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>
23 #include <linux/rbtree.h>
24 #include <linux/socket.h>
26 #include <linux/atomic.h>
27 #include <asm/types.h>
28 #include <linux/spinlock.h>
29 #include <linux/net.h>
30 #include <linux/textsearch.h>
31 #include <net/checksum.h>
32 #include <linux/rcupdate.h>
33 #include <linux/hrtimer.h>
34 #include <linux/dma-mapping.h>
35 #include <linux/netdev_features.h>
36 #include <linux/sched.h>
37 #include <net/flow_dissector.h>
39 /* A. Checksumming of received packets by device.
43 * Device failed to checksum this packet e.g. due to lack of capabilities.
44 * The packet contains full (though not verified) checksum in packet but
45 * not in skb->csum. Thus, skb->csum is undefined in this case.
47 * CHECKSUM_UNNECESSARY:
49 * The hardware you're dealing with doesn't calculate the full checksum
50 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
51 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
52 * if their checksums are okay. skb->csum is still undefined in this case
53 * though. It is a bad option, but, unfortunately, nowadays most vendors do
54 * this. Apparently with the secret goal to sell you new devices, when you
55 * will add new protocol to your host, f.e. IPv6 8)
57 * CHECKSUM_UNNECESSARY is applicable to following protocols:
59 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
60 * zero UDP checksum for either IPv4 or IPv6, the networking stack
61 * may perform further validation in this case.
62 * GRE: only if the checksum is present in the header.
63 * SCTP: indicates the CRC in SCTP header has been validated.
65 * skb->csum_level indicates the number of consecutive checksums found in
66 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
67 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
68 * and a device is able to verify the checksums for UDP (possibly zero),
69 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
70 * two. If the device were only able to verify the UDP checksum and not
71 * GRE, either because it doesn't support GRE checksum of because GRE
72 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
73 * not considered in this case).
77 * This is the most generic way. The device supplied checksum of the _whole_
78 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
79 * hardware doesn't need to parse L3/L4 headers to implement this.
81 * Note: Even if device supports only some protocols, but is able to produce
82 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
86 * A checksum is set up to be offloaded to a device as described in the
87 * output description for CHECKSUM_PARTIAL. This may occur on a packet
88 * received directly from another Linux OS, e.g., a virtualized Linux kernel
89 * on the same host, or it may be set in the input path in GRO or remote
90 * checksum offload. For the purposes of checksum verification, the checksum
91 * referred to by skb->csum_start + skb->csum_offset and any preceding
92 * checksums in the packet are considered verified. Any checksums in the
93 * packet that are after the checksum being offloaded are not considered to
96 * B. Checksumming on output.
100 * The skb was already checksummed by the protocol, or a checksum is not
105 * The device is required to checksum the packet as seen by hard_start_xmit()
106 * from skb->csum_start up to the end, and to record/write the checksum at
107 * offset skb->csum_start + skb->csum_offset.
109 * The device must show its capabilities in dev->features, set up at device
110 * setup time, e.g. netdev_features.h:
112 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
113 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
114 * IPv4. Sigh. Vendors like this way for an unknown reason.
115 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
116 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
117 * NETIF_F_... - Well, you get the picture.
119 * CHECKSUM_UNNECESSARY:
121 * Normally, the device will do per protocol specific checksumming. Protocol
122 * implementations that do not want the NIC to perform the checksum
123 * calculation should use this flag in their outgoing skbs.
125 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
126 * offload. Correspondingly, the FCoE protocol driver
127 * stack should use CHECKSUM_UNNECESSARY.
129 * Any questions? No questions, good. --ANK
132 /* Don't change this without changing skb_csum_unnecessary! */
133 #define CHECKSUM_NONE 0
134 #define CHECKSUM_UNNECESSARY 1
135 #define CHECKSUM_COMPLETE 2
136 #define CHECKSUM_PARTIAL 3
138 /* Maximum value in skb->csum_level */
139 #define SKB_MAX_CSUM_LEVEL 3
141 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
142 #define SKB_WITH_OVERHEAD(X) \
143 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
144 #define SKB_MAX_ORDER(X, ORDER) \
145 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
146 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
147 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
149 /* return minimum truesize of one skb containing X bytes of data */
150 #define SKB_TRUESIZE(X) ((X) + \
151 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
152 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
156 struct pipe_inode_info;
160 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
161 struct nf_conntrack {
166 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
167 struct nf_bridge_info {
170 BRNF_PROTO_UNCHANGED,
176 struct net_device *physindev;
178 struct net_device *physoutdev;
179 char neigh_header[8];
185 struct sk_buff_head {
186 /* These two members must be first. */
187 struct sk_buff *next;
188 struct sk_buff *prev;
196 /* To allow 64K frame to be packed as single skb without frag_list we
197 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
198 * buffers which do not start on a page boundary.
200 * Since GRO uses frags we allocate at least 16 regardless of page
203 #if (65536/PAGE_SIZE + 1) < 16
204 #define MAX_SKB_FRAGS 16UL
206 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
209 typedef struct skb_frag_struct skb_frag_t;
211 struct skb_frag_struct {
215 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
224 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
229 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
234 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
239 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
244 #define HAVE_HW_TIME_STAMP
247 * struct skb_shared_hwtstamps - hardware time stamps
248 * @hwtstamp: hardware time stamp transformed into duration
249 * since arbitrary point in time
251 * Software time stamps generated by ktime_get_real() are stored in
254 * hwtstamps can only be compared against other hwtstamps from
257 * This structure is attached to packets as part of the
258 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
260 struct skb_shared_hwtstamps {
264 /* Definitions for tx_flags in struct skb_shared_info */
266 /* generate hardware time stamp */
267 SKBTX_HW_TSTAMP = 1 << 0,
269 /* generate software time stamp when queueing packet to NIC */
270 SKBTX_SW_TSTAMP = 1 << 1,
272 /* device driver is going to provide hardware time stamp */
273 SKBTX_IN_PROGRESS = 1 << 2,
275 /* device driver supports TX zero-copy buffers */
276 SKBTX_DEV_ZEROCOPY = 1 << 3,
278 /* generate wifi status information (where possible) */
279 SKBTX_WIFI_STATUS = 1 << 4,
281 /* This indicates at least one fragment might be overwritten
282 * (as in vmsplice(), sendfile() ...)
283 * If we need to compute a TX checksum, we'll need to copy
284 * all frags to avoid possible bad checksum
286 SKBTX_SHARED_FRAG = 1 << 5,
288 /* generate software time stamp when entering packet scheduling */
289 SKBTX_SCHED_TSTAMP = 1 << 6,
291 /* generate software timestamp on peer data acknowledgment */
292 SKBTX_ACK_TSTAMP = 1 << 7,
295 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
296 SKBTX_SCHED_TSTAMP | \
298 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
301 * The callback notifies userspace to release buffers when skb DMA is done in
302 * lower device, the skb last reference should be 0 when calling this.
303 * The zerocopy_success argument is true if zero copy transmit occurred,
304 * false on data copy or out of memory error caused by data copy attempt.
305 * The ctx field is used to track device context.
306 * The desc field is used to track userspace buffer index.
309 void (*callback)(struct ubuf_info *, bool zerocopy_success);
314 /* This data is invariant across clones and lives at
315 * the end of the header data, ie. at skb->end.
317 struct skb_shared_info {
318 unsigned char nr_frags;
320 unsigned short gso_size;
321 /* Warning: this field is not always filled in (UFO)! */
322 unsigned short gso_segs;
323 unsigned short gso_type;
324 struct sk_buff *frag_list;
325 struct skb_shared_hwtstamps hwtstamps;
330 * Warning : all fields before dataref are cleared in __alloc_skb()
334 /* Intermediate layers must ensure that destructor_arg
335 * remains valid until skb destructor */
336 void * destructor_arg;
338 /* must be last field, see pskb_expand_head() */
339 skb_frag_t frags[MAX_SKB_FRAGS];
342 /* We divide dataref into two halves. The higher 16 bits hold references
343 * to the payload part of skb->data. The lower 16 bits hold references to
344 * the entire skb->data. A clone of a headerless skb holds the length of
345 * the header in skb->hdr_len.
347 * All users must obey the rule that the skb->data reference count must be
348 * greater than or equal to the payload reference count.
350 * Holding a reference to the payload part means that the user does not
351 * care about modifications to the header part of skb->data.
