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 /* A. Checksumming of received packets by device.
41 * Device failed to checksum this packet e.g. due to lack of capabilities.
42 * The packet contains full (though not verified) checksum in packet but
43 * not in skb->csum. Thus, skb->csum is undefined in this case.
45 * CHECKSUM_UNNECESSARY:
47 * The hardware you're dealing with doesn't calculate the full checksum
48 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
49 * for specific protocols e.g. TCP/UDP/SCTP, then, for such packets it will
50 * set CHECKSUM_UNNECESSARY if their checksums are okay. skb->csum is still
51 * undefined in this case though. It is a bad option, but, unfortunately,
52 * nowadays most vendors do this. Apparently with the secret goal to sell
53 * you new devices, when you will add new protocol to your host, f.e. IPv6 8)
57 * This is the most generic way. The device supplied checksum of the _whole_
58 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
59 * hardware doesn't need to parse L3/L4 headers to implement this.
61 * Note: Even if device supports only some protocols, but is able to produce
62 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
66 * This is identical to the case for output below. This may occur on a packet
67 * received directly from another Linux OS, e.g., a virtualized Linux kernel
68 * on the same host. The packet can be treated in the same way as
69 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
70 * checksum must be filled in by the OS or the hardware.
72 * B. Checksumming on output.
76 * The skb was already checksummed by the protocol, or a checksum is not
81 * The device is required to checksum the packet as seen by hard_start_xmit()
82 * from skb->csum_start up to the end, and to record/write the checksum at
83 * offset skb->csum_start + skb->csum_offset.
85 * The device must show its capabilities in dev->features, set up at device
86 * setup time, e.g. netdev_features.h:
88 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
89 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
90 * IPv4. Sigh. Vendors like this way for an unknown reason.
91 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
92 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
93 * NETIF_F_... - Well, you get the picture.
95 * CHECKSUM_UNNECESSARY:
97 * Normally, the device will do per protocol specific checksumming. Protocol
98 * implementations that do not want the NIC to perform the checksum
99 * calculation should use this flag in their outgoing skbs.
101 * NETIF_F_FCOE_CRC - This indicates that 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
108 /* Don't change this without changing skb_csum_unnecessary! */
109 #define CHECKSUM_NONE 0
110 #define CHECKSUM_UNNECESSARY 1
111 #define CHECKSUM_COMPLETE 2
112 #define CHECKSUM_PARTIAL 3
114 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
115 ~(SMP_CACHE_BYTES - 1))
116 #define SKB_WITH_OVERHEAD(X) \
117 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
118 #define SKB_MAX_ORDER(X, ORDER) \
119 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
120 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
121 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
123 /* return minimum truesize of one skb containing X bytes of data */
124 #define SKB_TRUESIZE(X) ((X) + \
125 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
126 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
130 struct pipe_inode_info;
132 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
133 struct nf_conntrack {
138 #ifdef CONFIG_BRIDGE_NETFILTER
139 struct nf_bridge_info {
142 struct net_device *physindev;
143 struct net_device *physoutdev;
144 unsigned long data[32 / sizeof(unsigned long)];
148 struct sk_buff_head {
149 /* These two members must be first. */
150 struct sk_buff *next;
151 struct sk_buff *prev;
159 /* To allow 64K frame to be packed as single skb without frag_list we
160 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
161 * buffers which do not start on a page boundary.
163 * Since GRO uses frags we allocate at least 16 regardless of page
166 #if (65536/PAGE_SIZE + 1) < 16
167 #define MAX_SKB_FRAGS 16UL
169 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
172 typedef struct skb_frag_struct skb_frag_t;
174 struct skb_frag_struct {
178 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
187 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
192 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
197 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
202 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
207 #define HAVE_HW_TIME_STAMP
210 * struct skb_shared_hwtstamps - hardware time stamps
211 * @hwtstamp: hardware time stamp transformed into duration
212 * since arbitrary point in time
213 * @syststamp: hwtstamp transformed to system time base
215 * Software time stamps generated by ktime_get_real() are stored in
216 * skb->tstamp. The relation between the different kinds of time
217 * stamps is as follows:
219 * syststamp and tstamp can be compared against each other in
220 * arbitrary combinations. The accuracy of a
221 * syststamp/tstamp/"syststamp from other device" comparison is
222 * limited by the accuracy of the transformation into system time
223 * base. This depends on the device driver and its underlying
226 * hwtstamps can only be compared against other hwtstamps from
229 * This structure is attached to packets as part of the
230 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
232 struct skb_shared_hwtstamps {
237 /* Definitions for tx_flags in struct skb_shared_info */
239 /* generate hardware time stamp */
240 SKBTX_HW_TSTAMP = 1 << 0,
242 /* generate software time stamp */
243 SKBTX_SW_TSTAMP = 1 << 1,
245 /* device driver is going to provide hardware time stamp */
246 SKBTX_IN_PROGRESS = 1 << 2,
248 /* device driver supports TX zero-copy buffers */
249 SKBTX_DEV_ZEROCOPY = 1 << 3,
251 /* generate wifi status information (where possible) */
252 SKBTX_WIFI_STATUS = 1 << 4,
254 /* This indicates at least one fragment might be overwritten
255 * (as in vmsplice(), sendfile() ...)
256 * If we need to compute a TX checksum, we'll need to copy
257 * all frags to avoid possible bad checksum
259 SKBTX_SHARED_FRAG = 1 << 5,
263 * The callback notifies userspace to release buffers when skb DMA is done in
264 * lower device, the skb last reference should be 0 when calling this.
265 * The zerocopy_success argument is true if zero copy transmit occurred,
266 * false on data copy or out of memory error caused by data copy attempt.
267 * The ctx field is used to track device context.
268 * The desc field is used to track userspace buffer index.
271 void (*callback)(struct ubuf_info *, bool zerocopy_success);
276 /* This data is invariant across clones and lives at
277 * the end of the header data, ie. at skb->end.
279 struct skb_shared_info {
280 unsigned char nr_frags;
282 unsigned short gso_size;
283 /* Warning: this field is not always filled in (UFO)! */
284 unsigned short gso_segs;
285 unsigned short gso_type;
286 struct sk_buff *frag_list;
287 struct skb_shared_hwtstamps hwtstamps;
291 * Warning : all fields before dataref are cleared in __alloc_skb()
295 /* Intermediate layers must ensure that destructor_arg
296 * remains valid until skb destructor */
297 void * destructor_arg;
299 /* must be last field, see pskb_expand_head() */
300 skb_frag_t frags[MAX_SKB_FRAGS];
303 /* We divide dataref into two halves. The higher 16 bits hold references
304 * to the payload part of skb->data. The lower 16 bits hold references to
305 * the entire skb->data. A clone of a headerless skb holds the length of
306 * the header in skb->hdr_len.
308 * All users must obey the rule that the skb->data reference count must be
309 * greater than or equal to the payload reference count.
311 * Holding a reference to the payload part means that the user does not
312 * care about modifications to the header part of skb->data.
