2 * Definitions for the 'struct sk_buff' memory handlers.
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/kernel.h>
18 #include <linux/kmemcheck.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/bug.h>
22 #include <linux/cache.h>
23 #include <linux/rbtree.h>
24 #include <linux/socket.h>
25 #include <linux/refcount.h>
27 #include <linux/atomic.h>
28 #include <asm/types.h>
29 #include <linux/spinlock.h>
30 #include <linux/net.h>
31 #include <linux/textsearch.h>
32 #include <net/checksum.h>
33 #include <linux/rcupdate.h>
34 #include <linux/hrtimer.h>
35 #include <linux/dma-mapping.h>
36 #include <linux/netdev_features.h>
37 #include <linux/sched.h>
38 #include <linux/sched/clock.h>
39 #include <net/flow_dissector.h>
40 #include <linux/splice.h>
41 #include <linux/in6.h>
42 #include <linux/if_packet.h>
45 /* The interface for checksum offload between the stack and networking drivers
48 * A. IP checksum related features
50 * Drivers advertise checksum offload capabilities in the features of a device.
51 * From the stack's point of view these are capabilities offered by the driver,
52 * a driver typically only advertises features that it is capable of offloading
55 * The checksum related features are:
57 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
58 * IP (one's complement) checksum for any combination
59 * of protocols or protocol layering. The checksum is
60 * computed and set in a packet per the CHECKSUM_PARTIAL
61 * interface (see below).
63 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
64 * TCP or UDP packets over IPv4. These are specifically
65 * unencapsulated packets of the form IPv4|TCP or
66 * IPv4|UDP where the Protocol field in the IPv4 header
67 * is TCP or UDP. The IPv4 header may contain IP options
68 * This feature cannot be set in features for a device
69 * with NETIF_F_HW_CSUM also set. This feature is being
70 * DEPRECATED (see below).
72 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
73 * TCP or UDP packets over IPv6. These are specifically
74 * unencapsulated packets of the form IPv6|TCP or
75 * IPv4|UDP where the Next Header field in the IPv6
76 * header is either TCP or UDP. IPv6 extension headers
77 * are not supported with this feature. This feature
78 * cannot be set in features for a device with
79 * NETIF_F_HW_CSUM also set. This feature is being
80 * DEPRECATED (see below).
82 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
83 * This flag is used only used to disable the RX checksum
84 * feature for a device. The stack will accept receive
85 * checksum indication in packets received on a device
86 * regardless of whether NETIF_F_RXCSUM is set.
88 * B. Checksumming of received packets by device. Indication of checksum
89 * verification is in set skb->ip_summed. Possible values are:
93 * Device did not checksum this packet e.g. due to lack of capabilities.
94 * The packet contains full (though not verified) checksum in packet but
95 * not in skb->csum. Thus, skb->csum is undefined in this case.
97 * CHECKSUM_UNNECESSARY:
99 * The hardware you're dealing with doesn't calculate the full checksum
100 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
101 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
102 * if their checksums are okay. skb->csum is still undefined in this case
103 * though. A driver or device must never modify the checksum field in the
104 * packet even if checksum is verified.
106 * CHECKSUM_UNNECESSARY is applicable to following protocols:
107 * TCP: IPv6 and IPv4.
108 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
109 * zero UDP checksum for either IPv4 or IPv6, the networking stack
110 * may perform further validation in this case.
111 * GRE: only if the checksum is present in the header.
112 * SCTP: indicates the CRC in SCTP header has been validated.
113 * FCOE: indicates the CRC in FC frame has been validated.
115 * skb->csum_level indicates the number of consecutive checksums found in
116 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
117 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
118 * and a device is able to verify the checksums for UDP (possibly zero),
119 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
120 * two. If the device were only able to verify the UDP checksum and not
121 * GRE, either because it doesn't support GRE checksum of because GRE
122 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
123 * not considered in this case).
127 * This is the most generic way. The device supplied checksum of the _whole_
128 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
129 * hardware doesn't need to parse L3/L4 headers to implement this.
132 * - Even if device supports only some protocols, but is able to produce
133 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
134 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
138 * A checksum is set up to be offloaded to a device as described in the
139 * output description for CHECKSUM_PARTIAL. This may occur on a packet
140 * received directly from another Linux OS, e.g., a virtualized Linux kernel
141 * on the same host, or it may be set in the input path in GRO or remote
142 * checksum offload. For the purposes of checksum verification, the checksum
143 * referred to by skb->csum_start + skb->csum_offset and any preceding
144 * checksums in the packet are considered verified. Any checksums in the
145 * packet that are after the checksum being offloaded are not considered to
148 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
149 * in the skb->ip_summed for a packet. Values are:
153 * The driver is required to checksum the packet as seen by hard_start_xmit()
154 * from skb->csum_start up to the end, and to record/write the checksum at
155 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
156 * csum_start and csum_offset values are valid values given the length and
157 * offset of the packet, however they should not attempt to validate that the
158 * checksum refers to a legitimate transport layer checksum-- it is the
159 * purview of the stack to validate that csum_start and csum_offset are set
162 * When the stack requests checksum offload for a packet, the driver MUST
163 * ensure that the checksum is set correctly. A driver can either offload the
164 * checksum calculation to the device, or call skb_checksum_help (in the case
165 * that the device does not support offload for a particular checksum).
167 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
168 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
169 * checksum offload capability.
170 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
171 * on network device checksumming capabilities: if a packet does not match
172 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
173 * csum_not_inet, see item D.) is called to resolve the checksum.
177 * The skb was already checksummed by the protocol, or a checksum is not
180 * CHECKSUM_UNNECESSARY:
182 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
186 * Not used in checksum output. If a driver observes a packet with this value
187 * set in skbuff, if should treat as CHECKSUM_NONE being set.
189 * D. Non-IP checksum (CRC) offloads
191 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
192 * offloading the SCTP CRC in a packet. To perform this offload the stack
193 * will set set csum_start and csum_offset accordingly, set ip_summed to
194 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
195 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
196 * A driver that supports both IP checksum offload and SCTP CRC32c offload
197 * must verify which offload is configured for a packet by testing the
198 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
199 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
201 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
202 * offloading the FCOE CRC in a packet. To perform this offload the stack
203 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
204 * accordingly. Note the there is no indication in the skbuff that the
205 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
206 * both IP checksum offload and FCOE CRC offload must verify which offload
207 * is configured for a packet presumably by inspecting packet headers.
209 * E. Checksumming on output with GSO.
211 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
212 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
213 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
214 * part of the GSO operation is implied. If a checksum is being offloaded
215 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
216 * are set to refer to the outermost checksum being offload (two offloaded
217 * checksums are possible with UDP encapsulation).
220 /* Don't change this without changing skb_csum_unnecessary! */
221 #define CHECKSUM_NONE 0
222 #define CHECKSUM_UNNECESSARY 1
223 #define CHECKSUM_COMPLETE 2
224 #define CHECKSUM_PARTIAL 3
226 /* Maximum value in skb->csum_level */
227 #define SKB_MAX_CSUM_LEVEL 3
229 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
230 #define SKB_WITH_OVERHEAD(X) \
231 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
232 #define SKB_MAX_ORDER(X, ORDER) \
233 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
234 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
235 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
237 /* return minimum truesize of one skb containing X bytes of data */
238 #define SKB_TRUESIZE(X) ((X) + \
239 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
240 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
244 struct pipe_inode_info
;
248 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
249 struct nf_conntrack
{
254 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
255 struct nf_bridge_info
{
258 BRNF_PROTO_UNCHANGED
,
266 struct net_device
*physindev
;
268 /* always valid & non-NULL from FORWARD on, for physdev match */
269 struct net_device
*physoutdev
;
271 /* prerouting: detect dnat in orig/reply direction */
273 struct in6_addr ipv6_daddr
;
275 /* after prerouting + nat detected: store original source
276 * mac since neigh resolution overwrites it, only used while
277 * skb is out in neigh layer.
279 char neigh_header
[8];
284 struct sk_buff_head
{
285 /* These two members must be first. */
286 struct sk_buff
*next
;
287 struct sk_buff
*prev
;
295 /* To allow 64K frame to be packed as single skb without frag_list we
296 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
297 * buffers which do not start on a page boundary.
299 * Since GRO uses frags we allocate at least 16 regardless of page
302 #if (65536/PAGE_SIZE + 1) < 16
303 #define MAX_SKB_FRAGS 16UL
305 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
307 extern int sysctl_max_skb_frags
;
309 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
310 * segment using its current segmentation instead.
