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/compiler.h>
19 #include <linux/time.h>
20 #include <linux/bug.h>
21 #include <linux/cache.h>
22 #include <linux/rbtree.h>
23 #include <linux/socket.h>
24 #include <linux/refcount.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 <linux/sched/clock.h>
38 #include <net/flow_dissector.h>
39 #include <linux/splice.h>
40 #include <linux/in6.h>
41 #include <linux/if_packet.h>
44 /* The interface for checksum offload between the stack and networking drivers
47 * A. IP checksum related features
49 * Drivers advertise checksum offload capabilities in the features of a device.
50 * From the stack's point of view these are capabilities offered by the driver,
51 * a driver typically only advertises features that it is capable of offloading
54 * The checksum related features are:
56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
57 * IP (one's complement) checksum for any combination
58 * of protocols or protocol layering. The checksum is
59 * computed and set in a packet per the CHECKSUM_PARTIAL
60 * interface (see below).
62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63 * TCP or UDP packets over IPv4. These are specifically
64 * unencapsulated packets of the form IPv4|TCP or
65 * IPv4|UDP where the Protocol field in the IPv4 header
66 * is TCP or UDP. The IPv4 header may contain IP options
67 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
74 * IPv4|UDP where the Next Header field in the IPv6
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
82 * This flag is used only used to disable the RX checksum
83 * feature for a device. The stack will accept receive
84 * checksum indication in packets received on a device
85 * regardless of whether NETIF_F_RXCSUM is set.
87 * B. Checksumming of received packets by device. Indication of checksum
88 * verification is in set skb->ip_summed. Possible values are:
92 * Device did not checksum this packet e.g. due to lack of capabilities.
93 * The packet contains full (though not verified) checksum in packet but
94 * not in skb->csum. Thus, skb->csum is undefined in this case.
96 * CHECKSUM_UNNECESSARY:
98 * The hardware you're dealing with doesn't calculate the full checksum
99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101 * if their checksums are okay. skb->csum is still undefined in this case
102 * though. A driver or device must never modify the checksum field in the
103 * packet even if checksum is verified.
105 * CHECKSUM_UNNECESSARY is applicable to following protocols:
106 * TCP: IPv6 and IPv4.
107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108 * zero UDP checksum for either IPv4 or IPv6, the networking stack
109 * may perform further validation in this case.
110 * GRE: only if the checksum is present in the header.
111 * SCTP: indicates the CRC in SCTP header has been validated.
112 * FCOE: indicates the CRC in FC frame has been validated.
114 * skb->csum_level indicates the number of consecutive checksums found in
115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
117 * and a device is able to verify the checksums for UDP (possibly zero),
118 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
119 * two. If the device were only able to verify the UDP checksum and not
120 * GRE, either because it doesn't support GRE checksum of because GRE
121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
122 * not considered in this case).
126 * This is the most generic way. The device supplied checksum of the _whole_
127 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
128 * hardware doesn't need to parse L3/L4 headers to implement this.
131 * - Even if device supports only some protocols, but is able to produce
132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
137 * A checksum is set up to be offloaded to a device as described in the
138 * output description for CHECKSUM_PARTIAL. This may occur on a packet
139 * received directly from another Linux OS, e.g., a virtualized Linux kernel
140 * on the same host, or it may be set in the input path in GRO or remote
141 * checksum offload. For the purposes of checksum verification, the checksum
142 * referred to by skb->csum_start + skb->csum_offset and any preceding
143 * checksums in the packet are considered verified. Any checksums in the
144 * packet that are after the checksum being offloaded are not considered to
147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
148 * in the skb->ip_summed for a packet. Values are:
152 * The driver is required to checksum the packet as seen by hard_start_xmit()
153 * from skb->csum_start up to the end, and to record/write the checksum at
154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
155 * csum_start and csum_offset values are valid values given the length and
156 * offset of the packet, however they should not attempt to validate that the
157 * checksum refers to a legitimate transport layer checksum-- it is the
158 * purview of the stack to validate that csum_start and csum_offset are set
161 * When the stack requests checksum offload for a packet, the driver MUST
162 * ensure that the checksum is set correctly. A driver can either offload the
163 * checksum calculation to the device, or call skb_checksum_help (in the case
164 * that the device does not support offload for a particular checksum).
166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
168 * checksum offload capability.
169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
170 * on network device checksumming capabilities: if a packet does not match
171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
172 * csum_not_inet, see item D.) is called to resolve the checksum.
176 * The skb was already checksummed by the protocol, or a checksum is not
179 * CHECKSUM_UNNECESSARY:
181 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
185 * Not used in checksum output. If a driver observes a packet with this value
186 * set in skbuff, if should treat as CHECKSUM_NONE being set.
188 * D. Non-IP checksum (CRC) offloads
190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191 * offloading the SCTP CRC in a packet. To perform this offload the stack
192 * will set set csum_start and csum_offset accordingly, set ip_summed to
193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195 * A driver that supports both IP checksum offload and SCTP CRC32c offload
196 * must verify which offload is configured for a packet by testing the
197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201 * offloading the FCOE CRC in a packet. To perform this offload the stack
202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
203 * accordingly. Note the there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
205 * both IP checksum offload and FCOE CRC offload must verify which offload
206 * is configured for a packet presumably by inspecting packet headers.
208 * E. Checksumming on output with GSO.
210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213 * part of the GSO operation is implied. If a checksum is being offloaded
214 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
215 * are set to refer to the outermost checksum being offload (two offloaded
216 * checksums are possible with UDP encapsulation).
219 /* Don't change this without changing skb_csum_unnecessary! */
220 #define CHECKSUM_NONE 0
221 #define CHECKSUM_UNNECESSARY 1
222 #define CHECKSUM_COMPLETE 2
223 #define CHECKSUM_PARTIAL 3
225 /* Maximum value in skb->csum_level */
226 #define SKB_MAX_CSUM_LEVEL 3
228 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
229 #define SKB_WITH_OVERHEAD(X) \
230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
231 #define SKB_MAX_ORDER(X, ORDER) \
232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
233 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
234 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
236 /* return minimum truesize of one skb containing X bytes of data */
237 #define SKB_TRUESIZE(X) ((X) + \
238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
243 struct pipe_inode_info
;
247 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
248 struct nf_conntrack
{
253 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
254 struct nf_bridge_info
{
257 BRNF_PROTO_UNCHANGED
,
265 struct net_device
*physindev
;
267 /* always valid & non-NULL from FORWARD on, for physdev match */
268 struct net_device
*physoutdev
;
270 /* prerouting: detect dnat in orig/reply direction */
272 struct in6_addr ipv6_daddr
;
274 /* after prerouting + nat detected: store original source
275 * mac since neigh resolution overwrites it, only used while
276 * skb is out in neigh layer.
278 char neigh_header
[8];
283 struct sk_buff_head
{
284 /* These two members must be first. */
285 struct sk_buff
*next
;
286 struct sk_buff
*prev
;
294 /* To allow 64K frame to be packed as single skb without frag_list we
295 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
296 * buffers which do not start on a page boundary.
298 * Since GRO uses frags we allocate at least 16 regardless of page
301 #if (65536/PAGE_SIZE + 1) < 16
302 #define MAX_SKB_FRAGS 16UL
304 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
306 extern int sysctl_max_skb_frags
;
308 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
309 * segment using its current segmentation instead.
311 #define GSO_BY_FRAGS 0xFFFF
313 typedef struct skb_frag_struct skb_frag_t
;
315 struct skb_frag_struct
{
319 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
328 static inline unsigned int skb_frag_size(const skb_frag_t
*frag
)
333 static inline void skb_frag_size_set(skb_frag_t
*frag
, unsigned int size
)
338 static inline void skb_frag_size_add(skb_frag_t
*frag
, int delta
)
343 static inline void skb_frag_size_sub(skb_frag_t
*frag
, int delta
)
348 static inline bool skb_frag_must_loop(struct page
*p
)
350 #if defined(CONFIG_HIGHMEM)
358 * skb_frag_foreach_page - loop over pages in a fragment
360 * @f: skb frag to operate on
361 * @f_off: offset from start of f->page.p
362 * @f_len: length from f_off to loop over
363 * @p: (temp var) current page
364 * @p_off: (temp var) offset from start of current page,
365 * non-zero only on first page.
366 * @p_len: (temp var) length in current page,
367 * < PAGE_SIZE only on first and last page.
368 * @copied: (temp var) length so far, excluding current p_len.
370 * A fragment can hold a compound page, in which case per-page
371 * operations, notably kmap_atomic, must be called for each
374 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
375 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
376 p_off = (f_off) & (PAGE_SIZE - 1), \
377 p_len = skb_frag_must_loop(p) ? \
378 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
381 copied += p_len, p++, p_off = 0, \
382 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
384 #define HAVE_HW_TIME_STAMP
387 * struct skb_shared_hwtstamps - hardware time stamps
388 * @hwtstamp: hardware time stamp transformed into duration
389 * since arbitrary point in time
391 * Software time stamps generated by ktime_get_real() are stored in
394 * hwtstamps can only be compared against other hwtstamps from
397 * This structure is attached to packets as part of the
398 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
400 struct skb_shared_hwtstamps
{
404 /* Definitions for tx_flags in struct skb_shared_info */
406 /* generate hardware time stamp */
407 SKBTX_HW_TSTAMP
= 1 << 0,
409 /* generate software time stamp when queueing packet to NIC */
410 SKBTX_SW_TSTAMP
= 1 << 1,
412 /* device driver is going to provide hardware time stamp */
413 SKBTX_IN_PROGRESS
= 1 << 2,
415 /* device driver supports TX zero-copy buffers */
416 SKBTX_DEV_ZEROCOPY
= 1 << 3,
418 /* generate wifi status information (where possible) */
419 SKBTX_WIFI_STATUS
= 1 << 4,
421 /* This indicates at least one fragment might be overwritten
422 * (as in vmsplice(), sendfile() ...)
423 * If we need to compute a TX checksum, we'll need to copy
424 * all frags to avoid possible bad checksum
426 SKBTX_SHARED_FRAG
= 1 << 5,
428 /* generate software time stamp when entering packet scheduling */
429 SKBTX_SCHED_TSTAMP
= 1 << 6,
432 #define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
433 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
435 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
438 * The callback notifies userspace to release buffers when skb DMA is done in
439 * lower device, the skb last reference should be 0 when calling this.
440 * The zerocopy_success argument is true if zero copy transmit occurred,
441 * false on data copy or out of memory error caused by data copy attempt.
442 * The ctx field is used to track device context.
443 * The desc field is used to track userspace buffer index.
446 void (*callback
)(struct ubuf_info
*, bool zerocopy_success
);
462 struct user_struct
*user
;
467 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
469 struct ubuf_info
*sock_zerocopy_alloc(struct sock
*sk
, size_t size
);
470 struct ubuf_info
*sock_zerocopy_realloc(struct sock
*sk
, size_t size
,
471 struct ubuf_info
*uarg
);
473 static inline void sock_zerocopy_get(struct ubuf_info
*uarg
)
475 refcount_inc(&uarg
->refcnt
);
478 void sock_zerocopy_put(struct ubuf_info
*uarg
);
479 void sock_zerocopy_put_abort(struct ubuf_info
*uarg
);
481 void sock_zerocopy_callback(struct ubuf_info
*uarg
, bool success
);
483 int skb_zerocopy_iter_stream(struct sock
*sk
, struct sk_buff
*skb
,
484 struct msghdr
*msg
, int len
,
485 struct ubuf_info
*uarg
);
487 /* This data is invariant across clones and lives at
488 * the end of the header data, ie. at skb->end.
490 struct skb_shared_info
{
495 unsigned short gso_size
;
496 /* Warning: this field is not always filled in (UFO)! */
497 unsigned short gso_segs
;
498 struct sk_buff
*frag_list
;
499 struct skb_shared_hwtstamps hwtstamps
;
500 unsigned int gso_type
;
504 * Warning : all fields before dataref are cleared in __alloc_skb()
508 /* Intermediate layers must ensure that destructor_arg
509 * remains valid until skb destructor */
510 void * destructor_arg
;
512 /* must be last field, see pskb_expand_head() */
513 skb_frag_t frags
[MAX_SKB_FRAGS
];
516 /* We divide dataref into two halves. The higher 16 bits hold references
517 * to the payload part of skb->data. The lower 16 bits hold references to
518 * the entire skb->data. A clone of a headerless skb holds the length of
519 * the header in skb->hdr_len.
521 * All users must obey the rule that the skb->data reference count must be
522 * greater than or equal to the payload reference count.
524 * Holding a reference to the payload part means that the user does not
525 * care about modifications to the header part of skb->data.
