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_IPIP = 1 << 7,
323 SKB_GSO_SIT = 1 << 8,
325 SKB_GSO_UDP_TUNNEL = 1 << 9,
327 SKB_GSO_MPLS = 1 << 10,
330 #if BITS_PER_LONG > 32
331 #define NET_SKBUFF_DATA_USES_OFFSET 1
334 #ifdef NET_SKBUFF_DATA_USES_OFFSET
335 typedef unsigned int sk_buff_data_t;
337 typedef unsigned char *sk_buff_data_t;
341 * struct sk_buff - socket buffer
342 * @next: Next buffer in list
343 * @prev: Previous buffer in list
344 * @tstamp: Time we arrived
345 * @sk: Socket we are owned by
346 * @dev: Device we arrived on/are leaving by
347 * @cb: Control buffer. Free for use by every layer. Put private vars here
348 * @_skb_refdst: destination entry (with norefcount bit)
349 * @sp: the security path, used for xfrm
350 * @len: Length of actual data
351 * @data_len: Data length
352 * @mac_len: Length of link layer header
353 * @hdr_len: writable header length of cloned skb
354 * @csum: Checksum (must include start/offset pair)
355 * @csum_start: Offset from skb->head where checksumming should start
356 * @csum_offset: Offset from csum_start where checksum should be stored
357 * @priority: Packet queueing priority
358 * @local_df: allow local fragmentation
359 * @cloned: Head may be cloned (check refcnt to be sure)
360 * @ip_summed: Driver fed us an IP checksum
361 * @nohdr: Payload reference only, must not modify header
362 * @nfctinfo: Relationship of this skb to the connection
363 * @pkt_type: Packet class
364 * @fclone: skbuff clone status
365 * @ipvs_property: skbuff is owned by ipvs
366 * @peeked: this packet has been seen already, so stats have been
367 * done for it, don't do them again
368 * @nf_trace: netfilter packet trace flag
369 * @protocol: Packet protocol from driver
370 * @destructor: Destruct function
371 * @nfct: Associated connection, if any
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 * @napi_id: id of the NAPI struct this skb came from
388 * @secmark: security marking
389 * @mark: Generic packet mark
390 * @dropcount: total number of sk_receive_queue overflows
391 * @vlan_proto: vlan encapsulation protocol
392 * @vlan_tci: vlan tag control information
393 * @inner_protocol: Protocol (encapsulation)
394 * @inner_transport_header: Inner transport layer header (encapsulation)
395 * @inner_network_header: Network layer header (encapsulation)
396 * @inner_mac_header: Link layer header (encapsulation)
397 * @transport_header: Transport layer header
398 * @network_header: Network layer header
399 * @mac_header: Link layer header
400 * @tail: Tail pointer
402 * @head: Head of buffer
403 * @data: Data head pointer
404 * @truesize: Buffer size
405 * @users: User count - see {datagram,tcp}.c
409 /* These two members must be first. */
410 struct sk_buff *next;
411 struct sk_buff *prev;
416 struct net_device *dev;
419 * This is the control buffer. It is free to use for every
420 * layer. Please put your private variables there. If you
421 * want to keep them across layers you have to do a skb_clone()
422 * first. This is owned by whoever has the skb queued ATM.
424 char cb[48] __aligned(8);
426 unsigned long _skb_refdst;
442 kmemcheck_bitfield_begin(flags1);
453 kmemcheck_bitfield_end(flags1);
456 void (*destructor)(struct sk_buff *skb);
457 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
458 struct nf_conntrack *nfct;
460 #ifdef CONFIG_BRIDGE_NETFILTER
461 struct nf_bridge_info *nf_bridge;
471 #ifdef CONFIG_NET_SCHED
472 __u16 tc_index; /* traffic control index */
473 #ifdef CONFIG_NET_CLS_ACT
474 __u16 tc_verd; /* traffic control verdict */
479 kmemcheck_bitfield_begin(flags2);
480 #ifdef CONFIG_IPV6_NDISC_NODETYPE
481 __u8 ndisc_nodetype:2;
486 __u8 wifi_acked_valid:1;
490 /* Encapsulation protocol and NIC drivers should use
491 * this flag to indicate to each other if the skb contains
492 * encapsulated packet or not and maybe use the inner packet
495 __u8 encapsulation:1;
496 /* 6/8 bit hole (depending on ndisc_nodetype presence) */
497 kmemcheck_bitfield_end(flags2);
499 #if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
501 unsigned int napi_id;
502 dma_cookie_t dma_cookie;
505 #ifdef CONFIG_NETWORK_SECMARK
511 __u32 reserved_tailroom;
514 __be16 inner_protocol;
515 __u16 inner_transport_header;
516 __u16 inner_network_header;
517 __u16 inner_mac_header;
518 __u16 transport_header;
519 __u16 network_header;
521 /* These elements must be at the end, see alloc_skb() for details. */
526 unsigned int truesize;
532 * Handling routines are only of interest to the kernel
534 #include <linux/slab.h>
537 #define SKB_ALLOC_FCLONE 0x01
538 #define SKB_ALLOC_RX 0x02
540 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
541 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
543 return unlikely(skb->pfmemalloc);
547 * skb might have a dst pointer attached, refcounted or not.
548 * _skb_refdst low order bit is set if refcount was _not_ taken
550 #define SKB_DST_NOREF 1UL
551 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
554 * skb_dst - returns skb dst_entry
557 * Returns skb dst_entry, regardless of reference taken or not.
559 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
561 /* If refdst was not refcounted, check we still are in a
562 * rcu_read_lock section
564 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
565 !rcu_read_lock_held() &&
566 !rcu_read_lock_bh_held());
567 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
571 * skb_dst_set - sets skb dst
575 * Sets skb dst, assuming a reference was taken on dst and should
576 * be released by skb_dst_drop()
578 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
580 skb->_skb_refdst = (unsigned long)dst;
583 void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
587 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
591 * Sets skb dst, assuming a reference was not taken on dst.
592 * If dst entry is cached, we do not take reference and dst_release
593 * will be avoided by refdst_drop. If dst entry is not cached, we take
594 * reference, so that last dst_release can destroy the dst immediately.
596 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
598 __skb_dst_set_noref(skb, dst, false);
602 * skb_dst_set_noref_force - sets skb dst, without taking reference
606 * Sets skb dst, assuming a reference was not taken on dst.
607 * No reference is taken and no dst_release will be called. While for
608 * cached dsts deferred reclaim is a basic feature, for entries that are
609 * not cached it is caller's job to guarantee that last dst_release for
610 * provided dst happens when nobody uses it, eg. after a RCU grace period.
