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 __u32 skb_get_hash_raw(const struct sk_buff *skb)
779 static inline void skb_clear_hash(struct sk_buff *skb)
785 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
791 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
793 to->rxhash = from->rxhash;
794 to->l4_rxhash = from->l4_rxhash;
797 #ifdef NET_SKBUFF_DATA_USES_OFFSET
798 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
800 return skb->head + skb->end;
803 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
808 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
813 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
815 return skb->end - skb->head;
820 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
822 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
824 return &skb_shinfo(skb)->hwtstamps;
828 * skb_queue_empty - check if a queue is empty
831 * Returns true if the queue is empty, false otherwise.
833 static inline int skb_queue_empty(const struct sk_buff_head *list)
835 return list->next == (const struct sk_buff *) list;
839 * skb_queue_is_last - check if skb is the last entry in the queue
843 * Returns true if @skb is the last buffer on the list.
845 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
846 const struct sk_buff *skb)
848 return skb->next == (const struct sk_buff *) list;
852 * skb_queue_is_first - check if skb is the first entry in the queue
856 * Returns true if @skb is the first buffer on the list.
858 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
859 const struct sk_buff *skb)
861 return skb->prev == (const struct sk_buff *) list;
865 * skb_queue_next - return the next packet in the queue
867 * @skb: current buffer
869 * Return the next packet in @list after @skb. It is only valid to
870 * call this if skb_queue_is_last() evaluates to false.
872 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
873 const struct sk_buff *skb)
875 /* This BUG_ON may seem severe, but if we just return then we
876 * are going to dereference garbage.
878 BUG_ON(skb_queue_is_last(list, skb));
883 * skb_queue_prev - return the prev packet in the queue
885 * @skb: current buffer
887 * Return the prev packet in @list before @skb. It is only valid to
888 * call this if skb_queue_is_first() evaluates to false.
890 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
891 const struct sk_buff *skb)
893 /* This BUG_ON may seem severe, but if we just return then we
894 * are going to dereference garbage.
896 BUG_ON(skb_queue_is_first(list, skb));
901 * skb_get - reference buffer
902 * @skb: buffer to reference
904 * Makes another reference to a socket buffer and returns a pointer
907 static inline struct sk_buff *skb_get(struct sk_buff *skb)
909 atomic_inc(&skb->users);
914 * If users == 1, we are the only owner and are can avoid redundant
919 * skb_cloned - is the buffer a clone
920 * @skb: buffer to check
922 * Returns true if the buffer was generated with skb_clone() and is
923 * one of multiple shared copies of the buffer. Cloned buffers are
924 * shared data so must not be written to under normal circumstances.
926 static inline int skb_cloned(const struct sk_buff *skb)
928 return skb->cloned &&
929 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
932 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
934 might_sleep_if(pri & __GFP_WAIT);
937 return pskb_expand_head(skb, 0, 0, pri);
943 * skb_header_cloned - is the header a clone
944 * @skb: buffer to check
946 * Returns true if modifying the header part of the buffer requires
947 * the data to be copied.
949 static inline int skb_header_cloned(const struct sk_buff *skb)
956 dataref = atomic_read(&skb_shinfo(skb)->dataref);
957 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
962 * skb_header_release - release reference to header
963 * @skb: buffer to operate on
965 * Drop a reference to the header part of the buffer. This is done
966 * by acquiring a payload reference. You must not read from the header
967 * part of skb->data after this.
969 static inline void skb_header_release(struct sk_buff *skb)
973 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
977 * skb_shared - is the buffer shared
978 * @skb: buffer to check
980 * Returns true if more than one person has a reference to this
983 static inline int skb_shared(const struct sk_buff *skb)
985 return atomic_read(&skb->users) != 1;
989 * skb_share_check - check if buffer is shared and if so clone it
990 * @skb: buffer to check
991 * @pri: priority for memory allocation
993 * If the buffer is shared the buffer is cloned and the old copy
994 * drops a reference. A new clone with a single reference is returned.
995 * If the buffer is not shared the original buffer is returned. When
996 * being called from interrupt status or with spinlocks held pri must
999 * NULL is returned on a memory allocation failure.
1001 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1003 might_sleep_if(pri & __GFP_WAIT);
1004 if (skb_shared(skb)) {
1005 struct sk_buff *nskb = skb_clone(skb, pri);
1017 * Copy shared buffers into a new sk_buff. We effectively do COW on
1018 * packets to handle cases where we have a local reader and forward
1019 * and a couple of other messy ones. The normal one is tcpdumping
1020 * a packet thats being forwarded.
1024 * skb_unshare - make a copy of a shared buffer
1025 * @skb: buffer to check
1026 * @pri: priority for memory allocation
1028 * If the socket buffer is a clone then this function creates a new
1029 * copy of the data, drops a reference count on the old copy and returns
1030 * the new copy with the reference count at 1. If the buffer is not a clone
1031 * the original buffer is returned. When called with a spinlock held or
1032 * from interrupt state @pri must be %GFP_ATOMIC
1034 * %NULL is returned on a memory allocation failure.
1036 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1039 might_sleep_if(pri & __GFP_WAIT);
1040 if (skb_cloned(skb)) {
1041 struct sk_buff *nskb = skb_copy(skb, pri);
1042 kfree_skb(skb); /* Free our shared copy */
1049 * skb_peek - peek at the head of an &sk_buff_head
1050 * @list_: list to peek at
1052 * Peek an &sk_buff. Unlike most other operations you _MUST_
1053 * be careful with this one. A peek leaves the buffer on the
1054 * list and someone else may run off with it. You must hold
1055 * the appropriate locks or have a private queue to do this.