314 #define SKB_DATAREF_SHIFT 16
315 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
319 SKB_FCLONE_UNAVAILABLE,
325 SKB_GSO_TCPV4 = 1 << 0,
326 SKB_GSO_UDP = 1 << 1,
328 /* This indicates the skb is from an untrusted source. */
329 SKB_GSO_DODGY = 1 << 2,
331 /* This indicates the tcp segment has CWR set. */
332 SKB_GSO_TCP_ECN = 1 << 3,
334 SKB_GSO_TCPV6 = 1 << 4,
336 SKB_GSO_FCOE = 1 << 5,
338 SKB_GSO_GRE = 1 << 6,
340 SKB_GSO_IPIP = 1 << 7,
342 SKB_GSO_SIT = 1 << 8,
344 SKB_GSO_UDP_TUNNEL = 1 << 9,
346 SKB_GSO_MPLS = 1 << 10,
349 #if BITS_PER_LONG > 32
350 #define NET_SKBUFF_DATA_USES_OFFSET 1
353 #ifdef NET_SKBUFF_DATA_USES_OFFSET
354 typedef unsigned int sk_buff_data_t;
356 typedef unsigned char *sk_buff_data_t;
360 * struct sk_buff - socket buffer
361 * @next: Next buffer in list
362 * @prev: Previous buffer in list
363 * @tstamp: Time we arrived
364 * @sk: Socket we are owned by
365 * @dev: Device we arrived on/are leaving by
366 * @cb: Control buffer. Free for use by every layer. Put private vars here
367 * @_skb_refdst: destination entry (with norefcount bit)
368 * @sp: the security path, used for xfrm
369 * @len: Length of actual data
370 * @data_len: Data length
371 * @mac_len: Length of link layer header
372 * @hdr_len: writable header length of cloned skb
373 * @csum: Checksum (must include start/offset pair)
374 * @csum_start: Offset from skb->head where checksumming should start
375 * @csum_offset: Offset from csum_start where checksum should be stored
376 * @priority: Packet queueing priority
377 * @local_df: allow local fragmentation
378 * @cloned: Head may be cloned (check refcnt to be sure)
379 * @ip_summed: Driver fed us an IP checksum
380 * @nohdr: Payload reference only, must not modify header
381 * @nfctinfo: Relationship of this skb to the connection
382 * @pkt_type: Packet class
383 * @fclone: skbuff clone status
384 * @ipvs_property: skbuff is owned by ipvs
385 * @peeked: this packet has been seen already, so stats have been
386 * done for it, don't do them again
387 * @nf_trace: netfilter packet trace flag
388 * @protocol: Packet protocol from driver
389 * @destructor: Destruct function
390 * @nfct: Associated connection, if any
391 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
392 * @skb_iif: ifindex of device we arrived on
393 * @tc_index: Traffic control index
394 * @tc_verd: traffic control verdict
395 * @rxhash: the packet hash computed on receive
396 * @queue_mapping: Queue mapping for multiqueue devices
397 * @ndisc_nodetype: router type (from link layer)
398 * @ooo_okay: allow the mapping of a socket to a queue to be changed
399 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
401 * @wifi_acked_valid: wifi_acked was set
402 * @wifi_acked: whether frame was acked on wifi or not
403 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
404 * @dma_cookie: a cookie to one of several possible DMA operations
405 * done by skb DMA functions
406 * @napi_id: id of the NAPI struct this skb came from
407 * @secmark: security marking
408 * @mark: Generic packet mark
409 * @dropcount: total number of sk_receive_queue overflows
410 * @vlan_proto: vlan encapsulation protocol
411 * @vlan_tci: vlan tag control information
412 * @inner_protocol: Protocol (encapsulation)
413 * @inner_transport_header: Inner transport layer header (encapsulation)
414 * @inner_network_header: Network layer header (encapsulation)
415 * @inner_mac_header: Link layer header (encapsulation)
416 * @transport_header: Transport layer header
417 * @network_header: Network layer header
418 * @mac_header: Link layer header
419 * @tail: Tail pointer
421 * @head: Head of buffer
422 * @data: Data head pointer
423 * @truesize: Buffer size
424 * @users: User count - see {datagram,tcp}.c
428 /* These two members must be first. */
429 struct sk_buff *next;
430 struct sk_buff *prev;
435 struct net_device *dev;
438 * This is the control buffer. It is free to use for every
439 * layer. Please put your private variables there. If you
440 * want to keep them across layers you have to do a skb_clone()
441 * first. This is owned by whoever has the skb queued ATM.
443 char cb[48] __aligned(8);
445 unsigned long _skb_refdst;
461 kmemcheck_bitfield_begin(flags1);
472 kmemcheck_bitfield_end(flags1);
475 void (*destructor)(struct sk_buff *skb);
476 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
477 struct nf_conntrack *nfct;
479 #ifdef CONFIG_BRIDGE_NETFILTER
480 struct nf_bridge_info *nf_bridge;
490 #ifdef CONFIG_NET_SCHED
491 __u16 tc_index; /* traffic control index */
492 #ifdef CONFIG_NET_CLS_ACT
493 __u16 tc_verd; /* traffic control verdict */
498 kmemcheck_bitfield_begin(flags2);
499 #ifdef CONFIG_IPV6_NDISC_NODETYPE
500 __u8 ndisc_nodetype:2;
505 __u8 wifi_acked_valid:1;
509 /* Encapsulation protocol and NIC drivers should use
510 * this flag to indicate to each other if the skb contains
511 * encapsulated packet or not and maybe use the inner packet
514 __u8 encapsulation:1;
515 /* 6/8 bit hole (depending on ndisc_nodetype presence) */
516 kmemcheck_bitfield_end(flags2);
518 #if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
520 unsigned int napi_id;
521 dma_cookie_t dma_cookie;
524 #ifdef CONFIG_NETWORK_SECMARK
530 __u32 reserved_tailroom;
533 __be16 inner_protocol;
534 __u16 inner_transport_header;
535 __u16 inner_network_header;
536 __u16 inner_mac_header;
537 __u16 transport_header;
538 __u16 network_header;
540 /* These elements must be at the end, see alloc_skb() for details. */
545 unsigned int truesize;
551 * Handling routines are only of interest to the kernel
553 #include <linux/slab.h>
556 #define SKB_ALLOC_FCLONE 0x01
557 #define SKB_ALLOC_RX 0x02
559 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
560 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
562 return unlikely(skb->pfmemalloc);
566 * skb might have a dst pointer attached, refcounted or not.
567 * _skb_refdst low order bit is set if refcount was _not_ taken
569 #define SKB_DST_NOREF 1UL
570 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
573 * skb_dst - returns skb dst_entry
576 * Returns skb dst_entry, regardless of reference taken or not.
578 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
580 /* If refdst was not refcounted, check we still are in a
581 * rcu_read_lock section
583 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
584 !rcu_read_lock_held() &&
585 !rcu_read_lock_bh_held());
586 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
590 * skb_dst_set - sets skb dst
594 * Sets skb dst, assuming a reference was taken on dst and should
595 * be released by skb_dst_drop()
597 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
599 skb->_skb_refdst = (unsigned long)dst;
602 void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
606 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
610 * Sets skb dst, assuming a reference was not taken on dst.
611 * If dst entry is cached, we do not take reference and dst_release
612 * will be avoided by refdst_drop. If dst entry is not cached, we take
613 * reference, so that last dst_release can destroy the dst immediately.
615 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
617 __skb_dst_set_noref(skb, dst, false);
621 * skb_dst_set_noref_force - sets skb dst, without taking reference
625 * Sets skb dst, assuming a reference was not taken on dst.
626 * No reference is taken and no dst_release will be called. While for
627 * cached dsts deferred reclaim is a basic feature, for entries that are
628 * not cached it is caller's job to guarantee that last dst_release for
629 * provided dst happens when nobody uses it, eg. after a RCU grace period.