612 static inline void skb_dst_set_noref_force(struct sk_buff *skb,
613 struct dst_entry *dst)
615 __skb_dst_set_noref(skb, dst, true);
619 * skb_dst_is_noref - Test if skb dst isn't refcounted
622 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
624 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
627 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
629 return (struct rtable *)skb_dst(skb);
632 void kfree_skb(struct sk_buff *skb);
633 void kfree_skb_list(struct sk_buff *segs);
634 void skb_tx_error(struct sk_buff *skb);
635 void consume_skb(struct sk_buff *skb);
636 void __kfree_skb(struct sk_buff *skb);
637 extern struct kmem_cache *skbuff_head_cache;
639 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
640 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
641 bool *fragstolen, int *delta_truesize);
643 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
645 struct sk_buff *build_skb(void *data, unsigned int frag_size);
646 static inline struct sk_buff *alloc_skb(unsigned int size,
649 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
652 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
655 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
658 struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
659 static inline struct sk_buff *alloc_skb_head(gfp_t priority)
661 return __alloc_skb_head(priority, -1);
664 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
665 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
666 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
667 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
668 struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
670 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
671 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
672 unsigned int headroom);
673 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
674 int newtailroom, gfp_t priority);
675 int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
677 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
678 int skb_pad(struct sk_buff *skb, int pad);
679 #define dev_kfree_skb(a) consume_skb(a)
681 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
682 int getfrag(void *from, char *to, int offset,
683 int len, int odd, struct sk_buff *skb),
684 void *from, int length);
686 struct skb_seq_state {
690 __u32 stepped_offset;
691 struct sk_buff *root_skb;
692 struct sk_buff *cur_skb;
696 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
697 unsigned int to, struct skb_seq_state *st);
698 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
699 struct skb_seq_state *st);
700 void skb_abort_seq_read(struct skb_seq_state *st);
702 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
703 unsigned int to, struct ts_config *config,
704 struct ts_state *state);
706 void __skb_get_rxhash(struct sk_buff *skb);
707 static inline __u32 skb_get_rxhash(struct sk_buff *skb)
710 __skb_get_rxhash(skb);
715 #ifdef NET_SKBUFF_DATA_USES_OFFSET
716 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
718 return skb->head + skb->end;
721 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
726 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
731 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
733 return skb->end - skb->head;
738 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
740 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
742 return &skb_shinfo(skb)->hwtstamps;
746 * skb_queue_empty - check if a queue is empty
749 * Returns true if the queue is empty, false otherwise.
751 static inline int skb_queue_empty(const struct sk_buff_head *list)
753 return list->next == (struct sk_buff *)list;
757 * skb_queue_is_last - check if skb is the last entry in the queue
761 * Returns true if @skb is the last buffer on the list.
763 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
764 const struct sk_buff *skb)
766 return skb->next == (struct sk_buff *)list;
770 * skb_queue_is_first - check if skb is the first entry in the queue
774 * Returns true if @skb is the first buffer on the list.
776 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
777 const struct sk_buff *skb)
779 return skb->prev == (struct sk_buff *)list;
783 * skb_queue_next - return the next packet in the queue
785 * @skb: current buffer
787 * Return the next packet in @list after @skb. It is only valid to
788 * call this if skb_queue_is_last() evaluates to false.
790 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
791 const struct sk_buff *skb)
793 /* This BUG_ON may seem severe, but if we just return then we
794 * are going to dereference garbage.
796 BUG_ON(skb_queue_is_last(list, skb));
801 * skb_queue_prev - return the prev packet in the queue
803 * @skb: current buffer
805 * Return the prev packet in @list before @skb. It is only valid to
806 * call this if skb_queue_is_first() evaluates to false.
808 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
809 const struct sk_buff *skb)
811 /* This BUG_ON may seem severe, but if we just return then we
812 * are going to dereference garbage.
814 BUG_ON(skb_queue_is_first(list, skb));
819 * skb_get - reference buffer
820 * @skb: buffer to reference
822 * Makes another reference to a socket buffer and returns a pointer
825 static inline struct sk_buff *skb_get(struct sk_buff *skb)
827 atomic_inc(&skb->users);
832 * If users == 1, we are the only owner and are can avoid redundant
837 * skb_cloned - is the buffer a clone
838 * @skb: buffer to check
840 * Returns true if the buffer was generated with skb_clone() and is
841 * one of multiple shared copies of the buffer. Cloned buffers are
842 * shared data so must not be written to under normal circumstances.
844 static inline int skb_cloned(const struct sk_buff *skb)
846 return skb->cloned &&
847 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
850 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
852 might_sleep_if(pri & __GFP_WAIT);
855 return pskb_expand_head(skb, 0, 0, pri);
861 * skb_header_cloned - is the header a clone
862 * @skb: buffer to check
864 * Returns true if modifying the header part of the buffer requires
865 * the data to be copied.
867 static inline int skb_header_cloned(const struct sk_buff *skb)
874 dataref = atomic_read(&skb_shinfo(skb)->dataref);
875 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
880 * skb_header_release - release reference to header
881 * @skb: buffer to operate on
883 * Drop a reference to the header part of the buffer. This is done
884 * by acquiring a payload reference. You must not read from the header
885 * part of skb->data after this.
887 static inline void skb_header_release(struct sk_buff *skb)
891 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
895 * skb_shared - is the buffer shared
896 * @skb: buffer to check
898 * Returns true if more than one person has a reference to this
901 static inline int skb_shared(const struct sk_buff *skb)
903 return atomic_read(&skb->users) != 1;
907 * skb_share_check - check if buffer is shared and if so clone it
908 * @skb: buffer to check
909 * @pri: priority for memory allocation
911 * If the buffer is shared the buffer is cloned and the old copy
912 * drops a reference. A new clone with a single reference is returned.
913 * If the buffer is not shared the original buffer is returned. When
914 * being called from interrupt status or with spinlocks held pri must
917 * NULL is returned on a memory allocation failure.
919 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
921 might_sleep_if(pri & __GFP_WAIT);
922 if (skb_shared(skb)) {
923 struct sk_buff *nskb = skb_clone(skb, pri);
935 * Copy shared buffers into a new sk_buff. We effectively do COW on
936 * packets to handle cases where we have a local reader and forward
937 * and a couple of other messy ones. The normal one is tcpdumping
938 * a packet thats being forwarded.