353 #define SKB_DATAREF_SHIFT 16
354 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
358 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
359 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
360 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
364 SKB_GSO_TCPV4 = 1 << 0,
365 SKB_GSO_UDP = 1 << 1,
367 /* This indicates the skb is from an untrusted source. */
368 SKB_GSO_DODGY = 1 << 2,
370 /* This indicates the tcp segment has CWR set. */
371 SKB_GSO_TCP_ECN = 1 << 3,
373 SKB_GSO_TCPV6 = 1 << 4,
375 SKB_GSO_FCOE = 1 << 5,
377 SKB_GSO_GRE = 1 << 6,
379 SKB_GSO_GRE_CSUM = 1 << 7,
381 SKB_GSO_IPIP = 1 << 8,
383 SKB_GSO_SIT = 1 << 9,
385 SKB_GSO_UDP_TUNNEL = 1 << 10,
387 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
389 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
392 #if BITS_PER_LONG > 32
393 #define NET_SKBUFF_DATA_USES_OFFSET 1
396 #ifdef NET_SKBUFF_DATA_USES_OFFSET
397 typedef unsigned int sk_buff_data_t;
399 typedef unsigned char *sk_buff_data_t;
403 * struct skb_mstamp - multi resolution time stamps
404 * @stamp_us: timestamp in us resolution
405 * @stamp_jiffies: timestamp in jiffies
418 * skb_mstamp_get - get current timestamp
419 * @cl: place to store timestamps
421 static inline void skb_mstamp_get(struct skb_mstamp *cl)
423 u64 val = local_clock();
425 do_div(val, NSEC_PER_USEC);
426 cl->stamp_us = (u32)val;
427 cl->stamp_jiffies = (u32)jiffies;
431 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
432 * @t1: pointer to newest sample
433 * @t0: pointer to oldest sample
435 static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
436 const struct skb_mstamp *t0)
438 s32 delta_us = t1->stamp_us - t0->stamp_us;
439 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
441 /* If delta_us is negative, this might be because interval is too big,
442 * or local_clock() drift is too big : fallback using jiffies.
445 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
447 delta_us = jiffies_to_usecs(delta_jiffies);
454 * struct sk_buff - socket buffer
455 * @next: Next buffer in list
456 * @prev: Previous buffer in list
457 * @tstamp: Time we arrived/left
458 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
459 * @sk: Socket we are owned by
460 * @dev: Device we arrived on/are leaving by
461 * @cb: Control buffer. Free for use by every layer. Put private vars here
462 * @_skb_refdst: destination entry (with norefcount bit)
463 * @sp: the security path, used for xfrm
464 * @len: Length of actual data
465 * @data_len: Data length
466 * @mac_len: Length of link layer header
467 * @hdr_len: writable header length of cloned skb
468 * @csum: Checksum (must include start/offset pair)
469 * @csum_start: Offset from skb->head where checksumming should start
470 * @csum_offset: Offset from csum_start where checksum should be stored
471 * @priority: Packet queueing priority
472 * @ignore_df: allow local fragmentation
473 * @cloned: Head may be cloned (check refcnt to be sure)
474 * @ip_summed: Driver fed us an IP checksum
475 * @nohdr: Payload reference only, must not modify header
476 * @nfctinfo: Relationship of this skb to the connection
477 * @pkt_type: Packet class
478 * @fclone: skbuff clone status
479 * @ipvs_property: skbuff is owned by ipvs
480 * @peeked: this packet has been seen already, so stats have been
481 * done for it, don't do them again
482 * @nf_trace: netfilter packet trace flag
483 * @protocol: Packet protocol from driver
484 * @destructor: Destruct function
485 * @nfct: Associated connection, if any
486 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
487 * @skb_iif: ifindex of device we arrived on
488 * @tc_index: Traffic control index
489 * @tc_verd: traffic control verdict
490 * @hash: the packet hash
491 * @queue_mapping: Queue mapping for multiqueue devices
492 * @xmit_more: More SKBs are pending for this queue
493 * @ndisc_nodetype: router type (from link layer)
494 * @ooo_okay: allow the mapping of a socket to a queue to be changed
495 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
497 * @sw_hash: indicates hash was computed in software stack
498 * @wifi_acked_valid: wifi_acked was set
499 * @wifi_acked: whether frame was acked on wifi or not
500 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
501 * @napi_id: id of the NAPI struct this skb came from
502 * @secmark: security marking
503 * @mark: Generic packet mark
504 * @vlan_proto: vlan encapsulation protocol
505 * @vlan_tci: vlan tag control information
506 * @inner_protocol: Protocol (encapsulation)
507 * @inner_transport_header: Inner transport layer header (encapsulation)
508 * @inner_network_header: Network layer header (encapsulation)
509 * @inner_mac_header: Link layer header (encapsulation)
510 * @transport_header: Transport layer header
511 * @network_header: Network layer header
512 * @mac_header: Link layer header
513 * @tail: Tail pointer
515 * @head: Head of buffer
516 * @data: Data head pointer
517 * @truesize: Buffer size
518 * @users: User count - see {datagram,tcp}.c
524 /* These two members must be first. */
525 struct sk_buff *next;
526 struct sk_buff *prev;
530 struct skb_mstamp skb_mstamp;
533 struct rb_node rbnode; /* used in netem & tcp stack */
536 struct net_device *dev;
539 * This is the control buffer. It is free to use for every
540 * layer. Please put your private variables there. If you
541 * want to keep them across layers you have to do a skb_clone()
542 * first. This is owned by whoever has the skb queued ATM.
544 char cb[48] __aligned(8);
546 unsigned long _skb_refdst;
547 void (*destructor)(struct sk_buff *skb);
551 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
552 struct nf_conntrack *nfct;
554 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
555 struct nf_bridge_info *nf_bridge;
562 /* Following fields are _not_ copied in __copy_skb_header()
563 * Note that queue_mapping is here mostly to fill a hole.
565 kmemcheck_bitfield_begin(flags1);
574 kmemcheck_bitfield_end(flags1);
576 /* fields enclosed in headers_start/headers_end are copied
577 * using a single memcpy() in __copy_skb_header()
580 __u32 headers_start[0];
583 /* if you move pkt_type around you also must adapt those constants */
584 #ifdef __BIG_ENDIAN_BITFIELD
585 #define PKT_TYPE_MAX (7 << 5)
587 #define PKT_TYPE_MAX 7
589 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
591 __u8 __pkt_type_offset[0];
602 __u8 wifi_acked_valid:1;
606 /* Indicates the inner headers are valid in the skbuff. */
607 __u8 encapsulation:1;
608 __u8 encap_hdr_csum:1;
610 __u8 csum_complete_sw:1;
614 #ifdef CONFIG_IPV6_NDISC_NODETYPE
615 __u8 ndisc_nodetype:2;
617 __u8 ipvs_property:1;
618 __u8 inner_protocol_type:1;
619 __u8 remcsum_offload:1;
620 /* 3 or 5 bit hole */
622 #ifdef CONFIG_NET_SCHED
623 __u16 tc_index; /* traffic control index */
624 #ifdef CONFIG_NET_CLS_ACT
625 __u16 tc_verd; /* traffic control verdict */
641 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
643 unsigned int napi_id;
644 unsigned int sender_cpu;
647 #ifdef CONFIG_NETWORK_SECMARK
652 __u32 reserved_tailroom;
656 __be16 inner_protocol;
660 __u16 inner_transport_header;
661 __u16 inner_network_header;
662 __u16 inner_mac_header;
665 __u16 transport_header;
666 __u16 network_header;
670 __u32 headers_end[0];
673 /* These elements must be at the end, see alloc_skb() for details. */
678 unsigned int truesize;
684 * Handling routines are only of interest to the kernel
686 #include <linux/slab.h>
689 #define SKB_ALLOC_FCLONE 0x01
690 #define SKB_ALLOC_RX 0x02
691 #define SKB_ALLOC_NAPI 0x04
693 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
694 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
696 return unlikely(skb->pfmemalloc);
700 * skb might have a dst pointer attached, refcounted or not.
701 * _skb_refdst low order bit is set if refcount was _not_ taken
703 #define SKB_DST_NOREF 1UL
704 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
707 * skb_dst - returns skb dst_entry
710 * Returns skb dst_entry, regardless of reference taken or not.
712 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
714 /* If refdst was not refcounted, check we still are in a
715 * rcu_read_lock section
717 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
718 !rcu_read_lock_held() &&
719 !rcu_read_lock_bh_held());
720 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
724 * skb_dst_set - sets skb dst
728 * Sets skb dst, assuming a reference was taken on dst and should
729 * be released by skb_dst_drop()
731 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
733 skb->_skb_refdst = (unsigned long)dst;
737 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
741 * Sets skb dst, assuming a reference was not taken on dst.
742 * If dst entry is cached, we do not take reference and dst_release
743 * will be avoided by refdst_drop. If dst entry is not cached, we take
744 * reference, so that last dst_release can destroy the dst immediately.
746 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
748 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
749 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
753 * skb_dst_is_noref - Test if skb dst isn't refcounted
756 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
758 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
761 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
763 return (struct rtable *)skb_dst(skb);
766 void kfree_skb(struct sk_buff *skb);
767 void kfree_skb_list(struct sk_buff *segs);
768 void skb_tx_error(struct sk_buff *skb);
769 void consume_skb(struct sk_buff *skb);
770 void __kfree_skb(struct sk_buff *skb);
771 extern struct kmem_cache *skbuff_head_cache;
773 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
774 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
775 bool *fragstolen, int *delta_truesize);
777 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
779 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
780 struct sk_buff *build_skb(void *data, unsigned int frag_size);
781 static inline struct sk_buff *alloc_skb(unsigned int size,
784 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
787 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
788 unsigned long data_len,
793 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
794 struct sk_buff_fclones {
803 * skb_fclone_busy - check if fclone is busy
806 * Returns true is skb is a fast clone, and its clone is not freed.