1057 * Returns %NULL for an empty list or a pointer to the head element.
1058 * The reference count is not incremented and the reference is therefore
1059 * volatile. Use with caution.
1061 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1063 struct sk_buff *skb = list_->next;
1065 if (skb == (struct sk_buff *)list_)
1071 * skb_peek_next - peek skb following the given one from a queue
1072 * @skb: skb to start from
1073 * @list_: list to peek at
1075 * Returns %NULL when the end of the list is met or a pointer to the
1076 * next element. The reference count is not incremented and the
1077 * reference is therefore volatile. Use with caution.
1079 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1080 const struct sk_buff_head *list_)
1082 struct sk_buff *next = skb->next;
1084 if (next == (struct sk_buff *)list_)
1090 * skb_peek_tail - peek at the tail of an &sk_buff_head
1091 * @list_: list to peek at
1093 * Peek an &sk_buff. Unlike most other operations you _MUST_
1094 * be careful with this one. A peek leaves the buffer on the
1095 * list and someone else may run off with it. You must hold
1096 * the appropriate locks or have a private queue to do this.
1098 * Returns %NULL for an empty list or a pointer to the tail element.
1099 * The reference count is not incremented and the reference is therefore
1100 * volatile. Use with caution.
1102 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1104 struct sk_buff *skb = list_->prev;
1106 if (skb == (struct sk_buff *)list_)
1113 * skb_queue_len - get queue length
1114 * @list_: list to measure
1116 * Return the length of an &sk_buff queue.
1118 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1124 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1125 * @list: queue to initialize
1127 * This initializes only the list and queue length aspects of
1128 * an sk_buff_head object. This allows to initialize the list
1129 * aspects of an sk_buff_head without reinitializing things like
1130 * the spinlock. It can also be used for on-stack sk_buff_head
1131 * objects where the spinlock is known to not be used.
1133 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1135 list->prev = list->next = (struct sk_buff *)list;
1140 * This function creates a split out lock class for each invocation;
1141 * this is needed for now since a whole lot of users of the skb-queue
1142 * infrastructure in drivers have different locking usage (in hardirq)
1143 * than the networking core (in softirq only). In the long run either the
1144 * network layer or drivers should need annotation to consolidate the
1145 * main types of usage into 3 classes.
1147 static inline void skb_queue_head_init(struct sk_buff_head *list)
1149 spin_lock_init(&list->lock);
1150 __skb_queue_head_init(list);
1153 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1154 struct lock_class_key *class)
1156 skb_queue_head_init(list);
1157 lockdep_set_class(&list->lock, class);
1161 * Insert an sk_buff on a list.
1163 * The "__skb_xxxx()" functions are the non-atomic ones that
1164 * can only be called with interrupts disabled.
1166 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1167 struct sk_buff_head *list);
1168 static inline void __skb_insert(struct sk_buff *newsk,
1169 struct sk_buff *prev, struct sk_buff *next,
1170 struct sk_buff_head *list)
1174 next->prev = prev->next = newsk;
1178 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1179 struct sk_buff *prev,
1180 struct sk_buff *next)
1182 struct sk_buff *first = list->next;
1183 struct sk_buff *last = list->prev;
1193 * skb_queue_splice - join two skb lists, this is designed for stacks
1194 * @list: the new list to add
1195 * @head: the place to add it in the first list
1197 static inline void skb_queue_splice(const struct sk_buff_head *list,
1198 struct sk_buff_head *head)
1200 if (!skb_queue_empty(list)) {
1201 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1202 head->qlen += list->qlen;
1207 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1208 * @list: the new list to add
1209 * @head: the place to add it in the first list
1211 * The list at @list is reinitialised
1213 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1214 struct sk_buff_head *head)
1216 if (!skb_queue_empty(list)) {
1217 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1218 head->qlen += list->qlen;
1219 __skb_queue_head_init(list);
1224 * skb_queue_splice_tail - join two skb lists, each list being a queue
1225 * @list: the new list to add
1226 * @head: the place to add it in the first list
1228 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1229 struct sk_buff_head *head)
1231 if (!skb_queue_empty(list)) {
1232 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1233 head->qlen += list->qlen;
1238 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1239 * @list: the new list to add
1240 * @head: the place to add it in the first list
1242 * Each of the lists is a queue.
1243 * The list at @list is reinitialised
1245 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1246 struct sk_buff_head *head)
1248 if (!skb_queue_empty(list)) {
1249 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1250 head->qlen += list->qlen;
1251 __skb_queue_head_init(list);
1256 * __skb_queue_after - queue a buffer at the list head
1257 * @list: list to use
1258 * @prev: place after this buffer
1259 * @newsk: buffer to queue
1261 * Queue a buffer int the middle of a list. This function takes no locks
1262 * and you must therefore hold required locks before calling it.
1264 * A buffer cannot be placed on two lists at the same time.
1266 static inline void __skb_queue_after(struct sk_buff_head *list,
1267 struct sk_buff *prev,
1268 struct sk_buff *newsk)
1270 __skb_insert(newsk, prev, prev->next, list);
1273 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1274 struct sk_buff_head *list);
1276 static inline void __skb_queue_before(struct sk_buff_head *list,
1277 struct sk_buff *next,
1278 struct sk_buff *newsk)
1280 __skb_insert(newsk, next->prev, next, list);
1284 * __skb_queue_head - queue a buffer at the list head
1285 * @list: list to use
1286 * @newsk: buffer to queue
1288 * Queue a buffer at the start of a list. This function takes no locks
1289 * and you must therefore hold required locks before calling it.
1291 * A buffer cannot be placed on two lists at the same time.