312 #define GSO_BY_FRAGS 0xFFFF
314 typedef struct skb_frag_struct skb_frag_t
;
316 struct skb_frag_struct
{
320 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
329 static inline unsigned int skb_frag_size(const skb_frag_t
*frag
)
334 static inline void skb_frag_size_set(skb_frag_t
*frag
, unsigned int size
)
339 static inline void skb_frag_size_add(skb_frag_t
*frag
, int delta
)
344 static inline void skb_frag_size_sub(skb_frag_t
*frag
, int delta
)
349 static inline bool skb_frag_must_loop(struct page
*p
)
351 #if defined(CONFIG_HIGHMEM)
359 * skb_frag_foreach_page - loop over pages in a fragment
361 * @f: skb frag to operate on
362 * @f_off: offset from start of f->page.p
363 * @f_len: length from f_off to loop over
364 * @p: (temp var) current page
365 * @p_off: (temp var) offset from start of current page,
366 * non-zero only on first page.
367 * @p_len: (temp var) length in current page,
368 * < PAGE_SIZE only on first and last page.
369 * @copied: (temp var) length so far, excluding current p_len.
371 * A fragment can hold a compound page, in which case per-page
372 * operations, notably kmap_atomic, must be called for each
375 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
376 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
377 p_off = (f_off) & (PAGE_SIZE - 1), \
378 p_len = skb_frag_must_loop(p) ? \
379 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
382 copied += p_len, p++, p_off = 0, \
383 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
385 #define HAVE_HW_TIME_STAMP
388 * struct skb_shared_hwtstamps - hardware time stamps
389 * @hwtstamp: hardware time stamp transformed into duration
390 * since arbitrary point in time
392 * Software time stamps generated by ktime_get_real() are stored in
395 * hwtstamps can only be compared against other hwtstamps from
398 * This structure is attached to packets as part of the
399 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
401 struct skb_shared_hwtstamps
{
405 /* Definitions for tx_flags in struct skb_shared_info */
407 /* generate hardware time stamp */
408 SKBTX_HW_TSTAMP
= 1 << 0,
410 /* generate software time stamp when queueing packet to NIC */
411 SKBTX_SW_TSTAMP
= 1 << 1,
413 /* device driver is going to provide hardware time stamp */
414 SKBTX_IN_PROGRESS
= 1 << 2,
416 /* device driver supports TX zero-copy buffers */
417 SKBTX_DEV_ZEROCOPY
= 1 << 3,
419 /* generate wifi status information (where possible) */
420 SKBTX_WIFI_STATUS
= 1 << 4,
422 /* This indicates at least one fragment might be overwritten
423 * (as in vmsplice(), sendfile() ...)
424 * If we need to compute a TX checksum, we'll need to copy
425 * all frags to avoid possible bad checksum
427 SKBTX_SHARED_FRAG
= 1 << 5,
429 /* generate software time stamp when entering packet scheduling */
430 SKBTX_SCHED_TSTAMP
= 1 << 6,
433 #define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
434 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
436 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
439 * The callback notifies userspace to release buffers when skb DMA is done in
440 * lower device, the skb last reference should be 0 when calling this.
441 * The zerocopy_success argument is true if zero copy transmit occurred,
442 * false on data copy or out of memory error caused by data copy attempt.
443 * The ctx field is used to track device context.
444 * The desc field is used to track userspace buffer index.
447 void (*callback
)(struct ubuf_info
*, bool zerocopy_success
);
463 struct user_struct
*user
;
468 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
470 struct ubuf_info
*sock_zerocopy_alloc(struct sock
*sk
, size_t size
);
471 struct ubuf_info
*sock_zerocopy_realloc(struct sock
*sk
, size_t size
,
472 struct ubuf_info
*uarg
);
474 static inline void sock_zerocopy_get(struct ubuf_info
*uarg
)
476 refcount_inc(&uarg
->refcnt
);
479 void sock_zerocopy_put(struct ubuf_info
*uarg
);
480 void sock_zerocopy_put_abort(struct ubuf_info
*uarg
);
482 void sock_zerocopy_callback(struct ubuf_info
*uarg
, bool success
);
484 int skb_zerocopy_iter_stream(struct sock
*sk
, struct sk_buff
*skb
,
485 struct msghdr
*msg
, int len
,
486 struct ubuf_info
*uarg
);
488 /* This data is invariant across clones and lives at
489 * the end of the header data, ie. at skb->end.
491 struct skb_shared_info
{
496 unsigned short gso_size
;
497 /* Warning: this field is not always filled in (UFO)! */
498 unsigned short gso_segs
;
499 struct sk_buff
*frag_list
;
500 struct skb_shared_hwtstamps hwtstamps
;
501 unsigned int gso_type
;
506 * Warning : all fields before dataref are cleared in __alloc_skb()
510 /* Intermediate layers must ensure that destructor_arg
511 * remains valid until skb destructor */
512 void * destructor_arg
;
514 /* must be last field, see pskb_expand_head() */
515 skb_frag_t frags
[MAX_SKB_FRAGS
];
518 /* We divide dataref into two halves. The higher 16 bits hold references
519 * to the payload part of skb->data. The lower 16 bits hold references to
520 * the entire skb->data. A clone of a headerless skb holds the length of
521 * the header in skb->hdr_len.
523 * All users must obey the rule that the skb->data reference count must be
524 * greater than or equal to the payload reference count.
526 * Holding a reference to the payload part means that the user does not
527 * care about modifications to the header part of skb->data.
529 #define SKB_DATAREF_SHIFT 16
530 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
534 SKB_FCLONE_UNAVAILABLE
, /* skb has no fclone (from head_cache) */
535 SKB_FCLONE_ORIG
, /* orig skb (from fclone_cache) */
536 SKB_FCLONE_CLONE
, /* companion fclone skb (from fclone_cache) */
540 SKB_GSO_TCPV4
= 1 << 0,
542 /* This indicates the skb is from an untrusted source. */
543 SKB_GSO_DODGY
= 1 << 1,
545 /* This indicates the tcp segment has CWR set. */
546 SKB_GSO_TCP_ECN
= 1 << 2,
548 SKB_GSO_TCP_FIXEDID
= 1 << 3,
550 SKB_GSO_TCPV6
= 1 << 4,
552 SKB_GSO_FCOE
= 1 << 5,
554 SKB_GSO_GRE
= 1 << 6,
556 SKB_GSO_GRE_CSUM
= 1 << 7,
558 SKB_GSO_IPXIP4
= 1 << 8,
560 SKB_GSO_IPXIP6
= 1 << 9,
562 SKB_GSO_UDP_TUNNEL
= 1 << 10,
564 SKB_GSO_UDP_TUNNEL_CSUM
= 1 << 11,
566 SKB_GSO_PARTIAL
= 1 << 12,
568 SKB_GSO_TUNNEL_REMCSUM
= 1 << 13,
570 SKB_GSO_SCTP
= 1 << 14,
572 SKB_GSO_ESP
= 1 << 15,
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 * @ndisc_nodetype: router type (from link layer)
629 * @ooo_okay: allow the mapping of a socket to a queue to be changed
630 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
632 * @sw_hash: indicates hash was computed in software stack
633 * @wifi_acked_valid: wifi_acked was set
634 * @wifi_acked: whether frame was acked on wifi or not
635 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
636 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
637 * @dst_pending_confirm: need to confirm neighbour
638 * @napi_id: id of the NAPI struct this skb came from
639 * @secmark: security marking
640 * @mark: Generic packet mark
641 * @vlan_proto: vlan encapsulation protocol
642 * @vlan_tci: vlan tag control information
643 * @inner_protocol: Protocol (encapsulation)
644 * @inner_transport_header: Inner transport layer header (encapsulation)
645 * @inner_network_header: Network layer header (encapsulation)
646 * @inner_mac_header: Link layer header (encapsulation)
647 * @transport_header: Transport layer header
648 * @network_header: Network layer header
649 * @mac_header: Link layer header
650 * @tail: Tail pointer
652 * @head: Head of buffer
653 * @data: Data head pointer
654 * @truesize: Buffer size
655 * @users: User count - see {datagram,tcp}.c
661 /* These two members must be first. */
662 struct sk_buff
*next
;
663 struct sk_buff
*prev
;
666 struct net_device
*dev
;
667 /* Some protocols might use this space to store information,
668 * while device pointer would be NULL.
669 * UDP receive path is one user.
671 unsigned long dev_scratch
;
674 struct rb_node rbnode
; /* used in netem & tcp stack */
683 * This is the control buffer. It is free to use for every
684 * layer. Please put your private variables there. If you
685 * want to keep them across layers you have to do a skb_clone()
686 * first. This is owned by whoever has the skb queued ATM.