807 * Some drivers call skb_orphan() in their ndo_start_xmit(),
808 * so we also check that this didnt happen.
810 static inline bool skb_fclone_busy(const struct sock *sk,
811 const struct sk_buff *skb)
813 const struct sk_buff_fclones *fclones;
815 fclones = container_of(skb, struct sk_buff_fclones, skb1);
817 return skb->fclone == SKB_FCLONE_ORIG &&
818 atomic_read(&fclones->fclone_ref) > 1 &&
819 fclones->skb2.sk == sk;
822 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
825 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
828 struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
829 static inline struct sk_buff *alloc_skb_head(gfp_t priority)
831 return __alloc_skb_head(priority, -1);
834 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
835 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
836 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
837 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
838 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
839 gfp_t gfp_mask, bool fclone);
840 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
843 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
846 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
847 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
848 unsigned int headroom);
849 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
850 int newtailroom, gfp_t priority);
851 int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
852 int offset, int len);
853 int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
855 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
856 int skb_pad(struct sk_buff *skb, int pad);
857 #define dev_kfree_skb(a) consume_skb(a)
859 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
860 int getfrag(void *from, char *to, int offset,
861 int len, int odd, struct sk_buff *skb),
862 void *from, int length);
864 struct skb_seq_state {
868 __u32 stepped_offset;
869 struct sk_buff *root_skb;
870 struct sk_buff *cur_skb;
874 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
875 unsigned int to, struct skb_seq_state *st);
876 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
877 struct skb_seq_state *st);
878 void skb_abort_seq_read(struct skb_seq_state *st);
880 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
881 unsigned int to, struct ts_config *config);
884 * Packet hash types specify the type of hash in skb_set_hash.
886 * Hash types refer to the protocol layer addresses which are used to
887 * construct a packet's hash. The hashes are used to differentiate or identify
888 * flows of the protocol layer for the hash type. Hash types are either
889 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
891 * Properties of hashes:
893 * 1) Two packets in different flows have different hash values
894 * 2) Two packets in the same flow should have the same hash value
896 * A hash at a higher layer is considered to be more specific. A driver should
897 * set the most specific hash possible.
899 * A driver cannot indicate a more specific hash than the layer at which a hash
900 * was computed. For instance an L3 hash cannot be set as an L4 hash.
902 * A driver may indicate a hash level which is less specific than the
903 * actual layer the hash was computed on. For instance, a hash computed
904 * at L4 may be considered an L3 hash. This should only be done if the
905 * driver can't unambiguously determine that the HW computed the hash at
906 * the higher layer. Note that the "should" in the second property above
909 enum pkt_hash_types {
910 PKT_HASH_TYPE_NONE, /* Undefined type */
911 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
912 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
913 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
917 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
919 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
924 static inline __u32 skb_get_hash(struct sk_buff *skb)
926 if (!skb->l4_hash && !skb->sw_hash)
932 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
934 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
939 static inline void skb_clear_hash(struct sk_buff *skb)
946 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
952 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
954 to->hash = from->hash;
955 to->sw_hash = from->sw_hash;
956 to->l4_hash = from->l4_hash;
959 static inline void skb_sender_cpu_clear(struct sk_buff *skb)
966 #ifdef NET_SKBUFF_DATA_USES_OFFSET
967 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
969 return skb->head + skb->end;
972 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
977 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
982 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
984 return skb->end - skb->head;
989 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
991 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
993 return &skb_shinfo(skb)->hwtstamps;
997 * skb_queue_empty - check if a queue is empty
1000 * Returns true if the queue is empty, false otherwise.
1002 static inline int skb_queue_empty(const struct sk_buff_head *list)
1004 return list->next == (const struct sk_buff *) list;
1008 * skb_queue_is_last - check if skb is the last entry in the queue
1012 * Returns true if @skb is the last buffer on the list.
1014 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1015 const struct sk_buff *skb)
1017 return skb->next == (const struct sk_buff *) list;
1021 * skb_queue_is_first - check if skb is the first entry in the queue
1025 * Returns true if @skb is the first buffer on the list.
1027 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1028 const struct sk_buff *skb)
1030 return skb->prev == (const struct sk_buff *) list;
1034 * skb_queue_next - return the next packet in the queue
1036 * @skb: current buffer
1038 * Return the next packet in @list after @skb. It is only valid to
1039 * call this if skb_queue_is_last() evaluates to false.
1041 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1042 const struct sk_buff *skb)
1044 /* This BUG_ON may seem severe, but if we just return then we
1045 * are going to dereference garbage.
1047 BUG_ON(skb_queue_is_last(list, skb));
1052 * skb_queue_prev - return the prev packet in the queue
1054 * @skb: current buffer
1056 * Return the prev packet in @list before @skb. It is only valid to
1057 * call this if skb_queue_is_first() evaluates to false.
1059 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1060 const struct sk_buff *skb)
1062 /* This BUG_ON may seem severe, but if we just return then we
1063 * are going to dereference garbage.
1065 BUG_ON(skb_queue_is_first(list, skb));
1070 * skb_get - reference buffer
1071 * @skb: buffer to reference
1073 * Makes another reference to a socket buffer and returns a pointer
1076 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1078 atomic_inc(&skb->users);
1083 * If users == 1, we are the only owner and are can avoid redundant
1088 * skb_cloned - is the buffer a clone
1089 * @skb: buffer to check
1091 * Returns true if the buffer was generated with skb_clone() and is
1092 * one of multiple shared copies of the buffer. Cloned buffers are
1093 * shared data so must not be written to under normal circumstances.
1095 static inline int skb_cloned(const struct sk_buff *skb)
1097 return skb->cloned &&
1098 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1101 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1103 might_sleep_if(pri & __GFP_WAIT);
1105 if (skb_cloned(skb))
1106 return pskb_expand_head(skb, 0, 0, pri);
1112 * skb_header_cloned - is the header a clone
1113 * @skb: buffer to check
1115 * Returns true if modifying the header part of the buffer requires
1116 * the data to be copied.
1118 static inline int skb_header_cloned(const struct sk_buff *skb)
1125 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1126 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1127 return dataref != 1;
1131 * skb_header_release - release reference to header
1132 * @skb: buffer to operate on
1134 * Drop a reference to the header part of the buffer. This is done
1135 * by acquiring a payload reference. You must not read from the header
1136 * part of skb->data after this.
1137 * Note : Check if you can use __skb_header_release() instead.
1139 static inline void skb_header_release(struct sk_buff *skb)
1143 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1147 * __skb_header_release - release reference to header
1148 * @skb: buffer to operate on
1150 * Variant of skb_header_release() assuming skb is private to caller.
1151 * We can avoid one atomic operation.
1153 static inline void __skb_header_release(struct sk_buff *skb)
1156 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1161 * skb_shared - is the buffer shared
1162 * @skb: buffer to check
1164 * Returns true if more than one person has a reference to this
1167 static inline int skb_shared(const struct sk_buff *skb)
1169 return atomic_read(&skb->users) != 1;
1173 * skb_share_check - check if buffer is shared and if so clone it
1174 * @skb: buffer to check
1175 * @pri: priority for memory allocation
1177 * If the buffer is shared the buffer is cloned and the old copy
1178 * drops a reference. A new clone with a single reference is returned.
1179 * If the buffer is not shared the original buffer is returned. When
1180 * being called from interrupt status or with spinlocks held pri must
1183 * NULL is returned on a memory allocation failure.
1185 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1187 might_sleep_if(pri & __GFP_WAIT);
1188 if (skb_shared(skb)) {
1189 struct sk_buff *nskb = skb_clone(skb, pri);
1201 * Copy shared buffers into a new sk_buff. We effectively do COW on
1202 * packets to handle cases where we have a local reader and forward
1203 * and a couple of other messy ones. The normal one is tcpdumping
1204 * a packet thats being forwarded.
1208 * skb_unshare - make a copy of a shared buffer
1209 * @skb: buffer to check
1210 * @pri: priority for memory allocation
1212 * If the socket buffer is a clone then this function creates a new
1213 * copy of the data, drops a reference count on the old copy and returns
1214 * the new copy with the reference count at 1. If the buffer is not a clone
1215 * the original buffer is returned. When called with a spinlock held or
1216 * from interrupt state @pri must be %GFP_ATOMIC
1218 * %NULL is returned on a memory allocation failure.
1220 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1223 might_sleep_if(pri & __GFP_WAIT);
1224 if (skb_cloned(skb)) {
1225 struct sk_buff *nskb = skb_copy(skb, pri);
1227 /* Free our shared copy */
1238 * skb_peek - peek at the head of an &sk_buff_head
1239 * @list_: list to peek at
1241 * Peek an &sk_buff. Unlike most other operations you _MUST_
1242 * be careful with this one. A peek leaves the buffer on the
1243 * list and someone else may run off with it. You must hold
1244 * the appropriate locks or have a private queue to do this.
1246 * Returns %NULL for an empty list or a pointer to the head element.
1247 * The reference count is not incremented and the reference is therefore
1248 * volatile. Use with caution.