1293 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1294 static inline void __skb_queue_head(struct sk_buff_head *list,
1295 struct sk_buff *newsk)
1297 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1301 * __skb_queue_tail - queue a buffer at the list tail
1302 * @list: list to use
1303 * @newsk: buffer to queue
1305 * Queue a buffer at the end of a list. This function takes no locks
1306 * and you must therefore hold required locks before calling it.
1308 * A buffer cannot be placed on two lists at the same time.
1310 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1311 static inline void __skb_queue_tail(struct sk_buff_head *list,
1312 struct sk_buff *newsk)
1314 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1318 * remove sk_buff from list. _Must_ be called atomically, and with
1321 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1322 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1324 struct sk_buff *next, *prev;
1329 skb->next = skb->prev = NULL;
1335 * __skb_dequeue - remove from the head of the queue
1336 * @list: list to dequeue from
1338 * Remove the head of the list. This function does not take any locks
1339 * so must be used with appropriate locks held only. The head item is
1340 * returned or %NULL if the list is empty.
1342 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1343 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1345 struct sk_buff *skb = skb_peek(list);
1347 __skb_unlink(skb, list);
1352 * __skb_dequeue_tail - remove from the tail of the queue
1353 * @list: list to dequeue from
1355 * Remove the tail of the list. This function does not take any locks
1356 * so must be used with appropriate locks held only. The tail item is
1357 * returned or %NULL if the list is empty.
1359 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1360 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1362 struct sk_buff *skb = skb_peek_tail(list);
1364 __skb_unlink(skb, list);
1369 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1371 return skb->data_len;
1374 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1376 return skb->len - skb->data_len;
1379 static inline int skb_pagelen(const struct sk_buff *skb)
1383 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1384 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1385 return len + skb_headlen(skb);
1389 * __skb_fill_page_desc - initialise a paged fragment in an skb
1390 * @skb: buffer containing fragment to be initialised
1391 * @i: paged fragment index to initialise
1392 * @page: the page to use for this fragment
1393 * @off: the offset to the data with @page
1394 * @size: the length of the data
1396 * Initialises the @i'th fragment of @skb to point to &size bytes at
1397 * offset @off within @page.
1399 * Does not take any additional reference on the fragment.
1401 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1402 struct page *page, int off, int size)
1404 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1407 * Propagate page->pfmemalloc to the skb if we can. The problem is
1408 * that not all callers have unique ownership of the page. If
1409 * pfmemalloc is set, we check the mapping as a mapping implies
1410 * page->index is set (index and pfmemalloc share space).
1411 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1412 * do not lose pfmemalloc information as the pages would not be
1413 * allocated using __GFP_MEMALLOC.
1415 frag->page.p = page;
1416 frag->page_offset = off;
1417 skb_frag_size_set(frag, size);
1419 page = compound_head(page);
1420 if (page->pfmemalloc && !page->mapping)
1421 skb->pfmemalloc = true;
1425 * skb_fill_page_desc - initialise a paged fragment in an skb
1426 * @skb: buffer containing fragment to be initialised
1427 * @i: paged fragment index to initialise
1428 * @page: the page to use for this fragment
1429 * @off: the offset to the data with @page
1430 * @size: the length of the data
1432 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1433 * @skb to point to @size bytes at offset @off within @page. In
1434 * addition updates @skb such that @i is the last fragment.
1436 * Does not take any additional reference on the fragment.
1438 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1439 struct page *page, int off, int size)
1441 __skb_fill_page_desc(skb, i, page, off, size);
1442 skb_shinfo(skb)->nr_frags = i + 1;
1445 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1446 int size, unsigned int truesize);
1448 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1449 unsigned int truesize);
1451 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1452 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1453 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1455 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1456 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1458 return skb->head + skb->tail;
1461 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1463 skb->tail = skb->data - skb->head;
1466 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1468 skb_reset_tail_pointer(skb);
1469 skb->tail += offset;
1472 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1473 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1478 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1480 skb->tail = skb->data;
1483 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1485 skb->tail = skb->data + offset;
1488 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1491 * Add data to an sk_buff
1493 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1494 unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1495 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1497 unsigned char *tmp = skb_tail_pointer(skb);
1498 SKB_LINEAR_ASSERT(skb);
1504 unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1505 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1512 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1513 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1516 BUG_ON(skb->len < skb->data_len);
1517 return skb->data += len;
1520 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1522 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1525 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1527 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1529 if (len > skb_headlen(skb) &&
1530 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1533 return skb->data += len;
1536 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1538 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1541 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1543 if (likely(len <= skb_headlen(skb)))
1545 if (unlikely(len > skb->len))
1547 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1551 * skb_headroom - bytes at buffer head
1552 * @skb: buffer to check
1554 * Return the number of bytes of free space at the head of an &sk_buff.
1556 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1558 return skb->data - skb->head;
1562 * skb_tailroom - bytes at buffer end
1563 * @skb: buffer to check
1565 * Return the number of bytes of free space at the tail of an sk_buff
1567 static inline int skb_tailroom(const struct sk_buff *skb)
1569 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1573 * skb_availroom - bytes at buffer end
1574 * @skb: buffer to check
1576 * Return the number of bytes of free space at the tail of an sk_buff
1577 * allocated by sk_stream_alloc()
1579 static inline int skb_availroom(const struct sk_buff *skb)
1581 if (skb_is_nonlinear(skb))
1584 return skb->end - skb->tail - skb->reserved_tailroom;
1588 * skb_reserve - adjust headroom
1589 * @skb: buffer to alter
1590 * @len: bytes to move
1592 * Increase the headroom of an empty &sk_buff by reducing the tail
1593 * room. This is only allowed for an empty buffer.
