 #ifdef CONFIG_NET_DMA
495 dma_cookie_t dma_cookie;
497 #ifdef CONFIG_NETWORK_SECMARK
506 sk_buff_data_t inner_transport_header;
507 sk_buff_data_t inner_network_header;
508 sk_buff_data_t transport_header;
509 sk_buff_data_t network_header;
510 sk_buff_data_t mac_header;
511 /* These elements must be at the end, see alloc_skb() for details. */
516 unsigned int truesize;
522 * Handling routines are only of interest to the kernel
524 #include <linux/slab.h>
527 #define SKB_ALLOC_FCLONE 0x01
528 #define SKB_ALLOC_RX 0x02
530 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
531 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
533 return unlikely(skb->pfmemalloc);
537 * skb might have a dst pointer attached, refcounted or not.
538 * _skb_refdst low order bit is set if refcount was _not_ taken
540 #define SKB_DST_NOREF 1UL
541 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
544 * skb_dst - returns skb dst_entry
547 * Returns skb dst_entry, regardless of reference taken or not.
549 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
551 /* If refdst was not refcounted, check we still are in a
552 * rcu_read_lock section
554 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
555 !rcu_read_lock_held() &&
556 !rcu_read_lock_bh_held());
557 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
561 * skb_dst_set - sets skb dst
565 * Sets skb dst, assuming a reference was taken on dst and should
566 * be released by skb_dst_drop()
568 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
570 skb->_skb_refdst = (unsigned long)dst;
573 extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
576 * skb_dst_is_noref - Test if skb dst isn't refcounted
579 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
581 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
584 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
586 return (struct rtable *)skb_dst(skb);
589 extern void kfree_skb(struct sk_buff *skb);
590 extern void skb_tx_error(struct sk_buff *skb);
591 extern void consume_skb(struct sk_buff *skb);
592 extern void __kfree_skb(struct sk_buff *skb);
593 extern struct kmem_cache *skbuff_head_cache;
595 extern void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
596 extern bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
597 bool *fragstolen, int *delta_truesize);
599 extern struct sk_buff *__alloc_skb(unsigned int size,
600 gfp_t priority, int flags, int node);
601 extern struct sk_buff *build_skb(void *data, unsigned int frag_size);
602 static inline struct sk_buff *alloc_skb(unsigned int size,
605 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
608 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
611 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
614 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
615 extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
616 extern struct sk_buff *skb_clone(struct sk_buff *skb,
618 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
620 extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
621 int headroom, gfp_t gfp_mask);
623 extern int pskb_expand_head(struct sk_buff *skb,
624 int nhead, int ntail,
626 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
627 unsigned int headroom);
628 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
629 int newheadroom, int newtailroom,
631 extern int skb_to_sgvec(struct sk_buff *skb,
632 struct scatterlist *sg, int offset,
634 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
635 struct sk_buff **trailer);
636 extern int skb_pad(struct sk_buff *skb, int pad);
637 #define dev_kfree_skb(a) consume_skb(a)
639 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
640 int getfrag(void *from, char *to, int offset,
641 int len,int odd, struct sk_buff *skb),
642 void *from, int length);
644 struct skb_seq_state {
648 __u32 stepped_offset;
649 struct sk_buff *root_skb;
650 struct sk_buff *cur_skb;
654 extern void skb_prepare_seq_read(struct sk_buff *skb,
655 unsigned int from, unsigned int to,
656 struct skb_seq_state *st);
657 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
658 struct skb_seq_state *st);
659 extern void skb_abort_seq_read(struct skb_seq_state *st);
661 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
662 unsigned int to, struct ts_config *config,
663 struct ts_state *state);
665 extern void __skb_get_rxhash(struct sk_buff *skb);
666 static inline __u32 skb_get_rxhash(struct sk_buff *skb)
669 __skb_get_rxhash(skb);
674 #ifdef NET_SKBUFF_DATA_USES_OFFSET
675 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
677 return skb->head + skb->end;
680 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
685 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
690 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
692 return skb->end - skb->head;
697 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
699 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
701 return &skb_shinfo(skb)->hwtstamps;
705 * skb_queue_empty - check if a queue is empty
708 * Returns true if the queue is empty, false otherwise.
710 static inline int skb_queue_empty(const struct sk_buff_head *list)
712 return list->next == (struct sk_buff *)list;
716 * skb_queue_is_last - check if skb is the last entry in the queue
720 * Returns true if @skb is the last buffer on the list.
722 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
723 const struct sk_buff *skb)
725 return skb->next == (struct sk_buff *)list;
729 * skb_queue_is_first - check if skb is the first entry in the queue
733 * Returns true if @skb is the first buffer on the list.
735 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
736 const struct sk_buff *skb)
738 return skb->prev == (struct sk_buff *)list;
742 * skb_queue_next - return the next packet in the queue
744 * @skb: current buffer
746 * Return the next packet in @list after @skb. It is only valid to
747 * call this if skb_queue_is_last() evaluates to false.
749 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
750 const struct sk_buff *skb)
752 /* This BUG_ON may seem severe, but if we just return then we
753 * are going to dereference garbage.
755 BUG_ON(skb_queue_is_last(list, skb));
760 * skb_queue_prev - return the prev packet in the queue
762 * @skb: current buffer
764 * Return the prev packet in @list before @skb. It is only valid to
765 * call this if skb_queue_is_first() evaluates to false.
767 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
768 const struct sk_buff *skb)
770 /* This BUG_ON may seem severe, but if we just return then we
771 * are going to dereference garbage.
773 BUG_ON(skb_queue_is_first(list, skb));
778 * skb_get - reference buffer
779 * @skb: buffer to reference
781 * Makes another reference to a socket buffer and returns a pointer
784 static inline struct sk_buff *skb_get(struct sk_buff *skb)
786 atomic_inc(&skb->users);
791 * If users == 1, we are the only owner and are can avoid redundant
796 * skb_cloned - is the buffer a clone
797 * @skb: buffer to check
799 * Returns true if the buffer was generated with skb_clone() and is
800 * one of multiple shared copies of the buffer. Cloned buffers are
801 * shared data so must not be written to under normal circumstances.