688 char cb
[48] __aligned(8);
692 unsigned long _skb_refdst
;
693 void (*destructor
)(struct sk_buff
*skb
);
695 struct list_head tcp_tsorted_anchor
;
701 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
704 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
705 struct nf_bridge_info
*nf_bridge
;
712 /* Following fields are _not_ copied in __copy_skb_header()
713 * Note that queue_mapping is here mostly to fill a hole.
715 kmemcheck_bitfield_begin(flags1
);
718 /* if you move cloned around you also must adapt those constants */
719 #ifdef __BIG_ENDIAN_BITFIELD
720 #define CLONED_MASK (1 << 7)
722 #define CLONED_MASK 1
724 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
726 __u8 __cloned_offset
[0];
733 __unused
:1; /* one bit hole */
734 kmemcheck_bitfield_end(flags1
);
736 /* fields enclosed in headers_start/headers_end are copied
737 * using a single memcpy() in __copy_skb_header()
740 __u32 headers_start
[0];
743 /* if you move pkt_type around you also must adapt those constants */
744 #ifdef __BIG_ENDIAN_BITFIELD
745 #define PKT_TYPE_MAX (7 << 5)
747 #define PKT_TYPE_MAX 7
749 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
751 __u8 __pkt_type_offset
[0];
761 __u8 wifi_acked_valid
:1;
765 /* Indicates the inner headers are valid in the skbuff. */
766 __u8 encapsulation
:1;
767 __u8 encap_hdr_csum
:1;
769 __u8 csum_complete_sw
:1;
771 __u8 csum_not_inet
:1;
773 __u8 dst_pending_confirm
:1;
774 #ifdef CONFIG_IPV6_NDISC_NODETYPE
775 __u8 ndisc_nodetype
:2;
777 __u8 ipvs_property
:1;
778 __u8 inner_protocol_type
:1;
779 __u8 remcsum_offload
:1;
780 #ifdef CONFIG_NET_SWITCHDEV
781 __u8 offload_fwd_mark
:1;
782 __u8 offload_mr_fwd_mark
:1;
784 #ifdef CONFIG_NET_CLS_ACT
785 __u8 tc_skip_classify
:1;
786 __u8 tc_at_ingress
:1;
787 __u8 tc_redirected
:1;
788 __u8 tc_from_ingress
:1;
791 #ifdef CONFIG_NET_SCHED
792 __u16 tc_index
; /* traffic control index */
807 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
809 unsigned int napi_id
;
810 unsigned int sender_cpu
;
813 #ifdef CONFIG_NETWORK_SECMARK
819 __u32 reserved_tailroom
;
823 __be16 inner_protocol
;
827 __u16 inner_transport_header
;
828 __u16 inner_network_header
;
829 __u16 inner_mac_header
;
832 __u16 transport_header
;
833 __u16 network_header
;
837 __u32 headers_end
[0];
840 /* These elements must be at the end, see alloc_skb() for details. */
845 unsigned int truesize
;
851 * Handling routines are only of interest to the kernel
853 #include <linux/slab.h>
856 #define SKB_ALLOC_FCLONE 0x01
857 #define SKB_ALLOC_RX 0x02
858 #define SKB_ALLOC_NAPI 0x04
860 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
861 static inline bool skb_pfmemalloc(const struct sk_buff
*skb
)
863 return unlikely(skb
->pfmemalloc
);
867 * skb might have a dst pointer attached, refcounted or not.
868 * _skb_refdst low order bit is set if refcount was _not_ taken
870 #define SKB_DST_NOREF 1UL
871 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
873 #define SKB_NFCT_PTRMASK ~(7UL)
875 * skb_dst - returns skb dst_entry
878 * Returns skb dst_entry, regardless of reference taken or not.
880 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
882 /* If refdst was not refcounted, check we still are in a
883 * rcu_read_lock section
885 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
886 !rcu_read_lock_held() &&
887 !rcu_read_lock_bh_held());
888 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
892 * skb_dst_set - sets skb dst
896 * Sets skb dst, assuming a reference was taken on dst and should
897 * be released by skb_dst_drop()
899 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
901 skb
->_skb_refdst
= (unsigned long)dst
;
905 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
909 * Sets skb dst, assuming a reference was not taken on dst.
910 * If dst entry is cached, we do not take reference and dst_release
911 * will be avoided by refdst_drop. If dst entry is not cached, we take
912 * reference, so that last dst_release can destroy the dst immediately.
914 static inline void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
)
916 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
917 skb
->_skb_refdst
= (unsigned long)dst
| SKB_DST_NOREF
;
921 * skb_dst_is_noref - Test if skb dst isn't refcounted
924 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
926 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
929 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
931 return (struct rtable
*)skb_dst(skb
);
934 /* For mangling skb->pkt_type from user space side from applications
935 * such as nft, tc, etc, we only allow a conservative subset of
936 * possible pkt_types to be set.
938 static inline bool skb_pkt_type_ok(u32 ptype
)
940 return ptype
<= PACKET_OTHERHOST
;
943 static inline unsigned int skb_napi_id(const struct sk_buff
*skb
)
945 #ifdef CONFIG_NET_RX_BUSY_POLL
952 /* decrement the reference count and return true if we can free the skb */
953 static inline bool skb_unref(struct sk_buff
*skb
)
957 if (likely(refcount_read(&skb
->users
) == 1))
959 else if (likely(!refcount_dec_and_test(&skb
->users
)))
965 void skb_release_head_state(struct sk_buff
*skb
);
966 void kfree_skb(struct sk_buff
*skb
);
967 void kfree_skb_list(struct sk_buff
*segs
);
968 void skb_tx_error(struct sk_buff
*skb
);
969 void consume_skb(struct sk_buff
*skb
);
970 void __consume_stateless_skb(struct sk_buff
*skb
);
971 void __kfree_skb(struct sk_buff
*skb
);
972 extern struct kmem_cache
*skbuff_head_cache
;
974 void kfree_skb_partial(struct sk_buff
*skb
, bool head_stolen
);
975 bool skb_try_coalesce(struct sk_buff
*to
, struct sk_buff
*from
,
976 bool *fragstolen
, int *delta_truesize
);
978 struct sk_buff
*__alloc_skb(unsigned int size
, gfp_t priority
, int flags
,
980 struct sk_buff
*__build_skb(void *data
, unsigned int frag_size
);
981 struct sk_buff
*build_skb(void *data
, unsigned int frag_size
);
982 static inline struct sk_buff
*alloc_skb(unsigned int size
,
985 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
988 struct sk_buff
*alloc_skb_with_frags(unsigned long header_len
,
989 unsigned long data_len
,
994 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
995 struct sk_buff_fclones
{
1000 refcount_t fclone_ref
;
1004 * skb_fclone_busy - check if fclone is busy
1008 * Returns true if skb is a fast clone, and its clone is not freed.
1009 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1010 * so we also check that this didnt happen.
1012 static inline bool skb_fclone_busy(const struct sock
*sk
,
1013 const struct sk_buff
*skb
)
1015 const struct sk_buff_fclones
*fclones
;
1017 fclones
= container_of(skb
, struct sk_buff_fclones
, skb1
);
1019 return skb
->fclone
== SKB_FCLONE_ORIG
&&
1020 refcount_read(&fclones
->fclone_ref
) > 1 &&
1021 fclones
->skb2
.sk
== sk
;
1024 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
1027 return __alloc_skb(size
, priority
, SKB_ALLOC_FCLONE
, NUMA_NO_NODE
);
1030 struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
1031 int skb_copy_ubufs(struct sk_buff
*skb
, gfp_t gfp_mask
);
1032 struct sk_buff
*skb_clone(struct sk_buff
*skb
, gfp_t priority
);
1033 struct sk_buff
*skb_copy(const struct sk_buff
*skb
, gfp_t priority
);
1034 struct sk_buff
*__pskb_copy_fclone(struct sk_buff
*skb
, int headroom
,
1035 gfp_t gfp_mask
, bool fclone
);
1036 static inline struct sk_buff
*__pskb_copy(struct sk_buff
*skb
, int headroom
,
1039 return __pskb_copy_fclone(skb
, headroom
, gfp_mask
, false);
1042 int pskb_expand_head(struct sk_buff
*skb
, int nhead
, int ntail
, gfp_t gfp_mask
);
1043 struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
1044 unsigned int headroom
);
1045 struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
, int newheadroom
,
1046 int newtailroom
, gfp_t priority
);
1047 int __must_check
skb_to_sgvec_nomark(struct sk_buff
*skb
, struct scatterlist
*sg
,
1048 int offset
, int len
);
1049 int __must_check
skb_to_sgvec(struct sk_buff
*skb
, struct scatterlist
*sg
,
1050 int offset
, int len
);
1051 int skb_cow_data(struct sk_buff
*skb
, int tailbits
, struct sk_buff
**trailer
);
1052 int __skb_pad(struct sk_buff
*skb
, int pad
, bool free_on_error
);
1055 * skb_pad - zero pad the tail of an skb
1056 * @skb: buffer to pad
1057 * @pad: space to pad
1059 * Ensure that a buffer is followed by a padding area that is zero
1060 * filled. Used by network drivers which may DMA or transfer data
1061 * beyond the buffer end onto the wire.