942 * skb_unshare - make a copy of a shared buffer
943 * @skb: buffer to check
944 * @pri: priority for memory allocation
946 * If the socket buffer is a clone then this function creates a new
947 * copy of the data, drops a reference count on the old copy and returns
948 * the new copy with the reference count at 1. If the buffer is not a clone
949 * the original buffer is returned. When called with a spinlock held or
950 * from interrupt state @pri must be %GFP_ATOMIC
952 * %NULL is returned on a memory allocation failure.
954 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
957 might_sleep_if(pri & __GFP_WAIT);
958 if (skb_cloned(skb)) {
959 struct sk_buff *nskb = skb_copy(skb, pri);
960 kfree_skb(skb); /* Free our shared copy */
967 * skb_peek - peek at the head 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 head 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(const struct sk_buff_head *list_)
981 struct sk_buff *skb = list_->next;
983 if (skb == (struct sk_buff *)list_)
989 * skb_peek_next - peek skb following the given one from a queue
990 * @skb: skb to start from
991 * @list_: list to peek at
993 * Returns %NULL when the end of the list is met or a pointer to the
994 * next element. The reference count is not incremented and the
995 * reference is therefore volatile. Use with caution.
997 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
998 const struct sk_buff_head *list_)
1000 struct sk_buff *next = skb->next;
1002 if (next == (struct sk_buff *)list_)
1008 * skb_peek_tail - peek at the tail of an &sk_buff_head
1009 * @list_: list to peek at
1011 * Peek an &sk_buff. Unlike most other operations you _MUST_
1012 * be careful with this one. A peek leaves the buffer on the
1013 * list and someone else may run off with it. You must hold
1014 * the appropriate locks or have a private queue to do this.
1016 * Returns %NULL for an empty list or a pointer to the tail element.
1017 * The reference count is not incremented and the reference is therefore
1018 * volatile. Use with caution.
1020 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1022 struct sk_buff *skb = list_->prev;
1024 if (skb == (struct sk_buff *)list_)
1031 * skb_queue_len - get queue length
1032 * @list_: list to measure
1034 * Return the length of an &sk_buff queue.
1036 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1042 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1043 * @list: queue to initialize
1045 * This initializes only the list and queue length aspects of
1046 * an sk_buff_head object. This allows to initialize the list
1047 * aspects of an sk_buff_head without reinitializing things like
1048 * the spinlock. It can also be used for on-stack sk_buff_head
1049 * objects where the spinlock is known to not be used.
1051 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1053 list->prev = list->next = (struct sk_buff *)list;
1058 * This function creates a split out lock class for each invocation;
1059 * this is needed for now since a whole lot of users of the skb-queue
1060 * infrastructure in drivers have different locking usage (in hardirq)
1061 * than the networking core (in softirq only). In the long run either the
1062 * network layer or drivers should need annotation to consolidate the
1063 * main types of usage into 3 classes.
1065 static inline void skb_queue_head_init(struct sk_buff_head *list)
1067 spin_lock_init(&list->lock);
1068 __skb_queue_head_init(list);
1071 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1072 struct lock_class_key *class)
1074 skb_queue_head_init(list);
1075 lockdep_set_class(&list->lock, class);
1079 * Insert an sk_buff on a list.
1081 * The "__skb_xxxx()" functions are the non-atomic ones that
1082 * can only be called with interrupts disabled.
1084 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1085 struct sk_buff_head *list);
1086 static inline void __skb_insert(struct sk_buff *newsk,
1087 struct sk_buff *prev, struct sk_buff *next,
1088 struct sk_buff_head *list)
1092 next->prev = prev->next = newsk;
1096 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1097 struct sk_buff *prev,
1098 struct sk_buff *next)
1100 struct sk_buff *first = list->next;
1101 struct sk_buff *last = list->prev;
1111 * skb_queue_splice - join two skb lists, this is designed for stacks
1112 * @list: the new list to add
1113 * @head: the place to add it in the first list
1115 static inline void skb_queue_splice(const struct sk_buff_head *list,
1116 struct sk_buff_head *head)
1118 if (!skb_queue_empty(list)) {
1119 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1120 head->qlen += list->qlen;
1125 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1126 * @list: the new list to add
1127 * @head: the place to add it in the first list
1129 * The list at @list is reinitialised
1131 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1132 struct sk_buff_head *head)
1134 if (!skb_queue_empty(list)) {
1135 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1136 head->qlen += list->qlen;
1137 __skb_queue_head_init(list);
1142 * skb_queue_splice_tail - join two skb lists, each list being a queue
1143 * @list: the new list to add
1144 * @head: the place to add it in the first list
1146 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1147 struct sk_buff_head *head)
1149 if (!skb_queue_empty(list)) {
1150 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1151 head->qlen += list->qlen;
1156 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1157 * @list: the new list to add
1158 * @head: the place to add it in the first list
1160 * Each of the lists is a queue.
1161 * The list at @list is reinitialised
1163 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1164 struct sk_buff_head *head)
1166 if (!skb_queue_empty(list)) {
1167 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1168 head->qlen += list->qlen;
1169 __skb_queue_head_init(list);
1174 * __skb_queue_after - queue a buffer at the list head
1175 * @list: list to use
1176 * @prev: place after this buffer
1177 * @newsk: buffer to queue
1179 * Queue a buffer int the middle of a list. This function takes no locks
1180 * and you must therefore hold required locks before calling it.
1182 * A buffer cannot be placed on two lists at the same time.
1184 static inline void __skb_queue_after(struct sk_buff_head *list,
1185 struct sk_buff *prev,
1186 struct sk_buff *newsk)
1188 __skb_insert(newsk, prev, prev->next, list);
1191 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1192 struct sk_buff_head *list);
1194 static inline void __skb_queue_before(struct sk_buff_head *list,
1195 struct sk_buff *next,
1196 struct sk_buff *newsk)
1198 __skb_insert(newsk, next->prev, next, list);
1202 * __skb_queue_head - queue a buffer at the list head
1203 * @list: list to use
1204 * @newsk: buffer to queue
1206 * Queue a buffer at the start of a list. This function takes no locks
1207 * and you must therefore hold required locks before calling it.
1209 * A buffer cannot be placed on two lists at the same time.
1211 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1212 static inline void __skb_queue_head(struct sk_buff_head *list,
1213 struct sk_buff *newsk)
1215 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1219 * __skb_queue_tail - queue a buffer at the list tail
1220 * @list: list to use
1221 * @newsk: buffer to queue
1223 * Queue a buffer at the end of a list. This function takes no locks
1224 * and you must therefore hold required locks before calling it.