803 static inline int skb_cloned(const struct sk_buff *skb)
805 return skb->cloned &&
806 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
809 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
811 might_sleep_if(pri & __GFP_WAIT);
814 return pskb_expand_head(skb, 0, 0, pri);
820 * skb_header_cloned - is the header a clone
821 * @skb: buffer to check
823 * Returns true if modifying the header part of the buffer requires
824 * the data to be copied.
826 static inline int skb_header_cloned(const struct sk_buff *skb)
833 dataref = atomic_read(&skb_shinfo(skb)->dataref);
834 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
839 * skb_header_release - release reference to header
840 * @skb: buffer to operate on
842 * Drop a reference to the header part of the buffer. This is done
843 * by acquiring a payload reference. You must not read from the header
844 * part of skb->data after this.
846 static inline void skb_header_release(struct sk_buff *skb)
850 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
854 * skb_shared - is the buffer shared
855 * @skb: buffer to check
857 * Returns true if more than one person has a reference to this
860 static inline int skb_shared(const struct sk_buff *skb)
862 return atomic_read(&skb->users) != 1;
866 * skb_share_check - check if buffer is shared and if so clone it
867 * @skb: buffer to check
868 * @pri: priority for memory allocation
870 * If the buffer is shared the buffer is cloned and the old copy
871 * drops a reference. A new clone with a single reference is returned.
872 * If the buffer is not shared the original buffer is returned. When
873 * being called from interrupt status or with spinlocks held pri must
876 * NULL is returned on a memory allocation failure.
878 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
880 might_sleep_if(pri & __GFP_WAIT);
881 if (skb_shared(skb)) {
882 struct sk_buff *nskb = skb_clone(skb, pri);
894 * Copy shared buffers into a new sk_buff. We effectively do COW on
895 * packets to handle cases where we have a local reader and forward
896 * and a couple of other messy ones. The normal one is tcpdumping
897 * a packet thats being forwarded.
901 * skb_unshare - make a copy of a shared buffer
902 * @skb: buffer to check
903 * @pri: priority for memory allocation
905 * If the socket buffer is a clone then this function creates a new
906 * copy of the data, drops a reference count on the old copy and returns
907 * the new copy with the reference count at 1. If the buffer is not a clone
908 * the original buffer is returned. When called with a spinlock held or
909 * from interrupt state @pri must be %GFP_ATOMIC
911 * %NULL is returned on a memory allocation failure.
913 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
916 might_sleep_if(pri & __GFP_WAIT);
917 if (skb_cloned(skb)) {
918 struct sk_buff *nskb = skb_copy(skb, pri);
919 kfree_skb(skb); /* Free our shared copy */
926 * skb_peek - peek at the head of an &sk_buff_head
927 * @list_: list to peek at
929 * Peek an &sk_buff. Unlike most other operations you _MUST_
930 * be careful with this one. A peek leaves the buffer on the
931 * list and someone else may run off with it. You must hold
932 * the appropriate locks or have a private queue to do this.
934 * Returns %NULL for an empty list or a pointer to the head element.
935 * The reference count is not incremented and the reference is therefore
936 * volatile. Use with caution.
938 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
940 struct sk_buff *skb = list_->next;
942 if (skb == (struct sk_buff *)list_)
948 * skb_peek_next - peek skb following the given one from a queue
949 * @skb: skb to start from
950 * @list_: list to peek at
952 * Returns %NULL when the end of the list is met or a pointer to the
953 * next element. The reference count is not incremented and the
954 * reference is therefore volatile. Use with caution.
956 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
957 const struct sk_buff_head *list_)
959 struct sk_buff *next = skb->next;
961 if (next == (struct sk_buff *)list_)
967 * skb_peek_tail - peek at the tail of an &sk_buff_head
968 * @list_: list to peek at
970 * Peek an &sk_buff. Unlike most other operations you _MUST_
971 * be careful with this one. A peek leaves the buffer on the
972 * list and someone else may run off with it. You must hold
973 * the appropriate locks or have a private queue to do this.
975 * Returns %NULL for an empty list or a pointer to the tail element.
976 * The reference count is not incremented and the reference is therefore
977 * volatile. Use with caution.
979 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
981 struct sk_buff *skb = list_->prev;
983 if (skb == (struct sk_buff *)list_)
990 * skb_queue_len - get queue length
991 * @list_: list to measure
993 * Return the length of an &sk_buff queue.
995 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1001 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1002 * @list: queue to initialize
1004 * This initializes only the list and queue length aspects of
1005 * an sk_buff_head object. This allows to initialize the list
1006 * aspects of an sk_buff_head without reinitializing things like
1007 * the spinlock. It can also be used for on-stack sk_buff_head
1008 * objects where the spinlock is known to not be used.
1010 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1012 list->prev = list->next = (struct sk_buff *)list;
1017 * This function creates a split out lock class for each invocation;
1018 * this is needed for now since a whole lot of users of the skb-queue
1019 * infrastructure in drivers have different locking usage (in hardirq)
1020 * than the networking core (in softirq only). In the long run either the
1021 * network layer or drivers should need annotation to consolidate the
1022 * main types of usage into 3 classes.
1024 static inline void skb_queue_head_init(struct sk_buff_head *list)
1026 spin_lock_init(&list->lock);
1027 __skb_queue_head_init(list);
1030 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1031 struct lock_class_key *class)
1033 skb_queue_head_init(list);
1034 lockdep_set_class(&list->lock, class);
1038 * Insert an sk_buff on a list.
1040 * The "__skb_xxxx()" functions are the non-atomic ones that
1041 * can only be called with interrupts disabled.