1595 static inline void skb_reserve(struct sk_buff *skb, int len)
1601 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1603 skb->inner_mac_header = skb->mac_header;
1604 skb->inner_network_header = skb->network_header;
1605 skb->inner_transport_header = skb->transport_header;
1608 static inline void skb_reset_mac_len(struct sk_buff *skb)
1610 skb->mac_len = skb->network_header - skb->mac_header;
1613 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1616 return skb->head + skb->inner_transport_header;
1619 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1621 skb->inner_transport_header = skb->data - skb->head;
1624 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1627 skb_reset_inner_transport_header(skb);
1628 skb->inner_transport_header += offset;
1631 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1633 return skb->head + skb->inner_network_header;
1636 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1638 skb->inner_network_header = skb->data - skb->head;
1641 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1644 skb_reset_inner_network_header(skb);
1645 skb->inner_network_header += offset;
1648 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1650 return skb->head + skb->inner_mac_header;
1653 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1655 skb->inner_mac_header = skb->data - skb->head;
1658 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1661 skb_reset_inner_mac_header(skb);
1662 skb->inner_mac_header += offset;
1664 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1666 return skb->transport_header != (typeof(skb->transport_header))~0U;
1669 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1671 return skb->head + skb->transport_header;
1674 static inline void skb_reset_transport_header(struct sk_buff *skb)
1676 skb->transport_header = skb->data - skb->head;
1679 static inline void skb_set_transport_header(struct sk_buff *skb,
1682 skb_reset_transport_header(skb);
1683 skb->transport_header += offset;
1686 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1688 return skb->head + skb->network_header;
1691 static inline void skb_reset_network_header(struct sk_buff *skb)
1693 skb->network_header = skb->data - skb->head;
1696 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1698 skb_reset_network_header(skb);
1699 skb->network_header += offset;
1702 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1704 return skb->head + skb->mac_header;
1707 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1709 return skb->mac_header != (typeof(skb->mac_header))~0U;
1712 static inline void skb_reset_mac_header(struct sk_buff *skb)
1714 skb->mac_header = skb->data - skb->head;
1717 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1719 skb_reset_mac_header(skb);
1720 skb->mac_header += offset;
1723 static inline void skb_pop_mac_header(struct sk_buff *skb)
1725 skb->mac_header = skb->network_header;
1728 static inline void skb_probe_transport_header(struct sk_buff *skb,
1729 const int offset_hint)
1731 struct flow_keys keys;
1733 if (skb_transport_header_was_set(skb))
1735 else if (skb_flow_dissect(skb, &keys))
1736 skb_set_transport_header(skb, keys.thoff);
1738 skb_set_transport_header(skb, offset_hint);
1741 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1743 if (skb_mac_header_was_set(skb)) {
1744 const unsigned char *old_mac = skb_mac_header(skb);
1746 skb_set_mac_header(skb, -skb->mac_len);
1747 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1751 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1753 return skb->csum_start - skb_headroom(skb);
1756 static inline int skb_transport_offset(const struct sk_buff *skb)
1758 return skb_transport_header(skb) - skb->data;
1761 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1763 return skb->transport_header - skb->network_header;
1766 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1768 return skb->inner_transport_header - skb->inner_network_header;
1771 static inline int skb_network_offset(const struct sk_buff *skb)
1773 return skb_network_header(skb) - skb->data;
1776 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1778 return skb_inner_network_header(skb) - skb->data;
1781 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1783 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1787 * CPUs often take a performance hit when accessing unaligned memory
1788 * locations. The actual performance hit varies, it can be small if the
1789 * hardware handles it or large if we have to take an exception and fix it
1792 * Since an ethernet header is 14 bytes network drivers often end up with
1793 * the IP header at an unaligned offset. The IP header can be aligned by
1794 * shifting the start of the packet by 2 bytes. Drivers should do this
1797 * skb_reserve(skb, NET_IP_ALIGN);
1799 * The downside to this alignment of the IP header is that the DMA is now
1800 * unaligned. On some architectures the cost of an unaligned DMA is high
1801 * and this cost outweighs the gains made by aligning the IP header.
1803 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1806 #ifndef NET_IP_ALIGN
1807 #define NET_IP_ALIGN 2
1811 * The networking layer reserves some headroom in skb data (via
1812 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1813 * the header has to grow. In the default case, if the header has to grow
1814 * 32 bytes or less we avoid the reallocation.
1816 * Unfortunately this headroom changes the DMA alignment of the resulting
1817 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1818 * on some architectures. An architecture can override this value,
1819 * perhaps setting it to a cacheline in size (since that will maintain
1820 * cacheline alignment of the DMA). It must be a power of 2.
1822 * Various parts of the networking layer expect at least 32 bytes of
1823 * headroom, you should not reduce this.
1825 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1826 * to reduce average number of cache lines per packet.
1827 * get_rps_cpus() for example only access one 64 bytes aligned block :
1828 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1831 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1834 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1836 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1838 if (unlikely(skb_is_nonlinear(skb))) {
1843 skb_set_tail_pointer(skb, len);
1846 void skb_trim(struct sk_buff *skb, unsigned int len);
1848 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1851 return ___pskb_trim(skb, len);
1852 __skb_trim(skb, len);
1856 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1858 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1862 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1863 * @skb: buffer to alter
1866 * This is identical to pskb_trim except that the caller knows that
1867 * the skb is not cloned so we should never get an error due to out-
1870 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1872 int err = pskb_trim(skb, len);
1877 * skb_orphan - orphan a buffer
1878 * @skb: buffer to orphan
1880 * If a buffer currently has an owner then we call the owner's
1881 * destructor function and make the @skb unowned. The buffer continues
1882 * to exist but is no longer charged to its former owner.