527 #define SKB_DATAREF_SHIFT 16
528 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
532 SKB_FCLONE_UNAVAILABLE
, /* skb has no fclone (from head_cache) */
533 SKB_FCLONE_ORIG
, /* orig skb (from fclone_cache) */
534 SKB_FCLONE_CLONE
, /* companion fclone skb (from fclone_cache) */
538 SKB_GSO_TCPV4
= 1 << 0,
540 /* This indicates the skb is from an untrusted source. */
541 SKB_GSO_DODGY
= 1 << 1,
543 /* This indicates the tcp segment has CWR set. */
544 SKB_GSO_TCP_ECN
= 1 << 2,
546 SKB_GSO_TCP_FIXEDID
= 1 << 3,
548 SKB_GSO_TCPV6
= 1 << 4,
550 SKB_GSO_FCOE
= 1 << 5,
552 SKB_GSO_GRE
= 1 << 6,
554 SKB_GSO_GRE_CSUM
= 1 << 7,
556 SKB_GSO_IPXIP4
= 1 << 8,
558 SKB_GSO_IPXIP6
= 1 << 9,
560 SKB_GSO_UDP_TUNNEL
= 1 << 10,
562 SKB_GSO_UDP_TUNNEL_CSUM
= 1 << 11,
564 SKB_GSO_PARTIAL
= 1 << 12,
566 SKB_GSO_TUNNEL_REMCSUM
= 1 << 13,
568 SKB_GSO_SCTP
= 1 << 14,
570 SKB_GSO_ESP
= 1 << 15,
572 SKB_GSO_UDP
= 1 << 16,
575 #if BITS_PER_LONG > 32
576 #define NET_SKBUFF_DATA_USES_OFFSET 1
579 #ifdef NET_SKBUFF_DATA_USES_OFFSET
580 typedef unsigned int sk_buff_data_t
;
582 typedef unsigned char *sk_buff_data_t
;
586 * struct sk_buff - socket buffer
587 * @next: Next buffer in list
588 * @prev: Previous buffer in list
589 * @tstamp: Time we arrived/left
590 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
591 * @sk: Socket we are owned by
592 * @dev: Device we arrived on/are leaving by
593 * @cb: Control buffer. Free for use by every layer. Put private vars here
594 * @_skb_refdst: destination entry (with norefcount bit)
595 * @sp: the security path, used for xfrm
596 * @len: Length of actual data
597 * @data_len: Data length
598 * @mac_len: Length of link layer header
599 * @hdr_len: writable header length of cloned skb
600 * @csum: Checksum (must include start/offset pair)
601 * @csum_start: Offset from skb->head where checksumming should start
602 * @csum_offset: Offset from csum_start where checksum should be stored
603 * @priority: Packet queueing priority
604 * @ignore_df: allow local fragmentation
605 * @cloned: Head may be cloned (check refcnt to be sure)
606 * @ip_summed: Driver fed us an IP checksum
607 * @nohdr: Payload reference only, must not modify header
608 * @pkt_type: Packet class
609 * @fclone: skbuff clone status
610 * @ipvs_property: skbuff is owned by ipvs
611 * @tc_skip_classify: do not classify packet. set by IFB device
612 * @tc_at_ingress: used within tc_classify to distinguish in/egress
613 * @tc_redirected: packet was redirected by a tc action
614 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
615 * @peeked: this packet has been seen already, so stats have been
616 * done for it, don't do them again
617 * @nf_trace: netfilter packet trace flag
618 * @protocol: Packet protocol from driver
619 * @destructor: Destruct function
620 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
621 * @_nfct: Associated connection, if any (with nfctinfo bits)
622 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
623 * @skb_iif: ifindex of device we arrived on
624 * @tc_index: Traffic control index
625 * @hash: the packet hash
626 * @queue_mapping: Queue mapping for multiqueue devices
627 * @xmit_more: More SKBs are pending for this queue
628 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
629 * @ndisc_nodetype: router type (from link layer)
630 * @ooo_okay: allow the mapping of a socket to a queue to be changed
631 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
633 * @sw_hash: indicates hash was computed in software stack
634 * @wifi_acked_valid: wifi_acked was set
635 * @wifi_acked: whether frame was acked on wifi or not
636 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
637 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
638 * @dst_pending_confirm: need to confirm neighbour
639 * @napi_id: id of the NAPI struct this skb came from
640 * @secmark: security marking
641 * @mark: Generic packet mark
642 * @vlan_proto: vlan encapsulation protocol
643 * @vlan_tci: vlan tag control information
644 * @inner_protocol: Protocol (encapsulation)
645 * @inner_transport_header: Inner transport layer header (encapsulation)
646 * @inner_network_header: Network layer header (encapsulation)
647 * @inner_mac_header: Link layer header (encapsulation)
648 * @transport_header: Transport layer header
649 * @network_header: Network layer header
650 * @mac_header: Link layer header
651 * @tail: Tail pointer
653 * @head: Head of buffer
654 * @data: Data head pointer
655 * @truesize: Buffer size
656 * @users: User count - see {datagram,tcp}.c
662 /* These two members must be first. */
663 struct sk_buff
*next
;
664 struct sk_buff
*prev
;
667 struct net_device
*dev
;
668 /* Some protocols might use this space to store information,
669 * while device pointer would be NULL.
670 * UDP receive path is one user.
672 unsigned long dev_scratch
;
675 struct rb_node rbnode
; /* used in netem, ip4 defrag, and tcp stack */
676 struct list_head list
;
681 int ip_defrag_offset
;
689 * This is the control buffer. It is free to use for every
690 * layer. Please put your private variables there. If you
691 * want to keep them across layers you have to do a skb_clone()
692 * first. This is owned by whoever has the skb queued ATM.
694 char cb
[48] __aligned(8);
698 unsigned long _skb_refdst
;
699 void (*destructor
)(struct sk_buff
*skb
);
701 struct list_head tcp_tsorted_anchor
;
707 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
710 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
711 struct nf_bridge_info
*nf_bridge
;
718 /* Following fields are _not_ copied in __copy_skb_header()
719 * Note that queue_mapping is here mostly to fill a hole.
723 /* if you move cloned around you also must adapt those constants */
724 #ifdef __BIG_ENDIAN_BITFIELD
725 #define CLONED_MASK (1 << 7)
727 #define CLONED_MASK 1
729 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
731 __u8 __cloned_offset
[0];
740 /* fields enclosed in headers_start/headers_end are copied
741 * using a single memcpy() in __copy_skb_header()
744 __u32 headers_start
[0];
747 /* if you move pkt_type around you also must adapt those constants */
748 #ifdef __BIG_ENDIAN_BITFIELD
749 #define PKT_TYPE_MAX (7 << 5)
751 #define PKT_TYPE_MAX 7
753 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
755 __u8 __pkt_type_offset
[0];
764 __u8 wifi_acked_valid
:1;
767 /* Indicates the inner headers are valid in the skbuff. */
768 __u8 encapsulation
:1;
769 __u8 encap_hdr_csum
:1;
772 __u8 csum_complete_sw
:1;
774 __u8 csum_not_inet
:1;
775 __u8 dst_pending_confirm
:1;
776 #ifdef CONFIG_IPV6_NDISC_NODETYPE
777 __u8 ndisc_nodetype
:2;
779 __u8 ipvs_property
:1;
781 __u8 inner_protocol_type
:1;
782 __u8 remcsum_offload
:1;
783 #ifdef CONFIG_NET_SWITCHDEV
784 __u8 offload_fwd_mark
:1;
785 __u8 offload_mr_fwd_mark
:1;
787 #ifdef CONFIG_NET_CLS_ACT
788 __u8 tc_skip_classify
:1;
789 __u8 tc_at_ingress
:1;
790 __u8 tc_redirected
:1;
791 __u8 tc_from_ingress
:1;
794 #ifdef CONFIG_NET_SCHED
795 __u16 tc_index
; /* traffic control index */
810 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
812 unsigned int napi_id
;
813 unsigned int sender_cpu
;
816 #ifdef CONFIG_NETWORK_SECMARK
822 __u32 reserved_tailroom
;
826 __be16 inner_protocol
;
830 __u16 inner_transport_header
;
831 __u16 inner_network_header
;
832 __u16 inner_mac_header
;
835 __u16 transport_header
;
836 __u16 network_header
;
840 __u32 headers_end
[0];
843 /* These elements must be at the end, see alloc_skb() for details. */
848 unsigned int truesize
;
854 * Handling routines are only of interest to the kernel
856 #include <linux/slab.h>
859 #define SKB_ALLOC_FCLONE 0x01
860 #define SKB_ALLOC_RX 0x02
861 #define SKB_ALLOC_NAPI 0x04
863 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
864 static inline bool skb_pfmemalloc(const struct sk_buff
*skb
)
866 return unlikely(skb
->pfmemalloc
);
870 * skb might have a dst pointer attached, refcounted or not.
871 * _skb_refdst low order bit is set if refcount was _not_ taken
873 #define SKB_DST_NOREF 1UL
874 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
876 #define SKB_NFCT_PTRMASK ~(7UL)
878 * skb_dst - returns skb dst_entry
881 * Returns skb dst_entry, regardless of reference taken or not.
883 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
885 /* If refdst was not refcounted, check we still are in a
886 * rcu_read_lock section
888 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
889 !rcu_read_lock_held() &&
890 !rcu_read_lock_bh_held());
891 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
895 * skb_dst_set - sets skb dst
899 * Sets skb dst, assuming a reference was taken on dst and should
900 * be released by skb_dst_drop()
902 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
904 skb
->_skb_refdst
= (unsigned long)dst
;
908 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
912 * Sets skb dst, assuming a reference was not taken on dst.
913 * If dst entry is cached, we do not take reference and dst_release
914 * will be avoided by refdst_drop. If dst entry is not cached, we take
915 * reference, so that last dst_release can destroy the dst immediately.
917 static inline void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
)
919 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
920 skb
->_skb_refdst
= (unsigned long)dst
| SKB_DST_NOREF
;
924 * skb_dst_is_noref - Test if skb dst isn't refcounted
927 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
929 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
932 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
934 return (struct rtable
*)skb_dst(skb
);
937 /* For mangling skb->pkt_type from user space side from applications
938 * such as nft, tc, etc, we only allow a conservative subset of
939 * possible pkt_types to be set.
941 static inline bool skb_pkt_type_ok(u32 ptype
)
943 return ptype
<= PACKET_OTHERHOST
;
946 static inline unsigned int skb_napi_id(const struct sk_buff
*skb
)
948 #ifdef CONFIG_NET_RX_BUSY_POLL
955 /* decrement the reference count and return true if we can free the skb */
956 static inline bool skb_unref(struct sk_buff
*skb
)
960 if (likely(refcount_read(&skb
->users
) == 1))
962 else if (likely(!refcount_dec_and_test(&skb
->users
)))
968 void skb_release_head_state(struct sk_buff
*skb
);
969 void kfree_skb(struct sk_buff
*skb
);
970 void kfree_skb_list(struct sk_buff
*segs
);
971 void skb_tx_error(struct sk_buff
*skb
);
972 void consume_skb(struct sk_buff
*skb
);
973 void __consume_stateless_skb(struct sk_buff
*skb
);
974 void __kfree_skb(struct sk_buff
*skb
);
975 extern struct kmem_cache
*skbuff_head_cache
;
977 void kfree_skb_partial(struct sk_buff
*skb
, bool head_stolen
);
978 bool skb_try_coalesce(struct sk_buff
*to
, struct sk_buff
*from
,
979 bool *fragstolen
, int *delta_truesize
);
981 struct sk_buff
*__alloc_skb(unsigned int size
, gfp_t priority
, int flags
,
983 struct sk_buff
*__build_skb(void *data
, unsigned int frag_size
);
984 struct sk_buff
*build_skb(void *data
, unsigned int frag_size
);
985 static inline struct sk_buff
*alloc_skb(unsigned int size
,
988 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
991 struct sk_buff
*alloc_skb_with_frags(unsigned long header_len
,
992 unsigned long data_len
,
997 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
998 struct sk_buff_fclones
{
1001 struct sk_buff skb2
;
1003 refcount_t fclone_ref
;
1007 * skb_fclone_busy - check if fclone is busy
1011 * Returns true if skb is a fast clone, and its clone is not freed.
1012 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1013 * so we also check that this didnt happen.
1015 static inline bool skb_fclone_busy(const struct sock
*sk
,
1016 const struct sk_buff
*skb
)
1018 const struct sk_buff_fclones
*fclones
;
1020 fclones
= container_of(skb
, struct sk_buff_fclones
, skb1
);
1022 return skb
->fclone
== SKB_FCLONE_ORIG
&&
1023 refcount_read(&fclones
->fclone_ref
) > 1 &&
1024 fclones
->skb2
.sk
== sk
;
1027 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
1030 return __alloc_skb(size
, priority
, SKB_ALLOC_FCLONE
, NUMA_NO_NODE
);
1033 struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
1034 int skb_copy_ubufs(struct sk_buff
*skb
, gfp_t gfp_mask
);
1035 struct sk_buff
*skb_clone(struct sk_buff
*skb
, gfp_t priority
);
1036 struct sk_buff
*skb_copy(const struct sk_buff
*skb
, gfp_t priority
);
1037 struct sk_buff
*__pskb_copy_fclone(struct sk_buff
*skb
, int headroom
,
1038 gfp_t gfp_mask
, bool fclone
);
1039 static inline struct sk_buff
*__pskb_copy(struct sk_buff
*skb
, int headroom
,
1042 return __pskb_copy_fclone(skb
, headroom
, gfp_mask
, false);
1045 int pskb_expand_head(struct sk_buff
*skb
, int nhead
, int ntail
, gfp_t gfp_mask
);
1046 struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
1047 unsigned int headroom
);
1048 struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
, int newheadroom
,
1049 int newtailroom
, gfp_t priority
);
1050 int __must_check
skb_to_sgvec_nomark(struct sk_buff
*skb
, struct scatterlist
*sg
,
1051 int offset
, int len
);
1052 int __must_check
skb_to_sgvec(struct sk_buff
*skb
, struct scatterlist
*sg
,
1053 int offset
, int len
);
1054 int skb_cow_data(struct sk_buff
*skb
, int tailbits
, struct sk_buff
**trailer
);
1055 int __skb_pad(struct sk_buff
*skb
, int pad
, bool free_on_error
);
1058 * skb_pad - zero pad the tail of an skb
1059 * @skb: buffer to pad
1060 * @pad: space to pad
1062 * Ensure that a buffer is followed by a padding area that is zero
1063 * filled. Used by network drivers which may DMA or transfer data
1064 * beyond the buffer end onto the wire.
1066 * May return error in out of memory cases. The skb is freed on error.
1068 static inline int skb_pad(struct sk_buff
*skb
, int pad
)
1070 return __skb_pad(skb
, pad
, true);
1072 #define dev_kfree_skb(a) consume_skb(a)
1074 int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
1075 int getfrag(void *from
, char *to
, int offset
,
1076 int len
, int odd
, struct sk_buff
*skb
),
1077 void *from
, int length
);
1079 int skb_append_pagefrags(struct sk_buff
*skb
, struct page
*page
,
1080 int offset
, size_t size
);
1082 struct skb_seq_state
{
1086 __u32 stepped_offset
;
1087 struct sk_buff
*root_skb
;
1088 struct sk_buff
*cur_skb
;
1092 void skb_prepare_seq_read(struct sk_buff
*skb
, unsigned int from
,
1093 unsigned int to
, struct skb_seq_state
*st
);
1094 unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
1095 struct skb_seq_state
*st
);
1096 void skb_abort_seq_read(struct skb_seq_state
*st
);
1098 unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
1099 unsigned int to
, struct ts_config
*config
);
1102 * Packet hash types specify the type of hash in skb_set_hash.