1043 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
1044 static inline void __skb_insert(struct sk_buff *newsk,
1045 struct sk_buff *prev, struct sk_buff *next,
1046 struct sk_buff_head *list)
1050 next->prev = prev->next = newsk;
1054 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1055 struct sk_buff *prev,
1056 struct sk_buff *next)
1058 struct sk_buff *first = list->next;
1059 struct sk_buff *last = list->prev;
1069 * skb_queue_splice - join two skb lists, this is designed for stacks
1070 * @list: the new list to add
1071 * @head: the place to add it in the first list
1073 static inline void skb_queue_splice(const struct sk_buff_head *list,
1074 struct sk_buff_head *head)
1076 if (!skb_queue_empty(list)) {
1077 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1078 head->qlen += list->qlen;
1083 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1084 * @list: the new list to add
1085 * @head: the place to add it in the first list
1087 * The list at @list is reinitialised
1089 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1090 struct sk_buff_head *head)
1092 if (!skb_queue_empty(list)) {
1093 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1094 head->qlen += list->qlen;
1095 __skb_queue_head_init(list);
1100 * skb_queue_splice_tail - join two skb lists, each list being a queue
1101 * @list: the new list to add
1102 * @head: the place to add it in the first list
1104 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1105 struct sk_buff_head *head)
1107 if (!skb_queue_empty(list)) {
1108 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1109 head->qlen += list->qlen;
1114 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1115 * @list: the new list to add
1116 * @head: the place to add it in the first list
1118 * Each of the lists is a queue.
1119 * The list at @list is reinitialised
1121 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1122 struct sk_buff_head *head)
1124 if (!skb_queue_empty(list)) {
1125 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1126 head->qlen += list->qlen;
1127 __skb_queue_head_init(list);
1132 * __skb_queue_after - queue a buffer at the list head
1133 * @list: list to use
1134 * @prev: place after this buffer
1135 * @newsk: buffer to queue
1137 * Queue a buffer int the middle of a list. This function takes no locks
1138 * and you must therefore hold required locks before calling it.
1140 * A buffer cannot be placed on two lists at the same time.
1142 static inline void __skb_queue_after(struct sk_buff_head *list,
1143 struct sk_buff *prev,
1144 struct sk_buff *newsk)
1146 __skb_insert(newsk, prev, prev->next, list);
1149 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1150 struct sk_buff_head *list);
1152 static inline void __skb_queue_before(struct sk_buff_head *list,
1153 struct sk_buff *next,
1154 struct sk_buff *newsk)
1156 __skb_insert(newsk, next->prev, next, list);
1160 * __skb_queue_head - queue a buffer at the list head
1161 * @list: list to use
1162 * @newsk: buffer to queue
1164 * Queue a buffer at the start of a list. This function takes no locks
1165 * and you must therefore hold required locks before calling it.
1167 * A buffer cannot be placed on two lists at the same time.
1169 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1170 static inline void __skb_queue_head(struct sk_buff_head *list,
1171 struct sk_buff *newsk)
1173 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1177 * __skb_queue_tail - queue a buffer at the list tail
1178 * @list: list to use
1179 * @newsk: buffer to queue
1181 * Queue a buffer at the end of a list. This function takes no locks
1182 * and you must therefore hold required locks before calling it.
1184 * A buffer cannot be placed on two lists at the same time.
1186 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1187 static inline void __skb_queue_tail(struct sk_buff_head *list,
1188 struct sk_buff *newsk)
1190 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1194 * remove sk_buff from list. _Must_ be called atomically, and with
1197 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1198 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1200 struct sk_buff *next, *prev;
1205 skb->next = skb->prev = NULL;
1211 * __skb_dequeue - remove from the head of the queue
1212 * @list: list to dequeue from
1214 * Remove the head of the list. This function does not take any locks
1215 * so must be used with appropriate locks held only. The head item is
1216 * returned or %NULL if the list is empty.
1218 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1219 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1221 struct sk_buff *skb = skb_peek(list);
1223 __skb_unlink(skb, list);
1228 * __skb_dequeue_tail - remove from the tail of the queue
1229 * @list: list to dequeue from
1231 * Remove the tail of the list. This function does not take any locks
1232 * so must be used with appropriate locks held only. The tail item is
1233 * returned or %NULL if the list is empty.
1235 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1236 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1238 struct sk_buff *skb = skb_peek_tail(list);
1240 __skb_unlink(skb, list);
1245 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1247 return skb->data_len;
1250 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1252 return skb->len - skb->data_len;
1255 static inline int skb_pagelen(const struct sk_buff *skb)
1259 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1260 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1261 return len + skb_headlen(skb);
1265 * __skb_fill_page_desc - initialise a paged fragment in an skb
1266 * @skb: buffer containing fragment to be initialised
1267 * @i: paged fragment index to initialise
1268 * @page: the page to use for this fragment
1269 * @off: the offset to the data with @page
1270 * @size: the length of the data
1272 * Initialises the @i'th fragment of @skb to point to &size bytes at
1273 * offset @off within @page.
1275 * Does not take any additional reference on the fragment.
1277 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1278 struct page *page, int off, int size)
1280 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1283 * Propagate page->pfmemalloc to the skb if we can. The problem is
1284 * that not all callers have unique ownership of the page. If
1285 * pfmemalloc is set, we check the mapping as a mapping implies
1286 * page->index is set (index and pfmemalloc share space).
1287 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1288 * do not lose pfmemalloc information as the pages would not be
1289 * allocated using __GFP_MEMALLOC.
1291 if (page->pfmemalloc && !page->mapping)
1292 skb->pfmemalloc = true;
1293 frag->page.p = page;
1294 frag->page_offset = off;
1295 skb_frag_size_set(frag, size);
1299 * skb_fill_page_desc - initialise a paged fragment in an skb
1300 * @skb: buffer containing fragment to be initialised
1301 * @i: paged fragment index to initialise
1302 * @page: the page to use for this fragment
1303 * @off: the offset to the data with @page
1304 * @size: the length of the data
1306 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1307 * @skb to point to &size bytes at offset @off within @page. In
1308 * addition updates @skb such that @i is the last fragment.