1884 static inline void skb_orphan(struct sk_buff *skb)
1886 if (skb->destructor) {
1887 skb->destructor(skb);
1888 skb->destructor = NULL;
1896 * skb_orphan_frags - orphan the frags contained in a buffer
1897 * @skb: buffer to orphan frags from
1898 * @gfp_mask: allocation mask for replacement pages
1900 * For each frag in the SKB which needs a destructor (i.e. has an
1901 * owner) create a copy of that frag and release the original
1902 * page by calling the destructor.
1904 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1906 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1908 return skb_copy_ubufs(skb, gfp_mask);
1912 * __skb_queue_purge - empty a list
1913 * @list: list to empty
1915 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1916 * the list and one reference dropped. This function does not take the
1917 * list lock and the caller must hold the relevant locks to use it.
1919 void skb_queue_purge(struct sk_buff_head *list);
1920 static inline void __skb_queue_purge(struct sk_buff_head *list)
1922 struct sk_buff *skb;
1923 while ((skb = __skb_dequeue(list)) != NULL)
1927 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1928 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1929 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1931 void *netdev_alloc_frag(unsigned int fragsz);
1933 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1937 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1938 * @dev: network device to receive on
1939 * @length: length to allocate
1941 * Allocate a new &sk_buff and assign it a usage count of one. The
1942 * buffer has unspecified headroom built in. Users should allocate
1943 * the headroom they think they need without accounting for the
1944 * built in space. The built in space is used for optimisations.
1946 * %NULL is returned if there is no free memory. Although this function
1947 * allocates memory it can be called from an interrupt.
1949 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1950 unsigned int length)
1952 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1955 /* legacy helper around __netdev_alloc_skb() */
1956 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1959 return __netdev_alloc_skb(NULL, length, gfp_mask);
1962 /* legacy helper around netdev_alloc_skb() */
1963 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1965 return netdev_alloc_skb(NULL, length);
1969 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1970 unsigned int length, gfp_t gfp)
1972 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
1974 if (NET_IP_ALIGN && skb)
1975 skb_reserve(skb, NET_IP_ALIGN);
1979 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1980 unsigned int length)
1982 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1986 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1987 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1988 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1989 * @order: size of the allocation
1991 * Allocate a new page.
1993 * %NULL is returned if there is no free memory.
1995 static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1996 struct sk_buff *skb,
2001 gfp_mask |= __GFP_COLD;
2003 if (!(gfp_mask & __GFP_NOMEMALLOC))
2004 gfp_mask |= __GFP_MEMALLOC;
2006 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2007 if (skb && page && page->pfmemalloc)
2008 skb->pfmemalloc = true;
2014 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2015 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2016 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2018 * Allocate a new page.
2020 * %NULL is returned if there is no free memory.
2022 static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2023 struct sk_buff *skb)
2025 return __skb_alloc_pages(gfp_mask, skb, 0);
2029 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2030 * @page: The page that was allocated from skb_alloc_page
2031 * @skb: The skb that may need pfmemalloc set
2033 static inline void skb_propagate_pfmemalloc(struct page *page,
2034 struct sk_buff *skb)
2036 if (page && page->pfmemalloc)
2037 skb->pfmemalloc = true;
2041 * skb_frag_page - retrieve the page refered to by a paged fragment
2042 * @frag: the paged fragment
2044 * Returns the &struct page associated with @frag.
2046 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2048 return frag->page.p;
2052 * __skb_frag_ref - take an addition reference on a paged fragment.
2053 * @frag: the paged fragment
2055 * Takes an additional reference on the paged fragment @frag.
2057 static inline void __skb_frag_ref(skb_frag_t *frag)
2059 get_page(skb_frag_page(frag));
2063 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2065 * @f: the fragment offset.
2067 * Takes an additional reference on the @f'th paged fragment of @skb.
2069 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2071 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2075 * __skb_frag_unref - release a reference on a paged fragment.
2076 * @frag: the paged fragment
2078 * Releases a reference on the paged fragment @frag.
2080 static inline void __skb_frag_unref(skb_frag_t *frag)
2082 put_page(skb_frag_page(frag));
2086 * skb_frag_unref - release a reference on a paged fragment of an skb.
2088 * @f: the fragment offset
2090 * Releases a reference on the @f'th paged fragment of @skb.
2092 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2094 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2098 * skb_frag_address - gets the address of the data contained in a paged fragment
2099 * @frag: the paged fragment buffer
2101 * Returns the address of the data within @frag. The page must already
2104 static inline void *skb_frag_address(const skb_frag_t *frag)
2106 return page_address(skb_frag_page(frag)) + frag->page_offset;
2110 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2111 * @frag: the paged fragment buffer
2113 * Returns the address of the data within @frag. Checks that the page
2114 * is mapped and returns %NULL otherwise.
2116 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2118 void *ptr = page_address(skb_frag_page(frag));
2122 return ptr + frag->page_offset;
2126 * __skb_frag_set_page - sets the page contained in a paged fragment
2127 * @frag: the paged fragment
2128 * @page: the page to set
2130 * Sets the fragment @frag to contain @page.
2132 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2134 frag->page.p = page;
2138 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2140 * @f: the fragment offset
2141 * @page: the page to set
2143 * Sets the @f'th fragment of @skb to contain @page.
2145 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2148 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2151 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2154 * skb_frag_dma_map - maps a paged fragment via the DMA API
2155 * @dev: the device to map the fragment to
2156 * @frag: the paged fragment to map
2157 * @offset: the offset within the fragment (starting at the
2158 * fragment's own offset)
2159 * @size: the number of bytes to map
2160 * @dir: the direction of the mapping (%PCI_DMA_*)
2162 * Maps the page associated with @frag to @device.