1250 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1252 struct sk_buff *skb = list_->next;
1254 if (skb == (struct sk_buff *)list_)
1260 * skb_peek_next - peek skb following the given one from a queue
1261 * @skb: skb to start from
1262 * @list_: list to peek at
1264 * Returns %NULL when the end of the list is met or a pointer to the
1265 * next element. The reference count is not incremented and the
1266 * reference is therefore volatile. Use with caution.
1268 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1269 const struct sk_buff_head *list_)
1271 struct sk_buff *next = skb->next;
1273 if (next == (struct sk_buff *)list_)
1279 * skb_peek_tail - peek at the tail of an &sk_buff_head
1280 * @list_: list to peek at
1282 * Peek an &sk_buff. Unlike most other operations you _MUST_
1283 * be careful with this one. A peek leaves the buffer on the
1284 * list and someone else may run off with it. You must hold
1285 * the appropriate locks or have a private queue to do this.
1287 * Returns %NULL for an empty list or a pointer to the tail element.
1288 * The reference count is not incremented and the reference is therefore
1289 * volatile. Use with caution.
1291 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1293 struct sk_buff *skb = list_->prev;
1295 if (skb == (struct sk_buff *)list_)
1302 * skb_queue_len - get queue length
1303 * @list_: list to measure
1305 * Return the length of an &sk_buff queue.
1307 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1313 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1314 * @list: queue to initialize
1316 * This initializes only the list and queue length aspects of
1317 * an sk_buff_head object. This allows to initialize the list
1318 * aspects of an sk_buff_head without reinitializing things like
1319 * the spinlock. It can also be used for on-stack sk_buff_head
1320 * objects where the spinlock is known to not be used.
1322 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1324 list->prev = list->next = (struct sk_buff *)list;
1329 * This function creates a split out lock class for each invocation;
1330 * this is needed for now since a whole lot of users of the skb-queue
1331 * infrastructure in drivers have different locking usage (in hardirq)
1332 * than the networking core (in softirq only). In the long run either the
1333 * network layer or drivers should need annotation to consolidate the
1334 * main types of usage into 3 classes.
1336 static inline void skb_queue_head_init(struct sk_buff_head *list)
1338 spin_lock_init(&list->lock);
1339 __skb_queue_head_init(list);
1342 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1343 struct lock_class_key *class)
1345 skb_queue_head_init(list);
1346 lockdep_set_class(&list->lock, class);
1350 * Insert an sk_buff on a list.
1352 * The "__skb_xxxx()" functions are the non-atomic ones that
1353 * can only be called with interrupts disabled.
1355 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1356 struct sk_buff_head *list);
1357 static inline void __skb_insert(struct sk_buff *newsk,
1358 struct sk_buff *prev, struct sk_buff *next,
1359 struct sk_buff_head *list)
1363 next->prev = prev->next = newsk;
1367 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1368 struct sk_buff *prev,
1369 struct sk_buff *next)
1371 struct sk_buff *first = list->next;
1372 struct sk_buff *last = list->prev;
1382 * skb_queue_splice - join two skb lists, this is designed for stacks
1383 * @list: the new list to add
1384 * @head: the place to add it in the first list
1386 static inline void skb_queue_splice(const struct sk_buff_head *list,
1387 struct sk_buff_head *head)
1389 if (!skb_queue_empty(list)) {
1390 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1391 head->qlen += list->qlen;
1396 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1397 * @list: the new list to add
1398 * @head: the place to add it in the first list
1400 * The list at @list is reinitialised
1402 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1403 struct sk_buff_head *head)
1405 if (!skb_queue_empty(list)) {
1406 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1407 head->qlen += list->qlen;
1408 __skb_queue_head_init(list);
1413 * skb_queue_splice_tail - join two skb lists, each list being a queue
1414 * @list: the new list to add
1415 * @head: the place to add it in the first list
1417 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1418 struct sk_buff_head *head)
1420 if (!skb_queue_empty(list)) {
1421 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1422 head->qlen += list->qlen;
1427 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1428 * @list: the new list to add
1429 * @head: the place to add it in the first list
1431 * Each of the lists is a queue.
1432 * The list at @list is reinitialised
1434 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1435 struct sk_buff_head *head)
1437 if (!skb_queue_empty(list)) {
1438 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1439 head->qlen += list->qlen;
1440 __skb_queue_head_init(list);
1445 * __skb_queue_after - queue a buffer at the list head
1446 * @list: list to use
1447 * @prev: place after this buffer
1448 * @newsk: buffer to queue
1450 * Queue a buffer int the middle of a list. This function takes no locks
1451 * and you must therefore hold required locks before calling it.
1453 * A buffer cannot be placed on two lists at the same time.
1455 static inline void __skb_queue_after(struct sk_buff_head *list,
1456 struct sk_buff *prev,
1457 struct sk_buff *newsk)
1459 __skb_insert(newsk, prev, prev->next, list);
1462 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1463 struct sk_buff_head *list);
1465 static inline void __skb_queue_before(struct sk_buff_head *list,
1466 struct sk_buff *next,
1467 struct sk_buff *newsk)
1469 __skb_insert(newsk, next->prev, next, list);
1473 * __skb_queue_head - queue a buffer at the list head
1474 * @list: list to use
1475 * @newsk: buffer to queue
1477 * Queue a buffer at the start of a list. This function takes no locks
1478 * and you must therefore hold required locks before calling it.
1480 * A buffer cannot be placed on two lists at the same time.
1482 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1483 static inline void __skb_queue_head(struct sk_buff_head *list,
1484 struct sk_buff *newsk)
1486 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1490 * __skb_queue_tail - queue a buffer at the list tail
1491 * @list: list to use
1492 * @newsk: buffer to queue
1494 * Queue a buffer at the end of a list. This function takes no locks
1495 * and you must therefore hold required locks before calling it.
1497 * A buffer cannot be placed on two lists at the same time.
1499 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1500 static inline void __skb_queue_tail(struct sk_buff_head *list,
1501 struct sk_buff *newsk)
1503 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1507 * remove sk_buff from list. _Must_ be called atomically, and with
1510 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1511 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1513 struct sk_buff *next, *prev;
1518 skb->next = skb->prev = NULL;
1524 * __skb_dequeue - remove from the head of the queue
1525 * @list: list to dequeue from
1527 * Remove the head of the list. This function does not take any locks
1528 * so must be used with appropriate locks held only. The head item is
1529 * returned or %NULL if the list is empty.
1531 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1532 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1534 struct sk_buff *skb = skb_peek(list);
1536 __skb_unlink(skb, list);
1541 * __skb_dequeue_tail - remove from the tail of the queue
1542 * @list: list to dequeue from
1544 * Remove the tail of the list. This function does not take any locks
1545 * so must be used with appropriate locks held only. The tail item is
1546 * returned or %NULL if the list is empty.
1548 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1549 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1551 struct sk_buff *skb = skb_peek_tail(list);
1553 __skb_unlink(skb, list);
1558 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1560 return skb->data_len;
1563 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1565 return skb->len - skb->data_len;
1568 static inline int skb_pagelen(const struct sk_buff *skb)
1572 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1573 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1574 return len + skb_headlen(skb);
1578 * __skb_fill_page_desc - initialise a paged fragment in an skb
1579 * @skb: buffer containing fragment to be initialised
1580 * @i: paged fragment index to initialise
1581 * @page: the page to use for this fragment
1582 * @off: the offset to the data with @page
1583 * @size: the length of the data
1585 * Initialises the @i'th fragment of @skb to point to &size bytes at
1586 * offset @off within @page.
1588 * Does not take any additional reference on the fragment.
1590 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1591 struct page *page, int off, int size)
1593 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1596 * Propagate page->pfmemalloc to the skb if we can. The problem is
1597 * that not all callers have unique ownership of the page. If
1598 * pfmemalloc is set, we check the mapping as a mapping implies
1599 * page->index is set (index and pfmemalloc share space).
1600 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1601 * do not lose pfmemalloc information as the pages would not be
1602 * allocated using __GFP_MEMALLOC.
1604 frag->page.p = page;
1605 frag->page_offset = off;
1606 skb_frag_size_set(frag, size);
1608 page = compound_head(page);
1609 if (page->pfmemalloc && !page->mapping)
1610 skb->pfmemalloc = true;
1614 * skb_fill_page_desc - initialise a paged fragment in an skb
1615 * @skb: buffer containing fragment to be initialised
1616 * @i: paged fragment index to initialise
1617 * @page: the page to use for this fragment
1618 * @off: the offset to the data with @page
1619 * @size: the length of the data
1621 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1622 * @skb to point to @size bytes at offset @off within @page. In
1623 * addition updates @skb such that @i is the last fragment.
1625 * Does not take any additional reference on the fragment.