1226 * A buffer cannot be placed on two lists at the same time.
1228 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1229 static inline void __skb_queue_tail(struct sk_buff_head *list,
1230 struct sk_buff *newsk)
1232 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1236 * remove sk_buff from list. _Must_ be called atomically, and with
1239 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1240 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1242 struct sk_buff *next, *prev;
1247 skb->next = skb->prev = NULL;
1253 * __skb_dequeue - remove from the head of the queue
1254 * @list: list to dequeue from
1256 * Remove the head of the list. This function does not take any locks
1257 * so must be used with appropriate locks held only. The head item is
1258 * returned or %NULL if the list is empty.
1260 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1261 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1263 struct sk_buff *skb = skb_peek(list);
1265 __skb_unlink(skb, list);
1270 * __skb_dequeue_tail - remove from the tail of the queue
1271 * @list: list to dequeue from
1273 * Remove the tail of the list. This function does not take any locks
1274 * so must be used with appropriate locks held only. The tail item is
1275 * returned or %NULL if the list is empty.
1277 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1278 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1280 struct sk_buff *skb = skb_peek_tail(list);
1282 __skb_unlink(skb, list);
1287 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1289 return skb->data_len;
1292 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1294 return skb->len - skb->data_len;
1297 static inline int skb_pagelen(const struct sk_buff *skb)
1301 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1302 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1303 return len + skb_headlen(skb);
1307 * __skb_fill_page_desc - initialise a paged fragment in an skb
1308 * @skb: buffer containing fragment to be initialised
1309 * @i: paged fragment index to initialise
1310 * @page: the page to use for this fragment
1311 * @off: the offset to the data with @page
1312 * @size: the length of the data
1314 * Initialises the @i'th fragment of @skb to point to &size bytes at
1315 * offset @off within @page.
1317 * Does not take any additional reference on the fragment.
1319 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1320 struct page *page, int off, int size)
1322 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1325 * Propagate page->pfmemalloc to the skb if we can. The problem is
1326 * that not all callers have unique ownership of the page. If
1327 * pfmemalloc is set, we check the mapping as a mapping implies
1328 * page->index is set (index and pfmemalloc share space).
1329 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1330 * do not lose pfmemalloc information as the pages would not be
1331 * allocated using __GFP_MEMALLOC.
1333 frag->page.p = page;
1334 frag->page_offset = off;
1335 skb_frag_size_set(frag, size);
1337 page = compound_head(page);
1338 if (page->pfmemalloc && !page->mapping)
1339 skb->pfmemalloc = true;
1343 * skb_fill_page_desc - initialise a paged fragment in an skb
1344 * @skb: buffer containing fragment to be initialised
1345 * @i: paged fragment index to initialise
1346 * @page: the page to use for this fragment
1347 * @off: the offset to the data with @page
1348 * @size: the length of the data
1350 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1351 * @skb to point to @size bytes at offset @off within @page. In
1352 * addition updates @skb such that @i is the last fragment.
1354 * Does not take any additional reference on the fragment.
1356 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1357 struct page *page, int off, int size)
1359 __skb_fill_page_desc(skb, i, page, off, size);
1360 skb_shinfo(skb)->nr_frags = i + 1;
1363 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1364 int size, unsigned int truesize);
1366 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1367 unsigned int truesize);
1369 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1370 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1371 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1373 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1374 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1376 return skb->head + skb->tail;
1379 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1381 skb->tail = skb->data - skb->head;
1384 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1386 skb_reset_tail_pointer(skb);
1387 skb->tail += offset;
1390 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1391 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1396 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1398 skb->tail = skb->data;
1401 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1403 skb->tail = skb->data + offset;
1406 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1409 * Add data to an sk_buff
1411 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1412 unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1413 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1415 unsigned char *tmp = skb_tail_pointer(skb);
1416 SKB_LINEAR_ASSERT(skb);
1422 unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1423 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1430 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1431 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1434 BUG_ON(skb->len < skb->data_len);
1435 return skb->data += len;
1438 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1440 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1443 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1445 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1447 if (len > skb_headlen(skb) &&
1448 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1451 return skb->data += len;
1454 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1456 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1459 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1461 if (likely(len <= skb_headlen(skb)))
1463 if (unlikely(len > skb->len))
1465 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1469 * skb_headroom - bytes at buffer head
1470 * @skb: buffer to check
1472 * Return the number of bytes of free space at the head of an &sk_buff.
1474 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1476 return skb->data - skb->head;
1480 * skb_tailroom - bytes at buffer end
1481 * @skb: buffer to check
1483 * Return the number of bytes of free space at the tail of an sk_buff
1485 static inline int skb_tailroom(const struct sk_buff *skb)
1487 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1491 * skb_availroom - bytes at buffer end
1492 * @skb: buffer to check
1494 * Return the number of bytes of free space at the tail of an sk_buff
1495 * allocated by sk_stream_alloc()
1497 static inline int skb_availroom(const struct sk_buff *skb)
1499 if (skb_is_nonlinear(skb))
1502 return skb->end - skb->tail - skb->reserved_tailroom;
1506 * skb_reserve - adjust headroom
1507 * @skb: buffer to alter
1508 * @len: bytes to move
1510 * Increase the headroom of an empty &sk_buff by reducing the tail
1511 * room. This is only allowed for an empty buffer.