1104 * Hash types refer to the protocol layer addresses which are used to
1105 * construct a packet's hash. The hashes are used to differentiate or identify
1106 * flows of the protocol layer for the hash type. Hash types are either
1107 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1109 * Properties of hashes:
1111 * 1) Two packets in different flows have different hash values
1112 * 2) Two packets in the same flow should have the same hash value
1114 * A hash at a higher layer is considered to be more specific. A driver should
1115 * set the most specific hash possible.
1117 * A driver cannot indicate a more specific hash than the layer at which a hash
1118 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1120 * A driver may indicate a hash level which is less specific than the
1121 * actual layer the hash was computed on. For instance, a hash computed
1122 * at L4 may be considered an L3 hash. This should only be done if the
1123 * driver can't unambiguously determine that the HW computed the hash at
1124 * the higher layer. Note that the "should" in the second property above
1127 enum pkt_hash_types
{
1128 PKT_HASH_TYPE_NONE
, /* Undefined type */
1129 PKT_HASH_TYPE_L2
, /* Input: src_MAC, dest_MAC */
1130 PKT_HASH_TYPE_L3
, /* Input: src_IP, dst_IP */
1131 PKT_HASH_TYPE_L4
, /* Input: src_IP, dst_IP, src_port, dst_port */
1134 static inline void skb_clear_hash(struct sk_buff
*skb
)
1141 static inline void skb_clear_hash_if_not_l4(struct sk_buff
*skb
)
1144 skb_clear_hash(skb
);
1148 __skb_set_hash(struct sk_buff
*skb
, __u32 hash
, bool is_sw
, bool is_l4
)
1150 skb
->l4_hash
= is_l4
;
1151 skb
->sw_hash
= is_sw
;
1156 skb_set_hash(struct sk_buff
*skb
, __u32 hash
, enum pkt_hash_types type
)
1158 /* Used by drivers to set hash from HW */
1159 __skb_set_hash(skb
, hash
, false, type
== PKT_HASH_TYPE_L4
);
1163 __skb_set_sw_hash(struct sk_buff
*skb
, __u32 hash
, bool is_l4
)
1165 __skb_set_hash(skb
, hash
, true, is_l4
);
1168 void __skb_get_hash(struct sk_buff
*skb
);
1169 u32
__skb_get_hash_symmetric(const struct sk_buff
*skb
);
1170 u32
skb_get_poff(const struct sk_buff
*skb
);
1171 u32
__skb_get_poff(const struct sk_buff
*skb
, void *data
,
1172 const struct flow_keys
*keys
, int hlen
);
1173 __be32
__skb_flow_get_ports(const struct sk_buff
*skb
, int thoff
, u8 ip_proto
,
1174 void *data
, int hlen_proto
);
1176 static inline __be32
skb_flow_get_ports(const struct sk_buff
*skb
,
1177 int thoff
, u8 ip_proto
)
1179 return __skb_flow_get_ports(skb
, thoff
, ip_proto
, NULL
, 0);
1182 void skb_flow_dissector_init(struct flow_dissector
*flow_dissector
,
1183 const struct flow_dissector_key
*key
,
1184 unsigned int key_count
);
1186 bool __skb_flow_dissect(const struct sk_buff
*skb
,
1187 struct flow_dissector
*flow_dissector
,
1188 void *target_container
,
1189 void *data
, __be16 proto
, int nhoff
, int hlen
,
1190 unsigned int flags
);
1192 static inline bool skb_flow_dissect(const struct sk_buff
*skb
,
1193 struct flow_dissector
*flow_dissector
,
1194 void *target_container
, unsigned int flags
)
1196 return __skb_flow_dissect(skb
, flow_dissector
, target_container
,
1197 NULL
, 0, 0, 0, flags
);
1200 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff
*skb
,
1201 struct flow_keys
*flow
,
1204 memset(flow
, 0, sizeof(*flow
));
1205 return __skb_flow_dissect(skb
, &flow_keys_dissector
, flow
,
1206 NULL
, 0, 0, 0, flags
);
1209 static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys
*flow
,
1210 void *data
, __be16 proto
,
1211 int nhoff
, int hlen
,
1214 memset(flow
, 0, sizeof(*flow
));
1215 return __skb_flow_dissect(NULL
, &flow_keys_buf_dissector
, flow
,
1216 data
, proto
, nhoff
, hlen
, flags
);
1219 static inline __u32
skb_get_hash(struct sk_buff
*skb
)
1221 if (!skb
->l4_hash
&& !skb
->sw_hash
)
1222 __skb_get_hash(skb
);
1227 static inline __u32
skb_get_hash_flowi6(struct sk_buff
*skb
, const struct flowi6
*fl6
)
1229 if (!skb
->l4_hash
&& !skb
->sw_hash
) {
1230 struct flow_keys keys
;
1231 __u32 hash
= __get_hash_from_flowi6(fl6
, &keys
);
1233 __skb_set_sw_hash(skb
, hash
, flow_keys_have_l4(&keys
));
1239 __u32
skb_get_hash_perturb(const struct sk_buff
*skb
,
1240 const siphash_key_t
*perturb
);
1242 static inline __u32
skb_get_hash_raw(const struct sk_buff
*skb
)
1247 static inline void skb_copy_hash(struct sk_buff
*to
, const struct sk_buff
*from
)
1249 to
->hash
= from
->hash
;
1250 to
->sw_hash
= from
->sw_hash
;
1251 to
->l4_hash
= from
->l4_hash
;
1254 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1255 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
1257 return skb
->head
+ skb
->end
;
1260 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
1265 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
1270 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
1272 return skb
->end
- skb
->head
;
1277 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1279 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
1281 return &skb_shinfo(skb
)->hwtstamps
;
1284 static inline struct ubuf_info
*skb_zcopy(struct sk_buff
*skb
)
1286 bool is_zcopy
= skb
&& skb_shinfo(skb
)->tx_flags
& SKBTX_DEV_ZEROCOPY
;
1288 return is_zcopy
? skb_uarg(skb
) : NULL
;
1291 static inline void skb_zcopy_set(struct sk_buff
*skb
, struct ubuf_info
*uarg
)
1293 if (skb
&& uarg
&& !skb_zcopy(skb
)) {
1294 sock_zerocopy_get(uarg
);
1295 skb_shinfo(skb
)->destructor_arg
= uarg
;
1296 skb_shinfo(skb
)->tx_flags
|= SKBTX_ZEROCOPY_FRAG
;
1300 static inline void skb_zcopy_set_nouarg(struct sk_buff
*skb
, void *val
)
1302 skb_shinfo(skb
)->destructor_arg
= (void *)((uintptr_t) val
| 0x1UL
);
1303 skb_shinfo(skb
)->tx_flags
|= SKBTX_ZEROCOPY_FRAG
;
1306 static inline bool skb_zcopy_is_nouarg(struct sk_buff
*skb
)
1308 return (uintptr_t) skb_shinfo(skb
)->destructor_arg
& 0x1UL
;
1311 static inline void *skb_zcopy_get_nouarg(struct sk_buff
*skb
)
1313 return (void *)((uintptr_t) skb_shinfo(skb
)->destructor_arg
& ~0x1UL
);
1316 /* Release a reference on a zerocopy structure */
1317 static inline void skb_zcopy_clear(struct sk_buff
*skb
, bool zerocopy
)
1319 struct ubuf_info
*uarg
= skb_zcopy(skb
);
1322 if (skb_zcopy_is_nouarg(skb
)) {
1323 /* no notification callback */
1324 } else if (uarg
->callback
== sock_zerocopy_callback
) {
1325 uarg
->zerocopy
= uarg
->zerocopy
&& zerocopy
;
1326 sock_zerocopy_put(uarg
);
1328 uarg
->callback(uarg
, zerocopy
);
1331 skb_shinfo(skb
)->tx_flags
&= ~SKBTX_ZEROCOPY_FRAG
;
1335 /* Abort a zerocopy operation and revert zckey on error in send syscall */
1336 static inline void skb_zcopy_abort(struct sk_buff
*skb
)
1338 struct ubuf_info
*uarg
= skb_zcopy(skb
);
1341 sock_zerocopy_put_abort(uarg
);
1342 skb_shinfo(skb
)->tx_flags
&= ~SKBTX_ZEROCOPY_FRAG
;
1347 * skb_queue_empty - check if a queue is empty
1350 * Returns true if the queue is empty, false otherwise.
1352 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
1354 return list
->next
== (const struct sk_buff
*) list
;
1358 * skb_queue_empty_lockless - check if a queue is empty
1361 * Returns true if the queue is empty, false otherwise.
1362 * This variant can be used in lockless contexts.
1364 static inline bool skb_queue_empty_lockless(const struct sk_buff_head
*list
)
1366 return READ_ONCE(list
->next
) == (const struct sk_buff
*) list
;
1371 * skb_queue_is_last - check if skb is the last entry in the queue
1375 * Returns true if @skb is the last buffer on the list.
1377 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
1378 const struct sk_buff
*skb
)
1380 return skb
->next
== (const struct sk_buff
*) list
;
1384 * skb_queue_is_first - check if skb is the first entry in the queue
1388 * Returns true if @skb is the first buffer on the list.
1390 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
1391 const struct sk_buff
*skb
)
1393 return skb
->prev
== (const struct sk_buff
*) list
;
1397 * skb_queue_next - return the next packet in the queue
1399 * @skb: current buffer
1401 * Return the next packet in @list after @skb. It is only valid to
1402 * call this if skb_queue_is_last() evaluates to false.
1404 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
1405 const struct sk_buff
*skb
)
1407 /* This BUG_ON may seem severe, but if we just return then we
1408 * are going to dereference garbage.
1410 BUG_ON(skb_queue_is_last(list
, skb
));
1415 * skb_queue_prev - return the prev packet in the queue
1417 * @skb: current buffer
1419 * Return the prev packet in @list before @skb. It is only valid to
1420 * call this if skb_queue_is_first() evaluates to false.
1422 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
1423 const struct sk_buff
*skb
)
1425 /* This BUG_ON may seem severe, but if we just return then we
1426 * are going to dereference garbage.
1428 BUG_ON(skb_queue_is_first(list
, skb
));
1433 * skb_get - reference buffer
1434 * @skb: buffer to reference
1436 * Makes another reference to a socket buffer and returns a pointer
1439 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
1441 refcount_inc(&skb
->users
);
1446 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1450 * skb_cloned - is the buffer a clone
1451 * @skb: buffer to check
1453 * Returns true if the buffer was generated with skb_clone() and is
1454 * one of multiple shared copies of the buffer. Cloned buffers are
1455 * shared data so must not be written to under normal circumstances.
1457 static inline int skb_cloned(const struct sk_buff
*skb
)
1459 return skb
->cloned
&&
1460 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
1463 static inline int skb_unclone(struct sk_buff
*skb
, gfp_t pri
)
1465 might_sleep_if(gfpflags_allow_blocking(pri
));
1467 if (skb_cloned(skb
))
1468 return pskb_expand_head(skb
, 0, 0, pri
);
1474 * skb_header_cloned - is the header a clone
1475 * @skb: buffer to check
1477 * Returns true if modifying the header part of the buffer requires
1478 * the data to be copied.
1480 static inline int skb_header_cloned(const struct sk_buff
*skb
)
1487 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
1488 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
1489 return dataref
!= 1;
1492 static inline int skb_header_unclone(struct sk_buff
*skb
, gfp_t pri
)
1494 might_sleep_if(gfpflags_allow_blocking(pri
));
1496 if (skb_header_cloned(skb
))
1497 return pskb_expand_head(skb
, 0, 0, pri
);
1503 * __skb_header_release - release reference to header
1504 * @skb: buffer to operate on
1506 static inline void __skb_header_release(struct sk_buff
*skb
)
1509 atomic_set(&skb_shinfo(skb
)->dataref
, 1 + (1 << SKB_DATAREF_SHIFT
));
1514 * skb_shared - is the buffer shared
1515 * @skb: buffer to check
1517 * Returns true if more than one person has a reference to this
1520 static inline int skb_shared(const struct sk_buff
*skb
)
1522 return refcount_read(&skb
->users
) != 1;
1526 * skb_share_check - check if buffer is shared and if so clone it
1527 * @skb: buffer to check
1528 * @pri: priority for memory allocation
1530 * If the buffer is shared the buffer is cloned and the old copy
1531 * drops a reference. A new clone with a single reference is returned.
1532 * If the buffer is not shared the original buffer is returned. When
1533 * being called from interrupt status or with spinlocks held pri must
1536 * NULL is returned on a memory allocation failure.
1538 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
, gfp_t pri
)
1540 might_sleep_if(gfpflags_allow_blocking(pri
));
1541 if (skb_shared(skb
)) {
1542 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
1554 * Copy shared buffers into a new sk_buff. We effectively do COW on
1555 * packets to handle cases where we have a local reader and forward
1556 * and a couple of other messy ones. The normal one is tcpdumping
1557 * a packet thats being forwarded.
1561 * skb_unshare - make a copy of a shared buffer
1562 * @skb: buffer to check
1563 * @pri: priority for memory allocation
1565 * If the socket buffer is a clone then this function creates a new
1566 * copy of the data, drops a reference count on the old copy and returns
1567 * the new copy with the reference count at 1. If the buffer is not a clone
1568 * the original buffer is returned. When called with a spinlock held or
1569 * from interrupt state @pri must be %GFP_ATOMIC
1571 * %NULL is returned on a memory allocation failure.
1573 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
1576 might_sleep_if(gfpflags_allow_blocking(pri
));
1577 if (skb_cloned(skb
)) {
1578 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
1580 /* Free our shared copy */
1591 * skb_peek - peek at the head of an &sk_buff_head
1592 * @list_: list to peek at
1594 * Peek an &sk_buff. Unlike most other operations you _MUST_
1595 * be careful with this one. A peek leaves the buffer on the
1596 * list and someone else may run off with it. You must hold
1597 * the appropriate locks or have a private queue to do this.
1599 * Returns %NULL for an empty list or a pointer to the head element.