2164 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2165 const skb_frag_t *frag,
2166 size_t offset, size_t size,
2167 enum dma_data_direction dir)
2169 return dma_map_page(dev, skb_frag_page(frag),
2170 frag->page_offset + offset, size, dir);
2173 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2176 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2180 * skb_clone_writable - is the header of a clone writable
2181 * @skb: buffer to check
2182 * @len: length up to which to write
2184 * Returns true if modifying the header part of the cloned buffer
2185 * does not requires the data to be copied.
2187 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2189 return !skb_header_cloned(skb) &&
2190 skb_headroom(skb) + len <= skb->hdr_len;
2193 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2198 if (headroom > skb_headroom(skb))
2199 delta = headroom - skb_headroom(skb);
2201 if (delta || cloned)
2202 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2208 * skb_cow - copy header of skb when it is required
2209 * @skb: buffer to cow
2210 * @headroom: needed headroom
2212 * If the skb passed lacks sufficient headroom or its data part
2213 * is shared, data is reallocated. If reallocation fails, an error
2214 * is returned and original skb is not changed.
2216 * The result is skb with writable area skb->head...skb->tail
2217 * and at least @headroom of space at head.
2219 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2221 return __skb_cow(skb, headroom, skb_cloned(skb));
2225 * skb_cow_head - skb_cow but only making the head writable
2226 * @skb: buffer to cow
2227 * @headroom: needed headroom
2229 * This function is identical to skb_cow except that we replace the
2230 * skb_cloned check by skb_header_cloned. It should be used when
2231 * you only need to push on some header and do not need to modify
2234 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2236 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2240 * skb_padto - pad an skbuff up to a minimal size
2241 * @skb: buffer to pad
2242 * @len: minimal length
2244 * Pads up a buffer to ensure the trailing bytes exist and are
2245 * blanked. If the buffer already contains sufficient data it
2246 * is untouched. Otherwise it is extended. Returns zero on
2247 * success. The skb is freed on error.
2250 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2252 unsigned int size = skb->len;
2253 if (likely(size >= len))
2255 return skb_pad(skb, len - size);
2258 static inline int skb_add_data(struct sk_buff *skb,
2259 char __user *from, int copy)
2261 const int off = skb->len;
2263 if (skb->ip_summed == CHECKSUM_NONE) {
2265 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
2268 skb->csum = csum_block_add(skb->csum, csum, off);
2271 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2274 __skb_trim(skb, off);
2278 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2279 const struct page *page, int off)
2282 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2284 return page == skb_frag_page(frag) &&
2285 off == frag->page_offset + skb_frag_size(frag);
2290 static inline int __skb_linearize(struct sk_buff *skb)
2292 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2296 * skb_linearize - convert paged skb to linear one
2297 * @skb: buffer to linarize
2299 * If there is no free memory -ENOMEM is returned, otherwise zero
2300 * is returned and the old skb data released.
2302 static inline int skb_linearize(struct sk_buff *skb)
2304 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2308 * skb_has_shared_frag - can any frag be overwritten
2309 * @skb: buffer to test
2311 * Return true if the skb has at least one frag that might be modified
2312 * by an external entity (as in vmsplice()/sendfile())
2314 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2316 return skb_is_nonlinear(skb) &&
2317 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2321 * skb_linearize_cow - make sure skb is linear and writable
2322 * @skb: buffer to process
2324 * If there is no free memory -ENOMEM is returned, otherwise zero
2325 * is returned and the old skb data released.
2327 static inline int skb_linearize_cow(struct sk_buff *skb)
2329 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2330 __skb_linearize(skb) : 0;
2334 * skb_postpull_rcsum - update checksum for received skb after pull
2335 * @skb: buffer to update
2336 * @start: start of data before pull
2337 * @len: length of data pulled
2339 * After doing a pull on a received packet, you need to call this to
2340 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2341 * CHECKSUM_NONE so that it can be recomputed from scratch.
2344 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2345 const void *start, unsigned int len)
2347 if (skb->ip_summed == CHECKSUM_COMPLETE)
2348 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2351 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2354 * pskb_trim_rcsum - trim received skb and update checksum
2355 * @skb: buffer to trim
2358 * This is exactly the same as pskb_trim except that it ensures the
2359 * checksum of received packets are still valid after the operation.