1513 static inline void skb_reserve(struct sk_buff *skb, int len)
1519 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1521 skb->inner_mac_header = skb->mac_header;
1522 skb->inner_network_header = skb->network_header;
1523 skb->inner_transport_header = skb->transport_header;
1526 static inline void skb_reset_mac_len(struct sk_buff *skb)
1528 skb->mac_len = skb->network_header - skb->mac_header;
1531 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1534 return skb->head + skb->inner_transport_header;
1537 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1539 skb->inner_transport_header = skb->data - skb->head;
1542 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1545 skb_reset_inner_transport_header(skb);
1546 skb->inner_transport_header += offset;
1549 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1551 return skb->head + skb->inner_network_header;
1554 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1556 skb->inner_network_header = skb->data - skb->head;
1559 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1562 skb_reset_inner_network_header(skb);
1563 skb->inner_network_header += offset;
1566 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1568 return skb->head + skb->inner_mac_header;
1571 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1573 skb->inner_mac_header = skb->data - skb->head;
1576 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1579 skb_reset_inner_mac_header(skb);
1580 skb->inner_mac_header += offset;
1582 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1584 return skb->transport_header != (typeof(skb->transport_header))~0U;
1587 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1589 return skb->head + skb->transport_header;
1592 static inline void skb_reset_transport_header(struct sk_buff *skb)
1594 skb->transport_header = skb->data - skb->head;
1597 static inline void skb_set_transport_header(struct sk_buff *skb,
1600 skb_reset_transport_header(skb);
1601 skb->transport_header += offset;
1604 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1606 return skb->head + skb->network_header;
1609 static inline void skb_reset_network_header(struct sk_buff *skb)
1611 skb->network_header = skb->data - skb->head;
1614 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1616 skb_reset_network_header(skb);
1617 skb->network_header += offset;
1620 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1622 return skb->head + skb->mac_header;
1625 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1627 return skb->mac_header != (typeof(skb->mac_header))~0U;
1630 static inline void skb_reset_mac_header(struct sk_buff *skb)
1632 skb->mac_header = skb->data - skb->head;
1635 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1637 skb_reset_mac_header(skb);
1638 skb->mac_header += offset;
1641 static inline void skb_pop_mac_header(struct sk_buff *skb)
1643 skb->mac_header = skb->network_header;
1646 static inline void skb_probe_transport_header(struct sk_buff *skb,
1647 const int offset_hint)
1649 struct flow_keys keys;
1651 if (skb_transport_header_was_set(skb))
1653 else if (skb_flow_dissect(skb, &keys))
1654 skb_set_transport_header(skb, keys.thoff);
1656 skb_set_transport_header(skb, offset_hint);
1659 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1661 if (skb_mac_header_was_set(skb)) {
1662 const unsigned char *old_mac = skb_mac_header(skb);
1664 skb_set_mac_header(skb, -skb->mac_len);
1665 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1669 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1671 return skb->csum_start - skb_headroom(skb);
1674 static inline int skb_transport_offset(const struct sk_buff *skb)
1676 return skb_transport_header(skb) - skb->data;
1679 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1681 return skb->transport_header - skb->network_header;
1684 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1686 return skb->inner_transport_header - skb->inner_network_header;
1689 static inline int skb_network_offset(const struct sk_buff *skb)
1691 return skb_network_header(skb) - skb->data;
1694 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1696 return skb_inner_network_header(skb) - skb->data;
1699 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1701 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1705 * CPUs often take a performance hit when accessing unaligned memory
1706 * locations. The actual performance hit varies, it can be small if the
1707 * hardware handles it or large if we have to take an exception and fix it
1710 * Since an ethernet header is 14 bytes network drivers often end up with
1711 * the IP header at an unaligned offset. The IP header can be aligned by
1712 * shifting the start of the packet by 2 bytes. Drivers should do this
1715 * skb_reserve(skb, NET_IP_ALIGN);
1717 * The downside to this alignment of the IP header is that the DMA is now
1718 * unaligned. On some architectures the cost of an unaligned DMA is high
1719 * and this cost outweighs the gains made by aligning the IP header.
1721 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1724 #ifndef NET_IP_ALIGN
1725 #define NET_IP_ALIGN 2
1729 * The networking layer reserves some headroom in skb data (via
1730 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1731 * the header has to grow. In the default case, if the header has to grow
1732 * 32 bytes or less we avoid the reallocation.
1734 * Unfortunately this headroom changes the DMA alignment of the resulting
1735 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1736 * on some architectures. An architecture can override this value,
1737 * perhaps setting it to a cacheline in size (since that will maintain
1738 * cacheline alignment of the DMA). It must be a power of 2.
1740 * Various parts of the networking layer expect at least 32 bytes of
1741 * headroom, you should not reduce this.
1743 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1744 * to reduce average number of cache lines per packet.
1745 * get_rps_cpus() for example only access one 64 bytes aligned block :
1746 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1749 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1752 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1754 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1756 if (unlikely(skb_is_nonlinear(skb))) {
1761 skb_set_tail_pointer(skb, len);
1764 void skb_trim(struct sk_buff *skb, unsigned int len);
1766 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1769 return ___pskb_trim(skb, len);
1770 __skb_trim(skb, len);
1774 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1776 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1780 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1781 * @skb: buffer to alter
1784 * This is identical to pskb_trim except that the caller knows that
1785 * the skb is not cloned so we should never get an error due to out-
1788 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1790 int err = pskb_trim(skb, len);
1795 * skb_orphan - orphan a buffer
1796 * @skb: buffer to orphan
1798 * If a buffer currently has an owner then we call the owner's
1799 * destructor function and make the @skb unowned. The buffer continues
1800 * to exist but is no longer charged to its former owner.
1802 static inline void skb_orphan(struct sk_buff *skb)
1804 if (skb->destructor) {
1805 skb->destructor(skb);
1806 skb->destructor = NULL;
1814 * skb_orphan_frags - orphan the frags contained in a buffer
1815 * @skb: buffer to orphan frags from
1816 * @gfp_mask: allocation mask for replacement pages
1818 * For each frag in the SKB which needs a destructor (i.e. has an
1819 * owner) create a copy of that frag and release the original
1820 * page by calling the destructor.
1822 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1824 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1826 return skb_copy_ubufs(skb, gfp_mask);
1830 * __skb_queue_purge - empty a list
1831 * @list: list to empty
1833 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1834 * the list and one reference dropped. This function does not take the
1835 * list lock and the caller must hold the relevant locks to use it.
1837 void skb_queue_purge(struct sk_buff_head *list);
1838 static inline void __skb_queue_purge(struct sk_buff_head *list)
1840 struct sk_buff *skb;
1841 while ((skb = __skb_dequeue(list)) != NULL)
1845 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1846 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1847 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1849 void *netdev_alloc_frag(unsigned int fragsz);
1851 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1855 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1856 * @dev: network device to receive on
1857 * @length: length to allocate
1859 * Allocate a new &sk_buff and assign it a usage count of one. The
1860 * buffer has unspecified headroom built in. Users should allocate
1861 * the headroom they think they need without accounting for the
1862 * built in space. The built in space is used for optimisations.
1864 * %NULL is returned if there is no free memory. Although this function
1865 * allocates memory it can be called from an interrupt.
1867 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1868 unsigned int length)
1870 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1873 /* legacy helper around __netdev_alloc_skb() */
1874 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1877 return __netdev_alloc_skb(NULL, length, gfp_mask);
1880 /* legacy helper around netdev_alloc_skb() */
1881 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1883 return netdev_alloc_skb(NULL, length);
1887 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1888 unsigned int length, gfp_t gfp)
1890 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
1892 if (NET_IP_ALIGN && skb)
1893 skb_reserve(skb, NET_IP_ALIGN);
1897 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1898 unsigned int length)
1900 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1904 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1905 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1906 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1907 * @order: size of the allocation
1909 * Allocate a new page.