631 static inline void skb_dst_set_noref_force(struct sk_buff *skb,
632 struct dst_entry *dst)
634 __skb_dst_set_noref(skb, dst, true);
638 * skb_dst_is_noref - Test if skb dst isn't refcounted
641 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
643 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
646 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
648 return (struct rtable *)skb_dst(skb);
651 void kfree_skb(struct sk_buff *skb);
652 void kfree_skb_list(struct sk_buff *segs);
653 void skb_tx_error(struct sk_buff *skb);
654 void consume_skb(struct sk_buff *skb);
655 void __kfree_skb(struct sk_buff *skb);
656 extern struct kmem_cache *skbuff_head_cache;
658 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
659 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
660 bool *fragstolen, int *delta_truesize);
662 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
664 struct sk_buff *build_skb(void *data, unsigned int frag_size);
665 static inline struct sk_buff *alloc_skb(unsigned int size,
668 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
671 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
674 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
677 struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
678 static inline struct sk_buff *alloc_skb_head(gfp_t priority)
680 return __alloc_skb_head(priority, -1);
683 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
684 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
685 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
686 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
687 struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
689 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
690 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
691 unsigned int headroom);
692 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
693 int newtailroom, gfp_t priority);
694 int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
696 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
697 int skb_pad(struct sk_buff *skb, int pad);
698 #define dev_kfree_skb(a) consume_skb(a)
700 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
701 int getfrag(void *from, char *to, int offset,
702 int len, int odd, struct sk_buff *skb),
703 void *from, int length);
705 struct skb_seq_state {
709 __u32 stepped_offset;
710 struct sk_buff *root_skb;
711 struct sk_buff *cur_skb;
715 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
716 unsigned int to, struct skb_seq_state *st);
717 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
718 struct skb_seq_state *st);
719 void skb_abort_seq_read(struct skb_seq_state *st);
721 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
722 unsigned int to, struct ts_config *config,
723 struct ts_state *state);
726 * Packet hash types specify the type of hash in skb_set_hash.
728 * Hash types refer to the protocol layer addresses which are used to
729 * construct a packet's hash. The hashes are used to differentiate or identify
730 * flows of the protocol layer for the hash type. Hash types are either
731 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
733 * Properties of hashes:
735 * 1) Two packets in different flows have different hash values
736 * 2) Two packets in the same flow should have the same hash value
738 * A hash at a higher layer is considered to be more specific. A driver should
739 * set the most specific hash possible.
741 * A driver cannot indicate a more specific hash than the layer at which a hash
742 * was computed. For instance an L3 hash cannot be set as an L4 hash.
744 * A driver may indicate a hash level which is less specific than the
745 * actual layer the hash was computed on. For instance, a hash computed
746 * at L4 may be considered an L3 hash. This should only be done if the
747 * driver can't unambiguously determine that the HW computed the hash at
748 * the higher layer. Note that the "should" in the second property above
751 enum pkt_hash_types {
752 PKT_HASH_TYPE_NONE, /* Undefined type */
753 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
754 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
755 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
759 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
761 skb->l4_rxhash = (type == PKT_HASH_TYPE_L4);
765 void __skb_get_hash(struct sk_buff *skb);
766 static inline __u32 skb_get_hash(struct sk_buff *skb)
774 static inline void skb_clear_hash(struct sk_buff *skb)
780 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
786 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
788 to->rxhash = from->rxhash;
789 to->l4_rxhash = from->l4_rxhash;
792 #ifdef NET_SKBUFF_DATA_USES_OFFSET
793 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
795 return skb->head + skb->end;
798 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
803 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
808 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
810 return skb->end - skb->head;
815 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
817 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
819 return &skb_shinfo(skb)->hwtstamps;
823 * skb_queue_empty - check if a queue is empty
826 * Returns true if the queue is empty, false otherwise.
828 static inline int skb_queue_empty(const struct sk_buff_head *list)
830 return list->next == (struct sk_buff *)list;
834 * skb_queue_is_last - check if skb is the last entry in the queue
838 * Returns true if @skb is the last buffer on the list.
840 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
841 const struct sk_buff *skb)
843 return skb->next == (struct sk_buff *)list;
847 * skb_queue_is_first - check if skb is the first entry in the queue
851 * Returns true if @skb is the first buffer on the list.
853 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
854 const struct sk_buff *skb)
856 return skb->prev == (struct sk_buff *)list;
860 * skb_queue_next - return the next packet in the queue
862 * @skb: current buffer
864 * Return the next packet in @list after @skb. It is only valid to
865 * call this if skb_queue_is_last() evaluates to false.
867 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
868 const struct sk_buff *skb)
870 /* This BUG_ON may seem severe, but if we just return then we
871 * are going to dereference garbage.
873 BUG_ON(skb_queue_is_last(list, skb));
878 * skb_queue_prev - return the prev packet in the queue
880 * @skb: current buffer
882 * Return the prev packet in @list before @skb. It is only valid to
883 * call this if skb_queue_is_first() evaluates to false.
885 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
886 const struct sk_buff *skb)
888 /* This BUG_ON may seem severe, but if we just return then we
889 * are going to dereference garbage.
891 BUG_ON(skb_queue_is_first(list, skb));
896 * skb_get - reference buffer
897 * @skb: buffer to reference
899 * Makes another reference to a socket buffer and returns a pointer
902 static inline struct sk_buff *skb_get(struct sk_buff *skb)
904 atomic_inc(&skb->users);
909 * If users == 1, we are the only owner and are can avoid redundant
914 * skb_cloned - is the buffer a clone
915 * @skb: buffer to check
917 * Returns true if the buffer was generated with skb_clone() and is
918 * one of multiple shared copies of the buffer. Cloned buffers are
919 * shared data so must not be written to under normal circumstances.
921 static inline int skb_cloned(const struct sk_buff *skb)
923 return skb->cloned &&
924 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
927 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
929 might_sleep_if(pri & __GFP_WAIT);
932 return pskb_expand_head(skb, 0, 0, pri);
938 * skb_header_cloned - is the header a clone
939 * @skb: buffer to check
941 * Returns true if modifying the header part of the buffer requires
942 * the data to be copied.
944 static inline int skb_header_cloned(const struct sk_buff *skb)
951 dataref = atomic_read(&skb_shinfo(skb)->dataref);
952 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
957 * skb_header_release - release reference to header
958 * @skb: buffer to operate on
960 * Drop a reference to the header part of the buffer. This is done
961 * by acquiring a payload reference. You must not read from the header
962 * part of skb->data after this.
964 static inline void skb_header_release(struct sk_buff *skb)
968 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
972 * skb_shared - is the buffer shared
973 * @skb: buffer to check
975 * Returns true if more than one person has a reference to this
978 static inline int skb_shared(const struct sk_buff *skb)
980 return atomic_read(&skb->users) != 1;
984 * skb_share_check - check if buffer is shared and if so clone it
985 * @skb: buffer to check
986 * @pri: priority for memory allocation
988 * If the buffer is shared the buffer is cloned and the old copy
989 * drops a reference. A new clone with a single reference is returned.
990 * If the buffer is not shared the original buffer is returned. When
991 * being called from interrupt status or with spinlocks held pri must
994 * NULL is returned on a memory allocation failure.
996 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
998 might_sleep_if(pri & __GFP_WAIT);
999 if (skb_shared(skb)) {
1000 struct sk_buff *nskb = skb_clone(skb, pri);
1012 * Copy shared buffers into a new sk_buff. We effectively do COW on
1013 * packets to handle cases where we have a local reader and forward
1014 * and a couple of other messy ones. The normal one is tcpdumping
1015 * a packet thats being forwarded.
1019 * skb_unshare - make a copy of a shared buffer
1020 * @skb: buffer to check
1021 * @pri: priority for memory allocation
1023 * If the socket buffer is a clone then this function creates a new
1024 * copy of the data, drops a reference count on the old copy and returns
1025 * the new copy with the reference count at 1. If the buffer is not a clone
1026 * the original buffer is returned. When called with a spinlock held or
1027 * from interrupt state @pri must be %GFP_ATOMIC
1029 * %NULL is returned on a memory allocation failure.
1031 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1034 might_sleep_if(pri & __GFP_WAIT);
1035 if (skb_cloned(skb)) {
1036 struct sk_buff *nskb = skb_copy(skb, pri);
1037 kfree_skb(skb); /* Free our shared copy */
1044 * skb_peek - peek at the head of an &sk_buff_head
1045 * @list_: list to peek at
1047 * Peek an &sk_buff. Unlike most other operations you _MUST_
1048 * be careful with this one. A peek leaves the buffer on the
1049 * list and someone else may run off with it. You must hold
1050 * the appropriate locks or have a private queue to do this.