1063 * May return error in out of memory cases. The skb is freed on error.
1065 static inline int skb_pad(struct sk_buff
*skb
, int pad
)
1067 return __skb_pad(skb
, pad
, true);
1069 #define dev_kfree_skb(a) consume_skb(a)
1071 int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
1072 int getfrag(void *from
, char *to
, int offset
,
1073 int len
, int odd
, struct sk_buff
*skb
),
1074 void *from
, int length
);
1076 int skb_append_pagefrags(struct sk_buff
*skb
, struct page
*page
,
1077 int offset
, size_t size
);
1079 struct skb_seq_state
{
1083 __u32 stepped_offset
;
1084 struct sk_buff
*root_skb
;
1085 struct sk_buff
*cur_skb
;
1089 void skb_prepare_seq_read(struct sk_buff
*skb
, unsigned int from
,
1090 unsigned int to
, struct skb_seq_state
*st
);
1091 unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
1092 struct skb_seq_state
*st
);
1093 void skb_abort_seq_read(struct skb_seq_state
*st
);
1095 unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
1096 unsigned int to
, struct ts_config
*config
);
1099 * Packet hash types specify the type of hash in skb_set_hash.
1101 * Hash types refer to the protocol layer addresses which are used to
1102 * construct a packet's hash. The hashes are used to differentiate or identify
1103 * flows of the protocol layer for the hash type. Hash types are either
1104 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1106 * Properties of hashes:
1108 * 1) Two packets in different flows have different hash values
1109 * 2) Two packets in the same flow should have the same hash value
1111 * A hash at a higher layer is considered to be more specific. A driver should
1112 * set the most specific hash possible.
1114 * A driver cannot indicate a more specific hash than the layer at which a hash
1115 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1117 * A driver may indicate a hash level which is less specific than the
1118 * actual layer the hash was computed on. For instance, a hash computed
1119 * at L4 may be considered an L3 hash. This should only be done if the
1120 * driver can't unambiguously determine that the HW computed the hash at
1121 * the higher layer. Note that the "should" in the second property above
1124 enum pkt_hash_types
{
1125 PKT_HASH_TYPE_NONE
, /* Undefined type */
1126 PKT_HASH_TYPE_L2
, /* Input: src_MAC, dest_MAC */
1127 PKT_HASH_TYPE_L3
, /* Input: src_IP, dst_IP */
1128 PKT_HASH_TYPE_L4
, /* Input: src_IP, dst_IP, src_port, dst_port */
1131 static inline void skb_clear_hash(struct sk_buff
*skb
)
1138 static inline void skb_clear_hash_if_not_l4(struct sk_buff
*skb
)
1141 skb_clear_hash(skb
);
1145 __skb_set_hash(struct sk_buff
*skb
, __u32 hash
, bool is_sw
, bool is_l4
)
1147 skb
->l4_hash
= is_l4
;
1148 skb
->sw_hash
= is_sw
;
1153 skb_set_hash(struct sk_buff
*skb
, __u32 hash
, enum pkt_hash_types type
)
1155 /* Used by drivers to set hash from HW */
1156 __skb_set_hash(skb
, hash
, false, type
== PKT_HASH_TYPE_L4
);
1160 __skb_set_sw_hash(struct sk_buff
*skb
, __u32 hash
, bool is_l4
)
1162 __skb_set_hash(skb
, hash
, true, is_l4
);
1165 void __skb_get_hash(struct sk_buff
*skb
);
1166 u32
__skb_get_hash_symmetric(const struct sk_buff
*skb
);
1167 u32
skb_get_poff(const struct sk_buff
*skb
);
1168 u32
__skb_get_poff(const struct sk_buff
*skb
, void *data
,
1169 const struct flow_keys
*keys
, int hlen
);
1170 __be32
__skb_flow_get_ports(const struct sk_buff
*skb
, int thoff
, u8 ip_proto
,
1171 void *data
, int hlen_proto
);
1173 static inline __be32
skb_flow_get_ports(const struct sk_buff
*skb
,
1174 int thoff
, u8 ip_proto
)
1176 return __skb_flow_get_ports(skb
, thoff
, ip_proto
, NULL
, 0);
1179 void skb_flow_dissector_init(struct flow_dissector
*flow_dissector
,
1180 const struct flow_dissector_key
*key
,
1181 unsigned int key_count
);
1183 bool __skb_flow_dissect(const struct sk_buff
*skb
,
1184 struct flow_dissector
*flow_dissector
,
1185 void *target_container
,
1186 void *data
, __be16 proto
, int nhoff
, int hlen
,
1187 unsigned int flags
);
1189 static inline bool skb_flow_dissect(const struct sk_buff
*skb
,
1190 struct flow_dissector
*flow_dissector
,
1191 void *target_container
, unsigned int flags
)
1193 return __skb_flow_dissect(skb
, flow_dissector
, target_container
,
1194 NULL
, 0, 0, 0, flags
);
1197 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff
*skb
,
1198 struct flow_keys
*flow
,
1201 memset(flow
, 0, sizeof(*flow
));
1202 return __skb_flow_dissect(skb
, &flow_keys_dissector
, flow
,
1203 NULL
, 0, 0, 0, flags
);
1206 static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys
*flow
,
1207 void *data
, __be16 proto
,
1208 int nhoff
, int hlen
,
1211 memset(flow
, 0, sizeof(*flow
));
1212 return __skb_flow_dissect(NULL
, &flow_keys_buf_dissector
, flow
,
1213 data
, proto
, nhoff
, hlen
, flags
);
1216 static inline __u32
skb_get_hash(struct sk_buff
*skb
)
1218 if (!skb
->l4_hash
&& !skb
->sw_hash
)
1219 __skb_get_hash(skb
);
1224 static inline __u32
skb_get_hash_flowi6(struct sk_buff
*skb
, const struct flowi6
*fl6
)
1226 if (!skb
->l4_hash
&& !skb
->sw_hash
) {
1227 struct flow_keys keys
;
1228 __u32 hash
= __get_hash_from_flowi6(fl6
, &keys
);
1230 __skb_set_sw_hash(skb
, hash
, flow_keys_have_l4(&keys
));
1236 __u32
skb_get_hash_perturb(const struct sk_buff
*skb
, u32 perturb
);
1238 static inline __u32
skb_get_hash_raw(const struct sk_buff
*skb
)
1243 static inline void skb_copy_hash(struct sk_buff
*to
, const struct sk_buff
*from
)
1245 to
->hash
= from
->hash
;
1246 to
->sw_hash
= from
->sw_hash
;
1247 to
->l4_hash
= from
->l4_hash
;
1250 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1251 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
1253 return skb
->head
+ skb
->end
;
1256 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
1261 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
1266 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
1268 return skb
->end
- skb
->head
;
1273 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1275 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
1277 return &skb_shinfo(skb
)->hwtstamps
;
1280 static inline struct ubuf_info
*skb_zcopy(struct sk_buff
*skb
)
1282 bool is_zcopy
= skb
&& skb_shinfo(skb
)->tx_flags
& SKBTX_DEV_ZEROCOPY
;
1284 return is_zcopy
? skb_uarg(skb
) : NULL
;
1287 static inline void skb_zcopy_set(struct sk_buff
*skb
, struct ubuf_info
*uarg
)
1289 if (skb
&& uarg
&& !skb_zcopy(skb
)) {
1290 sock_zerocopy_get(uarg
);
1291 skb_shinfo(skb
)->destructor_arg
= uarg
;
1292 skb_shinfo(skb
)->tx_flags
|= SKBTX_ZEROCOPY_FRAG
;
1296 /* Release a reference on a zerocopy structure */
1297 static inline void skb_zcopy_clear(struct sk_buff
*skb
, bool zerocopy
)
1299 struct ubuf_info
*uarg
= skb_zcopy(skb
);
1302 if (uarg
->callback
== sock_zerocopy_callback
) {
1303 uarg
->zerocopy
= uarg
->zerocopy
&& zerocopy
;
1304 sock_zerocopy_put(uarg
);
1306 uarg
->callback(uarg
, zerocopy
);
1309 skb_shinfo(skb
)->tx_flags
&= ~SKBTX_ZEROCOPY_FRAG
;
1313 /* Abort a zerocopy operation and revert zckey on error in send syscall */
1314 static inline void skb_zcopy_abort(struct sk_buff
*skb
)
1316 struct ubuf_info
*uarg
= skb_zcopy(skb
);
1319 sock_zerocopy_put_abort(uarg
);
1320 skb_shinfo(skb
)->tx_flags
&= ~SKBTX_ZEROCOPY_FRAG
;
1325 * skb_queue_empty - check if a queue is empty
1328 * Returns true if the queue is empty, false otherwise.