1310 * Does not take any additional reference on the fragment.
1312 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1313 struct page *page, int off, int size)
1315 __skb_fill_page_desc(skb, i, page, off, size);
1316 skb_shinfo(skb)->nr_frags = i + 1;
1319 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1320 int off, int size, unsigned int truesize);
1322 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1323 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1324 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1326 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1327 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1329 return skb->head + skb->tail;
1332 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1334 skb->tail = skb->data - skb->head;
1337 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1339 skb_reset_tail_pointer(skb);
1340 skb->tail += offset;
1342 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1343 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1348 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1350 skb->tail = skb->data;
1353 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1355 skb->tail = skb->data + offset;
1358 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1361 * Add data to an sk_buff
1363 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1364 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1366 unsigned char *tmp = skb_tail_pointer(skb);
1367 SKB_LINEAR_ASSERT(skb);
1373 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1374 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1381 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1382 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1385 BUG_ON(skb->len < skb->data_len);
1386 return skb->data += len;
1389 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1391 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1394 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1396 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1398 if (len > skb_headlen(skb) &&
1399 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1402 return skb->data += len;
1405 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1407 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1410 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1412 if (likely(len <= skb_headlen(skb)))
1414 if (unlikely(len > skb->len))
1416 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1420 * skb_headroom - bytes at buffer head
1421 * @skb: buffer to check
1423 * Return the number of bytes of free space at the head of an &sk_buff.
1425 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1427 return skb->data - skb->head;
1431 * skb_tailroom - bytes at buffer end
1432 * @skb: buffer to check
1434 * Return the number of bytes of free space at the tail of an sk_buff
1436 static inline int skb_tailroom(const struct sk_buff *skb)
1438 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1442 * skb_availroom - bytes at buffer end
1443 * @skb: buffer to check
1445 * Return the number of bytes of free space at the tail of an sk_buff
1446 * allocated by sk_stream_alloc()
1448 static inline int skb_availroom(const struct sk_buff *skb)
1450 return skb_is_nonlinear(skb) ? 0 : skb->avail_size - skb->len;
1454 * skb_reserve - adjust headroom
1455 * @skb: buffer to alter
1456 * @len: bytes to move
1458 * Increase the headroom of an empty &sk_buff by reducing the tail
1459 * room. This is only allowed for an empty buffer.
1461 static inline void skb_reserve(struct sk_buff *skb, int len)
1467 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1469 skb->inner_network_header = skb->network_header;
1470 skb->inner_transport_header = skb->transport_header;
1473 static inline void skb_reset_mac_len(struct sk_buff *skb)
1475 skb->mac_len = skb->network_header - skb->mac_header;
1478 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1479 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1482 return skb->head + skb->inner_transport_header;
1485 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1487 skb->inner_transport_header = skb->data - skb->head;
1490 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1493 skb_reset_inner_transport_header(skb);
1494 skb->inner_transport_header += offset;
1497 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1499 return skb->head + skb->inner_network_header;
1502 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1504 skb->inner_network_header = skb->data - skb->head;
1507 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1510 skb_reset_inner_network_header(skb);
1511 skb->inner_network_header += offset;
1514 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1516 return skb->transport_header != ~0U;
1519 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1521 return skb->head + skb->transport_header;
1524 static inline void skb_reset_transport_header(struct sk_buff *skb)
1526 skb->transport_header = skb->data - skb->head;
1529 static inline void skb_set_transport_header(struct sk_buff *skb,
1532 skb_reset_transport_header(skb);
1533 skb->transport_header += offset;
1536 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1538 return skb->head + skb->network_header;
1541 static inline void skb_reset_network_header(struct sk_buff *skb)
1543 skb->network_header = skb->data - skb->head;
1546 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1548 skb_reset_network_header(skb);
1549 skb->network_header += offset;
1552 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1554 return skb->head + skb->mac_header;
1557 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1559 return skb->mac_header != ~0U;
1562 static inline void skb_reset_mac_header(struct sk_buff *skb)
1564 skb->mac_header = skb->data - skb->head;
1567 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1569 skb_reset_mac_header(skb);
1570 skb->mac_header += offset;
1573 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1574 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1577 return skb->inner_transport_header;
1580 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1582 skb->inner_transport_header = skb->data;
1585 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1588 skb->inner_transport_header = skb->data + offset;
1591 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1593 return skb->inner_network_header;
1596 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1598 skb->inner_network_header = skb->data;
1601 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1604 skb->inner_network_header = skb->data + offset;
1607 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1609 return skb->transport_header != NULL;
1612 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1614 return skb->transport_header;
1617 static inline void skb_reset_transport_header(struct sk_buff *skb)
1619 skb->transport_header = skb->data;
1622 static inline void skb_set_transport_header(struct sk_buff *skb,
1625 skb->transport_header = skb->data + offset;
1628 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1630 return skb->network_header;
1633 static inline void skb_reset_network_header(struct sk_buff *skb)
1635 skb->network_header = skb->data;
1638 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1640 skb->network_header = skb->data + offset;
1643 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1645 return skb->mac_header;
1648 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1650 return skb->mac_header != NULL;
1653 static inline void skb_reset_mac_header(struct sk_buff *skb)
1655 skb->mac_header = skb->data;
1658 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1660 skb->mac_header = skb->data + offset;
1662 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1664 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1666 if (skb_mac_header_was_set(skb)) {
1667 const unsigned char *old_mac = skb_mac_header(skb);
1669 skb_set_mac_header(skb, -skb->mac_len);
1670 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1674 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1676 return skb->csum_start - skb_headroom(skb);
1679 static inline int skb_transport_offset(const struct sk_buff *skb)
1681 return skb_transport_header(skb) - skb->data;
1684 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1686 return skb->transport_header - skb->network_header;
1689 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1691 return skb->inner_transport_header - skb->inner_network_header;
1694 static inline int skb_network_offset(const struct sk_buff *skb)
1696 return skb_network_header(skb) - skb->data;
1699 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1701 return skb_inner_network_header(skb) - skb->data;
1704 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1706 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1710 * CPUs often take a performance hit when accessing unaligned memory
1711 * locations. The actual performance hit varies, it can be small if the
1712 * hardware handles it or large if we have to take an exception and fix it
1715 * Since an ethernet header is 14 bytes network drivers often end up with
1716 * the IP header at an unaligned offset. The IP header can be aligned by
1717 * shifting the start of the packet by 2 bytes. Drivers should do this
1720 * skb_reserve(skb, NET_IP_ALIGN);
1722 * The downside to this alignment of the IP header is that the DMA is now
1723 * unaligned. On some architectures the cost of an unaligned DMA is high
1724 * and this cost outweighs the gains made by aligning the IP header.