1052 * Returns %NULL for an empty list or a pointer to the head element.
1053 * The reference count is not incremented and the reference is therefore
1054 * volatile. Use with caution.
1056 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1058 struct sk_buff *skb = list_->next;
1060 if (skb == (struct sk_buff *)list_)
1066 * skb_peek_next - peek skb following the given one from a queue
1067 * @skb: skb to start from
1068 * @list_: list to peek at
1070 * Returns %NULL when the end of the list is met or a pointer to the
1071 * next element. The reference count is not incremented and the
1072 * reference is therefore volatile. Use with caution.
1074 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1075 const struct sk_buff_head *list_)
1077 struct sk_buff *next = skb->next;
1079 if (next == (struct sk_buff *)list_)
1085 * skb_peek_tail - peek at the tail of an &sk_buff_head
1086 * @list_: list to peek at
1088 * Peek an &sk_buff. Unlike most other operations you _MUST_
1089 * be careful with this one. A peek leaves the buffer on the
1090 * list and someone else may run off with it. You must hold
1091 * the appropriate locks or have a private queue to do this.
1093 * Returns %NULL for an empty list or a pointer to the tail element.
1094 * The reference count is not incremented and the reference is therefore
1095 * volatile. Use with caution.
1097 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1099 struct sk_buff *skb = list_->prev;
1101 if (skb == (struct sk_buff *)list_)
1108 * skb_queue_len - get queue length
1109 * @list_: list to measure
1111 * Return the length of an &sk_buff queue.
1113 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1119 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1120 * @list: queue to initialize
1122 * This initializes only the list and queue length aspects of
1123 * an sk_buff_head object. This allows to initialize the list
1124 * aspects of an sk_buff_head without reinitializing things like
1125 * the spinlock. It can also be used for on-stack sk_buff_head
1126 * objects where the spinlock is known to not be used.
1128 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1130 list->prev = list->next = (struct sk_buff *)list;
1135 * This function creates a split out lock class for each invocation;
1136 * this is needed for now since a whole lot of users of the skb-queue
1137 * infrastructure in drivers have different locking usage (in hardirq)
1138 * than the networking core (in softirq only). In the long run either the
1139 * network layer or drivers should need annotation to consolidate the
1140 * main types of usage into 3 classes.
1142 static inline void skb_queue_head_init(struct sk_buff_head *list)
1144 spin_lock_init(&list->lock);
1145 __skb_queue_head_init(list);
1148 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1149 struct lock_class_key *class)
1151 skb_queue_head_init(list);
1152 lockdep_set_class(&list->lock, class);
1156 * Insert an sk_buff on a list.
1158 * The "__skb_xxxx()" functions are the non-atomic ones that
1159 * can only be called with interrupts disabled.
1161 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1162 struct sk_buff_head *list);
1163 static inline void __skb_insert(struct sk_buff *newsk,
1164 struct sk_buff *prev, struct sk_buff *next,
1165 struct sk_buff_head *list)
1169 next->prev = prev->next = newsk;
1173 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1174 struct sk_buff *prev,
1175 struct sk_buff *next)
1177 struct sk_buff *first = list->next;
1178 struct sk_buff *last = list->prev;
1188 * skb_queue_splice - join two skb lists, this is designed for stacks
1189 * @list: the new list to add
1190 * @head: the place to add it in the first list
1192 static inline void skb_queue_splice(const struct sk_buff_head *list,
1193 struct sk_buff_head *head)
1195 if (!skb_queue_empty(list)) {
1196 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1197 head->qlen += list->qlen;
1202 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1203 * @list: the new list to add
1204 * @head: the place to add it in the first list
1206 * The list at @list is reinitialised
1208 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1209 struct sk_buff_head *head)
1211 if (!skb_queue_empty(list)) {
1212 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1213 head->qlen += list->qlen;
1214 __skb_queue_head_init(list);
1219 * skb_queue_splice_tail - join two skb lists, each list being a queue
1220 * @list: the new list to add
1221 * @head: the place to add it in the first list
1223 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1224 struct sk_buff_head *head)
1226 if (!skb_queue_empty(list)) {
1227 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1228 head->qlen += list->qlen;
1233 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1234 * @list: the new list to add
1235 * @head: the place to add it in the first list
1237 * Each of the lists is a queue.
1238 * The list at @list is reinitialised
1240 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1241 struct sk_buff_head *head)
1243 if (!skb_queue_empty(list)) {
1244 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1245 head->qlen += list->qlen;
1246 __skb_queue_head_init(list);
1251 * __skb_queue_after - queue a buffer at the list head
1252 * @list: list to use
1253 * @prev: place after this buffer
1254 * @newsk: buffer to queue
1256 * Queue a buffer int the middle of a list. This function takes no locks
1257 * and you must therefore hold required locks before calling it.
1259 * A buffer cannot be placed on two lists at the same time.
1261 static inline void __skb_queue_after(struct sk_buff_head *list,
1262 struct sk_buff *prev,
1263 struct sk_buff *newsk)
1265 __skb_insert(newsk, prev, prev->next, list);
1268 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1269 struct sk_buff_head *list);
1271 static inline void __skb_queue_before(struct sk_buff_head *list,
1272 struct sk_buff *next,
1273 struct sk_buff *newsk)
1275 __skb_insert(newsk, next->prev, next, list);
1279 * __skb_queue_head - queue a buffer at the list head
1280 * @list: list to use
1281 * @newsk: buffer to queue
1283 * Queue a buffer at the start of a list. This function takes no locks
1284 * and you must therefore hold required locks before calling it.
1286 * A buffer cannot be placed on two lists at the same time.
1288 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1289 static inline void __skb_queue_head(struct sk_buff_head *list,
1290 struct sk_buff *newsk)
1292 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1296 * __skb_queue_tail - queue a buffer at the list tail
1297 * @list: list to use
1298 * @newsk: buffer to queue
1300 * Queue a buffer at the end of a list. This function takes no locks
1301 * and you must therefore hold required locks before calling it.
1303 * A buffer cannot be placed on two lists at the same time.
1305 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1306 static inline void __skb_queue_tail(struct sk_buff_head *list,
1307 struct sk_buff *newsk)
1309 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1313 * remove sk_buff from list. _Must_ be called atomically, and with
1316 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1317 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1319 struct sk_buff *next, *prev;
1324 skb->next = skb->prev = NULL;
1330 * __skb_dequeue - remove from the head of the queue
1331 * @list: list to dequeue from
1333 * Remove the head of the list. This function does not take any locks
1334 * so must be used with appropriate locks held only. The head item is
1335 * returned or %NULL if the list is empty.
1337 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1338 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1340 struct sk_buff *skb = skb_peek(list);
1342 __skb_unlink(skb, list);
1347 * __skb_dequeue_tail - remove from the tail of the queue
1348 * @list: list to dequeue from
1350 * Remove the tail of the list. This function does not take any locks
1351 * so must be used with appropriate locks held only. The tail item is
1352 * returned or %NULL if the list is empty.
1354 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1355 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1357 struct sk_buff *skb = skb_peek_tail(list);
1359 __skb_unlink(skb, list);
1364 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1366 return skb->data_len;
1369 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1371 return skb->len - skb->data_len;
1374 static inline int skb_pagelen(const struct sk_buff *skb)
1378 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1379 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1380 return len + skb_headlen(skb);
1384 * __skb_fill_page_desc - initialise a paged fragment in an skb
1385 * @skb: buffer containing fragment to be initialised
1386 * @i: paged fragment index to initialise
1387 * @page: the page to use for this fragment
1388 * @off: the offset to the data with @page
1389 * @size: the length of the data
1391 * Initialises the @i'th fragment of @skb to point to &size bytes at
1392 * offset @off within @page.