1330 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
1332 return list
->next
== (const struct sk_buff
*) list
;
1336 * skb_queue_is_last - check if skb is the last entry in the queue
1340 * Returns true if @skb is the last buffer on the list.
1342 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
1343 const struct sk_buff
*skb
)
1345 return skb
->next
== (const struct sk_buff
*) list
;
1349 * skb_queue_is_first - check if skb is the first entry in the queue
1353 * Returns true if @skb is the first buffer on the list.
1355 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
1356 const struct sk_buff
*skb
)
1358 return skb
->prev
== (const struct sk_buff
*) list
;
1362 * skb_queue_next - return the next packet in the queue
1364 * @skb: current buffer
1366 * Return the next packet in @list after @skb. It is only valid to
1367 * call this if skb_queue_is_last() evaluates to false.
1369 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
1370 const struct sk_buff
*skb
)
1372 /* This BUG_ON may seem severe, but if we just return then we
1373 * are going to dereference garbage.
1375 BUG_ON(skb_queue_is_last(list
, skb
));
1380 * skb_queue_prev - return the prev packet in the queue
1382 * @skb: current buffer
1384 * Return the prev packet in @list before @skb. It is only valid to
1385 * call this if skb_queue_is_first() evaluates to false.
1387 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
1388 const struct sk_buff
*skb
)
1390 /* This BUG_ON may seem severe, but if we just return then we
1391 * are going to dereference garbage.
1393 BUG_ON(skb_queue_is_first(list
, skb
));
1398 * skb_get - reference buffer
1399 * @skb: buffer to reference
1401 * Makes another reference to a socket buffer and returns a pointer
1404 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
1406 refcount_inc(&skb
->users
);
1411 * If users == 1, we are the only owner and are can avoid redundant
1416 * skb_cloned - is the buffer a clone
1417 * @skb: buffer to check
1419 * Returns true if the buffer was generated with skb_clone() and is
1420 * one of multiple shared copies of the buffer. Cloned buffers are
1421 * shared data so must not be written to under normal circumstances.
1423 static inline int skb_cloned(const struct sk_buff
*skb
)
1425 return skb
->cloned
&&
1426 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
1429 static inline int skb_unclone(struct sk_buff
*skb
, gfp_t pri
)
1431 might_sleep_if(gfpflags_allow_blocking(pri
));
1433 if (skb_cloned(skb
))
1434 return pskb_expand_head(skb
, 0, 0, pri
);
1440 * skb_header_cloned - is the header a clone
1441 * @skb: buffer to check
1443 * Returns true if modifying the header part of the buffer requires
1444 * the data to be copied.
1446 static inline int skb_header_cloned(const struct sk_buff
*skb
)
1453 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
1454 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
1455 return dataref
!= 1;
1458 static inline int skb_header_unclone(struct sk_buff
*skb
, gfp_t pri
)
1460 might_sleep_if(gfpflags_allow_blocking(pri
));
1462 if (skb_header_cloned(skb
))
1463 return pskb_expand_head(skb
, 0, 0, pri
);
1469 * __skb_header_release - release reference to header
1470 * @skb: buffer to operate on
1472 static inline void __skb_header_release(struct sk_buff
*skb
)
1475 atomic_set(&skb_shinfo(skb
)->dataref
, 1 + (1 << SKB_DATAREF_SHIFT
));
1480 * skb_shared - is the buffer shared
1481 * @skb: buffer to check
1483 * Returns true if more than one person has a reference to this
1486 static inline int skb_shared(const struct sk_buff
*skb
)
1488 return refcount_read(&skb
->users
) != 1;
1492 * skb_share_check - check if buffer is shared and if so clone it
1493 * @skb: buffer to check
1494 * @pri: priority for memory allocation
1496 * If the buffer is shared the buffer is cloned and the old copy
1497 * drops a reference. A new clone with a single reference is returned.
1498 * If the buffer is not shared the original buffer is returned. When
1499 * being called from interrupt status or with spinlocks held pri must
1502 * NULL is returned on a memory allocation failure.
1504 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
, gfp_t pri
)
1506 might_sleep_if(gfpflags_allow_blocking(pri
));
1507 if (skb_shared(skb
)) {
1508 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
1520 * Copy shared buffers into a new sk_buff. We effectively do COW on
1521 * packets to handle cases where we have a local reader and forward
1522 * and a couple of other messy ones. The normal one is tcpdumping
1523 * a packet thats being forwarded.
1527 * skb_unshare - make a copy of a shared buffer
1528 * @skb: buffer to check
1529 * @pri: priority for memory allocation
1531 * If the socket buffer is a clone then this function creates a new
1532 * copy of the data, drops a reference count on the old copy and returns
1533 * the new copy with the reference count at 1. If the buffer is not a clone
1534 * the original buffer is returned. When called with a spinlock held or
1535 * from interrupt state @pri must be %GFP_ATOMIC
1537 * %NULL is returned on a memory allocation failure.
1539 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
1542 might_sleep_if(gfpflags_allow_blocking(pri
));
1543 if (skb_cloned(skb
)) {
1544 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
1546 /* Free our shared copy */
1557 * skb_peek - peek at the head of an &sk_buff_head
1558 * @list_: list to peek at
1560 * Peek an &sk_buff. Unlike most other operations you _MUST_
1561 * be careful with this one. A peek leaves the buffer on the
1562 * list and someone else may run off with it. You must hold
1563 * the appropriate locks or have a private queue to do this.
1565 * Returns %NULL for an empty list or a pointer to the head element.
1566 * The reference count is not incremented and the reference is therefore
1567 * volatile. Use with caution.
1569 static inline struct sk_buff
*skb_peek(const struct sk_buff_head
*list_
)
1571 struct sk_buff
*skb
= list_
->next
;
1573 if (skb
== (struct sk_buff
*)list_
)
1579 * skb_peek_next - peek skb following the given one from a queue
1580 * @skb: skb to start from
1581 * @list_: list to peek at
1583 * Returns %NULL when the end of the list is met or a pointer to the
1584 * next element. The reference count is not incremented and the
1585 * reference is therefore volatile. Use with caution.
1587 static inline struct sk_buff
*skb_peek_next(struct sk_buff
*skb
,
1588 const struct sk_buff_head
*list_
)
1590 struct sk_buff
*next
= skb
->next
;
1592 if (next
== (struct sk_buff
*)list_
)
1598 * skb_peek_tail - peek at the tail of an &sk_buff_head
1599 * @list_: list to peek at
1601 * Peek an &sk_buff. Unlike most other operations you _MUST_
1602 * be careful with this one. A peek leaves the buffer on the
1603 * list and someone else may run off with it. You must hold
1604 * the appropriate locks or have a private queue to do this.
1606 * Returns %NULL for an empty list or a pointer to the tail element.
1607 * The reference count is not incremented and the reference is therefore
1608 * volatile. Use with caution.
1610 static inline struct sk_buff
*skb_peek_tail(const struct sk_buff_head
*list_
)
1612 struct sk_buff
*skb
= list_
->prev
;
1614 if (skb
== (struct sk_buff
*)list_
)
1621 * skb_queue_len - get queue length
1622 * @list_: list to measure
1624 * Return the length of an &sk_buff queue.
1626 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
1632 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1633 * @list: queue to initialize
1635 * This initializes only the list and queue length aspects of
1636 * an sk_buff_head object. This allows to initialize the list
1637 * aspects of an sk_buff_head without reinitializing things like
1638 * the spinlock. It can also be used for on-stack sk_buff_head
1639 * objects where the spinlock is known to not be used.
1641 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
1643 list
->prev
= list
->next
= (struct sk_buff
*)list
;
1648 * This function creates a split out lock class for each invocation;
1649 * this is needed for now since a whole lot of users of the skb-queue
1650 * infrastructure in drivers have different locking usage (in hardirq)
1651 * than the networking core (in softirq only). In the long run either the
1652 * network layer or drivers should need annotation to consolidate the
1653 * main types of usage into 3 classes.
1655 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
1657 spin_lock_init(&list
->lock
);
1658 __skb_queue_head_init(list
);
1661 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
1662 struct lock_class_key
*class)
1664 skb_queue_head_init(list
);
1665 lockdep_set_class(&list
->lock
, class);
1669 * Insert an sk_buff on a list.
1671 * The "__skb_xxxx()" functions are the non-atomic ones that
1672 * can only be called with interrupts disabled.