1627 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1628 struct page *page, int off, int size)
1630 __skb_fill_page_desc(skb, i, page, off, size);
1631 skb_shinfo(skb)->nr_frags = i + 1;
1634 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1635 int size, unsigned int truesize);
1637 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1638 unsigned int truesize);
1640 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1641 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1642 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1644 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1645 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1647 return skb->head + skb->tail;
1650 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1652 skb->tail = skb->data - skb->head;
1655 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1657 skb_reset_tail_pointer(skb);
1658 skb->tail += offset;
1661 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1662 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1667 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1669 skb->tail = skb->data;
1672 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1674 skb->tail = skb->data + offset;
1677 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1680 * Add data to an sk_buff
1682 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1683 unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1684 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1686 unsigned char *tmp = skb_tail_pointer(skb);
1687 SKB_LINEAR_ASSERT(skb);
1693 unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1694 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1701 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1702 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1705 BUG_ON(skb->len < skb->data_len);
1706 return skb->data += len;
1709 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1711 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1714 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1716 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1718 if (len > skb_headlen(skb) &&
1719 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1722 return skb->data += len;
1725 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1727 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1730 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1732 if (likely(len <= skb_headlen(skb)))
1734 if (unlikely(len > skb->len))
1736 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1740 * skb_headroom - bytes at buffer head
1741 * @skb: buffer to check
1743 * Return the number of bytes of free space at the head of an &sk_buff.
1745 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1747 return skb->data - skb->head;
1751 * skb_tailroom - bytes at buffer end
1752 * @skb: buffer to check
1754 * Return the number of bytes of free space at the tail of an sk_buff
1756 static inline int skb_tailroom(const struct sk_buff *skb)
1758 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1762 * skb_availroom - bytes at buffer end
1763 * @skb: buffer to check
1765 * Return the number of bytes of free space at the tail of an sk_buff
1766 * allocated by sk_stream_alloc()
1768 static inline int skb_availroom(const struct sk_buff *skb)
1770 if (skb_is_nonlinear(skb))
1773 return skb->end - skb->tail - skb->reserved_tailroom;
1777 * skb_reserve - adjust headroom
1778 * @skb: buffer to alter
1779 * @len: bytes to move
1781 * Increase the headroom of an empty &sk_buff by reducing the tail
1782 * room. This is only allowed for an empty buffer.
1784 static inline void skb_reserve(struct sk_buff *skb, int len)
1790 #define ENCAP_TYPE_ETHER 0
1791 #define ENCAP_TYPE_IPPROTO 1
1793 static inline void skb_set_inner_protocol(struct sk_buff *skb,
1796 skb->inner_protocol = protocol;
1797 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1800 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1803 skb->inner_ipproto = ipproto;
1804 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1807 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1809 skb->inner_mac_header = skb->mac_header;
1810 skb->inner_network_header = skb->network_header;
1811 skb->inner_transport_header = skb->transport_header;
1814 static inline void skb_reset_mac_len(struct sk_buff *skb)
1816 skb->mac_len = skb->network_header - skb->mac_header;
1819 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1822 return skb->head + skb->inner_transport_header;
1825 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1827 skb->inner_transport_header = skb->data - skb->head;
1830 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1833 skb_reset_inner_transport_header(skb);
1834 skb->inner_transport_header += offset;
1837 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1839 return skb->head + skb->inner_network_header;
1842 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1844 skb->inner_network_header = skb->data - skb->head;
1847 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1850 skb_reset_inner_network_header(skb);
1851 skb->inner_network_header += offset;
1854 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1856 return skb->head + skb->inner_mac_header;
1859 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1861 skb->inner_mac_header = skb->data - skb->head;
1864 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1867 skb_reset_inner_mac_header(skb);
1868 skb->inner_mac_header += offset;
1870 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1872 return skb->transport_header != (typeof(skb->transport_header))~0U;
1875 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1877 return skb->head + skb->transport_header;
1880 static inline void skb_reset_transport_header(struct sk_buff *skb)
1882 skb->transport_header = skb->data - skb->head;
1885 static inline void skb_set_transport_header(struct sk_buff *skb,
1888 skb_reset_transport_header(skb);
1889 skb->transport_header += offset;
1892 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1894 return skb->head + skb->network_header;
1897 static inline void skb_reset_network_header(struct sk_buff *skb)
1899 skb->network_header = skb->data - skb->head;
1902 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1904 skb_reset_network_header(skb);
1905 skb->network_header += offset;
1908 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1910 return skb->head + skb->mac_header;
1913 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1915 return skb->mac_header != (typeof(skb->mac_header))~0U;
1918 static inline void skb_reset_mac_header(struct sk_buff *skb)
1920 skb->mac_header = skb->data - skb->head;
1923 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1925 skb_reset_mac_header(skb);
1926 skb->mac_header += offset;
1929 static inline void skb_pop_mac_header(struct sk_buff *skb)
1931 skb->mac_header = skb->network_header;
1934 static inline void skb_probe_transport_header(struct sk_buff *skb,
1935 const int offset_hint)
1937 struct flow_keys keys;
1939 if (skb_transport_header_was_set(skb))
1941 else if (skb_flow_dissect_flow_keys(skb, &keys))
1942 skb_set_transport_header(skb, keys.basic.thoff);
1944 skb_set_transport_header(skb, offset_hint);
1947 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1949 if (skb_mac_header_was_set(skb)) {
1950 const unsigned char *old_mac = skb_mac_header(skb);
1952 skb_set_mac_header(skb, -skb->mac_len);
1953 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1957 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1959 return skb->csum_start - skb_headroom(skb);
1962 static inline int skb_transport_offset(const struct sk_buff *skb)
1964 return skb_transport_header(skb) - skb->data;
1967 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1969 return skb->transport_header - skb->network_header;
1972 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1974 return skb->inner_transport_header - skb->inner_network_header;
1977 static inline int skb_network_offset(const struct sk_buff *skb)
1979 return skb_network_header(skb) - skb->data;
1982 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1984 return skb_inner_network_header(skb) - skb->data;
1987 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1989 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1993 * CPUs often take a performance hit when accessing unaligned memory
1994 * locations. The actual performance hit varies, it can be small if the
1995 * hardware handles it or large if we have to take an exception and fix it
1998 * Since an ethernet header is 14 bytes network drivers often end up with
1999 * the IP header at an unaligned offset. The IP header can be aligned by
2000 * shifting the start of the packet by 2 bytes. Drivers should do this
2003 * skb_reserve(skb, NET_IP_ALIGN);
2005 * The downside to this alignment of the IP header is that the DMA is now
2006 * unaligned. On some architectures the cost of an unaligned DMA is high
2007 * and this cost outweighs the gains made by aligning the IP header.
2009 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2012 #ifndef NET_IP_ALIGN
2013 #define NET_IP_ALIGN 2
2017 * The networking layer reserves some headroom in skb data (via
2018 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2019 * the header has to grow. In the default case, if the header has to grow
2020 * 32 bytes or less we avoid the reallocation.
2022 * Unfortunately this headroom changes the DMA alignment of the resulting
2023 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2024 * on some architectures. An architecture can override this value,
2025 * perhaps setting it to a cacheline in size (since that will maintain
2026 * cacheline alignment of the DMA). It must be a power of 2.
2028 * Various parts of the networking layer expect at least 32 bytes of
2029 * headroom, you should not reduce this.
2031 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2032 * to reduce average number of cache lines per packet.
2033 * get_rps_cpus() for example only access one 64 bytes aligned block :
2034 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2037 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2040 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2042 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2044 if (unlikely(skb_is_nonlinear(skb))) {
2049 skb_set_tail_pointer(skb, len);
2052 void skb_trim(struct sk_buff *skb, unsigned int len);
2054 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2057 return ___pskb_trim(skb, len);
2058 __skb_trim(skb, len);
2062 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2064 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2068 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2069 * @skb: buffer to alter
2072 * This is identical to pskb_trim except that the caller knows that
2073 * the skb is not cloned so we should never get an error due to out-
2076 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2078 int err = pskb_trim(skb, len);
2083 * skb_orphan - orphan a buffer
2084 * @skb: buffer to orphan
2086 * If a buffer currently has an owner then we call the owner's
2087 * destructor function and make the @skb unowned. The buffer continues
2088 * to exist but is no longer charged to its former owner.
2090 static inline void skb_orphan(struct sk_buff *skb)
2092 if (skb->destructor) {
2093 skb->destructor(skb);
2094 skb->destructor = NULL;
2102 * skb_orphan_frags - orphan the frags contained in a buffer
2103 * @skb: buffer to orphan frags from
2104 * @gfp_mask: allocation mask for replacement pages
2106 * For each frag in the SKB which needs a destructor (i.e. has an
2107 * owner) create a copy of that frag and release the original
2108 * page by calling the destructor.
2110 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2112 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2114 return skb_copy_ubufs(skb, gfp_mask);
2118 * __skb_queue_purge - empty a list
2119 * @list: list to empty
2121 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2122 * the list and one reference dropped. This function does not take the
2123 * list lock and the caller must hold the relevant locks to use it.
2125 void skb_queue_purge(struct sk_buff_head *list);
2126 static inline void __skb_queue_purge(struct sk_buff_head *list)
2128 struct sk_buff *skb;
2129 while ((skb = __skb_dequeue(list)) != NULL)
2133 void *netdev_alloc_frag(unsigned int fragsz);
2135 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2139 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2140 * @dev: network device to receive on
2141 * @length: length to allocate
2143 * Allocate a new &sk_buff and assign it a usage count of one. The
2144 * buffer has unspecified headroom built in. Users should allocate
2145 * the headroom they think they need without accounting for the
2146 * built in space. The built in space is used for optimisations.