2362 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2364 if (likely(len >= skb->len))
2366 if (skb->ip_summed == CHECKSUM_COMPLETE)
2367 skb->ip_summed = CHECKSUM_NONE;
2368 return __pskb_trim(skb, len);
2371 #define skb_queue_walk(queue, skb) \
2372 for (skb = (queue)->next; \
2373 skb != (struct sk_buff *)(queue); \
2376 #define skb_queue_walk_safe(queue, skb, tmp) \
2377 for (skb = (queue)->next, tmp = skb->next; \
2378 skb != (struct sk_buff *)(queue); \
2379 skb = tmp, tmp = skb->next)
2381 #define skb_queue_walk_from(queue, skb) \
2382 for (; skb != (struct sk_buff *)(queue); \
2385 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2386 for (tmp = skb->next; \
2387 skb != (struct sk_buff *)(queue); \
2388 skb = tmp, tmp = skb->next)
2390 #define skb_queue_reverse_walk(queue, skb) \
2391 for (skb = (queue)->prev; \
2392 skb != (struct sk_buff *)(queue); \
2395 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2396 for (skb = (queue)->prev, tmp = skb->prev; \
2397 skb != (struct sk_buff *)(queue); \
2398 skb = tmp, tmp = skb->prev)
2400 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2401 for (tmp = skb->prev; \
2402 skb != (struct sk_buff *)(queue); \
2403 skb = tmp, tmp = skb->prev)
2405 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2407 return skb_shinfo(skb)->frag_list != NULL;
2410 static inline void skb_frag_list_init(struct sk_buff *skb)
2412 skb_shinfo(skb)->frag_list = NULL;
2415 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2417 frag->next = skb_shinfo(skb)->frag_list;
2418 skb_shinfo(skb)->frag_list = frag;
2421 #define skb_walk_frags(skb, iter) \
2422 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2424 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2425 int *peeked, int *off, int *err);
2426 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2428 unsigned int datagram_poll(struct file *file, struct socket *sock,
2429 struct poll_table_struct *wait);
2430 int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2431 struct iovec *to, int size);
2432 int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2434 int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2435 const struct iovec *from, int from_offset,
2437 int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2438 int offset, size_t count);
2439 int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2440 const struct iovec *to, int to_offset,
2442 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2443 void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2444 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
2445 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2446 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2447 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2448 int len, __wsum csum);
2449 int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2450 struct pipe_inode_info *pipe, unsigned int len,
2451 unsigned int flags);
2452 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2453 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2454 void skb_zerocopy(struct sk_buff *to, const struct sk_buff *from,
2456 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2457 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2458 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
2459 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
2460 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
2462 struct skb_checksum_ops {
2463 __wsum (*update)(const void *mem, int len, __wsum wsum);
2464 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2467 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2468 __wsum csum, const struct skb_checksum_ops *ops);
2469 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2472 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2473 int len, void *buffer)
2475 int hlen = skb_headlen(skb);
2477 if (hlen - offset >= len)
2478 return skb->data + offset;
2480 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2487 * skb_needs_linearize - check if we need to linearize a given skb
2488 * depending on the given device features.
2489 * @skb: socket buffer to check
2490 * @features: net device features
2492 * Returns true if either:
2493 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2494 * 2. skb is fragmented and the device does not support SG.
2496 static inline bool skb_needs_linearize(struct sk_buff *skb,
2497 netdev_features_t features)
2499 return skb_is_nonlinear(skb) &&
2500 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2501 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2504 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2506 const unsigned int len)
2508 memcpy(to, skb->data, len);
2511 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2512 const int offset, void *to,
2513 const unsigned int len)
2515 memcpy(to, skb->data + offset, len);
2518 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2520 const unsigned int len)
2522 memcpy(skb->data, from, len);
2525 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2528 const unsigned int len)
2530 memcpy(skb->data + offset, from, len);
2533 void skb_init(void);
2535 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2541 * skb_get_timestamp - get timestamp from a skb
2542 * @skb: skb to get stamp from
2543 * @stamp: pointer to struct timeval to store stamp in
2545 * Timestamps are stored in the skb as offsets to a base timestamp.
2546 * This function converts the offset back to a struct timeval and stores
2549 static inline void skb_get_timestamp(const struct sk_buff *skb,
2550 struct timeval *stamp)
2552 *stamp = ktime_to_timeval(skb->tstamp);
2555 static inline void skb_get_timestampns(const struct sk_buff *skb,
2556 struct timespec *stamp)
2558 *stamp = ktime_to_timespec(skb->tstamp);
2561 static inline void __net_timestamp(struct sk_buff *skb)
2563 skb->tstamp = ktime_get_real();
2566 static inline ktime_t net_timedelta(ktime_t t)
2568 return ktime_sub(ktime_get_real(), t);
2571 static inline ktime_t net_invalid_timestamp(void)
2573 return ktime_set(0, 0);
2576 void skb_timestamping_init(void);
2578 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2580 void skb_clone_tx_timestamp(struct sk_buff *skb);
2581 bool skb_defer_rx_timestamp(struct sk_buff *skb);
2583 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2585 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2589 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2594 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2597 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2599 * PHY drivers may accept clones of transmitted packets for
2600 * timestamping via their phy_driver.txtstamp method. These drivers
2601 * must call this function to return the skb back to the stack, with
2602 * or without a timestamp.
2604 * @skb: clone of the the original outgoing packet
2605 * @hwtstamps: hardware time stamps, may be NULL if not available
2608 void skb_complete_tx_timestamp(struct sk_buff *skb,
2609 struct skb_shared_hwtstamps *hwtstamps);
2612 * skb_tstamp_tx - queue clone of skb with send time stamps
2613 * @orig_skb: the original outgoing packet
2614 * @hwtstamps: hardware time stamps, may be NULL if not available
2616 * If the skb has a socket associated, then this function clones the
2617 * skb (thus sharing the actual data and optional structures), stores
2618 * the optional hardware time stamping information (if non NULL) or
2619 * generates a software time stamp (otherwise), then queues the clone
2620 * to the error queue of the socket. Errors are silently ignored.
2622 void skb_tstamp_tx(struct sk_buff *orig_skb,
2623 struct skb_shared_hwtstamps *hwtstamps);
2625 static inline void sw_tx_timestamp(struct sk_buff *skb)
2627 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2628 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2629 skb_tstamp_tx(skb, NULL);
2633 * skb_tx_timestamp() - Driver hook for transmit timestamping
2635 * Ethernet MAC Drivers should call this function in their hard_xmit()
2636 * function immediately before giving the sk_buff to the MAC hardware.
2638 * Specifically, one should make absolutely sure that this function is
2639 * called before TX completion of this packet can trigger. Otherwise
2640 * the packet could potentially already be freed.
2642 * @skb: A socket buffer.