1600 * The reference count is not incremented and the reference is therefore
1601 * volatile. Use with caution.
1603 static inline struct sk_buff
*skb_peek(const struct sk_buff_head
*list_
)
1605 struct sk_buff
*skb
= list_
->next
;
1607 if (skb
== (struct sk_buff
*)list_
)
1613 * skb_peek_next - peek skb following the given one from a queue
1614 * @skb: skb to start from
1615 * @list_: list to peek at
1617 * Returns %NULL when the end of the list is met or a pointer to the
1618 * next element. The reference count is not incremented and the
1619 * reference is therefore volatile. Use with caution.
1621 static inline struct sk_buff
*skb_peek_next(struct sk_buff
*skb
,
1622 const struct sk_buff_head
*list_
)
1624 struct sk_buff
*next
= skb
->next
;
1626 if (next
== (struct sk_buff
*)list_
)
1632 * skb_peek_tail - peek at the tail of an &sk_buff_head
1633 * @list_: list to peek at
1635 * Peek an &sk_buff. Unlike most other operations you _MUST_
1636 * be careful with this one. A peek leaves the buffer on the
1637 * list and someone else may run off with it. You must hold
1638 * the appropriate locks or have a private queue to do this.
1640 * Returns %NULL for an empty list or a pointer to the tail element.
1641 * The reference count is not incremented and the reference is therefore
1642 * volatile. Use with caution.
1644 static inline struct sk_buff
*skb_peek_tail(const struct sk_buff_head
*list_
)
1646 struct sk_buff
*skb
= READ_ONCE(list_
->prev
);
1648 if (skb
== (struct sk_buff
*)list_
)
1655 * skb_queue_len - get queue length
1656 * @list_: list to measure
1658 * Return the length of an &sk_buff queue.
1660 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
1666 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1667 * @list: queue to initialize
1669 * This initializes only the list and queue length aspects of
1670 * an sk_buff_head object. This allows to initialize the list
1671 * aspects of an sk_buff_head without reinitializing things like
1672 * the spinlock. It can also be used for on-stack sk_buff_head
1673 * objects where the spinlock is known to not be used.
1675 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
1677 list
->prev
= list
->next
= (struct sk_buff
*)list
;
1682 * This function creates a split out lock class for each invocation;
1683 * this is needed for now since a whole lot of users of the skb-queue
1684 * infrastructure in drivers have different locking usage (in hardirq)
1685 * than the networking core (in softirq only). In the long run either the
1686 * network layer or drivers should need annotation to consolidate the
1687 * main types of usage into 3 classes.
1689 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
1691 spin_lock_init(&list
->lock
);
1692 __skb_queue_head_init(list
);
1695 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
1696 struct lock_class_key
*class)
1698 skb_queue_head_init(list
);
1699 lockdep_set_class(&list
->lock
, class);
1703 * Insert an sk_buff on a list.
1705 * The "__skb_xxxx()" functions are the non-atomic ones that
1706 * can only be called with interrupts disabled.
1708 void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
,
1709 struct sk_buff_head
*list
);
1710 static inline void __skb_insert(struct sk_buff
*newsk
,
1711 struct sk_buff
*prev
, struct sk_buff
*next
,
1712 struct sk_buff_head
*list
)
1714 /* See skb_queue_empty_lockless() and skb_peek_tail()
1715 * for the opposite READ_ONCE()
1717 WRITE_ONCE(newsk
->next
, next
);
1718 WRITE_ONCE(newsk
->prev
, prev
);
1719 WRITE_ONCE(next
->prev
, newsk
);
1720 WRITE_ONCE(prev
->next
, newsk
);
1724 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
1725 struct sk_buff
*prev
,
1726 struct sk_buff
*next
)
1728 struct sk_buff
*first
= list
->next
;
1729 struct sk_buff
*last
= list
->prev
;
1731 WRITE_ONCE(first
->prev
, prev
);
1732 WRITE_ONCE(prev
->next
, first
);
1734 WRITE_ONCE(last
->next
, next
);
1735 WRITE_ONCE(next
->prev
, last
);
1739 * skb_queue_splice - join two skb lists, this is designed for stacks
1740 * @list: the new list to add
1741 * @head: the place to add it in the first list
1743 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
1744 struct sk_buff_head
*head
)
1746 if (!skb_queue_empty(list
)) {
1747 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1748 head
->qlen
+= list
->qlen
;
1753 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1754 * @list: the new list to add
1755 * @head: the place to add it in the first list
1757 * The list at @list is reinitialised
1759 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
1760 struct sk_buff_head
*head
)
1762 if (!skb_queue_empty(list
)) {
1763 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1764 head
->qlen
+= list
->qlen
;
1765 __skb_queue_head_init(list
);
1770 * skb_queue_splice_tail - join two skb lists, each list being a queue
1771 * @list: the new list to add
1772 * @head: the place to add it in the first list
1774 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
1775 struct sk_buff_head
*head
)
1777 if (!skb_queue_empty(list
)) {
1778 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1779 head
->qlen
+= list
->qlen
;
1784 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1785 * @list: the new list to add
1786 * @head: the place to add it in the first list
1788 * Each of the lists is a queue.
1789 * The list at @list is reinitialised
1791 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
1792 struct sk_buff_head
*head
)
1794 if (!skb_queue_empty(list
)) {
1795 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1796 head
->qlen
+= list
->qlen
;
1797 __skb_queue_head_init(list
);
1802 * __skb_queue_after - queue a buffer at the list head
1803 * @list: list to use
1804 * @prev: place after this buffer
1805 * @newsk: buffer to queue
1807 * Queue a buffer int the middle of a list. This function takes no locks
1808 * and you must therefore hold required locks before calling it.
1810 * A buffer cannot be placed on two lists at the same time.
1812 static inline void __skb_queue_after(struct sk_buff_head
*list
,
1813 struct sk_buff
*prev
,
1814 struct sk_buff
*newsk
)
1816 __skb_insert(newsk
, prev
, prev
->next
, list
);
1819 void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
1820 struct sk_buff_head
*list
);
1822 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1823 struct sk_buff
*next
,
1824 struct sk_buff
*newsk
)
1826 __skb_insert(newsk
, next
->prev
, next
, list
);
1830 * __skb_queue_head - queue a buffer at the list head
1831 * @list: list to use
1832 * @newsk: buffer to queue
1834 * Queue a buffer at the start of a list. This function takes no locks
1835 * and you must therefore hold required locks before calling it.
1837 * A buffer cannot be placed on two lists at the same time.
1839 void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1840 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1841 struct sk_buff
*newsk
)
1843 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1847 * __skb_queue_tail - queue a buffer at the list tail
1848 * @list: list to use
1849 * @newsk: buffer to queue
1851 * Queue a buffer at the end of a list. This function takes no locks
1852 * and you must therefore hold required locks before calling it.
1854 * A buffer cannot be placed on two lists at the same time.
1856 void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1857 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1858 struct sk_buff
*newsk
)
1860 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1864 * remove sk_buff from list. _Must_ be called atomically, and with
1867 void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1868 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1870 struct sk_buff
*next
, *prev
;
1875 skb
->next
= skb
->prev
= NULL
;
1876 WRITE_ONCE(next
->prev
, prev
);
1877 WRITE_ONCE(prev
->next
, next
);
1881 * __skb_dequeue - remove from the head of the queue
1882 * @list: list to dequeue from
1884 * Remove the head of the list. This function does not take any locks
1885 * so must be used with appropriate locks held only. The head item is
1886 * returned or %NULL if the list is empty.
1888 struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1889 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1891 struct sk_buff
*skb
= skb_peek(list
);
1893 __skb_unlink(skb
, list
);
1898 * __skb_dequeue_tail - remove from the tail of the queue
1899 * @list: list to dequeue from
1901 * Remove the tail of the list. This function does not take any locks
1902 * so must be used with appropriate locks held only. The tail item is
1903 * returned or %NULL if the list is empty.
1905 struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1906 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1908 struct sk_buff
*skb
= skb_peek_tail(list
);
1910 __skb_unlink(skb
, list
);
1915 static inline bool skb_is_nonlinear(const struct sk_buff
*skb
)
1917 return skb
->data_len
;
1920 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1922 return skb
->len
- skb
->data_len
;
1925 static inline unsigned int __skb_pagelen(const struct sk_buff
*skb
)
1927 unsigned int i
, len
= 0;
1929 for (i
= skb_shinfo(skb
)->nr_frags
- 1; (int)i
>= 0; i
--)
1930 len
+= skb_frag_size(&skb_shinfo(skb
)->frags
[i
]);
1934 static inline unsigned int skb_pagelen(const struct sk_buff
*skb
)
1936 return skb_headlen(skb
) + __skb_pagelen(skb
);
1940 * __skb_fill_page_desc - initialise a paged fragment in an skb
1941 * @skb: buffer containing fragment to be initialised
1942 * @i: paged fragment index to initialise
1943 * @page: the page to use for this fragment
1944 * @off: the offset to the data with @page
1945 * @size: the length of the data
1947 * Initialises the @i'th fragment of @skb to point to &size bytes at
1948 * offset @off within @page.
1950 * Does not take any additional reference on the fragment.
1952 static inline void __skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1953 struct page
*page
, int off
, int size
)
1955 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1958 * Propagate page pfmemalloc to the skb if we can. The problem is
1959 * that not all callers have unique ownership of the page but rely
1960 * on page_is_pfmemalloc doing the right thing(tm).
1962 frag
->page
.p
= page
;
1963 frag
->page_offset
= off
;
1964 skb_frag_size_set(frag
, size
);
1966 page
= compound_head(page
);
1967 if (page_is_pfmemalloc(page
))
1968 skb
->pfmemalloc
= true;
1972 * skb_fill_page_desc - initialise a paged fragment in an skb
1973 * @skb: buffer containing fragment to be initialised
1974 * @i: paged fragment index to initialise
1975 * @page: the page to use for this fragment
1976 * @off: the offset to the data with @page
1977 * @size: the length of the data
1979 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1980 * @skb to point to @size bytes at offset @off within @page. In
1981 * addition updates @skb such that @i is the last fragment.
1983 * Does not take any additional reference on the fragment.
1985 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1986 struct page
*page
, int off
, int size
)
1988 __skb_fill_page_desc(skb
, i
, page
, off
, size
);
1989 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1992 void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
, int off
,
1993 int size
, unsigned int truesize
);
1995 void skb_coalesce_rx_frag(struct sk_buff
*skb
, int i
, int size
,
1996 unsigned int truesize
);
1998 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1999 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
2000 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2002 #ifdef NET_SKBUFF_DATA_USES_OFFSET
2003 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
2005 return skb
->head
+ skb
->tail
;
2008 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
2010 skb
->tail
= skb
->data
- skb
->head
;
2013 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
2015 skb_reset_tail_pointer(skb
);
2016 skb
->tail
+= offset
;
2019 #else /* NET_SKBUFF_DATA_USES_OFFSET */
2020 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
2025 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
2027 skb
->tail
= skb
->data
;
2030 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
2032 skb
->tail
= skb
->data
+ offset
;
2035 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2038 * Add data to an sk_buff
2040 void *pskb_put(struct sk_buff
*skb
, struct sk_buff
*tail
, int len
);
2041 void *skb_put(struct sk_buff
*skb
, unsigned int len
);
2042 static inline void *__skb_put(struct sk_buff
*skb
, unsigned int len
)
2044 void *tmp
= skb_tail_pointer(skb
);
2045 SKB_LINEAR_ASSERT(skb
);
2051 static inline void *__skb_put_zero(struct sk_buff
*skb
, unsigned int len
)
2053 void *tmp
= __skb_put(skb
, len
);
2055 memset(tmp
, 0, len
);
2059 static inline void *__skb_put_data(struct sk_buff
*skb
, const void *data
,
2062 void *tmp
= __skb_put(skb
, len
);
2064 memcpy(tmp
, data
, len
);
2068 static inline void __skb_put_u8(struct sk_buff
*skb
, u8 val
)
2070 *(u8
*)__skb_put(skb
, 1) = val
;
2073 static inline void *skb_put_zero(struct sk_buff
*skb
, unsigned int len
)
2075 void *tmp
= skb_put(skb
, len
);
2077 memset(tmp
, 0, len
);
2082 static inline void *skb_put_data(struct sk_buff
*skb
, const void *data
,
2085 void *tmp
= skb_put(skb
, len
);
2087 memcpy(tmp
, data
, len
);
2092 static inline void skb_put_u8(struct sk_buff
*skb
, u8 val
)
2094 *(u8
*)skb_put(skb
, 1) = val
;
2097 void *skb_push(struct sk_buff
*skb
, unsigned int len
);
2098 static inline void *__skb_push(struct sk_buff
*skb
, unsigned int len
)
2105 void *skb_pull(struct sk_buff
*skb
, unsigned int len
);
2106 static inline void *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
2109 BUG_ON(skb
->len
< skb
->data_len
);
2110 return skb
->data
+= len
;
2113 static inline void *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
2115 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
2118 void *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
2120 static inline void *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
2122 if (len
> skb_headlen(skb
) &&
2123 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
2126 return skb
->data
+= len
;
2129 static inline void *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
2131 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
2134 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
2136 if (likely(len
<= skb_headlen(skb
)))
2138 if (unlikely(len
> skb
->len
))
2140 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
2143 void skb_condense(struct sk_buff
*skb
);
2146 * skb_headroom - bytes at buffer head
2147 * @skb: buffer to check
2149 * Return the number of bytes of free space at the head of an &sk_buff.
2151 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
2153 return skb
->data
- skb
->head
;
2157 * skb_tailroom - bytes at buffer end
2158 * @skb: buffer to check
2160 * Return the number of bytes of free space at the tail of an sk_buff
2162 static inline int skb_tailroom(const struct sk_buff
*skb
)
2164 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
2168 * skb_availroom - bytes at buffer end
2169 * @skb: buffer to check
2171 * Return the number of bytes of free space at the tail of an sk_buff
2172 * allocated by sk_stream_alloc()
2174 static inline int skb_availroom(const struct sk_buff
*skb
)
2176 if (skb_is_nonlinear(skb
))
2179 return skb
->end
- skb
->tail
- skb
->reserved_tailroom
;
2183 * skb_reserve - adjust headroom
2184 * @skb: buffer to alter
2185 * @len: bytes to move
2187 * Increase the headroom of an empty &sk_buff by reducing the tail
2188 * room. This is only allowed for an empty buffer.