2148 * %NULL is returned if there is no free memory. Although this function
2149 * allocates memory it can be called from an interrupt.
2151 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2152 unsigned int length)
2154 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2157 /* legacy helper around __netdev_alloc_skb() */
2158 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2161 return __netdev_alloc_skb(NULL, length, gfp_mask);
2164 /* legacy helper around netdev_alloc_skb() */
2165 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2167 return netdev_alloc_skb(NULL, length);
2171 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2172 unsigned int length, gfp_t gfp)
2174 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2176 if (NET_IP_ALIGN && skb)
2177 skb_reserve(skb, NET_IP_ALIGN);
2181 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2182 unsigned int length)
2184 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2187 static inline void skb_free_frag(void *addr)
2189 __free_page_frag(addr);
2192 void *napi_alloc_frag(unsigned int fragsz);
2193 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2194 unsigned int length, gfp_t gfp_mask);
2195 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2196 unsigned int length)
2198 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2202 * __dev_alloc_pages - allocate page for network Rx
2203 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2204 * @order: size of the allocation
2206 * Allocate a new page.
2208 * %NULL is returned if there is no free memory.
2210 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2213 /* This piece of code contains several assumptions.
2214 * 1. This is for device Rx, therefor a cold page is preferred.
2215 * 2. The expectation is the user wants a compound page.
2216 * 3. If requesting a order 0 page it will not be compound
2217 * due to the check to see if order has a value in prep_new_page
2218 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2219 * code in gfp_to_alloc_flags that should be enforcing this.
2221 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2223 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2226 static inline struct page *dev_alloc_pages(unsigned int order)
2228 return __dev_alloc_pages(GFP_ATOMIC, order);
2232 * __dev_alloc_page - allocate a page for network Rx
2233 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2235 * Allocate a new page.
2237 * %NULL is returned if there is no free memory.
2239 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2241 return __dev_alloc_pages(gfp_mask, 0);
2244 static inline struct page *dev_alloc_page(void)
2246 return __dev_alloc_page(GFP_ATOMIC);
2250 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2251 * @page: The page that was allocated from skb_alloc_page
2252 * @skb: The skb that may need pfmemalloc set
2254 static inline void skb_propagate_pfmemalloc(struct page *page,
2255 struct sk_buff *skb)
2257 if (page && page->pfmemalloc)
2258 skb->pfmemalloc = true;
2262 * skb_frag_page - retrieve the page referred to by a paged fragment
2263 * @frag: the paged fragment
2265 * Returns the &struct page associated with @frag.
2267 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2269 return frag->page.p;
2273 * __skb_frag_ref - take an addition reference on a paged fragment.
2274 * @frag: the paged fragment
2276 * Takes an additional reference on the paged fragment @frag.
2278 static inline void __skb_frag_ref(skb_frag_t *frag)
2280 get_page(skb_frag_page(frag));
2284 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2286 * @f: the fragment offset.
2288 * Takes an additional reference on the @f'th paged fragment of @skb.
2290 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2292 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2296 * __skb_frag_unref - release a reference on a paged fragment.
2297 * @frag: the paged fragment
2299 * Releases a reference on the paged fragment @frag.
2301 static inline void __skb_frag_unref(skb_frag_t *frag)
2303 put_page(skb_frag_page(frag));
2307 * skb_frag_unref - release a reference on a paged fragment of an skb.
2309 * @f: the fragment offset
2311 * Releases a reference on the @f'th paged fragment of @skb.
2313 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2315 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2319 * skb_frag_address - gets the address of the data contained in a paged fragment
2320 * @frag: the paged fragment buffer
2322 * Returns the address of the data within @frag. The page must already
2325 static inline void *skb_frag_address(const skb_frag_t *frag)
2327 return page_address(skb_frag_page(frag)) + frag->page_offset;
2331 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2332 * @frag: the paged fragment buffer
2334 * Returns the address of the data within @frag. Checks that the page
2335 * is mapped and returns %NULL otherwise.
2337 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2339 void *ptr = page_address(skb_frag_page(frag));
2343 return ptr + frag->page_offset;
2347 * __skb_frag_set_page - sets the page contained in a paged fragment
2348 * @frag: the paged fragment
2349 * @page: the page to set
2351 * Sets the fragment @frag to contain @page.
2353 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2355 frag->page.p = page;
2359 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2361 * @f: the fragment offset
2362 * @page: the page to set
2364 * Sets the @f'th fragment of @skb to contain @page.
2366 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2369 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2372 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2375 * skb_frag_dma_map - maps a paged fragment via the DMA API
2376 * @dev: the device to map the fragment to
2377 * @frag: the paged fragment to map
2378 * @offset: the offset within the fragment (starting at the
2379 * fragment's own offset)
2380 * @size: the number of bytes to map
2381 * @dir: the direction of the mapping (%PCI_DMA_*)
2383 * Maps the page associated with @frag to @device.
2385 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2386 const skb_frag_t *frag,
2387 size_t offset, size_t size,
2388 enum dma_data_direction dir)
2390 return dma_map_page(dev, skb_frag_page(frag),
2391 frag->page_offset + offset, size, dir);
2394 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2397 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2401 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2404 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2409 * skb_clone_writable - is the header of a clone writable
2410 * @skb: buffer to check
2411 * @len: length up to which to write
2413 * Returns true if modifying the header part of the cloned buffer
2414 * does not requires the data to be copied.
2416 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2418 return !skb_header_cloned(skb) &&
2419 skb_headroom(skb) + len <= skb->hdr_len;
2422 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2427 if (headroom > skb_headroom(skb))
2428 delta = headroom - skb_headroom(skb);
2430 if (delta || cloned)
2431 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2437 * skb_cow - copy header of skb when it is required
2438 * @skb: buffer to cow
2439 * @headroom: needed headroom
2441 * If the skb passed lacks sufficient headroom or its data part
2442 * is shared, data is reallocated. If reallocation fails, an error
2443 * is returned and original skb is not changed.
2445 * The result is skb with writable area skb->head...skb->tail
2446 * and at least @headroom of space at head.
2448 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2450 return __skb_cow(skb, headroom, skb_cloned(skb));
2454 * skb_cow_head - skb_cow but only making the head writable
2455 * @skb: buffer to cow
2456 * @headroom: needed headroom
2458 * This function is identical to skb_cow except that we replace the
2459 * skb_cloned check by skb_header_cloned. It should be used when
2460 * you only need to push on some header and do not need to modify
2463 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2465 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2469 * skb_padto - pad an skbuff up to a minimal size
2470 * @skb: buffer to pad
2471 * @len: minimal length
2473 * Pads up a buffer to ensure the trailing bytes exist and are
2474 * blanked. If the buffer already contains sufficient data it
2475 * is untouched. Otherwise it is extended. Returns zero on
2476 * success. The skb is freed on error.
2478 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2480 unsigned int size = skb->len;
2481 if (likely(size >= len))
2483 return skb_pad(skb, len - size);
2487 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2488 * @skb: buffer to pad
2489 * @len: minimal length
2491 * Pads up a buffer to ensure the trailing bytes exist and are
2492 * blanked. If the buffer already contains sufficient data it
2493 * is untouched. Otherwise it is extended. Returns zero on
2494 * success. The skb is freed on error.
2496 static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2498 unsigned int size = skb->len;
2500 if (unlikely(size < len)) {
2502 if (skb_pad(skb, len))
2504 __skb_put(skb, len);
2509 static inline int skb_add_data(struct sk_buff *skb,
2510 struct iov_iter *from, int copy)
2512 const int off = skb->len;
2514 if (skb->ip_summed == CHECKSUM_NONE) {
2516 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2517 &csum, from) == copy) {
2518 skb->csum = csum_block_add(skb->csum, csum, off);
2521 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
2524 __skb_trim(skb, off);
2528 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2529 const struct page *page, int off)
2532 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2534 return page == skb_frag_page(frag) &&
2535 off == frag->page_offset + skb_frag_size(frag);
2540 static inline int __skb_linearize(struct sk_buff *skb)
2542 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2546 * skb_linearize - convert paged skb to linear one
2547 * @skb: buffer to linarize
2549 * If there is no free memory -ENOMEM is returned, otherwise zero
2550 * is returned and the old skb data released.
2552 static inline int skb_linearize(struct sk_buff *skb)
2554 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2558 * skb_has_shared_frag - can any frag be overwritten
2559 * @skb: buffer to test
2561 * Return true if the skb has at least one frag that might be modified
2562 * by an external entity (as in vmsplice()/sendfile())
2564 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2566 return skb_is_nonlinear(skb) &&
2567 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2571 * skb_linearize_cow - make sure skb is linear and writable
2572 * @skb: buffer to process
2574 * If there is no free memory -ENOMEM is returned, otherwise zero
2575 * is returned and the old skb data released.
2577 static inline int skb_linearize_cow(struct sk_buff *skb)
2579 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2580 __skb_linearize(skb) : 0;
2584 * skb_postpull_rcsum - update checksum for received skb after pull
2585 * @skb: buffer to update
2586 * @start: start of data before pull
2587 * @len: length of data pulled
2589 * After doing a pull on a received packet, you need to call this to
2590 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2591 * CHECKSUM_NONE so that it can be recomputed from scratch.