1394 * Does not take any additional reference on the fragment.
1396 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1397 struct page *page, int off, int size)
1399 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1402 * Propagate page->pfmemalloc to the skb if we can. The problem is
1403 * that not all callers have unique ownership of the page. If
1404 * pfmemalloc is set, we check the mapping as a mapping implies
1405 * page->index is set (index and pfmemalloc share space).
1406 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1407 * do not lose pfmemalloc information as the pages would not be
1408 * allocated using __GFP_MEMALLOC.
1410 frag->page.p = page;
1411 frag->page_offset = off;
1412 skb_frag_size_set(frag, size);
1414 page = compound_head(page);
1415 if (page->pfmemalloc && !page->mapping)
1416 skb->pfmemalloc = true;
1420 * skb_fill_page_desc - initialise a paged fragment in an skb
1421 * @skb: buffer containing fragment to be initialised
1422 * @i: paged fragment index to initialise
1423 * @page: the page to use for this fragment
1424 * @off: the offset to the data with @page
1425 * @size: the length of the data
1427 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1428 * @skb to point to @size bytes at offset @off within @page. In
1429 * addition updates @skb such that @i is the last fragment.
1431 * Does not take any additional reference on the fragment.
1433 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1434 struct page *page, int off, int size)
1436 __skb_fill_page_desc(skb, i, page, off, size);
1437 skb_shinfo(skb)->nr_frags = i + 1;
1440 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1441 int size, unsigned int truesize);
1443 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1444 unsigned int truesize);
1446 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1447 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1448 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1450 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1451 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1453 return skb->head + skb->tail;
1456 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1458 skb->tail = skb->data - skb->head;
1461 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1463 skb_reset_tail_pointer(skb);
1464 skb->tail += offset;
1467 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1468 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1473 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1475 skb->tail = skb->data;
1478 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1480 skb->tail = skb->data + offset;
1483 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1486 * Add data to an sk_buff
1488 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1489 unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1490 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1492 unsigned char *tmp = skb_tail_pointer(skb);
1493 SKB_LINEAR_ASSERT(skb);
1499 unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1500 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1507 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1508 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1511 BUG_ON(skb->len < skb->data_len);
1512 return skb->data += len;
1515 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1517 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1520 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1522 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1524 if (len > skb_headlen(skb) &&
1525 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1528 return skb->data += len;
1531 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1533 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1536 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1538 if (likely(len <= skb_headlen(skb)))
1540 if (unlikely(len > skb->len))
1542 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1546 * skb_headroom - bytes at buffer head
1547 * @skb: buffer to check
1549 * Return the number of bytes of free space at the head of an &sk_buff.
1551 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1553 return skb->data - skb->head;
1557 * skb_tailroom - bytes at buffer end
1558 * @skb: buffer to check
1560 * Return the number of bytes of free space at the tail of an sk_buff
1562 static inline int skb_tailroom(const struct sk_buff *skb)
1564 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1568 * skb_availroom - bytes at buffer end
1569 * @skb: buffer to check
1571 * Return the number of bytes of free space at the tail of an sk_buff
1572 * allocated by sk_stream_alloc()
1574 static inline int skb_availroom(const struct sk_buff *skb)
1576 if (skb_is_nonlinear(skb))
1579 return skb->end - skb->tail - skb->reserved_tailroom;
1583 * skb_reserve - adjust headroom
1584 * @skb: buffer to alter
1585 * @len: bytes to move
1587 * Increase the headroom of an empty &sk_buff by reducing the tail
1588 * room. This is only allowed for an empty buffer.
1590 static inline void skb_reserve(struct sk_buff *skb, int len)
1596 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1598 skb->inner_mac_header = skb->mac_header;
1599 skb->inner_network_header = skb->network_header;
1600 skb->inner_transport_header = skb->transport_header;
1603 static inline void skb_reset_mac_len(struct sk_buff *skb)
1605 skb->mac_len = skb->network_header - skb->mac_header;
1608 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1611 return skb->head + skb->inner_transport_header;
1614 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1616 skb->inner_transport_header = skb->data - skb->head;
1619 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1622 skb_reset_inner_transport_header(skb);
1623 skb->inner_transport_header += offset;
1626 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1628 return skb->head + skb->inner_network_header;
1631 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1633 skb->inner_network_header = skb->data - skb->head;
1636 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1639 skb_reset_inner_network_header(skb);
1640 skb->inner_network_header += offset;
1643 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1645 return skb->head + skb->inner_mac_header;
1648 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1650 skb->inner_mac_header = skb->data - skb->head;
1653 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1656 skb_reset_inner_mac_header(skb);
1657 skb->inner_mac_header += offset;
1659 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1661 return skb->transport_header != (typeof(skb->transport_header))~0U;
1664 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1666 return skb->head + skb->transport_header;
1669 static inline void skb_reset_transport_header(struct sk_buff *skb)
1671 skb->transport_header = skb->data - skb->head;
1674 static inline void skb_set_transport_header(struct sk_buff *skb,
1677 skb_reset_transport_header(skb);
1678 skb->transport_header += offset;
1681 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1683 return skb->head + skb->network_header;
1686 static inline void skb_reset_network_header(struct sk_buff *skb)
1688 skb->network_header = skb->data - skb->head;
1691 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1693 skb_reset_network_header(skb);
1694 skb->network_header += offset;
1697 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1699 return skb->head + skb->mac_header;
1702 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1704 return skb->mac_header != (typeof(skb->mac_header))~0U;
1707 static inline void skb_reset_mac_header(struct sk_buff *skb)
1709 skb->mac_header = skb->data - skb->head;
1712 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1714 skb_reset_mac_header(skb);
1715 skb->mac_header += offset;
1718 static inline void skb_pop_mac_header(struct sk_buff *skb)
1720 skb->mac_header = skb->network_header;
1723 static inline void skb_probe_transport_header(struct sk_buff *skb,
1724 const int offset_hint)
1726 struct flow_keys keys;
1728 if (skb_transport_header_was_set(skb))
1730 else if (skb_flow_dissect(skb, &keys))
1731 skb_set_transport_header(skb, keys.thoff);
1733 skb_set_transport_header(skb, offset_hint);
1736 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1738 if (skb_mac_header_was_set(skb)) {
1739 const unsigned char *old_mac = skb_mac_header(skb);
1741 skb_set_mac_header(skb, -skb->mac_len);
1742 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1746 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1748 return skb->csum_start - skb_headroom(skb);
1751 static inline int skb_transport_offset(const struct sk_buff *skb)
1753 return skb_transport_header(skb) - skb->data;
1756 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1758 return skb->transport_header - skb->network_header;
1761 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1763 return skb->inner_transport_header - skb->inner_network_header;
1766 static inline int skb_network_offset(const struct sk_buff *skb)
1768 return skb_network_header(skb) - skb->data;
1771 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1773 return skb_inner_network_header(skb) - skb->data;
1776 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1778 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1782 * CPUs often take a performance hit when accessing unaligned memory
1783 * locations. The actual performance hit varies, it can be small if the
1784 * hardware handles it or large if we have to take an exception and fix it
1787 * Since an ethernet header is 14 bytes network drivers often end up with
1788 * the IP header at an unaligned offset. The IP header can be aligned by
1789 * shifting the start of the packet by 2 bytes. Drivers should do this
1792 * skb_reserve(skb, NET_IP_ALIGN);
1794 * The downside to this alignment of the IP header is that the DMA is now
1795 * unaligned. On some architectures the cost of an unaligned DMA is high
1796 * and this cost outweighs the gains made by aligning the IP header.
1798 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1801 #ifndef NET_IP_ALIGN
1802 #define NET_IP_ALIGN 2
1806 * The networking layer reserves some headroom in skb data (via
1807 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1808 * the header has to grow. In the default case, if the header has to grow
1809 * 32 bytes or less we avoid the reallocation.