1726 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1729 #ifndef NET_IP_ALIGN
1730 #define NET_IP_ALIGN 2
1734 * The networking layer reserves some headroom in skb data (via
1735 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1736 * the header has to grow. In the default case, if the header has to grow
1737 * 32 bytes or less we avoid the reallocation.
1739 * Unfortunately this headroom changes the DMA alignment of the resulting
1740 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1741 * on some architectures. An architecture can override this value,
1742 * perhaps setting it to a cacheline in size (since that will maintain
1743 * cacheline alignment of the DMA). It must be a power of 2.
1745 * Various parts of the networking layer expect at least 32 bytes of
1746 * headroom, you should not reduce this.
1748 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1749 * to reduce average number of cache lines per packet.
1750 * get_rps_cpus() for example only access one 64 bytes aligned block :
1751 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1754 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1757 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1759 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1761 if (unlikely(skb_is_nonlinear(skb))) {
1766 skb_set_tail_pointer(skb, len);
1769 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1771 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1774 return ___pskb_trim(skb, len);
1775 __skb_trim(skb, len);
1779 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1781 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1785 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1786 * @skb: buffer to alter
1789 * This is identical to pskb_trim except that the caller knows that
1790 * the skb is not cloned so we should never get an error due to out-
1793 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1795 int err = pskb_trim(skb, len);
1800 * skb_orphan - orphan a buffer
1801 * @skb: buffer to orphan
1803 * If a buffer currently has an owner then we call the owner's
1804 * destructor function and make the @skb unowned. The buffer continues
1805 * to exist but is no longer charged to its former owner.
1807 static inline void skb_orphan(struct sk_buff *skb)
1809 if (skb->destructor)
1810 skb->destructor(skb);
1811 skb->destructor = NULL;
1816 * skb_orphan_frags - orphan the frags contained in a buffer
1817 * @skb: buffer to orphan frags from
1818 * @gfp_mask: allocation mask for replacement pages
1820 * For each frag in the SKB which needs a destructor (i.e. has an
1821 * owner) create a copy of that frag and release the original
1822 * page by calling the destructor.
1824 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1826 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1828 return skb_copy_ubufs(skb, gfp_mask);
1832 * __skb_queue_purge - empty a list
1833 * @list: list to empty
1835 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1836 * the list and one reference dropped. This function does not take the
1837 * list lock and the caller must hold the relevant locks to use it.
1839 extern void skb_queue_purge(struct sk_buff_head *list);
1840 static inline void __skb_queue_purge(struct sk_buff_head *list)
1842 struct sk_buff *skb;
1843 while ((skb = __skb_dequeue(list)) != NULL)
1847 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1848 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1849 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1851 extern void *netdev_alloc_frag(unsigned int fragsz);
1853 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1854 unsigned int length,
1858 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1859 * @dev: network device to receive on
1860 * @length: length to allocate
1862 * Allocate a new &sk_buff and assign it a usage count of one. The
1863 * buffer has unspecified headroom built in. Users should allocate
1864 * the headroom they think they need without accounting for the
1865 * built in space. The built in space is used for optimisations.
1867 * %NULL is returned if there is no free memory. Although this function
1868 * allocates memory it can be called from an interrupt.
1870 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1871 unsigned int length)
1873 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1876 /* legacy helper around __netdev_alloc_skb() */
1877 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1880 return __netdev_alloc_skb(NULL, length, gfp_mask);
1883 /* legacy helper around netdev_alloc_skb() */
1884 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1886 return netdev_alloc_skb(NULL, length);
1890 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1891 unsigned int length, gfp_t gfp)
1893 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
1895 if (NET_IP_ALIGN && skb)
1896 skb_reserve(skb, NET_IP_ALIGN);
1900 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1901 unsigned int length)
1903 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1907 * __skb_alloc_page - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1908 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1909 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1910 * @order: size of the allocation
1912 * Allocate a new page.
1914 * %NULL is returned if there is no free memory.
1916 static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1917 struct sk_buff *skb,
1922 gfp_mask |= __GFP_COLD;
1924 if (!(gfp_mask & __GFP_NOMEMALLOC))
1925 gfp_mask |= __GFP_MEMALLOC;
1927 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
1928 if (skb && page && page->pfmemalloc)
1929 skb->pfmemalloc = true;
1935 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
1936 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1937 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1939 * Allocate a new page.
1941 * %NULL is returned if there is no free memory.
1943 static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
1944 struct sk_buff *skb)
1946 return __skb_alloc_pages(gfp_mask, skb, 0);
1950 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
1951 * @page: The page that was allocated from skb_alloc_page
1952 * @skb: The skb that may need pfmemalloc set
1954 static inline void skb_propagate_pfmemalloc(struct page *page,
1955 struct sk_buff *skb)
1957 if (page && page->pfmemalloc)
1958 skb->pfmemalloc = true;
1962 * skb_frag_page - retrieve the page refered to by a paged fragment
1963 * @frag: the paged fragment
1965 * Returns the &struct page associated with @frag.
1967 static inline struct page *skb_frag_page(const skb_frag_t *frag)
1969 return frag->page.p;
1973 * __skb_frag_ref - take an addition reference on a paged fragment.
1974 * @frag: the paged fragment
1976 * Takes an additional reference on the paged fragment @frag.