2594 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2595 const void *start, unsigned int len)
2597 if (skb->ip_summed == CHECKSUM_COMPLETE)
2598 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2601 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2604 * pskb_trim_rcsum - trim received skb and update checksum
2605 * @skb: buffer to trim
2608 * This is exactly the same as pskb_trim except that it ensures the
2609 * checksum of received packets are still valid after the operation.
2612 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2614 if (likely(len >= skb->len))
2616 if (skb->ip_summed == CHECKSUM_COMPLETE)
2617 skb->ip_summed = CHECKSUM_NONE;
2618 return __pskb_trim(skb, len);
2621 #define skb_queue_walk(queue, skb) \
2622 for (skb = (queue)->next; \
2623 skb != (struct sk_buff *)(queue); \
2626 #define skb_queue_walk_safe(queue, skb, tmp) \
2627 for (skb = (queue)->next, tmp = skb->next; \
2628 skb != (struct sk_buff *)(queue); \
2629 skb = tmp, tmp = skb->next)
2631 #define skb_queue_walk_from(queue, skb) \
2632 for (; skb != (struct sk_buff *)(queue); \
2635 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2636 for (tmp = skb->next; \
2637 skb != (struct sk_buff *)(queue); \
2638 skb = tmp, tmp = skb->next)
2640 #define skb_queue_reverse_walk(queue, skb) \
2641 for (skb = (queue)->prev; \
2642 skb != (struct sk_buff *)(queue); \
2645 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2646 for (skb = (queue)->prev, tmp = skb->prev; \
2647 skb != (struct sk_buff *)(queue); \
2648 skb = tmp, tmp = skb->prev)
2650 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2651 for (tmp = skb->prev; \
2652 skb != (struct sk_buff *)(queue); \
2653 skb = tmp, tmp = skb->prev)
2655 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2657 return skb_shinfo(skb)->frag_list != NULL;
2660 static inline void skb_frag_list_init(struct sk_buff *skb)
2662 skb_shinfo(skb)->frag_list = NULL;
2665 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2667 frag->next = skb_shinfo(skb)->frag_list;
2668 skb_shinfo(skb)->frag_list = frag;
2671 #define skb_walk_frags(skb, iter) \
2672 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2674 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2675 int *peeked, int *off, int *err);
2676 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2678 unsigned int datagram_poll(struct file *file, struct socket *sock,
2679 struct poll_table_struct *wait);
2680 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2681 struct iov_iter *to, int size);
2682 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2683 struct msghdr *msg, int size)
2685 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
2687 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2688 struct msghdr *msg);
2689 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2690 struct iov_iter *from, int len);
2691 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
2692 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2693 void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2694 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
2695 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2696 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2697 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2698 int len, __wsum csum);
2699 int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2700 struct pipe_inode_info *pipe, unsigned int len,
2701 unsigned int flags);
2702 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2703 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2704 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2706 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2707 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2708 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
2709 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
2710 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
2711 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
2712 int skb_ensure_writable(struct sk_buff *skb, int write_len);
2713 int skb_vlan_pop(struct sk_buff *skb);
2714 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
2716 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2718 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
2721 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2723 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
2726 struct skb_checksum_ops {
2727 __wsum (*update)(const void *mem, int len, __wsum wsum);
2728 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2731 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2732 __wsum csum, const struct skb_checksum_ops *ops);
2733 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2736 static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2737 int len, void *data, int hlen, void *buffer)
2739 if (hlen - offset >= len)
2740 return data + offset;
2743 skb_copy_bits(skb, offset, buffer, len) < 0)
2749 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2750 int len, void *buffer)
2752 return __skb_header_pointer(skb, offset, len, skb->data,
2753 skb_headlen(skb), buffer);
2757 * skb_needs_linearize - check if we need to linearize a given skb
2758 * depending on the given device features.
2759 * @skb: socket buffer to check
2760 * @features: net device features
2762 * Returns true if either:
2763 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2764 * 2. skb is fragmented and the device does not support SG.
2766 static inline bool skb_needs_linearize(struct sk_buff *skb,
2767 netdev_features_t features)
2769 return skb_is_nonlinear(skb) &&
2770 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2771 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2774 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2776 const unsigned int len)
2778 memcpy(to, skb->data, len);
2781 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2782 const int offset, void *to,
2783 const unsigned int len)
2785 memcpy(to, skb->data + offset, len);
2788 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2790 const unsigned int len)
2792 memcpy(skb->data, from, len);
2795 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2798 const unsigned int len)
2800 memcpy(skb->data + offset, from, len);
2803 void skb_init(void);
2805 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2811 * skb_get_timestamp - get timestamp from a skb
2812 * @skb: skb to get stamp from
2813 * @stamp: pointer to struct timeval to store stamp in
2815 * Timestamps are stored in the skb as offsets to a base timestamp.
2816 * This function converts the offset back to a struct timeval and stores
2819 static inline void skb_get_timestamp(const struct sk_buff *skb,
2820 struct timeval *stamp)
2822 *stamp = ktime_to_timeval(skb->tstamp);
2825 static inline void skb_get_timestampns(const struct sk_buff *skb,
2826 struct timespec *stamp)
2828 *stamp = ktime_to_timespec(skb->tstamp);
2831 static inline void __net_timestamp(struct sk_buff *skb)
2833 skb->tstamp = ktime_get_real();
2836 static inline ktime_t net_timedelta(ktime_t t)
2838 return ktime_sub(ktime_get_real(), t);
2841 static inline ktime_t net_invalid_timestamp(void)
2843 return ktime_set(0, 0);
2846 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2848 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2850 void skb_clone_tx_timestamp(struct sk_buff *skb);
2851 bool skb_defer_rx_timestamp(struct sk_buff *skb);
2853 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2855 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2859 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2864 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2867 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2869 * PHY drivers may accept clones of transmitted packets for
2870 * timestamping via their phy_driver.txtstamp method. These drivers
2871 * must call this function to return the skb back to the stack, with
2872 * or without a timestamp.
2874 * @skb: clone of the the original outgoing packet
2875 * @hwtstamps: hardware time stamps, may be NULL if not available
2878 void skb_complete_tx_timestamp(struct sk_buff *skb,
2879 struct skb_shared_hwtstamps *hwtstamps);
2881 void __skb_tstamp_tx(struct sk_buff *orig_skb,
2882 struct skb_shared_hwtstamps *hwtstamps,
2883 struct sock *sk, int tstype);
2886 * skb_tstamp_tx - queue clone of skb with send time stamps
2887 * @orig_skb: the original outgoing packet
2888 * @hwtstamps: hardware time stamps, may be NULL if not available
2890 * If the skb has a socket associated, then this function clones the
2891 * skb (thus sharing the actual data and optional structures), stores
2892 * the optional hardware time stamping information (if non NULL) or
2893 * generates a software time stamp (otherwise), then queues the clone
2894 * to the error queue of the socket. Errors are silently ignored.
2896 void skb_tstamp_tx(struct sk_buff *orig_skb,
2897 struct skb_shared_hwtstamps *hwtstamps);
2899 static inline void sw_tx_timestamp(struct sk_buff *skb)
2901 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2902 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2903 skb_tstamp_tx(skb, NULL);
2907 * skb_tx_timestamp() - Driver hook for transmit timestamping
2909 * Ethernet MAC Drivers should call this function in their hard_xmit()
2910 * function immediately before giving the sk_buff to the MAC hardware.
2912 * Specifically, one should make absolutely sure that this function is
2913 * called before TX completion of this packet can trigger. Otherwise
2914 * the packet could potentially already be freed.
2916 * @skb: A socket buffer.
2918 static inline void skb_tx_timestamp(struct sk_buff *skb)
2920 skb_clone_tx_timestamp(skb);
2921 sw_tx_timestamp(skb);
2925 * skb_complete_wifi_ack - deliver skb with wifi status
2927 * @skb: the original outgoing packet
2928 * @acked: ack status
2931 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2933 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2934 __sum16 __skb_checksum_complete(struct sk_buff *skb);
2936 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2938 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
2940 (skb->ip_summed == CHECKSUM_PARTIAL &&
2941 skb_checksum_start_offset(skb) >= 0));
2945 * skb_checksum_complete - Calculate checksum of an entire packet
2946 * @skb: packet to process
2948 * This function calculates the checksum over the entire packet plus
2949 * the value of skb->csum. The latter can be used to supply the
2950 * checksum of a pseudo header as used by TCP/UDP. It returns the
2953 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2954 * this function can be used to verify that checksum on received
2955 * packets. In that case the function should return zero if the
2956 * checksum is correct. In particular, this function will return zero
2957 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2958 * hardware has already verified the correctness of the checksum.
2960 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2962 return skb_csum_unnecessary(skb) ?
2963 0 : __skb_checksum_complete(skb);
2966 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2968 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2969 if (skb->csum_level == 0)
2970 skb->ip_summed = CHECKSUM_NONE;
2976 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2978 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2979 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2981 } else if (skb->ip_summed == CHECKSUM_NONE) {
2982 skb->ip_summed = CHECKSUM_UNNECESSARY;
2983 skb->csum_level = 0;
2987 static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2989 /* Mark current checksum as bad (typically called from GRO
2990 * path). In the case that ip_summed is CHECKSUM_NONE
2991 * this must be the first checksum encountered in the packet.