1674 void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
,
1675 struct sk_buff_head
*list
);
1676 static inline void __skb_insert(struct sk_buff
*newsk
,
1677 struct sk_buff
*prev
, struct sk_buff
*next
,
1678 struct sk_buff_head
*list
)
1682 next
->prev
= prev
->next
= newsk
;
1686 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
1687 struct sk_buff
*prev
,
1688 struct sk_buff
*next
)
1690 struct sk_buff
*first
= list
->next
;
1691 struct sk_buff
*last
= list
->prev
;
1701 * skb_queue_splice - join two skb lists, this is designed for stacks
1702 * @list: the new list to add
1703 * @head: the place to add it in the first list
1705 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
1706 struct sk_buff_head
*head
)
1708 if (!skb_queue_empty(list
)) {
1709 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1710 head
->qlen
+= list
->qlen
;
1715 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1716 * @list: the new list to add
1717 * @head: the place to add it in the first list
1719 * The list at @list is reinitialised
1721 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
1722 struct sk_buff_head
*head
)
1724 if (!skb_queue_empty(list
)) {
1725 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1726 head
->qlen
+= list
->qlen
;
1727 __skb_queue_head_init(list
);
1732 * skb_queue_splice_tail - join two skb lists, each list being a queue
1733 * @list: the new list to add
1734 * @head: the place to add it in the first list
1736 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
1737 struct sk_buff_head
*head
)
1739 if (!skb_queue_empty(list
)) {
1740 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1741 head
->qlen
+= list
->qlen
;
1746 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1747 * @list: the new list to add
1748 * @head: the place to add it in the first list
1750 * Each of the lists is a queue.
1751 * The list at @list is reinitialised
1753 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
1754 struct sk_buff_head
*head
)
1756 if (!skb_queue_empty(list
)) {
1757 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1758 head
->qlen
+= list
->qlen
;
1759 __skb_queue_head_init(list
);
1764 * __skb_queue_after - queue a buffer at the list head
1765 * @list: list to use
1766 * @prev: place after this buffer
1767 * @newsk: buffer to queue
1769 * Queue a buffer int the middle of a list. This function takes no locks
1770 * and you must therefore hold required locks before calling it.
1772 * A buffer cannot be placed on two lists at the same time.
1774 static inline void __skb_queue_after(struct sk_buff_head
*list
,
1775 struct sk_buff
*prev
,
1776 struct sk_buff
*newsk
)
1778 __skb_insert(newsk
, prev
, prev
->next
, list
);
1781 void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
1782 struct sk_buff_head
*list
);
1784 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1785 struct sk_buff
*next
,
1786 struct sk_buff
*newsk
)
1788 __skb_insert(newsk
, next
->prev
, next
, list
);
1792 * __skb_queue_head - queue a buffer at the list head
1793 * @list: list to use
1794 * @newsk: buffer to queue
1796 * Queue a buffer at the start of a list. This function takes no locks
1797 * and you must therefore hold required locks before calling it.
1799 * A buffer cannot be placed on two lists at the same time.
1801 void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1802 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1803 struct sk_buff
*newsk
)
1805 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1809 * __skb_queue_tail - queue a buffer at the list tail
1810 * @list: list to use
1811 * @newsk: buffer to queue
1813 * Queue a buffer at the end of a list. This function takes no locks
1814 * and you must therefore hold required locks before calling it.
1816 * A buffer cannot be placed on two lists at the same time.
1818 void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1819 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1820 struct sk_buff
*newsk
)
1822 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1826 * remove sk_buff from list. _Must_ be called atomically, and with
1829 void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1830 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1832 struct sk_buff
*next
, *prev
;
1837 skb
->next
= skb
->prev
= NULL
;
1843 * __skb_dequeue - remove from the head of the queue
1844 * @list: list to dequeue from
1846 * Remove the head of the list. This function does not take any locks
1847 * so must be used with appropriate locks held only. The head item is
1848 * returned or %NULL if the list is empty.
1850 struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1851 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1853 struct sk_buff
*skb
= skb_peek(list
);
1855 __skb_unlink(skb
, list
);
1860 * __skb_dequeue_tail - remove from the tail of the queue
1861 * @list: list to dequeue from
1863 * Remove the tail of the list. This function does not take any locks
1864 * so must be used with appropriate locks held only. The tail item is
1865 * returned or %NULL if the list is empty.
1867 struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1868 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1870 struct sk_buff
*skb
= skb_peek_tail(list
);
1872 __skb_unlink(skb
, list
);
1877 static inline bool skb_is_nonlinear(const struct sk_buff
*skb
)
1879 return skb
->data_len
;
1882 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1884 return skb
->len
- skb
->data_len
;
1887 static inline unsigned int __skb_pagelen(const struct sk_buff
*skb
)
1889 unsigned int i
, len
= 0;
1891 for (i
= skb_shinfo(skb
)->nr_frags
- 1; (int)i
>= 0; i
--)
1892 len
+= skb_frag_size(&skb_shinfo(skb
)->frags
[i
]);
1896 static inline unsigned int skb_pagelen(const struct sk_buff
*skb
)
1898 return skb_headlen(skb
) + __skb_pagelen(skb
);
1902 * __skb_fill_page_desc - initialise a paged fragment in an skb
1903 * @skb: buffer containing fragment to be initialised
1904 * @i: paged fragment index to initialise
1905 * @page: the page to use for this fragment
1906 * @off: the offset to the data with @page
1907 * @size: the length of the data
1909 * Initialises the @i'th fragment of @skb to point to &size bytes at
1910 * offset @off within @page.
1912 * Does not take any additional reference on the fragment.
1914 static inline void __skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1915 struct page
*page
, int off
, int size
)
1917 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1920 * Propagate page pfmemalloc to the skb if we can. The problem is
1921 * that not all callers have unique ownership of the page but rely
1922 * on page_is_pfmemalloc doing the right thing(tm).
1924 frag
->page
.p
= page
;
1925 frag
->page_offset
= off
;
1926 skb_frag_size_set(frag
, size
);
1928 page
= compound_head(page
);
1929 if (page_is_pfmemalloc(page
))
1930 skb
->pfmemalloc
= true;
1934 * skb_fill_page_desc - initialise a paged fragment in an skb
1935 * @skb: buffer containing fragment to be initialised
1936 * @i: paged fragment index to initialise
1937 * @page: the page to use for this fragment
1938 * @off: the offset to the data with @page
1939 * @size: the length of the data
1941 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1942 * @skb to point to @size bytes at offset @off within @page. In
1943 * addition updates @skb such that @i is the last fragment.
1945 * Does not take any additional reference on the fragment.
1947 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1948 struct page
*page
, int off
, int size
)
1950 __skb_fill_page_desc(skb
, i
, page
, off
, size
);
1951 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1954 void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
, int off
,
1955 int size
, unsigned int truesize
);
1957 void skb_coalesce_rx_frag(struct sk_buff
*skb
, int i
, int size
,
1958 unsigned int truesize
);
1960 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1961 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1962 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1964 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1965 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1967 return skb
->head
+ skb
->tail
;
1970 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1972 skb
->tail
= skb
->data
- skb
->head
;
1975 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1977 skb_reset_tail_pointer(skb
);
1978 skb
->tail
+= offset
;
1981 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1982 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1987 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1989 skb
->tail
= skb
->data
;
1992 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1994 skb
->tail
= skb
->data
+ offset
;
1997 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2000 * Add data to an sk_buff
2002 void *pskb_put(struct sk_buff
*skb
, struct sk_buff
*tail
, int len
);
2003 void *skb_put(struct sk_buff
*skb
, unsigned int len
);
2004 static inline void *__skb_put(struct sk_buff
*skb
, unsigned int len
)
2006 void *tmp
= skb_tail_pointer(skb
);
2007 SKB_LINEAR_ASSERT(skb
);
2013 static inline void *__skb_put_zero(struct sk_buff
*skb
, unsigned int len
)
2015 void *tmp
= __skb_put(skb
, len
);
2017 memset(tmp
, 0, len
);
2021 static inline void *__skb_put_data(struct sk_buff
*skb
, const void *data
,
2024 void *tmp
= __skb_put(skb
, len
);
2026 memcpy(tmp
, data
, len
);
2030 static inline void __skb_put_u8(struct sk_buff
*skb
, u8 val
)
2032 *(u8
*)__skb_put(skb
, 1) = val
;
2035 static inline void *skb_put_zero(struct sk_buff
*skb
, unsigned int len
)
2037 void *tmp
= skb_put(skb
, len
);
2039 memset(tmp
, 0, len
);
2044 static inline void *skb_put_data(struct sk_buff
*skb
, const void *data
,
2047 void *tmp
= skb_put(skb
, len
);
2049 memcpy(tmp
, data
, len
);
2054 static inline void skb_put_u8(struct sk_buff
*skb
, u8 val
)
2056 *(u8
*)skb_put(skb
, 1) = val
;
2059 void *skb_push(struct sk_buff
*skb
, unsigned int len
);
2060 static inline void *__skb_push(struct sk_buff
*skb
, unsigned int len
)
2067 void *skb_pull(struct sk_buff
*skb
, unsigned int len
);
2068 static inline void *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
2071 BUG_ON(skb
->len
< skb
->data_len
);
2072 return skb
->data
+= len
;
2075 static inline void *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
2077 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
2080 void *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
2082 static inline void *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
2084 if (len
> skb_headlen(skb
) &&
2085 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
2088 return skb
->data
+= len
;
2091 static inline void *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
2093 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
2096 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
2098 if (likely(len
<= skb_headlen(skb
)))
2100 if (unlikely(len
> skb
->len
))
2102 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
2105 void skb_condense(struct sk_buff
*skb
);
2108 * skb_headroom - bytes at buffer head
2109 * @skb: buffer to check
2111 * Return the number of bytes of free space at the head of an &sk_buff.