1811 * Unfortunately this headroom changes the DMA alignment of the resulting
1812 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1813 * on some architectures. An architecture can override this value,
1814 * perhaps setting it to a cacheline in size (since that will maintain
1815 * cacheline alignment of the DMA). It must be a power of 2.
1817 * Various parts of the networking layer expect at least 32 bytes of
1818 * headroom, you should not reduce this.
1820 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1821 * to reduce average number of cache lines per packet.
1822 * get_rps_cpus() for example only access one 64 bytes aligned block :
1823 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1826 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1829 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1831 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1833 if (unlikely(skb_is_nonlinear(skb))) {
1838 skb_set_tail_pointer(skb, len);
1841 void skb_trim(struct sk_buff *skb, unsigned int len);
1843 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1846 return ___pskb_trim(skb, len);
1847 __skb_trim(skb, len);
1851 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1853 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1857 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1858 * @skb: buffer to alter
1861 * This is identical to pskb_trim except that the caller knows that
1862 * the skb is not cloned so we should never get an error due to out-
1865 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1867 int err = pskb_trim(skb, len);
1872 * skb_orphan - orphan a buffer
1873 * @skb: buffer to orphan
1875 * If a buffer currently has an owner then we call the owner's
1876 * destructor function and make the @skb unowned. The buffer continues
1877 * to exist but is no longer charged to its former owner.
1879 static inline void skb_orphan(struct sk_buff *skb)
1881 if (skb->destructor) {
1882 skb->destructor(skb);
1883 skb->destructor = NULL;
1891 * skb_orphan_frags - orphan the frags contained in a buffer
1892 * @skb: buffer to orphan frags from
1893 * @gfp_mask: allocation mask for replacement pages
1895 * For each frag in the SKB which needs a destructor (i.e. has an
1896 * owner) create a copy of that frag and release the original
1897 * page by calling the destructor.
1899 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1901 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1903 return skb_copy_ubufs(skb, gfp_mask);
1907 * __skb_queue_purge - empty a list
1908 * @list: list to empty
1910 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1911 * the list and one reference dropped. This function does not take the
1912 * list lock and the caller must hold the relevant locks to use it.
1914 void skb_queue_purge(struct sk_buff_head *list);
1915 static inline void __skb_queue_purge(struct sk_buff_head *list)
1917 struct sk_buff *skb;
1918 while ((skb = __skb_dequeue(list)) != NULL)
1922 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1923 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1924 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1926 void *netdev_alloc_frag(unsigned int fragsz);
1928 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1932 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1933 * @dev: network device to receive on
1934 * @length: length to allocate
1936 * Allocate a new &sk_buff and assign it a usage count of one. The
1937 * buffer has unspecified headroom built in. Users should allocate
1938 * the headroom they think they need without accounting for the
1939 * built in space. The built in space is used for optimisations.
1941 * %NULL is returned if there is no free memory. Although this function
1942 * allocates memory it can be called from an interrupt.
1944 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1945 unsigned int length)
1947 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1950 /* legacy helper around __netdev_alloc_skb() */
1951 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1954 return __netdev_alloc_skb(NULL, length, gfp_mask);
1957 /* legacy helper around netdev_alloc_skb() */
1958 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1960 return netdev_alloc_skb(NULL, length);
1964 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1965 unsigned int length, gfp_t gfp)
1967 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
1969 if (NET_IP_ALIGN && skb)
1970 skb_reserve(skb, NET_IP_ALIGN);
1974 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1975 unsigned int length)
1977 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1981 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1982 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1983 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1984 * @order: size of the allocation
1986 * Allocate a new page.
1988 * %NULL is returned if there is no free memory.
1990 static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1991 struct sk_buff *skb,
1996 gfp_mask |= __GFP_COLD;
1998 if (!(gfp_mask & __GFP_NOMEMALLOC))
1999 gfp_mask |= __GFP_MEMALLOC;
2001 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2002 if (skb && page && page->pfmemalloc)
2003 skb->pfmemalloc = true;
2009 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2010 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2011 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2013 * Allocate a new page.
2015 * %NULL is returned if there is no free memory.
2017 static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2018 struct sk_buff *skb)
2020 return __skb_alloc_pages(gfp_mask, skb, 0);
2024 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2025 * @page: The page that was allocated from skb_alloc_page
2026 * @skb: The skb that may need pfmemalloc set
2028 static inline void skb_propagate_pfmemalloc(struct page *page,
2029 struct sk_buff *skb)
2031 if (page && page->pfmemalloc)
2032 skb->pfmemalloc = true;
2036 * skb_frag_page - retrieve the page refered to by a paged fragment
2037 * @frag: the paged fragment
2039 * Returns the &struct page associated with @frag.
2041 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2043 return frag->page.p;
2047 * __skb_frag_ref - take an addition reference on a paged fragment.
2048 * @frag: the paged fragment
2050 * Takes an additional reference on the paged fragment @frag.
2052 static inline void __skb_frag_ref(skb_frag_t *frag)
2054 get_page(skb_frag_page(frag));
2058 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2060 * @f: the fragment offset.
2062 * Takes an additional reference on the @f'th paged fragment of @skb.
2064 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2066 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2070 * __skb_frag_unref - release a reference on a paged fragment.
2071 * @frag: the paged fragment
2073 * Releases a reference on the paged fragment @frag.
2075 static inline void __skb_frag_unref(skb_frag_t *frag)
2077 put_page(skb_frag_page(frag));
2081 * skb_frag_unref - release a reference on a paged fragment of an skb.
2083 * @f: the fragment offset
2085 * Releases a reference on the @f'th paged fragment of @skb.
2087 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2089 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2093 * skb_frag_address - gets the address of the data contained in a paged fragment
2094 * @frag: the paged fragment buffer
2096 * Returns the address of the data within @frag. The page must already
2099 static inline void *skb_frag_address(const skb_frag_t *frag)
2101 return page_address(skb_frag_page(frag)) + frag->page_offset;
2105 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2106 * @frag: the paged fragment buffer
2108 * Returns the address of the data within @frag. Checks that the page
2109 * is mapped and returns %NULL otherwise.
2111 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2113 void *ptr = page_address(skb_frag_page(frag));
2117 return ptr + frag->page_offset;
2121 * __skb_frag_set_page - sets the page contained in a paged fragment
2122 * @frag: the paged fragment
2123 * @page: the page to set
2125 * Sets the fragment @frag to contain @page.
2127 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2129 frag->page.p = page;
2133 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2135 * @f: the fragment offset
2136 * @page: the page to set
2138 * Sets the @f'th fragment of @skb to contain @page.
2140 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2143 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2146 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2149 * skb_frag_dma_map - maps a paged fragment via the DMA API
2150 * @dev: the device to map the fragment to
2151 * @frag: the paged fragment to map
2152 * @offset: the offset within the fragment (starting at the
2153 * fragment's own offset)
2154 * @size: the number of bytes to map
2155 * @dir: the direction of the mapping (%PCI_DMA_*)
2157 * Maps the page associated with @frag to @device.
2159 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2160 const skb_frag_t *frag,
2161 size_t offset, size_t size,
2162 enum dma_data_direction dir)
2164 return dma_map_page(dev, skb_frag_page(frag),
2165 frag->page_offset + offset, size, dir);
2168 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2171 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2175 * skb_clone_writable - is the header of a clone writable
2176 * @skb: buffer to check
2177 * @len: length up to which to write
2179 * Returns true if modifying the header part of the cloned buffer
2180 * does not requires the data to be copied.