1978 static inline void __skb_frag_ref(skb_frag_t *frag)
1980 get_page(skb_frag_page(frag));
1984 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1986 * @f: the fragment offset.
1988 * Takes an additional reference on the @f'th paged fragment of @skb.
1990 static inline void skb_frag_ref(struct sk_buff *skb, int f)
1992 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1996 * __skb_frag_unref - release a reference on a paged fragment.
1997 * @frag: the paged fragment
1999 * Releases a reference on the paged fragment @frag.
2001 static inline void __skb_frag_unref(skb_frag_t *frag)
2003 put_page(skb_frag_page(frag));
2007 * skb_frag_unref - release a reference on a paged fragment of an skb.
2009 * @f: the fragment offset
2011 * Releases a reference on the @f'th paged fragment of @skb.
2013 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2015 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2019 * skb_frag_address - gets the address of the data contained in a paged fragment
2020 * @frag: the paged fragment buffer
2022 * Returns the address of the data within @frag. The page must already
2025 static inline void *skb_frag_address(const skb_frag_t *frag)
2027 return page_address(skb_frag_page(frag)) + frag->page_offset;
2031 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2032 * @frag: the paged fragment buffer
2034 * Returns the address of the data within @frag. Checks that the page
2035 * is mapped and returns %NULL otherwise.
2037 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2039 void *ptr = page_address(skb_frag_page(frag));
2043 return ptr + frag->page_offset;
2047 * __skb_frag_set_page - sets the page contained in a paged fragment
2048 * @frag: the paged fragment
2049 * @page: the page to set
2051 * Sets the fragment @frag to contain @page.
2053 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2055 frag->page.p = page;
2059 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2061 * @f: the fragment offset
2062 * @page: the page to set
2064 * Sets the @f'th fragment of @skb to contain @page.
2066 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2069 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2073 * skb_frag_dma_map - maps a paged fragment via the DMA API
2074 * @dev: the device to map the fragment to
2075 * @frag: the paged fragment to map
2076 * @offset: the offset within the fragment (starting at the
2077 * fragment's own offset)
2078 * @size: the number of bytes to map
2079 * @dir: the direction of the mapping (%PCI_DMA_*)
2081 * Maps the page associated with @frag to @device.
2083 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2084 const skb_frag_t *frag,
2085 size_t offset, size_t size,
2086 enum dma_data_direction dir)
2088 return dma_map_page(dev, skb_frag_page(frag),
2089 frag->page_offset + offset, size, dir);
2092 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2095 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2099 * skb_clone_writable - is the header of a clone writable
2100 * @skb: buffer to check
2101 * @len: length up to which to write
2103 * Returns true if modifying the header part of the cloned buffer
2104 * does not requires the data to be copied.
2106 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2108 return !skb_header_cloned(skb) &&
2109 skb_headroom(skb) + len <= skb->hdr_len;
2112 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2117 if (headroom > skb_headroom(skb))
2118 delta = headroom - skb_headroom(skb);
2120 if (delta || cloned)
2121 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2127 * skb_cow - copy header of skb when it is required
2128 * @skb: buffer to cow
2129 * @headroom: needed headroom
2131 * If the skb passed lacks sufficient headroom or its data part
2132 * is shared, data is reallocated. If reallocation fails, an error
2133 * is returned and original skb is not changed.
2135 * The result is skb with writable area skb->head...skb->tail
2136 * and at least @headroom of space at head.
2138 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2140 return __skb_cow(skb, headroom, skb_cloned(skb));
2144 * skb_cow_head - skb_cow but only making the head writable
2145 * @skb: buffer to cow
2146 * @headroom: needed headroom
2148 * This function is identical to skb_cow except that we replace the
2149 * skb_cloned check by skb_header_cloned. It should be used when
2150 * you only need to push on some header and do not need to modify
2153 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2155 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2159 * skb_padto - pad an skbuff up to a minimal size
2160 * @skb: buffer to pad
2161 * @len: minimal length
2163 * Pads up a buffer to ensure the trailing bytes exist and are
2164 * blanked. If the buffer already contains sufficient data it
2165 * is untouched. Otherwise it is extended. Returns zero on
2166 * success. The skb is freed on error.
2169 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2171 unsigned int size = skb->len;
2172 if (likely(size >= len))
2174 return skb_pad(skb, len - size);
2177 static inline int skb_add_data(struct sk_buff *skb,
2178 char __user *from, int copy)
2180 const int off = skb->len;
2182 if (skb->ip_summed == CHECKSUM_NONE) {
2184 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
2187 skb->csum = csum_block_add(skb->csum, csum, off);
2190 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2193 __skb_trim(skb, off);
2197 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2198 const struct page *page, int off)
2201 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2203 return page == skb_frag_page(frag) &&
2204 off == frag->page_offset + skb_frag_size(frag);
2209 static inline int __skb_linearize(struct sk_buff *skb)
2211 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2215 * skb_linearize - convert paged skb to linear one
2216 * @skb: buffer to linarize
2218 * If there is no free memory -ENOMEM is returned, otherwise zero
2219 * is returned and the old skb data released.
2221 static inline int skb_linearize(struct sk_buff *skb)
2223 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2227 * skb_has_shared_frag - can any frag be overwritten
2228 * @skb: buffer to test
2230 * Return true if the skb has at least one frag that might be modified
2231 * by an external entity (as in vmsplice()/sendfile())
2233 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2235 return skb_is_nonlinear(skb) &&
2236 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2240 * skb_linearize_cow - make sure skb is linear and writable
2241 * @skb: buffer to process
2243 * If there is no free memory -ENOMEM is returned, otherwise zero
2244 * is returned and the old skb data released.
2246 static inline int skb_linearize_cow(struct sk_buff *skb)
2248 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2249 __skb_linearize(skb) : 0;
2253 * skb_postpull_rcsum - update checksum for received skb after pull
2254 * @skb: buffer to update
2255 * @start: start of data before pull
2256 * @len: length of data pulled
2258 * After doing a pull on a received packet, you need to call this to
2259 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2260 * CHECKSUM_NONE so that it can be recomputed from scratch.