2113 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
2115 return skb
->data
- skb
->head
;
2119 * skb_tailroom - bytes at buffer end
2120 * @skb: buffer to check
2122 * Return the number of bytes of free space at the tail of an sk_buff
2124 static inline int skb_tailroom(const struct sk_buff
*skb
)
2126 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
2130 * skb_availroom - bytes at buffer end
2131 * @skb: buffer to check
2133 * Return the number of bytes of free space at the tail of an sk_buff
2134 * allocated by sk_stream_alloc()
2136 static inline int skb_availroom(const struct sk_buff
*skb
)
2138 if (skb_is_nonlinear(skb
))
2141 return skb
->end
- skb
->tail
- skb
->reserved_tailroom
;
2145 * skb_reserve - adjust headroom
2146 * @skb: buffer to alter
2147 * @len: bytes to move
2149 * Increase the headroom of an empty &sk_buff by reducing the tail
2150 * room. This is only allowed for an empty buffer.
2152 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
2159 * skb_tailroom_reserve - adjust reserved_tailroom
2160 * @skb: buffer to alter
2161 * @mtu: maximum amount of headlen permitted
2162 * @needed_tailroom: minimum amount of reserved_tailroom
2164 * Set reserved_tailroom so that headlen can be as large as possible but
2165 * not larger than mtu and tailroom cannot be smaller than
2167 * The required headroom should already have been reserved before using
2170 static inline void skb_tailroom_reserve(struct sk_buff
*skb
, unsigned int mtu
,
2171 unsigned int needed_tailroom
)
2173 SKB_LINEAR_ASSERT(skb
);
2174 if (mtu
< skb_tailroom(skb
) - needed_tailroom
)
2175 /* use at most mtu */
2176 skb
->reserved_tailroom
= skb_tailroom(skb
) - mtu
;
2178 /* use up to all available space */
2179 skb
->reserved_tailroom
= needed_tailroom
;
2182 #define ENCAP_TYPE_ETHER 0
2183 #define ENCAP_TYPE_IPPROTO 1
2185 static inline void skb_set_inner_protocol(struct sk_buff
*skb
,
2188 skb
->inner_protocol
= protocol
;
2189 skb
->inner_protocol_type
= ENCAP_TYPE_ETHER
;
2192 static inline void skb_set_inner_ipproto(struct sk_buff
*skb
,
2195 skb
->inner_ipproto
= ipproto
;
2196 skb
->inner_protocol_type
= ENCAP_TYPE_IPPROTO
;
2199 static inline void skb_reset_inner_headers(struct sk_buff
*skb
)
2201 skb
->inner_mac_header
= skb
->mac_header
;
2202 skb
->inner_network_header
= skb
->network_header
;
2203 skb
->inner_transport_header
= skb
->transport_header
;
2206 static inline void skb_reset_mac_len(struct sk_buff
*skb
)
2208 skb
->mac_len
= skb
->network_header
- skb
->mac_header
;
2211 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2214 return skb
->head
+ skb
->inner_transport_header
;
2217 static inline int skb_inner_transport_offset(const struct sk_buff
*skb
)
2219 return skb_inner_transport_header(skb
) - skb
->data
;
2222 static inline void skb_reset_inner_transport_header(struct sk_buff
*skb
)
2224 skb
->inner_transport_header
= skb
->data
- skb
->head
;
2227 static inline void skb_set_inner_transport_header(struct sk_buff
*skb
,
2230 skb_reset_inner_transport_header(skb
);
2231 skb
->inner_transport_header
+= offset
;
2234 static inline unsigned char *skb_inner_network_header(const struct sk_buff
*skb
)
2236 return skb
->head
+ skb
->inner_network_header
;
2239 static inline void skb_reset_inner_network_header(struct sk_buff
*skb
)
2241 skb
->inner_network_header
= skb
->data
- skb
->head
;
2244 static inline void skb_set_inner_network_header(struct sk_buff
*skb
,
2247 skb_reset_inner_network_header(skb
);
2248 skb
->inner_network_header
+= offset
;
2251 static inline unsigned char *skb_inner_mac_header(const struct sk_buff
*skb
)
2253 return skb
->head
+ skb
->inner_mac_header
;
2256 static inline void skb_reset_inner_mac_header(struct sk_buff
*skb
)
2258 skb
->inner_mac_header
= skb
->data
- skb
->head
;
2261 static inline void skb_set_inner_mac_header(struct sk_buff
*skb
,
2264 skb_reset_inner_mac_header(skb
);
2265 skb
->inner_mac_header
+= offset
;
2267 static inline bool skb_transport_header_was_set(const struct sk_buff
*skb
)
2269 return skb
->transport_header
!= (typeof(skb
->transport_header
))~0U;
2272 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
2274 return skb
->head
+ skb
->transport_header
;
2277 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
2279 skb
->transport_header
= skb
->data
- skb
->head
;
2282 static inline void skb_set_transport_header(struct sk_buff
*skb
,
2285 skb_reset_transport_header(skb
);
2286 skb
->transport_header
+= offset
;
2289 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
2291 return skb
->head
+ skb
->network_header
;
2294 static inline void skb_reset_network_header(struct sk_buff
*skb
)
2296 skb
->network_header
= skb
->data
- skb
->head
;
2299 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
2301 skb_reset_network_header(skb
);
2302 skb
->network_header
+= offset
;
2305 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
2307 return skb
->head
+ skb
->mac_header
;
2310 static inline int skb_mac_offset(const struct sk_buff
*skb
)
2312 return skb_mac_header(skb
) - skb
->data
;
2315 static inline u32
skb_mac_header_len(const struct sk_buff
*skb
)
2317 return skb
->network_header
- skb
->mac_header
;
2320 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
2322 return skb
->mac_header
!= (typeof(skb
->mac_header
))~0U;
2325 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
2327 skb
->mac_header
= skb
->data
- skb
->head
;
2330 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
2332 skb_reset_mac_header(skb
);
2333 skb
->mac_header
+= offset
;
2336 static inline void skb_pop_mac_header(struct sk_buff
*skb
)
2338 skb
->mac_header
= skb
->network_header
;
2341 static inline void skb_probe_transport_header(struct sk_buff
*skb
,
2342 const int offset_hint
)
2344 struct flow_keys keys
;
2346 if (skb_transport_header_was_set(skb
))
2348 else if (skb_flow_dissect_flow_keys(skb
, &keys
, 0))
2349 skb_set_transport_header(skb
, keys
.control
.thoff
);
2351 skb_set_transport_header(skb
, offset_hint
);
2354 static inline void skb_mac_header_rebuild(struct sk_buff
*skb
)
2356 if (skb_mac_header_was_set(skb
)) {
2357 const unsigned char *old_mac
= skb_mac_header(skb
);
2359 skb_set_mac_header(skb
, -skb
->mac_len
);
2360 memmove(skb_mac_header(skb
), old_mac
, skb
->mac_len
);
2364 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
2366 return skb
->csum_start
- skb_headroom(skb
);
2369 static inline unsigned char *skb_checksum_start(const struct sk_buff
*skb
)
2371 return skb
->head
+ skb
->csum_start
;
2374 static inline int skb_transport_offset(const struct sk_buff
*skb
)
2376 return skb_transport_header(skb
) - skb
->data
;
2379 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
2381 return skb
->transport_header
- skb
->network_header
;
2384 static inline u32
skb_inner_network_header_len(const struct sk_buff
*skb
)
2386 return skb
->inner_transport_header
- skb
->inner_network_header
;
2389 static inline int skb_network_offset(const struct sk_buff
*skb
)
2391 return skb_network_header(skb
) - skb
->data
;
2394 static inline int skb_inner_network_offset(const struct sk_buff
*skb
)
2396 return skb_inner_network_header(skb
) - skb
->data
;
2399 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
2401 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
2405 * CPUs often take a performance hit when accessing unaligned memory
2406 * locations. The actual performance hit varies, it can be small if the
2407 * hardware handles it or large if we have to take an exception and fix it
2410 * Since an ethernet header is 14 bytes network drivers often end up with
2411 * the IP header at an unaligned offset. The IP header can be aligned by
2412 * shifting the start of the packet by 2 bytes. Drivers should do this
2415 * skb_reserve(skb, NET_IP_ALIGN);
2417 * The downside to this alignment of the IP header is that the DMA is now
2418 * unaligned. On some architectures the cost of an unaligned DMA is high
2419 * and this cost outweighs the gains made by aligning the IP header.