1911 * %NULL is returned if there is no free memory.
1913 static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1914 struct sk_buff *skb,
1919 gfp_mask |= __GFP_COLD;
1921 if (!(gfp_mask & __GFP_NOMEMALLOC))
1922 gfp_mask |= __GFP_MEMALLOC;
1924 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
1925 if (skb && page && page->pfmemalloc)
1926 skb->pfmemalloc = true;
1932 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
1933 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1934 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1936 * Allocate a new page.
1938 * %NULL is returned if there is no free memory.
1940 static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
1941 struct sk_buff *skb)
1943 return __skb_alloc_pages(gfp_mask, skb, 0);
1947 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
1948 * @page: The page that was allocated from skb_alloc_page
1949 * @skb: The skb that may need pfmemalloc set
1951 static inline void skb_propagate_pfmemalloc(struct page *page,
1952 struct sk_buff *skb)
1954 if (page && page->pfmemalloc)
1955 skb->pfmemalloc = true;
1959 * skb_frag_page - retrieve the page refered to by a paged fragment
1960 * @frag: the paged fragment
1962 * Returns the &struct page associated with @frag.
1964 static inline struct page *skb_frag_page(const skb_frag_t *frag)
1966 return frag->page.p;
1970 * __skb_frag_ref - take an addition reference on a paged fragment.
1971 * @frag: the paged fragment
1973 * Takes an additional reference on the paged fragment @frag.
1975 static inline void __skb_frag_ref(skb_frag_t *frag)
1977 get_page(skb_frag_page(frag));
1981 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1983 * @f: the fragment offset.
1985 * Takes an additional reference on the @f'th paged fragment of @skb.
1987 static inline void skb_frag_ref(struct sk_buff *skb, int f)
1989 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1993 * __skb_frag_unref - release a reference on a paged fragment.
1994 * @frag: the paged fragment
1996 * Releases a reference on the paged fragment @frag.
1998 static inline void __skb_frag_unref(skb_frag_t *frag)
2000 put_page(skb_frag_page(frag));
2004 * skb_frag_unref - release a reference on a paged fragment of an skb.
2006 * @f: the fragment offset
2008 * Releases a reference on the @f'th paged fragment of @skb.
2010 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2012 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2016 * skb_frag_address - gets the address of the data contained in a paged fragment
2017 * @frag: the paged fragment buffer
2019 * Returns the address of the data within @frag. The page must already
2022 static inline void *skb_frag_address(const skb_frag_t *frag)
2024 return page_address(skb_frag_page(frag)) + frag->page_offset;
2028 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2029 * @frag: the paged fragment buffer
2031 * Returns the address of the data within @frag. Checks that the page
2032 * is mapped and returns %NULL otherwise.
2034 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2036 void *ptr = page_address(skb_frag_page(frag));
2040 return ptr + frag->page_offset;
2044 * __skb_frag_set_page - sets the page contained in a paged fragment
2045 * @frag: the paged fragment
2046 * @page: the page to set
2048 * Sets the fragment @frag to contain @page.
2050 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2052 frag->page.p = page;
2056 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2058 * @f: the fragment offset
2059 * @page: the page to set
2061 * Sets the @f'th fragment of @skb to contain @page.
2063 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2066 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2069 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2072 * skb_frag_dma_map - maps a paged fragment via the DMA API
2073 * @dev: the device to map the fragment to
2074 * @frag: the paged fragment to map
2075 * @offset: the offset within the fragment (starting at the
2076 * fragment's own offset)
2077 * @size: the number of bytes to map
2078 * @dir: the direction of the mapping (%PCI_DMA_*)
2080 * Maps the page associated with @frag to @device.
2082 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2083 const skb_frag_t *frag,
2084 size_t offset, size_t size,
2085 enum dma_data_direction dir)
2087 return dma_map_page(dev, skb_frag_page(frag),
2088 frag->page_offset + offset, size, dir);
2091 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2094 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2098 * skb_clone_writable - is the header of a clone writable
2099 * @skb: buffer to check
2100 * @len: length up to which to write
2102 * Returns true if modifying the header part of the cloned buffer
2103 * does not requires the data to be copied.
2105 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2107 return !skb_header_cloned(skb) &&
2108 skb_headroom(skb) + len <= skb->hdr_len;
2111 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2116 if (headroom > skb_headroom(skb))
2117 delta = headroom - skb_headroom(skb);
2119 if (delta || cloned)
2120 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2126 * skb_cow - copy header of skb when it is required
2127 * @skb: buffer to cow
2128 * @headroom: needed headroom
2130 * If the skb passed lacks sufficient headroom or its data part
2131 * is shared, data is reallocated. If reallocation fails, an error
2132 * is returned and original skb is not changed.
2134 * The result is skb with writable area skb->head...skb->tail
2135 * and at least @headroom of space at head.
2137 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2139 return __skb_cow(skb, headroom, skb_cloned(skb));
2143 * skb_cow_head - skb_cow but only making the head writable
2144 * @skb: buffer to cow
2145 * @headroom: needed headroom
2147 * This function is identical to skb_cow except that we replace the
2148 * skb_cloned check by skb_header_cloned. It should be used when
2149 * you only need to push on some header and do not need to modify
2152 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2154 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2158 * skb_padto - pad an skbuff up to a minimal size
2159 * @skb: buffer to pad
2160 * @len: minimal length
2162 * Pads up a buffer to ensure the trailing bytes exist and are
2163 * blanked. If the buffer already contains sufficient data it
2164 * is untouched. Otherwise it is extended. Returns zero on
2165 * success. The skb is freed on error.
2168 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2170 unsigned int size = skb->len;
2171 if (likely(size >= len))
2173 return skb_pad(skb, len - size);
2176 static inline int skb_add_data(struct sk_buff *skb,
2177 char __user *from, int copy)
2179 const int off = skb->len;
2181 if (skb->ip_summed == CHECKSUM_NONE) {
2183 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
2186 skb->csum = csum_block_add(skb->csum, csum, off);
2189 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2192 __skb_trim(skb, off);
2196 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2197 const struct page *page, int off)
2200 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2202 return page == skb_frag_page(frag) &&
2203 off == frag->page_offset + skb_frag_size(frag);
2208 static inline int __skb_linearize(struct sk_buff *skb)
2210 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2214 * skb_linearize - convert paged skb to linear one
2215 * @skb: buffer to linarize
2217 * If there is no free memory -ENOMEM is returned, otherwise zero
2218 * is returned and the old skb data released.