2263 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2264 const void *start, unsigned int len)
2266 if (skb->ip_summed == CHECKSUM_COMPLETE)
2267 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2270 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2273 * pskb_trim_rcsum - trim received skb and update checksum
2274 * @skb: buffer to trim
2277 * This is exactly the same as pskb_trim except that it ensures the
2278 * checksum of received packets are still valid after the operation.
2281 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2283 if (likely(len >= skb->len))
2285 if (skb->ip_summed == CHECKSUM_COMPLETE)
2286 skb->ip_summed = CHECKSUM_NONE;
2287 return __pskb_trim(skb, len);
2290 #define skb_queue_walk(queue, skb) \
2291 for (skb = (queue)->next; \
2292 skb != (struct sk_buff *)(queue); \
2295 #define skb_queue_walk_safe(queue, skb, tmp) \
2296 for (skb = (queue)->next, tmp = skb->next; \
2297 skb != (struct sk_buff *)(queue); \
2298 skb = tmp, tmp = skb->next)
2300 #define skb_queue_walk_from(queue, skb) \
2301 for (; skb != (struct sk_buff *)(queue); \
2304 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2305 for (tmp = skb->next; \
2306 skb != (struct sk_buff *)(queue); \
2307 skb = tmp, tmp = skb->next)
2309 #define skb_queue_reverse_walk(queue, skb) \
2310 for (skb = (queue)->prev; \
2311 skb != (struct sk_buff *)(queue); \
2314 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2315 for (skb = (queue)->prev, tmp = skb->prev; \
2316 skb != (struct sk_buff *)(queue); \
2317 skb = tmp, tmp = skb->prev)
2319 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2320 for (tmp = skb->prev; \
2321 skb != (struct sk_buff *)(queue); \
2322 skb = tmp, tmp = skb->prev)
2324 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2326 return skb_shinfo(skb)->frag_list != NULL;
2329 static inline void skb_frag_list_init(struct sk_buff *skb)
2331 skb_shinfo(skb)->frag_list = NULL;
2334 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2336 frag->next = skb_shinfo(skb)->frag_list;
2337 skb_shinfo(skb)->frag_list = frag;
2340 #define skb_walk_frags(skb, iter) \
2341 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2343 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2344 int *peeked, int *off, int *err);
2345 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2346 int noblock, int *err);
2347 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2348 struct poll_table_struct *wait);
2349 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2350 int offset, struct iovec *to,
2352 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
2355 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2357 const struct iovec *from,
2360 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2362 const struct iovec *to,
2365 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2366 extern void skb_free_datagram_locked(struct sock *sk,
2367 struct sk_buff *skb);
2368 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
2369 unsigned int flags);
2370 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2371 int len, __wsum csum);
2372 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2374 extern int skb_store_bits(struct sk_buff *skb, int offset,
2375 const void *from, int len);
2376 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
2377 int offset, u8 *to, int len,
2379 extern int skb_splice_bits(struct sk_buff *skb,
2380 unsigned int offset,
2381 struct pipe_inode_info *pipe,
2383 unsigned int flags);
2384 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2385 extern void skb_split(struct sk_buff *skb,
2386 struct sk_buff *skb1, const u32 len);
2387 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2390 extern struct sk_buff *skb_segment(struct sk_buff *skb,
2391 netdev_features_t features);
2393 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2394 int len, void *buffer)
2396 int hlen = skb_headlen(skb);
2398 if (hlen - offset >= len)
2399 return skb->data + offset;
2401 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2407 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2409 const unsigned int len)
2411 memcpy(to, skb->data, len);
2414 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2415 const int offset, void *to,
2416 const unsigned int len)
2418 memcpy(to, skb->data + offset, len);
2421 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2423 const unsigned int len)
2425 memcpy(skb->data, from, len);
2428 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2431 const unsigned int len)
2433 memcpy(skb->data + offset, from, len);
2436 extern void skb_init(void);
2438 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2444 * skb_get_timestamp - get timestamp from a skb
2445 * @skb: skb to get stamp from
2446 * @stamp: pointer to struct timeval to store stamp in
2448 * Timestamps are stored in the skb as offsets to a base timestamp.
2449 * This function converts the offset back to a struct timeval and stores
2452 static inline void skb_get_timestamp(const struct sk_buff *skb,
2453 struct timeval *stamp)
2455 *stamp = ktime_to_timeval(skb->tstamp);
2458 static inline void skb_get_timestampns(const struct sk_buff *skb,
2459 struct timespec *stamp)
2461 *stamp = ktime_to_timespec(skb->tstamp);
2464 static inline void __net_timestamp(struct sk_buff *skb)
2466 skb->tstamp = ktime_get_real();
2469 static inline ktime_t net_timedelta(ktime_t t)
2471 return ktime_sub(ktime_get_real(), t);
2474 static inline ktime_t net_invalid_timestamp(void)
2476 return ktime_set(0, 0);
2479 extern void skb_timestamping_init(void);
2481 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2483 extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2484 extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2486 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2488 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2492 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2497 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2500 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2502 * PHY drivers may accept clones of transmitted packets for
2503 * timestamping via their phy_driver.txtstamp method. These drivers
2504 * must call this function to return the skb back to the stack, with
2505 * or without a timestamp.
2507 * @skb: clone of the the original outgoing packet
2508 * @hwtstamps: hardware time stamps, may be NULL if not available
2511 void skb_complete_tx_timestamp(struct sk_buff *skb,
2512 struct skb_shared_hwtstamps *hwtstamps);
2515 * skb_tstamp_tx - queue clone of skb with send time stamps
2516 * @orig_skb: the original outgoing packet
2517 * @hwtstamps: hardware time stamps, may be NULL if not available
2519 * If the skb has a socket associated, then this function clones the
2520 * skb (thus sharing the actual data and optional structures), stores
2521 * the optional hardware time stamping information (if non NULL) or
2522 * generates a software time stamp (otherwise), then queues the clone
2523 * to the error queue of the socket. Errors are silently ignored.