2644 static inline void skb_tx_timestamp(struct sk_buff *skb)
2646 skb_clone_tx_timestamp(skb);
2647 sw_tx_timestamp(skb);
2651 * skb_complete_wifi_ack - deliver skb with wifi status
2653 * @skb: the original outgoing packet
2654 * @acked: ack status
2657 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2659 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2660 __sum16 __skb_checksum_complete(struct sk_buff *skb);
2662 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2664 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2668 * skb_checksum_complete - Calculate checksum of an entire packet
2669 * @skb: packet to process
2671 * This function calculates the checksum over the entire packet plus
2672 * the value of skb->csum. The latter can be used to supply the
2673 * checksum of a pseudo header as used by TCP/UDP. It returns the
2676 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2677 * this function can be used to verify that checksum on received
2678 * packets. In that case the function should return zero if the
2679 * checksum is correct. In particular, this function will return zero
2680 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2681 * hardware has already verified the correctness of the checksum.
2683 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2685 return skb_csum_unnecessary(skb) ?
2686 0 : __skb_checksum_complete(skb);
2689 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2690 void nf_conntrack_destroy(struct nf_conntrack *nfct);
2691 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2693 if (nfct && atomic_dec_and_test(&nfct->use))
2694 nf_conntrack_destroy(nfct);
2696 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2699 atomic_inc(&nfct->use);
2702 #ifdef CONFIG_BRIDGE_NETFILTER
2703 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2705 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2708 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2711 atomic_inc(&nf_bridge->use);
2713 #endif /* CONFIG_BRIDGE_NETFILTER */
2714 static inline void nf_reset(struct sk_buff *skb)
2716 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2717 nf_conntrack_put(skb->nfct);
2720 #ifdef CONFIG_BRIDGE_NETFILTER
2721 nf_bridge_put(skb->nf_bridge);
2722 skb->nf_bridge = NULL;
2726 static inline void nf_reset_trace(struct sk_buff *skb)
2728 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2733 /* Note: This doesn't put any conntrack and bridge info in dst. */
2734 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2736 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2737 dst->nfct = src->nfct;
2738 nf_conntrack_get(src->nfct);
2739 dst->nfctinfo = src->nfctinfo;
2741 #ifdef CONFIG_BRIDGE_NETFILTER
2742 dst->nf_bridge = src->nf_bridge;
2743 nf_bridge_get(src->nf_bridge);
2745 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2746 dst->nf_trace = src->nf_trace;
2750 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2752 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2753 nf_conntrack_put(dst->nfct);
2755 #ifdef CONFIG_BRIDGE_NETFILTER
2756 nf_bridge_put(dst->nf_bridge);
2758 __nf_copy(dst, src);
2761 #ifdef CONFIG_NETWORK_SECMARK
2762 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2764 to->secmark = from->secmark;
2767 static inline void skb_init_secmark(struct sk_buff *skb)
2772 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2775 static inline void skb_init_secmark(struct sk_buff *skb)
2779 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2781 skb->queue_mapping = queue_mapping;
2784 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2786 return skb->queue_mapping;
2789 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2791 to->queue_mapping = from->queue_mapping;
2794 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2796 skb->queue_mapping = rx_queue + 1;
2799 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2801 return skb->queue_mapping - 1;
2804 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2806 return skb->queue_mapping != 0;
2809 u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2810 unsigned int num_tx_queues);
2812 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2821 /* Keeps track of mac header offset relative to skb->head.
2822 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2823 * For non-tunnel skb it points to skb_mac_header() and for
2824 * tunnel skb it points to outer mac header.
2825 * Keeps track of level of encapsulation of network headers.
2831 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2833 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2835 return (skb_mac_header(inner_skb) - inner_skb->head) -
2836 SKB_GSO_CB(inner_skb)->mac_offset;
2839 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
2841 int new_headroom, headroom;
2844 headroom = skb_headroom(skb);
2845 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
2849 new_headroom = skb_headroom(skb);
2850 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
2854 static inline bool skb_is_gso(const struct sk_buff *skb)
2856 return skb_shinfo(skb)->gso_size;
2859 /* Note: Should be called only if skb_is_gso(skb) is true */
2860 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
2862 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2865 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2867 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2869 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2870 * wanted then gso_type will be set. */
2871 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2873 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2874 unlikely(shinfo->gso_type == 0)) {
2875 __skb_warn_lro_forwarding(skb);
2881 static inline void skb_forward_csum(struct sk_buff *skb)
2883 /* Unfortunately we don't support this one. Any brave souls? */
2884 if (skb->ip_summed == CHECKSUM_COMPLETE)
2885 skb->ip_summed = CHECKSUM_NONE;
2889 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2890 * @skb: skb to check
2892 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2893 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2894 * use this helper, to document places where we make this assertion.
2896 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
2899 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2903 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2905 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
2907 u32 __skb_get_poff(const struct sk_buff *skb);
2910 * skb_head_is_locked - Determine if the skb->head is locked down
2911 * @skb: skb to check
2913 * The head on skbs build around a head frag can be removed if they are
2914 * not cloned. This function returns true if the skb head is locked down
2915 * due to either being allocated via kmalloc, or by being a clone with
2916 * multiple references to the head.
2918 static inline bool skb_head_is_locked(const struct sk_buff *skb)
2920 return !skb->head_frag || skb_cloned(skb);
2924 * skb_gso_network_seglen - Return length of individual segments of a gso packet
2928 * skb_gso_network_seglen is used to determine the real size of the
2929 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
2931 * The MAC/L2 header is not accounted for.
2933 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
2935 unsigned int hdr_len = skb_transport_header(skb) -
2936 skb_network_header(skb);
2937 return hdr_len + skb_gso_transport_seglen(skb);
2939 #endif /* __KERNEL__ */
2940 #endif /* _LINUX_SKBUFF_H */