2190 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
2197 * skb_tailroom_reserve - adjust reserved_tailroom
2198 * @skb: buffer to alter
2199 * @mtu: maximum amount of headlen permitted
2200 * @needed_tailroom: minimum amount of reserved_tailroom
2202 * Set reserved_tailroom so that headlen can be as large as possible but
2203 * not larger than mtu and tailroom cannot be smaller than
2205 * The required headroom should already have been reserved before using
2208 static inline void skb_tailroom_reserve(struct sk_buff
*skb
, unsigned int mtu
,
2209 unsigned int needed_tailroom
)
2211 SKB_LINEAR_ASSERT(skb
);
2212 if (mtu
< skb_tailroom(skb
) - needed_tailroom
)
2213 /* use at most mtu */
2214 skb
->reserved_tailroom
= skb_tailroom(skb
) - mtu
;
2216 /* use up to all available space */
2217 skb
->reserved_tailroom
= needed_tailroom
;
2220 #define ENCAP_TYPE_ETHER 0
2221 #define ENCAP_TYPE_IPPROTO 1
2223 static inline void skb_set_inner_protocol(struct sk_buff
*skb
,
2226 skb
->inner_protocol
= protocol
;
2227 skb
->inner_protocol_type
= ENCAP_TYPE_ETHER
;
2230 static inline void skb_set_inner_ipproto(struct sk_buff
*skb
,
2233 skb
->inner_ipproto
= ipproto
;
2234 skb
->inner_protocol_type
= ENCAP_TYPE_IPPROTO
;
2237 static inline void skb_reset_inner_headers(struct sk_buff
*skb
)
2239 skb
->inner_mac_header
= skb
->mac_header
;
2240 skb
->inner_network_header
= skb
->network_header
;
2241 skb
->inner_transport_header
= skb
->transport_header
;
2244 static inline void skb_reset_mac_len(struct sk_buff
*skb
)
2246 skb
->mac_len
= skb
->network_header
- skb
->mac_header
;
2249 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2252 return skb
->head
+ skb
->inner_transport_header
;
2255 static inline int skb_inner_transport_offset(const struct sk_buff
*skb
)
2257 return skb_inner_transport_header(skb
) - skb
->data
;
2260 static inline void skb_reset_inner_transport_header(struct sk_buff
*skb
)
2262 skb
->inner_transport_header
= skb
->data
- skb
->head
;
2265 static inline void skb_set_inner_transport_header(struct sk_buff
*skb
,
2268 skb_reset_inner_transport_header(skb
);
2269 skb
->inner_transport_header
+= offset
;
2272 static inline unsigned char *skb_inner_network_header(const struct sk_buff
*skb
)
2274 return skb
->head
+ skb
->inner_network_header
;
2277 static inline void skb_reset_inner_network_header(struct sk_buff
*skb
)
2279 skb
->inner_network_header
= skb
->data
- skb
->head
;
2282 static inline void skb_set_inner_network_header(struct sk_buff
*skb
,
2285 skb_reset_inner_network_header(skb
);
2286 skb
->inner_network_header
+= offset
;
2289 static inline unsigned char *skb_inner_mac_header(const struct sk_buff
*skb
)
2291 return skb
->head
+ skb
->inner_mac_header
;
2294 static inline void skb_reset_inner_mac_header(struct sk_buff
*skb
)
2296 skb
->inner_mac_header
= skb
->data
- skb
->head
;
2299 static inline void skb_set_inner_mac_header(struct sk_buff
*skb
,
2302 skb_reset_inner_mac_header(skb
);
2303 skb
->inner_mac_header
+= offset
;
2305 static inline bool skb_transport_header_was_set(const struct sk_buff
*skb
)
2307 return skb
->transport_header
!= (typeof(skb
->transport_header
))~0U;
2310 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
2312 return skb
->head
+ skb
->transport_header
;
2315 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
2317 skb
->transport_header
= skb
->data
- skb
->head
;
2320 static inline void skb_set_transport_header(struct sk_buff
*skb
,
2323 skb_reset_transport_header(skb
);
2324 skb
->transport_header
+= offset
;
2327 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
2329 return skb
->head
+ skb
->network_header
;
2332 static inline void skb_reset_network_header(struct sk_buff
*skb
)
2334 skb
->network_header
= skb
->data
- skb
->head
;
2337 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
2339 skb_reset_network_header(skb
);
2340 skb
->network_header
+= offset
;
2343 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
2345 return skb
->head
+ skb
->mac_header
;
2348 static inline int skb_mac_offset(const struct sk_buff
*skb
)
2350 return skb_mac_header(skb
) - skb
->data
;
2353 static inline u32
skb_mac_header_len(const struct sk_buff
*skb
)
2355 return skb
->network_header
- skb
->mac_header
;
2358 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
2360 return skb
->mac_header
!= (typeof(skb
->mac_header
))~0U;
2363 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
2365 skb
->mac_header
= skb
->data
- skb
->head
;
2368 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
2370 skb_reset_mac_header(skb
);
2371 skb
->mac_header
+= offset
;
2374 static inline void skb_pop_mac_header(struct sk_buff
*skb
)
2376 skb
->mac_header
= skb
->network_header
;
2379 static inline void skb_probe_transport_header(struct sk_buff
*skb
,
2380 const int offset_hint
)
2382 struct flow_keys keys
;
2384 if (skb_transport_header_was_set(skb
))
2386 else if (skb_flow_dissect_flow_keys(skb
, &keys
, 0))
2387 skb_set_transport_header(skb
, keys
.control
.thoff
);
2388 else if (offset_hint
>= 0)
2389 skb_set_transport_header(skb
, offset_hint
);
2392 static inline void skb_mac_header_rebuild(struct sk_buff
*skb
)
2394 if (skb_mac_header_was_set(skb
)) {
2395 const unsigned char *old_mac
= skb_mac_header(skb
);
2397 skb_set_mac_header(skb
, -skb
->mac_len
);
2398 memmove(skb_mac_header(skb
), old_mac
, skb
->mac_len
);
2402 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
2404 return skb
->csum_start
- skb_headroom(skb
);
2407 static inline unsigned char *skb_checksum_start(const struct sk_buff
*skb
)
2409 return skb
->head
+ skb
->csum_start
;
2412 static inline int skb_transport_offset(const struct sk_buff
*skb
)
2414 return skb_transport_header(skb
) - skb
->data
;
2417 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
2419 return skb
->transport_header
- skb
->network_header
;
2422 static inline u32
skb_inner_network_header_len(const struct sk_buff
*skb
)
2424 return skb
->inner_transport_header
- skb
->inner_network_header
;
2427 static inline int skb_network_offset(const struct sk_buff
*skb
)
2429 return skb_network_header(skb
) - skb
->data
;
2432 static inline int skb_inner_network_offset(const struct sk_buff
*skb
)
2434 return skb_inner_network_header(skb
) - skb
->data
;
2437 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
2439 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
2443 * CPUs often take a performance hit when accessing unaligned memory
2444 * locations. The actual performance hit varies, it can be small if the
2445 * hardware handles it or large if we have to take an exception and fix it
2448 * Since an ethernet header is 14 bytes network drivers often end up with
2449 * the IP header at an unaligned offset. The IP header can be aligned by
2450 * shifting the start of the packet by 2 bytes. Drivers should do this
2453 * skb_reserve(skb, NET_IP_ALIGN);
2455 * The downside to this alignment of the IP header is that the DMA is now
2456 * unaligned. On some architectures the cost of an unaligned DMA is high
2457 * and this cost outweighs the gains made by aligning the IP header.
2459 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2462 #ifndef NET_IP_ALIGN
2463 #define NET_IP_ALIGN 2
2467 * The networking layer reserves some headroom in skb data (via
2468 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2469 * the header has to grow. In the default case, if the header has to grow
2470 * 32 bytes or less we avoid the reallocation.
2472 * Unfortunately this headroom changes the DMA alignment of the resulting
2473 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2474 * on some architectures. An architecture can override this value,
2475 * perhaps setting it to a cacheline in size (since that will maintain
2476 * cacheline alignment of the DMA). It must be a power of 2.
2478 * Various parts of the networking layer expect at least 32 bytes of
2479 * headroom, you should not reduce this.
2481 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2482 * to reduce average number of cache lines per packet.
2483 * get_rps_cpus() for example only access one 64 bytes aligned block :
2484 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2487 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2490 int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
2492 static inline void __skb_set_length(struct sk_buff
*skb
, unsigned int len
)
2494 if (unlikely(skb_is_nonlinear(skb
))) {
2499 skb_set_tail_pointer(skb
, len
);
2502 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
2504 __skb_set_length(skb
, len
);
2507 void skb_trim(struct sk_buff
*skb
, unsigned int len
);
2509 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2512 return ___pskb_trim(skb
, len
);
2513 __skb_trim(skb
, len
);
2517 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2519 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
2523 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2524 * @skb: buffer to alter
2527 * This is identical to pskb_trim except that the caller knows that
2528 * the skb is not cloned so we should never get an error due to out-
2531 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
2533 int err
= pskb_trim(skb
, len
);
2537 static inline int __skb_grow(struct sk_buff
*skb
, unsigned int len
)
2539 unsigned int diff
= len
- skb
->len
;
2541 if (skb_tailroom(skb
) < diff
) {
2542 int ret
= pskb_expand_head(skb
, 0, diff
- skb_tailroom(skb
),
2547 __skb_set_length(skb
, len
);
2552 * skb_orphan - orphan a buffer
2553 * @skb: buffer to orphan
2555 * If a buffer currently has an owner then we call the owner's
2556 * destructor function and make the @skb unowned. The buffer continues
2557 * to exist but is no longer charged to its former owner.
2559 static inline void skb_orphan(struct sk_buff
*skb
)
2561 if (skb
->destructor
) {
2562 skb
->destructor(skb
);
2563 skb
->destructor
= NULL
;
2571 * skb_orphan_frags - orphan the frags contained in a buffer
2572 * @skb: buffer to orphan frags from
2573 * @gfp_mask: allocation mask for replacement pages
2575 * For each frag in the SKB which needs a destructor (i.e. has an
2576 * owner) create a copy of that frag and release the original
2577 * page by calling the destructor.
2579 static inline int skb_orphan_frags(struct sk_buff
*skb
, gfp_t gfp_mask
)
2581 if (likely(!skb_zcopy(skb
)))
2583 if (!skb_zcopy_is_nouarg(skb
) &&
2584 skb_uarg(skb
)->callback
== sock_zerocopy_callback
)
2586 return skb_copy_ubufs(skb
, gfp_mask
);
2589 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2590 static inline int skb_orphan_frags_rx(struct sk_buff
*skb
, gfp_t gfp_mask
)
2592 if (likely(!skb_zcopy(skb
)))
2594 return skb_copy_ubufs(skb
, gfp_mask
);
2598 * __skb_queue_purge - empty a list
2599 * @list: list to empty
2601 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2602 * the list and one reference dropped. This function does not take the
2603 * list lock and the caller must hold the relevant locks to use it.
2605 void skb_queue_purge(struct sk_buff_head
*list
);
2606 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
2608 struct sk_buff
*skb
;
2609 while ((skb
= __skb_dequeue(list
)) != NULL
)
2613 unsigned int skb_rbtree_purge(struct rb_root
*root
);
2615 void *netdev_alloc_frag(unsigned int fragsz
);
2617 struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
, unsigned int length
,
2621 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2622 * @dev: network device to receive on
2623 * @length: length to allocate
2625 * Allocate a new &sk_buff and assign it a usage count of one. The
2626 * buffer has unspecified headroom built in. Users should allocate
2627 * the headroom they think they need without accounting for the
2628 * built in space. The built in space is used for optimisations.
2630 * %NULL is returned if there is no free memory. Although this function
2631 * allocates memory it can be called from an interrupt.
2633 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
2634 unsigned int length
)
2636 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
2639 /* legacy helper around __netdev_alloc_skb() */
2640 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
2643 return __netdev_alloc_skb(NULL
, length
, gfp_mask
);
2646 /* legacy helper around netdev_alloc_skb() */
2647 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
2649 return netdev_alloc_skb(NULL
, length
);
2653 static inline struct sk_buff
*__netdev_alloc_skb_ip_align(struct net_device
*dev
,
2654 unsigned int length
, gfp_t gfp
)
2656 struct sk_buff
*skb
= __netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
, gfp
);
2658 if (NET_IP_ALIGN
&& skb
)
2659 skb_reserve(skb
, NET_IP_ALIGN
);
2663 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
2664 unsigned int length
)
2666 return __netdev_alloc_skb_ip_align(dev
, length
, GFP_ATOMIC
);
2669 static inline void skb_free_frag(void *addr
)
2671 page_frag_free(addr
);
2674 void *napi_alloc_frag(unsigned int fragsz
);
2675 struct sk_buff
*__napi_alloc_skb(struct napi_struct
*napi
,
2676 unsigned int length
, gfp_t gfp_mask
);
2677 static inline struct sk_buff
*napi_alloc_skb(struct napi_struct
*napi
,
2678 unsigned int length
)
2680 return __napi_alloc_skb(napi
, length
, GFP_ATOMIC
);
2682 void napi_consume_skb(struct sk_buff
*skb
, int budget
);
2684 void __kfree_skb_flush(void);
2685 void __kfree_skb_defer(struct sk_buff
*skb
);
2688 * __dev_alloc_pages - allocate page for network Rx
2689 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2690 * @order: size of the allocation
2692 * Allocate a new page.
2694 * %NULL is returned if there is no free memory.
2696 static inline struct page
*__dev_alloc_pages(gfp_t gfp_mask
,
2699 /* This piece of code contains several assumptions.
2700 * 1. This is for device Rx, therefor a cold page is preferred.
2701 * 2. The expectation is the user wants a compound page.
2702 * 3. If requesting a order 0 page it will not be compound
2703 * due to the check to see if order has a value in prep_new_page
2704 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2705 * code in gfp_to_alloc_flags that should be enforcing this.