2992 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2993 * checksum after the last one validated. For UDP, a zero
2994 * checksum can not be marked as bad.
2997 if (skb->ip_summed == CHECKSUM_NONE ||
2998 skb->ip_summed == CHECKSUM_UNNECESSARY)
3002 /* Check if we need to perform checksum complete validation.
3004 * Returns true if checksum complete is needed, false otherwise
3005 * (either checksum is unnecessary or zero checksum is allowed).
3007 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3011 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3012 skb->csum_valid = 1;
3013 __skb_decr_checksum_unnecessary(skb);
3020 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3023 #define CHECKSUM_BREAK 76
3025 /* Unset checksum-complete
3027 * Unset checksum complete can be done when packet is being modified
3028 * (uncompressed for instance) and checksum-complete value is
3031 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3033 if (skb->ip_summed == CHECKSUM_COMPLETE)
3034 skb->ip_summed = CHECKSUM_NONE;
3037 /* Validate (init) checksum based on checksum complete.
3040 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3041 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3042 * checksum is stored in skb->csum for use in __skb_checksum_complete
3043 * non-zero: value of invalid checksum
3046 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3050 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3051 if (!csum_fold(csum_add(psum, skb->csum))) {
3052 skb->csum_valid = 1;
3055 } else if (skb->csum_bad) {
3056 /* ip_summed == CHECKSUM_NONE in this case */
3057 return (__force __sum16)1;
3062 if (complete || skb->len <= CHECKSUM_BREAK) {
3065 csum = __skb_checksum_complete(skb);
3066 skb->csum_valid = !csum;
3073 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3078 /* Perform checksum validate (init). Note that this is a macro since we only
3079 * want to calculate the pseudo header which is an input function if necessary.
3080 * First we try to validate without any computation (checksum unnecessary) and
3081 * then calculate based on checksum complete calling the function to compute
3085 * 0: checksum is validated or try to in skb_checksum_complete
3086 * non-zero: value of invalid checksum
3088 #define __skb_checksum_validate(skb, proto, complete, \
3089 zero_okay, check, compute_pseudo) \
3091 __sum16 __ret = 0; \
3092 skb->csum_valid = 0; \
3093 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3094 __ret = __skb_checksum_validate_complete(skb, \
3095 complete, compute_pseudo(skb, proto)); \
3099 #define skb_checksum_init(skb, proto, compute_pseudo) \
3100 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3102 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3103 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3105 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3106 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3108 #define skb_checksum_validate_zero_check(skb, proto, check, \
3110 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3112 #define skb_checksum_simple_validate(skb) \
3113 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3115 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3117 return (skb->ip_summed == CHECKSUM_NONE &&
3118 skb->csum_valid && !skb->csum_bad);
3121 static inline void __skb_checksum_convert(struct sk_buff *skb,
3122 __sum16 check, __wsum pseudo)
3124 skb->csum = ~pseudo;
3125 skb->ip_summed = CHECKSUM_COMPLETE;
3128 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3130 if (__skb_checksum_convert_check(skb)) \
3131 __skb_checksum_convert(skb, check, \
3132 compute_pseudo(skb, proto)); \
3135 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3136 u16 start, u16 offset)
3138 skb->ip_summed = CHECKSUM_PARTIAL;
3139 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3140 skb->csum_offset = offset - start;
3143 /* Update skbuf and packet to reflect the remote checksum offload operation.
3144 * When called, ptr indicates the starting point for skb->csum when
3145 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3146 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3148 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
3149 int start, int offset, bool nopartial)
3154 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3158 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3159 __skb_checksum_complete(skb);
3160 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3163 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3165 /* Adjust skb->csum since we changed the packet */
3166 skb->csum = csum_add(skb->csum, delta);
3169 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3170 void nf_conntrack_destroy(struct nf_conntrack *nfct);
3171 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3173 if (nfct && atomic_dec_and_test(&nfct->use))
3174 nf_conntrack_destroy(nfct);
3176 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3179 atomic_inc(&nfct->use);
3182 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3183 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3185 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3188 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3191 atomic_inc(&nf_bridge->use);
3193 #endif /* CONFIG_BRIDGE_NETFILTER */
3194 static inline void nf_reset(struct sk_buff *skb)
3196 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3197 nf_conntrack_put(skb->nfct);
3200 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3201 nf_bridge_put(skb->nf_bridge);
3202 skb->nf_bridge = NULL;
3206 static inline void nf_reset_trace(struct sk_buff *skb)
3208 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3213 /* Note: This doesn't put any conntrack and bridge info in dst. */
3214 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3217 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3218 dst->nfct = src->nfct;
3219 nf_conntrack_get(src->nfct);
3221 dst->nfctinfo = src->nfctinfo;
3223 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3224 dst->nf_bridge = src->nf_bridge;
3225 nf_bridge_get(src->nf_bridge);
3227 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3229 dst->nf_trace = src->nf_trace;
3233 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3235 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3236 nf_conntrack_put(dst->nfct);
3238 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3239 nf_bridge_put(dst->nf_bridge);
3241 __nf_copy(dst, src, true);
3244 #ifdef CONFIG_NETWORK_SECMARK
3245 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3247 to->secmark = from->secmark;
3250 static inline void skb_init_secmark(struct sk_buff *skb)
3255 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3258 static inline void skb_init_secmark(struct sk_buff *skb)
3262 static inline bool skb_irq_freeable(const struct sk_buff *skb)
3264 return !skb->destructor &&
3265 #if IS_ENABLED(CONFIG_XFRM)
3268 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3271 !skb->_skb_refdst &&
3272 !skb_has_frag_list(skb);
3275 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3277 skb->queue_mapping = queue_mapping;
3280 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
3282 return skb->queue_mapping;
3285 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3287 to->queue_mapping = from->queue_mapping;
3290 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3292 skb->queue_mapping = rx_queue + 1;
3295 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
3297 return skb->queue_mapping - 1;
3300 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
3302 return skb->queue_mapping != 0;
3305 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3314 /* Keeps track of mac header offset relative to skb->head.
3315 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3316 * For non-tunnel skb it points to skb_mac_header() and for
3317 * tunnel skb it points to outer mac header.
3318 * Keeps track of level of encapsulation of network headers.
3325 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3327 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3329 return (skb_mac_header(inner_skb) - inner_skb->head) -
3330 SKB_GSO_CB(inner_skb)->mac_offset;
3333 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3335 int new_headroom, headroom;
3338 headroom = skb_headroom(skb);
3339 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3343 new_headroom = skb_headroom(skb);
3344 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3348 /* Compute the checksum for a gso segment. First compute the checksum value
3349 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3350 * then add in skb->csum (checksum from csum_start to end of packet).
3351 * skb->csum and csum_start are then updated to reflect the checksum of the
3352 * resultant packet starting from the transport header-- the resultant checksum
3353 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3356 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3358 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3359 skb_transport_offset(skb);
3362 partial = csum_partial(skb_transport_header(skb), plen, skb->csum);
3364 SKB_GSO_CB(skb)->csum_start -= plen;
3366 return csum_fold(partial);
3369 static inline bool skb_is_gso(const struct sk_buff *skb)
3371 return skb_shinfo(skb)->gso_size;
3374 /* Note: Should be called only if skb_is_gso(skb) is true */
3375 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
3377 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3380 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
3382 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3384 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3385 * wanted then gso_type will be set. */
3386 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3388 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3389 unlikely(shinfo->gso_type == 0)) {
3390 __skb_warn_lro_forwarding(skb);
3396 static inline void skb_forward_csum(struct sk_buff *skb)
3398 /* Unfortunately we don't support this one. Any brave souls? */
3399 if (skb->ip_summed == CHECKSUM_COMPLETE)
3400 skb->ip_summed = CHECKSUM_NONE;
3404 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3405 * @skb: skb to check
3407 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3408 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3409 * use this helper, to document places where we make this assertion.
3411 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
3414 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3418 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
3420 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3421 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3422 unsigned int transport_len,
3423 __sum16(*skb_chkf)(struct sk_buff *skb));
3426 * skb_head_is_locked - Determine if the skb->head is locked down
3427 * @skb: skb to check
3429 * The head on skbs build around a head frag can be removed if they are
3430 * not cloned. This function returns true if the skb head is locked down
3431 * due to either being allocated via kmalloc, or by being a clone with
3432 * multiple references to the head.
3434 static inline bool skb_head_is_locked(const struct sk_buff *skb)
3436 return !skb->head_frag || skb_cloned(skb);
3440 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3444 * skb_gso_network_seglen is used to determine the real size of the
3445 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3447 * The MAC/L2 header is not accounted for.
3449 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3451 unsigned int hdr_len = skb_transport_header(skb) -
3452 skb_network_header(skb);
3453 return hdr_len + skb_gso_transport_seglen(skb);
3455 #endif /* __KERNEL__ */
3456 #endif /* _LINUX_SKBUFF_H */