2182 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2184 return !skb_header_cloned(skb) &&
2185 skb_headroom(skb) + len <= skb->hdr_len;
2188 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2193 if (headroom > skb_headroom(skb))
2194 delta = headroom - skb_headroom(skb);
2196 if (delta || cloned)
2197 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2203 * skb_cow - copy header of skb when it is required
2204 * @skb: buffer to cow
2205 * @headroom: needed headroom
2207 * If the skb passed lacks sufficient headroom or its data part
2208 * is shared, data is reallocated. If reallocation fails, an error
2209 * is returned and original skb is not changed.
2211 * The result is skb with writable area skb->head...skb->tail
2212 * and at least @headroom of space at head.
2214 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2216 return __skb_cow(skb, headroom, skb_cloned(skb));
2220 * skb_cow_head - skb_cow but only making the head writable
2221 * @skb: buffer to cow
2222 * @headroom: needed headroom
2224 * This function is identical to skb_cow except that we replace the
2225 * skb_cloned check by skb_header_cloned. It should be used when
2226 * you only need to push on some header and do not need to modify
2229 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2231 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2235 * skb_padto - pad an skbuff up to a minimal size
2236 * @skb: buffer to pad
2237 * @len: minimal length
2239 * Pads up a buffer to ensure the trailing bytes exist and are
2240 * blanked. If the buffer already contains sufficient data it
2241 * is untouched. Otherwise it is extended. Returns zero on
2242 * success. The skb is freed on error.
2245 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2247 unsigned int size = skb->len;
2248 if (likely(size >= len))
2250 return skb_pad(skb, len - size);
2253 static inline int skb_add_data(struct sk_buff *skb,
2254 char __user *from, int copy)
2256 const int off = skb->len;
2258 if (skb->ip_summed == CHECKSUM_NONE) {
2260 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
2263 skb->csum = csum_block_add(skb->csum, csum, off);
2266 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2269 __skb_trim(skb, off);
2273 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2274 const struct page *page, int off)
2277 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2279 return page == skb_frag_page(frag) &&
2280 off == frag->page_offset + skb_frag_size(frag);
2285 static inline int __skb_linearize(struct sk_buff *skb)
2287 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2291 * skb_linearize - convert paged skb to linear one
2292 * @skb: buffer to linarize
2294 * If there is no free memory -ENOMEM is returned, otherwise zero
2295 * is returned and the old skb data released.
2297 static inline int skb_linearize(struct sk_buff *skb)
2299 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2303 * skb_has_shared_frag - can any frag be overwritten
2304 * @skb: buffer to test
2306 * Return true if the skb has at least one frag that might be modified
2307 * by an external entity (as in vmsplice()/sendfile())
2309 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2311 return skb_is_nonlinear(skb) &&
2312 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2316 * skb_linearize_cow - make sure skb is linear and writable
2317 * @skb: buffer to process
2319 * If there is no free memory -ENOMEM is returned, otherwise zero
2320 * is returned and the old skb data released.
2322 static inline int skb_linearize_cow(struct sk_buff *skb)
2324 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2325 __skb_linearize(skb) : 0;
2329 * skb_postpull_rcsum - update checksum for received skb after pull
2330 * @skb: buffer to update
2331 * @start: start of data before pull
2332 * @len: length of data pulled
2334 * After doing a pull on a received packet, you need to call this to
2335 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2336 * CHECKSUM_NONE so that it can be recomputed from scratch.
2339 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2340 const void *start, unsigned int len)
2342 if (skb->ip_summed == CHECKSUM_COMPLETE)
2343 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2346 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2349 * pskb_trim_rcsum - trim received skb and update checksum
2350 * @skb: buffer to trim
2353 * This is exactly the same as pskb_trim except that it ensures the
2354 * checksum of received packets are still valid after the operation.
2357 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2359 if (likely(len >= skb->len))
2361 if (skb->ip_summed == CHECKSUM_COMPLETE)
2362 skb->ip_summed = CHECKSUM_NONE;
2363 return __pskb_trim(skb, len);
2366 #define skb_queue_walk(queue, skb) \
2367 for (skb = (queue)->next; \
2368 skb != (struct sk_buff *)(queue); \
2371 #define skb_queue_walk_safe(queue, skb, tmp) \
2372 for (skb = (queue)->next, tmp = skb->next; \
2373 skb != (struct sk_buff *)(queue); \
2374 skb = tmp, tmp = skb->next)
2376 #define skb_queue_walk_from(queue, skb) \
2377 for (; skb != (struct sk_buff *)(queue); \
2380 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2381 for (tmp = skb->next; \
2382 skb != (struct sk_buff *)(queue); \
2383 skb = tmp, tmp = skb->next)
2385 #define skb_queue_reverse_walk(queue, skb) \
2386 for (skb = (queue)->prev; \
2387 skb != (struct sk_buff *)(queue); \
2390 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2391 for (skb = (queue)->prev, tmp = skb->prev; \
2392 skb != (struct sk_buff *)(queue); \
2393 skb = tmp, tmp = skb->prev)
2395 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2396 for (tmp = skb->prev; \
2397 skb != (struct sk_buff *)(queue); \
2398 skb = tmp, tmp = skb->prev)
2400 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2402 return skb_shinfo(skb)->frag_list != NULL;
2405 static inline void skb_frag_list_init(struct sk_buff *skb)
2407 skb_shinfo(skb)->frag_list = NULL;
2410 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2412 frag->next = skb_shinfo(skb)->frag_list;
2413 skb_shinfo(skb)->frag_list = frag;
2416 #define skb_walk_frags(skb, iter) \
2417 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2419 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2420 int *peeked, int *off, int *err);
2421 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2423 unsigned int datagram_poll(struct file *file, struct socket *sock,
2424 struct poll_table_struct *wait);
2425 int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2426 struct iovec *to, int size);
2427 int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2429 int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2430 const struct iovec *from, int from_offset,
2432 int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2433 int offset, size_t count);
2434 int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2435 const struct iovec *to, int to_offset,
2437 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2438 void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2439 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
2440 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2441 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2442 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2443 int len, __wsum csum);
2444 int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2445 struct pipe_inode_info *pipe, unsigned int len,
2446 unsigned int flags);
2447 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2448 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2449 void skb_zerocopy(struct sk_buff *to, const struct sk_buff *from,
2451 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2452 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2453 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
2454 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
2456 struct skb_checksum_ops {
2457 __wsum (*update)(const void *mem, int len, __wsum wsum);
2458 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2461 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2462 __wsum csum, const struct skb_checksum_ops *ops);
2463 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2466 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2467 int len, void *buffer)
2469 int hlen = skb_headlen(skb);
2471 if (hlen - offset >= len)
2472 return skb->data + offset;
2474 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2481 * skb_needs_linearize - check if we need to linearize a given skb
2482 * depending on the given device features.
2483 * @skb: socket buffer to check
2484 * @features: net device features
2486 * Returns true if either:
2487 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2488 * 2. skb is fragmented and the device does not support SG.
2490 static inline bool skb_needs_linearize(struct sk_buff *skb,
2491 netdev_features_t features)
2493 return skb_is_nonlinear(skb) &&
2494 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2495 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2498 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2500 const unsigned int len)
2502 memcpy(to, skb->data, len);
2505 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2506 const int offset, void *to,
2507 const unsigned int len)
2509 memcpy(to, skb->data + offset, len);
2512 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2514 const unsigned int len)
2516 memcpy(skb->data, from, len);
2519 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2522 const unsigned int len)
2524 memcpy(skb->data + offset, from, len);
2527 void skb_init(void);
2529 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2535 * skb_get_timestamp - get timestamp from a skb
2536 * @skb: skb to get stamp from
2537 * @stamp: pointer to struct timeval to store stamp in
2539 * Timestamps are stored in the skb as offsets to a base timestamp.