2707 gfp_mask
|= __GFP_COMP
| __GFP_MEMALLOC
;
2709 return alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, order
);
2712 static inline struct page
*dev_alloc_pages(unsigned int order
)
2714 return __dev_alloc_pages(GFP_ATOMIC
| __GFP_NOWARN
, order
);
2718 * __dev_alloc_page - allocate a page for network Rx
2719 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2721 * Allocate a new page.
2723 * %NULL is returned if there is no free memory.
2725 static inline struct page
*__dev_alloc_page(gfp_t gfp_mask
)
2727 return __dev_alloc_pages(gfp_mask
, 0);
2730 static inline struct page
*dev_alloc_page(void)
2732 return dev_alloc_pages(0);
2736 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2737 * @page: The page that was allocated from skb_alloc_page
2738 * @skb: The skb that may need pfmemalloc set
2740 static inline void skb_propagate_pfmemalloc(struct page
*page
,
2741 struct sk_buff
*skb
)
2743 if (page_is_pfmemalloc(page
))
2744 skb
->pfmemalloc
= true;
2748 * skb_frag_page - retrieve the page referred to by a paged fragment
2749 * @frag: the paged fragment
2751 * Returns the &struct page associated with @frag.
2753 static inline struct page
*skb_frag_page(const skb_frag_t
*frag
)
2755 return frag
->page
.p
;
2759 * __skb_frag_ref - take an addition reference on a paged fragment.
2760 * @frag: the paged fragment
2762 * Takes an additional reference on the paged fragment @frag.
2764 static inline void __skb_frag_ref(skb_frag_t
*frag
)
2766 get_page(skb_frag_page(frag
));
2770 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2772 * @f: the fragment offset.
2774 * Takes an additional reference on the @f'th paged fragment of @skb.
2776 static inline void skb_frag_ref(struct sk_buff
*skb
, int f
)
2778 __skb_frag_ref(&skb_shinfo(skb
)->frags
[f
]);
2782 * __skb_frag_unref - release a reference on a paged fragment.
2783 * @frag: the paged fragment
2785 * Releases a reference on the paged fragment @frag.
2787 static inline void __skb_frag_unref(skb_frag_t
*frag
)
2789 put_page(skb_frag_page(frag
));
2793 * skb_frag_unref - release a reference on a paged fragment of an skb.
2795 * @f: the fragment offset
2797 * Releases a reference on the @f'th paged fragment of @skb.
2799 static inline void skb_frag_unref(struct sk_buff
*skb
, int f
)
2801 __skb_frag_unref(&skb_shinfo(skb
)->frags
[f
]);
2805 * skb_frag_address - gets the address of the data contained in a paged fragment
2806 * @frag: the paged fragment buffer
2808 * Returns the address of the data within @frag. The page must already
2811 static inline void *skb_frag_address(const skb_frag_t
*frag
)
2813 return page_address(skb_frag_page(frag
)) + frag
->page_offset
;
2817 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2818 * @frag: the paged fragment buffer
2820 * Returns the address of the data within @frag. Checks that the page
2821 * is mapped and returns %NULL otherwise.
2823 static inline void *skb_frag_address_safe(const skb_frag_t
*frag
)
2825 void *ptr
= page_address(skb_frag_page(frag
));
2829 return ptr
+ frag
->page_offset
;
2833 * __skb_frag_set_page - sets the page contained in a paged fragment
2834 * @frag: the paged fragment
2835 * @page: the page to set
2837 * Sets the fragment @frag to contain @page.
2839 static inline void __skb_frag_set_page(skb_frag_t
*frag
, struct page
*page
)
2841 frag
->page
.p
= page
;
2845 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2847 * @f: the fragment offset
2848 * @page: the page to set
2850 * Sets the @f'th fragment of @skb to contain @page.
2852 static inline void skb_frag_set_page(struct sk_buff
*skb
, int f
,
2855 __skb_frag_set_page(&skb_shinfo(skb
)->frags
[f
], page
);
2858 bool skb_page_frag_refill(unsigned int sz
, struct page_frag
*pfrag
, gfp_t prio
);
2861 * skb_frag_dma_map - maps a paged fragment via the DMA API
2862 * @dev: the device to map the fragment to
2863 * @frag: the paged fragment to map
2864 * @offset: the offset within the fragment (starting at the
2865 * fragment's own offset)
2866 * @size: the number of bytes to map
2867 * @dir: the direction of the mapping (``PCI_DMA_*``)
2869 * Maps the page associated with @frag to @device.
2871 static inline dma_addr_t
skb_frag_dma_map(struct device
*dev
,
2872 const skb_frag_t
*frag
,
2873 size_t offset
, size_t size
,
2874 enum dma_data_direction dir
)
2876 return dma_map_page(dev
, skb_frag_page(frag
),
2877 frag
->page_offset
+ offset
, size
, dir
);
2880 static inline struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
2883 return __pskb_copy(skb
, skb_headroom(skb
), gfp_mask
);
2887 static inline struct sk_buff
*pskb_copy_for_clone(struct sk_buff
*skb
,
2890 return __pskb_copy_fclone(skb
, skb_headroom(skb
), gfp_mask
, true);
2895 * skb_clone_writable - is the header of a clone writable
2896 * @skb: buffer to check
2897 * @len: length up to which to write
2899 * Returns true if modifying the header part of the cloned buffer
2900 * does not requires the data to be copied.
2902 static inline int skb_clone_writable(const struct sk_buff
*skb
, unsigned int len
)
2904 return !skb_header_cloned(skb
) &&
2905 skb_headroom(skb
) + len
<= skb
->hdr_len
;
2908 static inline int skb_try_make_writable(struct sk_buff
*skb
,
2909 unsigned int write_len
)
2911 return skb_cloned(skb
) && !skb_clone_writable(skb
, write_len
) &&
2912 pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
);
2915 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
2920 if (headroom
> skb_headroom(skb
))
2921 delta
= headroom
- skb_headroom(skb
);
2923 if (delta
|| cloned
)
2924 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
2930 * skb_cow - copy header of skb when it is required
2931 * @skb: buffer to cow
2932 * @headroom: needed headroom
2934 * If the skb passed lacks sufficient headroom or its data part
2935 * is shared, data is reallocated. If reallocation fails, an error
2936 * is returned and original skb is not changed.
2938 * The result is skb with writable area skb->head...skb->tail
2939 * and at least @headroom of space at head.
2941 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
2943 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
2947 * skb_cow_head - skb_cow but only making the head writable
2948 * @skb: buffer to cow
2949 * @headroom: needed headroom
2951 * This function is identical to skb_cow except that we replace the
2952 * skb_cloned check by skb_header_cloned. It should be used when
2953 * you only need to push on some header and do not need to modify
2956 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
2958 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
2962 * skb_padto - pad an skbuff up to a minimal size
2963 * @skb: buffer to pad
2964 * @len: minimal length
2966 * Pads up a buffer to ensure the trailing bytes exist and are
2967 * blanked. If the buffer already contains sufficient data it
2968 * is untouched. Otherwise it is extended. Returns zero on
2969 * success. The skb is freed on error.
2971 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
2973 unsigned int size
= skb
->len
;
2974 if (likely(size
>= len
))
2976 return skb_pad(skb
, len
- size
);
2980 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2981 * @skb: buffer to pad
2982 * @len: minimal length
2983 * @free_on_error: free buffer on error
2985 * Pads up a buffer to ensure the trailing bytes exist and are
2986 * blanked. If the buffer already contains sufficient data it
2987 * is untouched. Otherwise it is extended. Returns zero on
2988 * success. The skb is freed on error if @free_on_error is true.
2990 static inline int __skb_put_padto(struct sk_buff
*skb
, unsigned int len
,
2993 unsigned int size
= skb
->len
;
2995 if (unlikely(size
< len
)) {
2997 if (__skb_pad(skb
, len
, free_on_error
))
2999 __skb_put(skb
, len
);
3005 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3006 * @skb: buffer to pad
3007 * @len: minimal length
3009 * Pads up a buffer to ensure the trailing bytes exist and are
3010 * blanked. If the buffer already contains sufficient data it
3011 * is untouched. Otherwise it is extended. Returns zero on
3012 * success. The skb is freed on error.
3014 static inline int skb_put_padto(struct sk_buff
*skb
, unsigned int len
)
3016 return __skb_put_padto(skb
, len
, true);
3019 static inline int skb_add_data(struct sk_buff
*skb
,
3020 struct iov_iter
*from
, int copy
)
3022 const int off
= skb
->len
;
3024 if (skb
->ip_summed
== CHECKSUM_NONE
) {
3026 if (csum_and_copy_from_iter_full(skb_put(skb
, copy
), copy
,
3028 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
3031 } else if (copy_from_iter_full(skb_put(skb
, copy
), copy
, from
))
3034 __skb_trim(skb
, off
);
3038 static inline bool skb_can_coalesce(struct sk_buff
*skb
, int i
,
3039 const struct page
*page
, int off
)
3044 const struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
3046 return page
== skb_frag_page(frag
) &&
3047 off
== frag
->page_offset
+ skb_frag_size(frag
);
3052 static inline int __skb_linearize(struct sk_buff
*skb
)
3054 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
3058 * skb_linearize - convert paged skb to linear one
3059 * @skb: buffer to linarize
3061 * If there is no free memory -ENOMEM is returned, otherwise zero
3062 * is returned and the old skb data released.
3064 static inline int skb_linearize(struct sk_buff
*skb
)
3066 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
3070 * skb_has_shared_frag - can any frag be overwritten
3071 * @skb: buffer to test
3073 * Return true if the skb has at least one frag that might be modified
3074 * by an external entity (as in vmsplice()/sendfile())
3076 static inline bool skb_has_shared_frag(const struct sk_buff
*skb
)
3078 return skb_is_nonlinear(skb
) &&
3079 skb_shinfo(skb
)->tx_flags
& SKBTX_SHARED_FRAG
;
3083 * skb_linearize_cow - make sure skb is linear and writable
3084 * @skb: buffer to process
3086 * If there is no free memory -ENOMEM is returned, otherwise zero
3087 * is returned and the old skb data released.
3089 static inline int skb_linearize_cow(struct sk_buff
*skb
)
3091 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
3092 __skb_linearize(skb
) : 0;
3095 static __always_inline
void
3096 __skb_postpull_rcsum(struct sk_buff
*skb
, const void *start
, unsigned int len
,
3099 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3100 skb
->csum
= csum_block_sub(skb
->csum
,
3101 csum_partial(start
, len
, 0), off
);
3102 else if (skb
->ip_summed
== CHECKSUM_PARTIAL
&&
3103 skb_checksum_start_offset(skb
) < 0)
3104 skb
->ip_summed
= CHECKSUM_NONE
;
3108 * skb_postpull_rcsum - update checksum for received skb after pull
3109 * @skb: buffer to update
3110 * @start: start of data before pull
3111 * @len: length of data pulled
3113 * After doing a pull on a received packet, you need to call this to
3114 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3115 * CHECKSUM_NONE so that it can be recomputed from scratch.
3117 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
3118 const void *start
, unsigned int len
)
3120 __skb_postpull_rcsum(skb
, start
, len
, 0);
3123 static __always_inline
void
3124 __skb_postpush_rcsum(struct sk_buff
*skb
, const void *start
, unsigned int len
,
3127 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3128 skb
->csum
= csum_block_add(skb
->csum
,
3129 csum_partial(start
, len
, 0), off
);
3133 * skb_postpush_rcsum - update checksum for received skb after push
3134 * @skb: buffer to update
3135 * @start: start of data after push
3136 * @len: length of data pushed
3138 * After doing a push on a received packet, you need to call this to
3139 * update the CHECKSUM_COMPLETE checksum.
3141 static inline void skb_postpush_rcsum(struct sk_buff
*skb
,
3142 const void *start
, unsigned int len
)
3144 __skb_postpush_rcsum(skb
, start
, len
, 0);
3147 void *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
3150 * skb_push_rcsum - push skb and update receive checksum
3151 * @skb: buffer to update
3152 * @len: length of data pulled
3154 * This function performs an skb_push on the packet and updates
3155 * the CHECKSUM_COMPLETE checksum. It should be used on
3156 * receive path processing instead of skb_push unless you know
3157 * that the checksum difference is zero (e.g., a valid IP header)
3158 * or you are setting ip_summed to CHECKSUM_NONE.
3160 static inline void *skb_push_rcsum(struct sk_buff
*skb
, unsigned int len
)
3163 skb_postpush_rcsum(skb
, skb
->data
, len
);
3167 int pskb_trim_rcsum_slow(struct sk_buff
*skb
, unsigned int len
);
3169 * pskb_trim_rcsum - trim received skb and update checksum
3170 * @skb: buffer to trim
3173 * This is exactly the same as pskb_trim except that it ensures the
3174 * checksum of received packets are still valid after the operation.
3175 * It can change skb pointers.