2220 static inline int skb_linearize(struct sk_buff *skb)
2222 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2226 * skb_has_shared_frag - can any frag be overwritten
2227 * @skb: buffer to test
2229 * Return true if the skb has at least one frag that might be modified
2230 * by an external entity (as in vmsplice()/sendfile())
2232 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2234 return skb_is_nonlinear(skb) &&
2235 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2239 * skb_linearize_cow - make sure skb is linear and writable
2240 * @skb: buffer to process
2242 * If there is no free memory -ENOMEM is returned, otherwise zero
2243 * is returned and the old skb data released.
2245 static inline int skb_linearize_cow(struct sk_buff *skb)
2247 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2248 __skb_linearize(skb) : 0;
2252 * skb_postpull_rcsum - update checksum for received skb after pull
2253 * @skb: buffer to update
2254 * @start: start of data before pull
2255 * @len: length of data pulled
2257 * After doing a pull on a received packet, you need to call this to
2258 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2259 * CHECKSUM_NONE so that it can be recomputed from scratch.
2262 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2263 const void *start, unsigned int len)
2265 if (skb->ip_summed == CHECKSUM_COMPLETE)
2266 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2269 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2272 * pskb_trim_rcsum - trim received skb and update checksum
2273 * @skb: buffer to trim
2276 * This is exactly the same as pskb_trim except that it ensures the
2277 * checksum of received packets are still valid after the operation.
2280 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2282 if (likely(len >= skb->len))
2284 if (skb->ip_summed == CHECKSUM_COMPLETE)
2285 skb->ip_summed = CHECKSUM_NONE;
2286 return __pskb_trim(skb, len);
2289 #define skb_queue_walk(queue, skb) \
2290 for (skb = (queue)->next; \
2291 skb != (struct sk_buff *)(queue); \
2294 #define skb_queue_walk_safe(queue, skb, tmp) \
2295 for (skb = (queue)->next, tmp = skb->next; \
2296 skb != (struct sk_buff *)(queue); \
2297 skb = tmp, tmp = skb->next)
2299 #define skb_queue_walk_from(queue, skb) \
2300 for (; skb != (struct sk_buff *)(queue); \
2303 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2304 for (tmp = skb->next; \
2305 skb != (struct sk_buff *)(queue); \
2306 skb = tmp, tmp = skb->next)
2308 #define skb_queue_reverse_walk(queue, skb) \
2309 for (skb = (queue)->prev; \
2310 skb != (struct sk_buff *)(queue); \
2313 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2314 for (skb = (queue)->prev, tmp = skb->prev; \
2315 skb != (struct sk_buff *)(queue); \
2316 skb = tmp, tmp = skb->prev)
2318 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2319 for (tmp = skb->prev; \
2320 skb != (struct sk_buff *)(queue); \
2321 skb = tmp, tmp = skb->prev)
2323 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2325 return skb_shinfo(skb)->frag_list != NULL;
2328 static inline void skb_frag_list_init(struct sk_buff *skb)
2330 skb_shinfo(skb)->frag_list = NULL;
2333 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2335 frag->next = skb_shinfo(skb)->frag_list;
2336 skb_shinfo(skb)->frag_list = frag;
2339 #define skb_walk_frags(skb, iter) \
2340 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2342 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2343 int *peeked, int *off, int *err);
2344 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2346 unsigned int datagram_poll(struct file *file, struct socket *sock,
2347 struct poll_table_struct *wait);
2348 int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2349 struct iovec *to, int size);
2350 int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2352 int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2353 const struct iovec *from, int from_offset,
2355 int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2356 int offset, size_t count);
2357 int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2358 const struct iovec *to, int to_offset,
2360 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2361 void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2362 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
2363 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2364 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2365 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2366 int len, __wsum csum);
2367 int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2368 struct pipe_inode_info *pipe, unsigned int len,
2369 unsigned int flags);
2370 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2371 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2372 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2373 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
2374 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
2376 struct skb_checksum_ops {
2377 __wsum (*update)(const void *mem, int len, __wsum wsum);
2378 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2381 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2382 __wsum csum, const struct skb_checksum_ops *ops);
2383 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2386 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2387 int len, void *buffer)
2389 int hlen = skb_headlen(skb);
2391 if (hlen - offset >= len)
2392 return skb->data + offset;
2394 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2400 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2402 const unsigned int len)
2404 memcpy(to, skb->data, len);
2407 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2408 const int offset, void *to,
2409 const unsigned int len)
2411 memcpy(to, skb->data + offset, len);
2414 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2416 const unsigned int len)
2418 memcpy(skb->data, from, len);
2421 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2424 const unsigned int len)
2426 memcpy(skb->data + offset, from, len);
2429 void skb_init(void);
2431 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2437 * skb_get_timestamp - get timestamp from a skb
2438 * @skb: skb to get stamp from
2439 * @stamp: pointer to struct timeval to store stamp in
2441 * Timestamps are stored in the skb as offsets to a base timestamp.
2442 * This function converts the offset back to a struct timeval and stores
2445 static inline void skb_get_timestamp(const struct sk_buff *skb,
2446 struct timeval *stamp)
2448 *stamp = ktime_to_timeval(skb->tstamp);
2451 static inline void skb_get_timestampns(const struct sk_buff *skb,
2452 struct timespec *stamp)
2454 *stamp = ktime_to_timespec(skb->tstamp);
2457 static inline void __net_timestamp(struct sk_buff *skb)
2459 skb->tstamp = ktime_get_real();
2462 static inline ktime_t net_timedelta(ktime_t t)
2464 return ktime_sub(ktime_get_real(), t);
2467 static inline ktime_t net_invalid_timestamp(void)
2469 return ktime_set(0, 0);
2472 void skb_timestamping_init(void);
2474 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2476 void skb_clone_tx_timestamp(struct sk_buff *skb);
2477 bool skb_defer_rx_timestamp(struct sk_buff *skb);
2479 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2481 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2485 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2490 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2493 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2495 * PHY drivers may accept clones of transmitted packets for
2496 * timestamping via their phy_driver.txtstamp method. These drivers
2497 * must call this function to return the skb back to the stack, with
2498 * or without a timestamp.