2540 * This function converts the offset back to a struct timeval and stores
2543 static inline void skb_get_timestamp(const struct sk_buff *skb,
2544 struct timeval *stamp)
2546 *stamp = ktime_to_timeval(skb->tstamp);
2549 static inline void skb_get_timestampns(const struct sk_buff *skb,
2550 struct timespec *stamp)
2552 *stamp = ktime_to_timespec(skb->tstamp);
2555 static inline void __net_timestamp(struct sk_buff *skb)
2557 skb->tstamp = ktime_get_real();
2560 static inline ktime_t net_timedelta(ktime_t t)
2562 return ktime_sub(ktime_get_real(), t);
2565 static inline ktime_t net_invalid_timestamp(void)
2567 return ktime_set(0, 0);
2570 void skb_timestamping_init(void);
2572 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2574 void skb_clone_tx_timestamp(struct sk_buff *skb);
2575 bool skb_defer_rx_timestamp(struct sk_buff *skb);
2577 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2579 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2583 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2588 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2591 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2593 * PHY drivers may accept clones of transmitted packets for
2594 * timestamping via their phy_driver.txtstamp method. These drivers
2595 * must call this function to return the skb back to the stack, with
2596 * or without a timestamp.
2598 * @skb: clone of the the original outgoing packet
2599 * @hwtstamps: hardware time stamps, may be NULL if not available
2602 void skb_complete_tx_timestamp(struct sk_buff *skb,
2603 struct skb_shared_hwtstamps *hwtstamps);
2606 * skb_tstamp_tx - queue clone of skb with send time stamps
2607 * @orig_skb: the original outgoing packet
2608 * @hwtstamps: hardware time stamps, may be NULL if not available
2610 * If the skb has a socket associated, then this function clones the
2611 * skb (thus sharing the actual data and optional structures), stores
2612 * the optional hardware time stamping information (if non NULL) or
2613 * generates a software time stamp (otherwise), then queues the clone
2614 * to the error queue of the socket. Errors are silently ignored.
2616 void skb_tstamp_tx(struct sk_buff *orig_skb,
2617 struct skb_shared_hwtstamps *hwtstamps);
2619 static inline void sw_tx_timestamp(struct sk_buff *skb)
2621 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2622 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2623 skb_tstamp_tx(skb, NULL);
2627 * skb_tx_timestamp() - Driver hook for transmit timestamping
2629 * Ethernet MAC Drivers should call this function in their hard_xmit()
2630 * function immediately before giving the sk_buff to the MAC hardware.
2632 * Specifically, one should make absolutely sure that this function is
2633 * called before TX completion of this packet can trigger. Otherwise
2634 * the packet could potentially already be freed.
2636 * @skb: A socket buffer.
2638 static inline void skb_tx_timestamp(struct sk_buff *skb)
2640 skb_clone_tx_timestamp(skb);
2641 sw_tx_timestamp(skb);
2645 * skb_complete_wifi_ack - deliver skb with wifi status
2647 * @skb: the original outgoing packet
2648 * @acked: ack status
2651 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2653 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2654 __sum16 __skb_checksum_complete(struct sk_buff *skb);
2656 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2658 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2662 * skb_checksum_complete - Calculate checksum of an entire packet
2663 * @skb: packet to process
2665 * This function calculates the checksum over the entire packet plus
2666 * the value of skb->csum. The latter can be used to supply the
2667 * checksum of a pseudo header as used by TCP/UDP. It returns the
2670 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2671 * this function can be used to verify that checksum on received
2672 * packets. In that case the function should return zero if the
2673 * checksum is correct. In particular, this function will return zero
2674 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2675 * hardware has already verified the correctness of the checksum.
2677 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2679 return skb_csum_unnecessary(skb) ?
2680 0 : __skb_checksum_complete(skb);
2683 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2684 void nf_conntrack_destroy(struct nf_conntrack *nfct);
2685 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2687 if (nfct && atomic_dec_and_test(&nfct->use))
2688 nf_conntrack_destroy(nfct);
2690 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2693 atomic_inc(&nfct->use);
2696 #ifdef CONFIG_BRIDGE_NETFILTER
2697 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2699 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2702 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2705 atomic_inc(&nf_bridge->use);
2707 #endif /* CONFIG_BRIDGE_NETFILTER */
2708 static inline void nf_reset(struct sk_buff *skb)
2710 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2711 nf_conntrack_put(skb->nfct);
2714 #ifdef CONFIG_BRIDGE_NETFILTER
2715 nf_bridge_put(skb->nf_bridge);
2716 skb->nf_bridge = NULL;
2720 static inline void nf_reset_trace(struct sk_buff *skb)
2722 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
2727 /* Note: This doesn't put any conntrack and bridge info in dst. */
2728 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2730 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2731 dst->nfct = src->nfct;
2732 nf_conntrack_get(src->nfct);
2733 dst->nfctinfo = src->nfctinfo;
2735 #ifdef CONFIG_BRIDGE_NETFILTER
2736 dst->nf_bridge = src->nf_bridge;
2737 nf_bridge_get(src->nf_bridge);
2741 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2743 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2744 nf_conntrack_put(dst->nfct);
2746 #ifdef CONFIG_BRIDGE_NETFILTER
2747 nf_bridge_put(dst->nf_bridge);
2749 __nf_copy(dst, src);
2752 #ifdef CONFIG_NETWORK_SECMARK
2753 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2755 to->secmark = from->secmark;
2758 static inline void skb_init_secmark(struct sk_buff *skb)
2763 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2766 static inline void skb_init_secmark(struct sk_buff *skb)
2770 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2772 skb->queue_mapping = queue_mapping;
2775 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2777 return skb->queue_mapping;
2780 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2782 to->queue_mapping = from->queue_mapping;
2785 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2787 skb->queue_mapping = rx_queue + 1;
2790 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2792 return skb->queue_mapping - 1;
2795 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2797 return skb->queue_mapping != 0;
2800 u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2801 unsigned int num_tx_queues);
2803 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2812 /* Keeps track of mac header offset relative to skb->head.
2813 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2814 * For non-tunnel skb it points to skb_mac_header() and for
2815 * tunnel skb it points to outer mac header.
2816 * Keeps track of level of encapsulation of network headers.
2822 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2824 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2826 return (skb_mac_header(inner_skb) - inner_skb->head) -
2827 SKB_GSO_CB(inner_skb)->mac_offset;
2830 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
2832 int new_headroom, headroom;
2835 headroom = skb_headroom(skb);
2836 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
2840 new_headroom = skb_headroom(skb);
2841 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
2845 static inline bool skb_is_gso(const struct sk_buff *skb)
2847 return skb_shinfo(skb)->gso_size;
2850 /* Note: Should be called only if skb_is_gso(skb) is true */
2851 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
2853 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2856 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2858 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2860 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2861 * wanted then gso_type will be set. */
2862 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2864 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2865 unlikely(shinfo->gso_type == 0)) {
2866 __skb_warn_lro_forwarding(skb);
2872 static inline void skb_forward_csum(struct sk_buff *skb)
2874 /* Unfortunately we don't support this one. Any brave souls? */
2875 if (skb->ip_summed == CHECKSUM_COMPLETE)
2876 skb->ip_summed = CHECKSUM_NONE;
2880 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2881 * @skb: skb to check
2883 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2884 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2885 * use this helper, to document places where we make this assertion.
2887 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
2890 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2894 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2896 u32 __skb_get_poff(const struct sk_buff *skb);
2899 * skb_head_is_locked - Determine if the skb->head is locked down
2900 * @skb: skb to check
2902 * The head on skbs build around a head frag can be removed if they are
2903 * not cloned. This function returns true if the skb head is locked down
2904 * due to either being allocated via kmalloc, or by being a clone with
2905 * multiple references to the head.
2907 static inline bool skb_head_is_locked(const struct sk_buff *skb)
2909 return !skb->head_frag || skb_cloned(skb);
2911 #endif /* __KERNEL__ */
2912 #endif /* _LINUX_SKBUFF_H */