3178 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
3180 if (likely(len
>= skb
->len
))
3182 return pskb_trim_rcsum_slow(skb
, len
);
3185 static inline int __skb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
3187 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3188 skb
->ip_summed
= CHECKSUM_NONE
;
3189 __skb_trim(skb
, len
);
3193 static inline int __skb_grow_rcsum(struct sk_buff
*skb
, unsigned int len
)
3195 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3196 skb
->ip_summed
= CHECKSUM_NONE
;
3197 return __skb_grow(skb
, len
);
3200 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3201 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3202 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3203 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3204 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3206 #define skb_queue_walk(queue, skb) \
3207 for (skb = (queue)->next; \
3208 skb != (struct sk_buff *)(queue); \
3211 #define skb_queue_walk_safe(queue, skb, tmp) \
3212 for (skb = (queue)->next, tmp = skb->next; \
3213 skb != (struct sk_buff *)(queue); \
3214 skb = tmp, tmp = skb->next)
3216 #define skb_queue_walk_from(queue, skb) \
3217 for (; skb != (struct sk_buff *)(queue); \
3220 #define skb_rbtree_walk(skb, root) \
3221 for (skb = skb_rb_first(root); skb != NULL; \
3222 skb = skb_rb_next(skb))
3224 #define skb_rbtree_walk_from(skb) \
3225 for (; skb != NULL; \
3226 skb = skb_rb_next(skb))
3228 #define skb_rbtree_walk_from_safe(skb, tmp) \
3229 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3232 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3233 for (tmp = skb->next; \
3234 skb != (struct sk_buff *)(queue); \
3235 skb = tmp, tmp = skb->next)
3237 #define skb_queue_reverse_walk(queue, skb) \
3238 for (skb = (queue)->prev; \
3239 skb != (struct sk_buff *)(queue); \
3242 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3243 for (skb = (queue)->prev, tmp = skb->prev; \
3244 skb != (struct sk_buff *)(queue); \
3245 skb = tmp, tmp = skb->prev)
3247 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3248 for (tmp = skb->prev; \
3249 skb != (struct sk_buff *)(queue); \
3250 skb = tmp, tmp = skb->prev)
3252 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
3254 return skb_shinfo(skb
)->frag_list
!= NULL
;
3257 static inline void skb_frag_list_init(struct sk_buff
*skb
)
3259 skb_shinfo(skb
)->frag_list
= NULL
;
3262 #define skb_walk_frags(skb, iter) \
3263 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3266 int __skb_wait_for_more_packets(struct sock
*sk
, int *err
, long *timeo_p
,
3267 const struct sk_buff
*skb
);
3268 struct sk_buff
*__skb_try_recv_from_queue(struct sock
*sk
,
3269 struct sk_buff_head
*queue
,
3271 void (*destructor
)(struct sock
*sk
,
3272 struct sk_buff
*skb
),
3273 int *peeked
, int *off
, int *err
,
3274 struct sk_buff
**last
);
3275 struct sk_buff
*__skb_try_recv_datagram(struct sock
*sk
, unsigned flags
,
3276 void (*destructor
)(struct sock
*sk
,
3277 struct sk_buff
*skb
),
3278 int *peeked
, int *off
, int *err
,
3279 struct sk_buff
**last
);
3280 struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
3281 void (*destructor
)(struct sock
*sk
,
3282 struct sk_buff
*skb
),
3283 int *peeked
, int *off
, int *err
);
3284 struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
, int noblock
,
3286 unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
3287 struct poll_table_struct
*wait
);
3288 int skb_copy_datagram_iter(const struct sk_buff
*from
, int offset
,
3289 struct iov_iter
*to
, int size
);
3290 static inline int skb_copy_datagram_msg(const struct sk_buff
*from
, int offset
,
3291 struct msghdr
*msg
, int size
)
3293 return skb_copy_datagram_iter(from
, offset
, &msg
->msg_iter
, size
);
3295 int skb_copy_and_csum_datagram_msg(struct sk_buff
*skb
, int hlen
,
3296 struct msghdr
*msg
);
3297 int skb_copy_datagram_from_iter(struct sk_buff
*skb
, int offset
,
3298 struct iov_iter
*from
, int len
);
3299 int zerocopy_sg_from_iter(struct sk_buff
*skb
, struct iov_iter
*frm
);
3300 void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
3301 void __skb_free_datagram_locked(struct sock
*sk
, struct sk_buff
*skb
, int len
);
3302 static inline void skb_free_datagram_locked(struct sock
*sk
,
3303 struct sk_buff
*skb
)
3305 __skb_free_datagram_locked(sk
, skb
, 0);
3307 int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
, unsigned int flags
);
3308 int skb_copy_bits(const struct sk_buff
*skb
, int offset
, void *to
, int len
);
3309 int skb_store_bits(struct sk_buff
*skb
, int offset
, const void *from
, int len
);
3310 __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
, int offset
, u8
*to
,
3311 int len
, __wsum csum
);
3312 int skb_splice_bits(struct sk_buff
*skb
, struct sock
*sk
, unsigned int offset
,
3313 struct pipe_inode_info
*pipe
, unsigned int len
,
3314 unsigned int flags
);
3315 int skb_send_sock_locked(struct sock
*sk
, struct sk_buff
*skb
, int offset
,
3317 int skb_send_sock(struct sock
*sk
, struct sk_buff
*skb
, int offset
, int len
);
3318 void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
3319 unsigned int skb_zerocopy_headlen(const struct sk_buff
*from
);
3320 int skb_zerocopy(struct sk_buff
*to
, struct sk_buff
*from
,
3322 void skb_split(struct sk_buff
*skb
, struct sk_buff
*skb1
, const u32 len
);
3323 int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
, int shiftlen
);
3324 void skb_scrub_packet(struct sk_buff
*skb
, bool xnet
);
3325 unsigned int skb_gso_transport_seglen(const struct sk_buff
*skb
);
3326 bool skb_gso_validate_mtu(const struct sk_buff
*skb
, unsigned int mtu
);
3327 bool skb_gso_validate_mac_len(const struct sk_buff
*skb
, unsigned int len
);
3328 struct sk_buff
*skb_segment(struct sk_buff
*skb
, netdev_features_t features
);
3329 struct sk_buff
*skb_vlan_untag(struct sk_buff
*skb
);
3330 int skb_ensure_writable(struct sk_buff
*skb
, int write_len
);
3331 int __skb_vlan_pop(struct sk_buff
*skb
, u16
*vlan_tci
);
3332 int skb_vlan_pop(struct sk_buff
*skb
);
3333 int skb_vlan_push(struct sk_buff
*skb
, __be16 vlan_proto
, u16 vlan_tci
);
3334 struct sk_buff
*pskb_extract(struct sk_buff
*skb
, int off
, int to_copy
,
3337 static inline int memcpy_from_msg(void *data
, struct msghdr
*msg
, int len
)
3339 return copy_from_iter_full(data
, len
, &msg
->msg_iter
) ? 0 : -EFAULT
;
3342 static inline int memcpy_to_msg(struct msghdr
*msg
, void *data
, int len
)
3344 return copy_to_iter(data
, len
, &msg
->msg_iter
) == len
? 0 : -EFAULT
;
3347 struct skb_checksum_ops
{
3348 __wsum (*update
)(const void *mem
, int len
, __wsum wsum
);
3349 __wsum (*combine
)(__wsum csum
, __wsum csum2
, int offset
, int len
);
3352 extern const struct skb_checksum_ops
*crc32c_csum_stub __read_mostly
;
3354 __wsum
__skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
3355 __wsum csum
, const struct skb_checksum_ops
*ops
);
3356 __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
3359 static inline void * __must_check
3360 __skb_header_pointer(const struct sk_buff
*skb
, int offset
,
3361 int len
, void *data
, int hlen
, void *buffer
)
3363 if (hlen
- offset
>= len
)
3364 return data
+ offset
;
3367 skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
3373 static inline void * __must_check
3374 skb_header_pointer(const struct sk_buff
*skb
, int offset
, int len
, void *buffer
)
3376 return __skb_header_pointer(skb
, offset
, len
, skb
->data
,
3377 skb_headlen(skb
), buffer
);
3381 * skb_needs_linearize - check if we need to linearize a given skb
3382 * depending on the given device features.
3383 * @skb: socket buffer to check
3384 * @features: net device features
3386 * Returns true if either:
3387 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3388 * 2. skb is fragmented and the device does not support SG.
3390 static inline bool skb_needs_linearize(struct sk_buff
*skb
,
3391 netdev_features_t features
)
3393 return skb_is_nonlinear(skb
) &&
3394 ((skb_has_frag_list(skb
) && !(features
& NETIF_F_FRAGLIST
)) ||
3395 (skb_shinfo(skb
)->nr_frags
&& !(features
& NETIF_F_SG
)));
3398 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
3400 const unsigned int len
)
3402 memcpy(to
, skb
->data
, len
);
3405 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
3406 const int offset
, void *to
,
3407 const unsigned int len
)
3409 memcpy(to
, skb
->data
+ offset
, len
);
3412 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
3414 const unsigned int len
)
3416 memcpy(skb
->data
, from
, len
);
3419 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
3422 const unsigned int len
)
3424 memcpy(skb
->data
+ offset
, from
, len
);
3427 void skb_init(void);
3429 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
3435 * skb_get_timestamp - get timestamp from a skb
3436 * @skb: skb to get stamp from
3437 * @stamp: pointer to struct timeval to store stamp in
3439 * Timestamps are stored in the skb as offsets to a base timestamp.
3440 * This function converts the offset back to a struct timeval and stores
3443 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
3444 struct timeval
*stamp
)
3446 *stamp
= ktime_to_timeval(skb
->tstamp
);
3449 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
3450 struct timespec
*stamp
)
3452 *stamp
= ktime_to_timespec(skb
->tstamp
);
3455 static inline void __net_timestamp(struct sk_buff
*skb
)
3457 skb
->tstamp
= ktime_get_real();
3460 static inline ktime_t
net_timedelta(ktime_t t
)
3462 return ktime_sub(ktime_get_real(), t
);
3465 static inline ktime_t
net_invalid_timestamp(void)
3470 static inline u8
skb_metadata_len(const struct sk_buff
*skb
)
3472 return skb_shinfo(skb
)->meta_len
;
3475 static inline void *skb_metadata_end(const struct sk_buff
*skb
)
3477 return skb_mac_header(skb
);
3480 static inline bool __skb_metadata_differs(const struct sk_buff
*skb_a
,
3481 const struct sk_buff
*skb_b
,
3484 const void *a
= skb_metadata_end(skb_a
);
3485 const void *b
= skb_metadata_end(skb_b
);
3486 /* Using more efficient varaiant than plain call to memcmp(). */
3487 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3491 #define __it(x, op) (x -= sizeof(u##op))
3492 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3493 case 32: diffs
|= __it_diff(a
, b
, 64);
3494 case 24: diffs
|= __it_diff(a
, b
, 64);
3495 case 16: diffs
|= __it_diff(a
, b
, 64);
3496 case 8: diffs
|= __it_diff(a
, b
, 64);
3498 case 28: diffs
|= __it_diff(a
, b
, 64);
3499 case 20: diffs
|= __it_diff(a
, b
, 64);
3500 case 12: diffs
|= __it_diff(a
, b
, 64);
3501 case 4: diffs
|= __it_diff(a
, b
, 32);
3506 return memcmp(a
- meta_len
, b
- meta_len
, meta_len
);
3510 static inline bool skb_metadata_differs(const struct sk_buff
*skb_a
,
3511 const struct sk_buff
*skb_b
)
3513 u8 len_a
= skb_metadata_len(skb_a
);
3514 u8 len_b
= skb_metadata_len(skb_b
);
3516 if (!(len_a
| len_b
))
3519 return len_a
!= len_b
?
3520 true : __skb_metadata_differs(skb_a
, skb_b
, len_a
);
3523 static inline void skb_metadata_set(struct sk_buff
*skb
, u8 meta_len
)
3525 skb_shinfo(skb
)->meta_len
= meta_len
;
3528 static inline void skb_metadata_clear(struct sk_buff
*skb
)
3530 skb_metadata_set(skb
, 0);
3533 struct sk_buff
*skb_clone_sk(struct sk_buff
*skb
);
3535 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3537 void skb_clone_tx_timestamp(struct sk_buff
*skb
);
3538 bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
3540 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3542 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
3546 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
3551 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3554 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3556 * PHY drivers may accept clones of transmitted packets for
3557 * timestamping via their phy_driver.txtstamp method. These drivers
3558 * must call this function to return the skb back to the stack with a
3561 * @skb: clone of the the original outgoing packet
3562 * @hwtstamps: hardware time stamps
3565 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
3566 struct skb_shared_hwtstamps
*hwtstamps
);
3568 void __skb_tstamp_tx(struct sk_buff
*orig_skb
,
3569 struct skb_shared_hwtstamps
*hwtstamps
,
3570 struct sock
*sk
, int tstype
);
3573 * skb_tstamp_tx - queue clone of skb with send time stamps
3574 * @orig_skb: the original outgoing packet
3575 * @hwtstamps: hardware time stamps, may be NULL if not available
3577 * If the skb has a socket associated, then this function clones the
3578 * skb (thus sharing the actual data and optional structures), stores
3579 * the optional hardware time stamping information (if non NULL) or
3580 * generates a software time stamp (otherwise), then queues the clone
3581 * to the error queue of the socket. Errors are silently ignored.
3583 void skb_tstamp_tx(struct sk_buff
*orig_skb
,
3584 struct skb_shared_hwtstamps
*hwtstamps
);
3587 * skb_tx_timestamp() - Driver hook for transmit timestamping
3589 * Ethernet MAC Drivers should call this function in their hard_xmit()
3590 * function immediately before giving the sk_buff to the MAC hardware.
3592 * Specifically, one should make absolutely sure that this function is
3593 * called before TX completion of this packet can trigger. Otherwise
3594 * the packet could potentially already be freed.
3596 * @skb: A socket buffer.
3598 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
3600 skb_clone_tx_timestamp(skb
);
3601 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
)
3602 skb_tstamp_tx(skb
, NULL
);
3606 * skb_complete_wifi_ack - deliver skb with wifi status
3608 * @skb: the original outgoing packet
3609 * @acked: ack status
3612 void skb_complete_wifi_ack(struct sk_buff
*skb
, bool acked
);
3614 __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
3615 __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
3617 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
3619 return ((skb
->ip_summed
== CHECKSUM_UNNECESSARY
) ||
3621 (skb
->ip_summed
== CHECKSUM_PARTIAL
&&
3622 skb_checksum_start_offset(skb
) >= 0));
3626 * skb_checksum_complete - Calculate checksum of an entire packet
3627 * @skb: packet to process
3629 * This function calculates the checksum over the entire packet plus
3630 * the value of skb->csum. The latter can be used to supply the
3631 * checksum of a pseudo header as used by TCP/UDP. It returns the
3634 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3635 * this function can be used to verify that checksum on received
3636 * packets. In that case the function should return zero if the
3637 * checksum is correct. In particular, this function will return zero
3638 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3639 * hardware has already verified the correctness of the checksum.
3641 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
3643 return skb_csum_unnecessary(skb
) ?