2500 * @skb: clone of the the original outgoing packet
2501 * @hwtstamps: hardware time stamps, may be NULL if not available
2504 void skb_complete_tx_timestamp(struct sk_buff *skb,
2505 struct skb_shared_hwtstamps *hwtstamps);
2508 * skb_tstamp_tx - queue clone of skb with send time stamps
2509 * @orig_skb: the original outgoing packet
2510 * @hwtstamps: hardware time stamps, may be NULL if not available
2512 * If the skb has a socket associated, then this function clones the
2513 * skb (thus sharing the actual data and optional structures), stores
2514 * the optional hardware time stamping information (if non NULL) or
2515 * generates a software time stamp (otherwise), then queues the clone
2516 * to the error queue of the socket. Errors are silently ignored.
2518 void skb_tstamp_tx(struct sk_buff *orig_skb,
2519 struct skb_shared_hwtstamps *hwtstamps);
2521 static inline void sw_tx_timestamp(struct sk_buff *skb)
2523 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2524 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2525 skb_tstamp_tx(skb, NULL);
2529 * skb_tx_timestamp() - Driver hook for transmit timestamping
2531 * Ethernet MAC Drivers should call this function in their hard_xmit()
2532 * function immediately before giving the sk_buff to the MAC hardware.
2534 * Specifically, one should make absolutely sure that this function is
2535 * called before TX completion of this packet can trigger. Otherwise
2536 * the packet could potentially already be freed.
2538 * @skb: A socket buffer.
2540 static inline void skb_tx_timestamp(struct sk_buff *skb)
2542 skb_clone_tx_timestamp(skb);
2543 sw_tx_timestamp(skb);
2547 * skb_complete_wifi_ack - deliver skb with wifi status
2549 * @skb: the original outgoing packet
2550 * @acked: ack status
2553 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2555 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2556 __sum16 __skb_checksum_complete(struct sk_buff *skb);
2558 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2560 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2564 * skb_checksum_complete - Calculate checksum of an entire packet
2565 * @skb: packet to process
2567 * This function calculates the checksum over the entire packet plus
2568 * the value of skb->csum. The latter can be used to supply the
2569 * checksum of a pseudo header as used by TCP/UDP. It returns the
2572 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2573 * this function can be used to verify that checksum on received
2574 * packets. In that case the function should return zero if the
2575 * checksum is correct. In particular, this function will return zero
2576 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2577 * hardware has already verified the correctness of the checksum.
2579 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2581 return skb_csum_unnecessary(skb) ?
2582 0 : __skb_checksum_complete(skb);
2585 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2586 void nf_conntrack_destroy(struct nf_conntrack *nfct);
2587 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2589 if (nfct && atomic_dec_and_test(&nfct->use))
2590 nf_conntrack_destroy(nfct);
2592 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2595 atomic_inc(&nfct->use);
2598 #ifdef CONFIG_BRIDGE_NETFILTER
2599 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2601 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2604 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2607 atomic_inc(&nf_bridge->use);
2609 #endif /* CONFIG_BRIDGE_NETFILTER */
2610 static inline void nf_reset(struct sk_buff *skb)
2612 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2613 nf_conntrack_put(skb->nfct);
2616 #ifdef CONFIG_BRIDGE_NETFILTER
2617 nf_bridge_put(skb->nf_bridge);
2618 skb->nf_bridge = NULL;
2622 static inline void nf_reset_trace(struct sk_buff *skb)
2624 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
2629 /* Note: This doesn't put any conntrack and bridge info in dst. */
2630 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2632 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2633 dst->nfct = src->nfct;
2634 nf_conntrack_get(src->nfct);
2635 dst->nfctinfo = src->nfctinfo;
2637 #ifdef CONFIG_BRIDGE_NETFILTER
2638 dst->nf_bridge = src->nf_bridge;
2639 nf_bridge_get(src->nf_bridge);
2643 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2645 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2646 nf_conntrack_put(dst->nfct);
2648 #ifdef CONFIG_BRIDGE_NETFILTER
2649 nf_bridge_put(dst->nf_bridge);
2651 __nf_copy(dst, src);
2654 #ifdef CONFIG_NETWORK_SECMARK
2655 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2657 to->secmark = from->secmark;
2660 static inline void skb_init_secmark(struct sk_buff *skb)
2665 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2668 static inline void skb_init_secmark(struct sk_buff *skb)
2672 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2674 skb->queue_mapping = queue_mapping;
2677 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2679 return skb->queue_mapping;
2682 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2684 to->queue_mapping = from->queue_mapping;
2687 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2689 skb->queue_mapping = rx_queue + 1;
2692 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2694 return skb->queue_mapping - 1;
2697 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2699 return skb->queue_mapping != 0;
2702 u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2703 unsigned int num_tx_queues);
2705 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2714 /* Keeps track of mac header offset relative to skb->head.
2715 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2716 * For non-tunnel skb it points to skb_mac_header() and for
2717 * tunnel skb it points to outer mac header.
2718 * Keeps track of level of encapsulation of network headers.
2724 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2726 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2728 return (skb_mac_header(inner_skb) - inner_skb->head) -
2729 SKB_GSO_CB(inner_skb)->mac_offset;
2732 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
2734 int new_headroom, headroom;
2737 headroom = skb_headroom(skb);
2738 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
2742 new_headroom = skb_headroom(skb);
2743 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
2747 static inline bool skb_is_gso(const struct sk_buff *skb)
2749 return skb_shinfo(skb)->gso_size;
2752 /* Note: Should be called only if skb_is_gso(skb) is true */
2753 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
2755 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2758 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2760 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2762 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2763 * wanted then gso_type will be set. */
2764 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2766 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2767 unlikely(shinfo->gso_type == 0)) {
2768 __skb_warn_lro_forwarding(skb);
2774 static inline void skb_forward_csum(struct sk_buff *skb)
2776 /* Unfortunately we don't support this one. Any brave souls? */
2777 if (skb->ip_summed == CHECKSUM_COMPLETE)
2778 skb->ip_summed = CHECKSUM_NONE;
2782 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2783 * @skb: skb to check
2785 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2786 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2787 * use this helper, to document places where we make this assertion.
2789 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
2792 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2796 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2798 u32 __skb_get_poff(const struct sk_buff *skb);
2801 * skb_head_is_locked - Determine if the skb->head is locked down
2802 * @skb: skb to check
2804 * The head on skbs build around a head frag can be removed if they are
2805 * not cloned. This function returns true if the skb head is locked down
2806 * due to either being allocated via kmalloc, or by being a clone with
2807 * multiple references to the head.
2809 static inline bool skb_head_is_locked(const struct sk_buff *skb)
2811 return !skb->head_frag || skb_cloned(skb);
2813 #endif /* __KERNEL__ */
2814 #endif /* _LINUX_SKBUFF_H */