2525 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2526 struct skb_shared_hwtstamps *hwtstamps);
2528 static inline void sw_tx_timestamp(struct sk_buff *skb)
2530 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2531 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2532 skb_tstamp_tx(skb, NULL);
2536 * skb_tx_timestamp() - Driver hook for transmit timestamping
2538 * Ethernet MAC Drivers should call this function in their hard_xmit()
2539 * function immediately before giving the sk_buff to the MAC hardware.
2541 * @skb: A socket buffer.
2543 static inline void skb_tx_timestamp(struct sk_buff *skb)
2545 skb_clone_tx_timestamp(skb);
2546 sw_tx_timestamp(skb);
2550 * skb_complete_wifi_ack - deliver skb with wifi status
2552 * @skb: the original outgoing packet
2553 * @acked: ack status
2556 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2558 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2559 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
2561 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2563 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2567 * skb_checksum_complete - Calculate checksum of an entire packet
2568 * @skb: packet to process
2570 * This function calculates the checksum over the entire packet plus
2571 * the value of skb->csum. The latter can be used to supply the
2572 * checksum of a pseudo header as used by TCP/UDP. It returns the
2575 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2576 * this function can be used to verify that checksum on received
2577 * packets. In that case the function should return zero if the
2578 * checksum is correct. In particular, this function will return zero
2579 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2580 * hardware has already verified the correctness of the checksum.
2582 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2584 return skb_csum_unnecessary(skb) ?
2585 0 : __skb_checksum_complete(skb);
2588 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2589 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
2590 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2592 if (nfct && atomic_dec_and_test(&nfct->use))
2593 nf_conntrack_destroy(nfct);
2595 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2598 atomic_inc(&nfct->use);
2601 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2602 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2605 atomic_inc(&skb->users);
2607 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2613 #ifdef CONFIG_BRIDGE_NETFILTER
2614 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2616 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2619 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2622 atomic_inc(&nf_bridge->use);
2624 #endif /* CONFIG_BRIDGE_NETFILTER */
2625 static inline void nf_reset(struct sk_buff *skb)
2627 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2628 nf_conntrack_put(skb->nfct);
2631 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2632 nf_conntrack_put_reasm(skb->nfct_reasm);
2633 skb->nfct_reasm = NULL;
2635 #ifdef CONFIG_BRIDGE_NETFILTER
2636 nf_bridge_put(skb->nf_bridge);
2637 skb->nf_bridge = NULL;
2641 /* Note: This doesn't put any conntrack and bridge info in dst. */
2642 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2644 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2645 dst->nfct = src->nfct;
2646 nf_conntrack_get(src->nfct);
2647 dst->nfctinfo = src->nfctinfo;
2649 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2650 dst->nfct_reasm = src->nfct_reasm;
2651 nf_conntrack_get_reasm(src->nfct_reasm);
2653 #ifdef CONFIG_BRIDGE_NETFILTER
2654 dst->nf_bridge = src->nf_bridge;
2655 nf_bridge_get(src->nf_bridge);
2659 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2661 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2662 nf_conntrack_put(dst->nfct);
2664 #ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
2665 nf_conntrack_put_reasm(dst->nfct_reasm);
2667 #ifdef CONFIG_BRIDGE_NETFILTER
2668 nf_bridge_put(dst->nf_bridge);
2670 __nf_copy(dst, src);
2673 #ifdef CONFIG_NETWORK_SECMARK
2674 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2676 to->secmark = from->secmark;
2679 static inline void skb_init_secmark(struct sk_buff *skb)
2684 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2687 static inline void skb_init_secmark(struct sk_buff *skb)
2691 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2693 skb->queue_mapping = queue_mapping;
2696 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2698 return skb->queue_mapping;
2701 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2703 to->queue_mapping = from->queue_mapping;
2706 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2708 skb->queue_mapping = rx_queue + 1;
2711 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2713 return skb->queue_mapping - 1;
2716 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2718 return skb->queue_mapping != 0;
2721 extern u16 __skb_tx_hash(const struct net_device *dev,
2722 const struct sk_buff *skb,
2723 unsigned int num_tx_queues);
2726 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2731 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2737 /* Keeps track of mac header offset relative to skb->head.
2738 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2739 * For non-tunnel skb it points to skb_mac_header() and for
2740 * tunnel skb it points to outer mac header. */
2744 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2746 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2748 return (skb_mac_header(inner_skb) - inner_skb->head) -
2749 SKB_GSO_CB(inner_skb)->mac_offset;
2752 static inline bool skb_is_gso(const struct sk_buff *skb)
2754 return skb_shinfo(skb)->gso_size;
2757 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
2759 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2762 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2764 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2766 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2767 * wanted then gso_type will be set. */
2768 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2770 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2771 unlikely(shinfo->gso_type == 0)) {
2772 __skb_warn_lro_forwarding(skb);
2778 static inline void skb_forward_csum(struct sk_buff *skb)
2780 /* Unfortunately we don't support this one. Any brave souls? */
2781 if (skb->ip_summed == CHECKSUM_COMPLETE)
2782 skb->ip_summed = CHECKSUM_NONE;
2786 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2787 * @skb: skb to check
2789 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2790 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2791 * use this helper, to document places where we make this assertion.
2793 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
2796 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2800 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2803 * skb_head_is_locked - Determine if the skb->head is locked down
2804 * @skb: skb to check
2806 * The head on skbs build around a head frag can be removed if they are
2807 * not cloned. This function returns true if the skb head is locked down
2808 * due to either being allocated via kmalloc, or by being a clone with
2809 * multiple references to the head.
2811 static inline bool skb_head_is_locked(const struct sk_buff *skb)
2813 return !skb->head_frag || skb_cloned(skb);
2815 #endif /* __KERNEL__ */
2816 #endif /* _LINUX_SKBUFF_H */