3644 0 : __skb_checksum_complete(skb
);
3647 static inline void __skb_decr_checksum_unnecessary(struct sk_buff
*skb
)
3649 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
3650 if (skb
->csum_level
== 0)
3651 skb
->ip_summed
= CHECKSUM_NONE
;
3657 static inline void __skb_incr_checksum_unnecessary(struct sk_buff
*skb
)
3659 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
3660 if (skb
->csum_level
< SKB_MAX_CSUM_LEVEL
)
3662 } else if (skb
->ip_summed
== CHECKSUM_NONE
) {
3663 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
3664 skb
->csum_level
= 0;
3668 /* Check if we need to perform checksum complete validation.
3670 * Returns true if checksum complete is needed, false otherwise
3671 * (either checksum is unnecessary or zero checksum is allowed).
3673 static inline bool __skb_checksum_validate_needed(struct sk_buff
*skb
,
3677 if (skb_csum_unnecessary(skb
) || (zero_okay
&& !check
)) {
3678 skb
->csum_valid
= 1;
3679 __skb_decr_checksum_unnecessary(skb
);
3686 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3689 #define CHECKSUM_BREAK 76
3691 /* Unset checksum-complete
3693 * Unset checksum complete can be done when packet is being modified
3694 * (uncompressed for instance) and checksum-complete value is
3697 static inline void skb_checksum_complete_unset(struct sk_buff
*skb
)
3699 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3700 skb
->ip_summed
= CHECKSUM_NONE
;
3703 /* Validate (init) checksum based on checksum complete.
3706 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3707 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3708 * checksum is stored in skb->csum for use in __skb_checksum_complete
3709 * non-zero: value of invalid checksum
3712 static inline __sum16
__skb_checksum_validate_complete(struct sk_buff
*skb
,
3716 if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
3717 if (!csum_fold(csum_add(psum
, skb
->csum
))) {
3718 skb
->csum_valid
= 1;
3725 if (complete
|| skb
->len
<= CHECKSUM_BREAK
) {
3728 csum
= __skb_checksum_complete(skb
);
3729 skb
->csum_valid
= !csum
;
3736 static inline __wsum
null_compute_pseudo(struct sk_buff
*skb
, int proto
)
3741 /* Perform checksum validate (init). Note that this is a macro since we only
3742 * want to calculate the pseudo header which is an input function if necessary.
3743 * First we try to validate without any computation (checksum unnecessary) and
3744 * then calculate based on checksum complete calling the function to compute
3748 * 0: checksum is validated or try to in skb_checksum_complete
3749 * non-zero: value of invalid checksum
3751 #define __skb_checksum_validate(skb, proto, complete, \
3752 zero_okay, check, compute_pseudo) \
3754 __sum16 __ret = 0; \
3755 skb->csum_valid = 0; \
3756 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3757 __ret = __skb_checksum_validate_complete(skb, \
3758 complete, compute_pseudo(skb, proto)); \
3762 #define skb_checksum_init(skb, proto, compute_pseudo) \
3763 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3765 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3766 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3768 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3769 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3771 #define skb_checksum_validate_zero_check(skb, proto, check, \
3773 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3775 #define skb_checksum_simple_validate(skb) \
3776 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3778 static inline bool __skb_checksum_convert_check(struct sk_buff
*skb
)
3780 return (skb
->ip_summed
== CHECKSUM_NONE
&& skb
->csum_valid
);
3783 static inline void __skb_checksum_convert(struct sk_buff
*skb
,
3784 __sum16 check
, __wsum pseudo
)
3786 skb
->csum
= ~pseudo
;
3787 skb
->ip_summed
= CHECKSUM_COMPLETE
;
3790 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3792 if (__skb_checksum_convert_check(skb)) \
3793 __skb_checksum_convert(skb, check, \
3794 compute_pseudo(skb, proto)); \
3797 static inline void skb_remcsum_adjust_partial(struct sk_buff
*skb
, void *ptr
,
3798 u16 start
, u16 offset
)
3800 skb
->ip_summed
= CHECKSUM_PARTIAL
;
3801 skb
->csum_start
= ((unsigned char *)ptr
+ start
) - skb
->head
;
3802 skb
->csum_offset
= offset
- start
;
3805 /* Update skbuf and packet to reflect the remote checksum offload operation.
3806 * When called, ptr indicates the starting point for skb->csum when
3807 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3808 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3810 static inline void skb_remcsum_process(struct sk_buff
*skb
, void *ptr
,
3811 int start
, int offset
, bool nopartial
)
3816 skb_remcsum_adjust_partial(skb
, ptr
, start
, offset
);
3820 if (unlikely(skb
->ip_summed
!= CHECKSUM_COMPLETE
)) {
3821 __skb_checksum_complete(skb
);
3822 skb_postpull_rcsum(skb
, skb
->data
, ptr
- (void *)skb
->data
);
3825 delta
= remcsum_adjust(ptr
, skb
->csum
, start
, offset
);
3827 /* Adjust skb->csum since we changed the packet */
3828 skb
->csum
= csum_add(skb
->csum
, delta
);
3831 static inline struct nf_conntrack
*skb_nfct(const struct sk_buff
*skb
)
3833 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3834 return (void *)(skb
->_nfct
& SKB_NFCT_PTRMASK
);
3840 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3841 void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
3842 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
3844 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
3845 nf_conntrack_destroy(nfct
);
3847 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
3850 atomic_inc(&nfct
->use
);
3853 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3854 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
3856 if (nf_bridge
&& refcount_dec_and_test(&nf_bridge
->use
))
3859 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
3862 refcount_inc(&nf_bridge
->use
);
3864 #endif /* CONFIG_BRIDGE_NETFILTER */
3865 static inline void nf_reset(struct sk_buff
*skb
)
3867 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3868 nf_conntrack_put(skb_nfct(skb
));
3871 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3872 nf_bridge_put(skb
->nf_bridge
);
3873 skb
->nf_bridge
= NULL
;
3877 static inline void nf_reset_trace(struct sk_buff
*skb
)
3879 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3884 static inline void ipvs_reset(struct sk_buff
*skb
)
3886 #if IS_ENABLED(CONFIG_IP_VS)
3887 skb
->ipvs_property
= 0;
3891 /* Note: This doesn't put any conntrack and bridge info in dst. */
3892 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
,
3895 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3896 dst
->_nfct
= src
->_nfct
;
3897 nf_conntrack_get(skb_nfct(src
));
3899 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3900 dst
->nf_bridge
= src
->nf_bridge
;
3901 nf_bridge_get(src
->nf_bridge
);
3903 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3905 dst
->nf_trace
= src
->nf_trace
;
3909 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
3911 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3912 nf_conntrack_put(skb_nfct(dst
));
3914 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3915 nf_bridge_put(dst
->nf_bridge
);
3917 __nf_copy(dst
, src
, true);
3920 #ifdef CONFIG_NETWORK_SECMARK
3921 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3923 to
->secmark
= from
->secmark
;
3926 static inline void skb_init_secmark(struct sk_buff
*skb
)
3931 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3934 static inline void skb_init_secmark(struct sk_buff
*skb
)
3938 static inline bool skb_irq_freeable(const struct sk_buff
*skb
)
3940 return !skb
->destructor
&&
3941 #if IS_ENABLED(CONFIG_XFRM)
3945 !skb
->_skb_refdst
&&
3946 !skb_has_frag_list(skb
);
3949 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
3951 skb
->queue_mapping
= queue_mapping
;
3954 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
3956 return skb
->queue_mapping
;
3959 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
3961 to
->queue_mapping
= from
->queue_mapping
;
3964 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
3966 skb
->queue_mapping
= rx_queue
+ 1;
3969 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
3971 return skb
->queue_mapping
- 1;
3974 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
3976 return skb
->queue_mapping
!= 0;
3979 static inline void skb_set_dst_pending_confirm(struct sk_buff
*skb
, u32 val
)
3981 skb
->dst_pending_confirm
= val
;
3984 static inline bool skb_get_dst_pending_confirm(const struct sk_buff
*skb
)
3986 return skb
->dst_pending_confirm
!= 0;
3989 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
3998 /* Keeps track of mac header offset relative to skb->head.
3999 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4000 * For non-tunnel skb it points to skb_mac_header() and for
4001 * tunnel skb it points to outer mac header.
4002 * Keeps track of level of encapsulation of network headers.
4013 #define SKB_SGO_CB_OFFSET 32
4014 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
4016 static inline int skb_tnl_header_len(const struct sk_buff
*inner_skb
)
4018 return (skb_mac_header(inner_skb
) - inner_skb
->head
) -
4019 SKB_GSO_CB(inner_skb
)->mac_offset
;
4022 static inline int gso_pskb_expand_head(struct sk_buff
*skb
, int extra
)
4024 int new_headroom
, headroom
;
4027 headroom
= skb_headroom(skb
);
4028 ret
= pskb_expand_head(skb
, extra
, 0, GFP_ATOMIC
);
4032 new_headroom
= skb_headroom(skb
);
4033 SKB_GSO_CB(skb
)->mac_offset
+= (new_headroom
- headroom
);
4037 static inline void gso_reset_checksum(struct sk_buff
*skb
, __wsum res
)
4039 /* Do not update partial checksums if remote checksum is enabled. */
4040 if (skb
->remcsum_offload
)
4043 SKB_GSO_CB(skb
)->csum
= res
;
4044 SKB_GSO_CB(skb
)->csum_start
= skb_checksum_start(skb
) - skb
->head
;
4047 /* Compute the checksum for a gso segment. First compute the checksum value
4048 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4049 * then add in skb->csum (checksum from csum_start to end of packet).
4050 * skb->csum and csum_start are then updated to reflect the checksum of the
4051 * resultant packet starting from the transport header-- the resultant checksum
4052 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4055 static inline __sum16
gso_make_checksum(struct sk_buff
*skb
, __wsum res
)
4057 unsigned char *csum_start
= skb_transport_header(skb
);
4058 int plen
= (skb
->head
+ SKB_GSO_CB(skb
)->csum_start
) - csum_start
;
4059 __wsum partial
= SKB_GSO_CB(skb
)->csum
;
4061 SKB_GSO_CB(skb
)->csum
= res
;
4062 SKB_GSO_CB(skb
)->csum_start
= csum_start
- skb
->head
;
4064 return csum_fold(csum_partial(csum_start
, plen
, partial
));
4067 static inline bool skb_is_gso(const struct sk_buff
*skb
)
4069 return skb_shinfo(skb
)->gso_size
;
4072 /* Note: Should be called only if skb_is_gso(skb) is true */
4073 static inline bool skb_is_gso_v6(const struct sk_buff
*skb
)
4075 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
4078 static inline void skb_gso_reset(struct sk_buff
*skb
)
4080 skb_shinfo(skb
)->gso_size
= 0;
4081 skb_shinfo(skb
)->gso_segs
= 0;
4082 skb_shinfo(skb
)->gso_type
= 0;
4085 void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
4087 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
4089 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4090 * wanted then gso_type will be set. */
4091 const struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
4093 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
4094 unlikely(shinfo
->gso_type
== 0)) {
4095 __skb_warn_lro_forwarding(skb
);
4101 static inline void skb_forward_csum(struct sk_buff
*skb
)
4103 /* Unfortunately we don't support this one. Any brave souls? */
4104 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
4105 skb
->ip_summed
= CHECKSUM_NONE
;
4109 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4110 * @skb: skb to check
4112 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4113 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4114 * use this helper, to document places where we make this assertion.
4116 static inline void skb_checksum_none_assert(const struct sk_buff
*skb
)
4119 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
4123 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
4125 int skb_checksum_setup(struct sk_buff
*skb
, bool recalculate
);
4126 struct sk_buff
*skb_checksum_trimmed(struct sk_buff
*skb
,
4127 unsigned int transport_len
,
4128 __sum16(*skb_chkf
)(struct sk_buff
*skb
));
4131 * skb_head_is_locked - Determine if the skb->head is locked down
4132 * @skb: skb to check
4134 * The head on skbs build around a head frag can be removed if they are
4135 * not cloned. This function returns true if the skb head is locked down
4136 * due to either being allocated via kmalloc, or by being a clone with
4137 * multiple references to the head.
4139 static inline bool skb_head_is_locked(const struct sk_buff
*skb
)
4141 return !skb
->head_frag
|| skb_cloned(skb
);
4145 * skb_gso_network_seglen - Return length of individual segments of a gso packet
4149 * skb_gso_network_seglen is used to determine the real size of the
4150 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
4152 * The MAC/L2 header is not accounted for.
4154 static inline unsigned int skb_gso_network_seglen(const struct sk_buff
*skb
)
4156 unsigned int hdr_len
= skb_transport_header(skb
) -
4157 skb_network_header(skb
);
4158 return hdr_len
+ skb_gso_transport_seglen(skb
);
4162 * skb_gso_mac_seglen - Return length of individual segments of a gso packet
4166 * skb_gso_mac_seglen is used to determine the real size of the
4167 * individual segments, including MAC/L2, Layer3 (IP, IPv6) and L4
4168 * headers (TCP/UDP).
4170 static inline unsigned int skb_gso_mac_seglen(const struct sk_buff
*skb
)
4172 unsigned int hdr_len
= skb_transport_header(skb
) - skb_mac_header(skb
);
4173 return hdr_len
+ skb_gso_transport_seglen(skb
);
4176 /* Local Checksum Offload.
4177 * Compute outer checksum based on the assumption that the
4178 * inner checksum will be offloaded later.
4179 * See Documentation/networking/checksum-offloads.txt for
4180 * explanation of how this works.
4181 * Fill in outer checksum adjustment (e.g. with sum of outer
4182 * pseudo-header) before calling.
4183 * Also ensure that inner checksum is in linear data area.
4185 static inline __wsum
lco_csum(struct sk_buff
*skb
)
4187 unsigned char *csum_start
= skb_checksum_start(skb
);
4188 unsigned char *l4_hdr
= skb_transport_header(skb
);
4191 /* Start with complement of inner checksum adjustment */
4192 partial
= ~csum_unfold(*(__force __sum16
*)(csum_start
+
4195 /* Add in checksum of our headers (incl. outer checksum
4196 * adjustment filled in by caller) and return result.
4198 return csum_partial(l4_hdr
, csum_start
- l4_hdr
, partial
);
4201 #endif /* __KERNEL__ */
4202 #endif /* _LINUX_SKBUFF_H */