2421 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2424 #ifndef NET_IP_ALIGN
2425 #define NET_IP_ALIGN 2
2429 * The networking layer reserves some headroom in skb data (via
2430 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2431 * the header has to grow. In the default case, if the header has to grow
2432 * 32 bytes or less we avoid the reallocation.
2434 * Unfortunately this headroom changes the DMA alignment of the resulting
2435 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2436 * on some architectures. An architecture can override this value,
2437 * perhaps setting it to a cacheline in size (since that will maintain
2438 * cacheline alignment of the DMA). It must be a power of 2.
2440 * Various parts of the networking layer expect at least 32 bytes of
2441 * headroom, you should not reduce this.
2443 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2444 * to reduce average number of cache lines per packet.
2445 * get_rps_cpus() for example only access one 64 bytes aligned block :
2446 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2449 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2452 int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
2454 static inline void __skb_set_length(struct sk_buff
*skb
, unsigned int len
)
2456 if (unlikely(skb_is_nonlinear(skb
))) {
2461 skb_set_tail_pointer(skb
, len
);
2464 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
2466 __skb_set_length(skb
, len
);
2469 void skb_trim(struct sk_buff
*skb
, unsigned int len
);
2471 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2474 return ___pskb_trim(skb
, len
);
2475 __skb_trim(skb
, len
);
2479 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2481 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
2485 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2486 * @skb: buffer to alter
2489 * This is identical to pskb_trim except that the caller knows that
2490 * the skb is not cloned so we should never get an error due to out-
2493 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
2495 int err
= pskb_trim(skb
, len
);
2499 static inline int __skb_grow(struct sk_buff
*skb
, unsigned int len
)
2501 unsigned int diff
= len
- skb
->len
;
2503 if (skb_tailroom(skb
) < diff
) {
2504 int ret
= pskb_expand_head(skb
, 0, diff
- skb_tailroom(skb
),
2509 __skb_set_length(skb
, len
);
2514 * skb_orphan - orphan a buffer
2515 * @skb: buffer to orphan
2517 * If a buffer currently has an owner then we call the owner's
2518 * destructor function and make the @skb unowned. The buffer continues
2519 * to exist but is no longer charged to its former owner.
2521 static inline void skb_orphan(struct sk_buff
*skb
)
2523 if (skb
->destructor
) {
2524 skb
->destructor(skb
);
2525 skb
->destructor
= NULL
;
2533 * skb_orphan_frags - orphan the frags contained in a buffer
2534 * @skb: buffer to orphan frags from
2535 * @gfp_mask: allocation mask for replacement pages
2537 * For each frag in the SKB which needs a destructor (i.e. has an
2538 * owner) create a copy of that frag and release the original
2539 * page by calling the destructor.
2541 static inline int skb_orphan_frags(struct sk_buff
*skb
, gfp_t gfp_mask
)
2543 if (likely(!skb_zcopy(skb
)))
2545 if (skb_uarg(skb
)->callback
== sock_zerocopy_callback
)
2547 return skb_copy_ubufs(skb
, gfp_mask
);
2550 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2551 static inline int skb_orphan_frags_rx(struct sk_buff
*skb
, gfp_t gfp_mask
)
2553 if (likely(!skb_zcopy(skb
)))
2555 return skb_copy_ubufs(skb
, gfp_mask
);
2559 * __skb_queue_purge - empty a list
2560 * @list: list to empty
2562 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2563 * the list and one reference dropped. This function does not take the
2564 * list lock and the caller must hold the relevant locks to use it.
2566 void skb_queue_purge(struct sk_buff_head
*list
);
2567 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
2569 struct sk_buff
*skb
;
2570 while ((skb
= __skb_dequeue(list
)) != NULL
)
2574 void skb_rbtree_purge(struct rb_root
*root
);
2576 void *netdev_alloc_frag(unsigned int fragsz
);
2578 struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
, unsigned int length
,
2582 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2583 * @dev: network device to receive on
2584 * @length: length to allocate
2586 * Allocate a new &sk_buff and assign it a usage count of one. The
2587 * buffer has unspecified headroom built in. Users should allocate
2588 * the headroom they think they need without accounting for the
2589 * built in space. The built in space is used for optimisations.
2591 * %NULL is returned if there is no free memory. Although this function
2592 * allocates memory it can be called from an interrupt.
2594 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
2595 unsigned int length
)
2597 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
2600 /* legacy helper around __netdev_alloc_skb() */
2601 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
2604 return __netdev_alloc_skb(NULL
, length
, gfp_mask
);
2607 /* legacy helper around netdev_alloc_skb() */
2608 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
2610 return netdev_alloc_skb(NULL
, length
);
2614 static inline struct sk_buff
*__netdev_alloc_skb_ip_align(struct net_device
*dev
,
2615 unsigned int length
, gfp_t gfp
)
2617 struct sk_buff
*skb
= __netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
, gfp
);
2619 if (NET_IP_ALIGN
&& skb
)
2620 skb_reserve(skb
, NET_IP_ALIGN
);
2624 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
2625 unsigned int length
)
2627 return __netdev_alloc_skb_ip_align(dev
, length
, GFP_ATOMIC
);
2630 static inline void skb_free_frag(void *addr
)
2632 page_frag_free(addr
);
2635 void *napi_alloc_frag(unsigned int fragsz
);
2636 struct sk_buff
*__napi_alloc_skb(struct napi_struct
*napi
,
2637 unsigned int length
, gfp_t gfp_mask
);
2638 static inline struct sk_buff
*napi_alloc_skb(struct napi_struct
*napi
,
2639 unsigned int length
)
2641 return __napi_alloc_skb(napi
, length
, GFP_ATOMIC
);
2643 void napi_consume_skb(struct sk_buff
*skb
, int budget
);
2645 void __kfree_skb_flush(void);
2646 void __kfree_skb_defer(struct sk_buff
*skb
);
2649 * __dev_alloc_pages - allocate page for network Rx
2650 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2651 * @order: size of the allocation
2653 * Allocate a new page.
2655 * %NULL is returned if there is no free memory.
2657 static inline struct page
*__dev_alloc_pages(gfp_t gfp_mask
,
2660 /* This piece of code contains several assumptions.
2661 * 1. This is for device Rx, therefor a cold page is preferred.
2662 * 2. The expectation is the user wants a compound page.
2663 * 3. If requesting a order 0 page it will not be compound
2664 * due to the check to see if order has a value in prep_new_page
2665 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2666 * code in gfp_to_alloc_flags that should be enforcing this.
2668 gfp_mask
|= __GFP_COLD
| __GFP_COMP
| __GFP_MEMALLOC
;
2670 return alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, order
);
2673 static inline struct page
*dev_alloc_pages(unsigned int order
)
2675 return __dev_alloc_pages(GFP_ATOMIC
| __GFP_NOWARN
, order
);
2679 * __dev_alloc_page - allocate a page for network Rx
2680 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2682 * Allocate a new page.
2684 * %NULL is returned if there is no free memory.
2686 static inline struct page
*__dev_alloc_page(gfp_t gfp_mask
)
2688 return __dev_alloc_pages(gfp_mask
, 0);
2691 static inline struct page
*dev_alloc_page(void)
2693 return dev_alloc_pages(0);
2697 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2698 * @page: The page that was allocated from skb_alloc_page
2699 * @skb: The skb that may need pfmemalloc set
2701 static inline void skb_propagate_pfmemalloc(struct page
*page
,
2702 struct sk_buff
*skb
)
2704 if (page_is_pfmemalloc(page
))
2705 skb
->pfmemalloc
= true;
2709 * skb_frag_page - retrieve the page referred to by a paged fragment
2710 * @frag: the paged fragment
2712 * Returns the &struct page associated with @frag.
2714 static inline struct page
*skb_frag_page(const skb_frag_t
*frag
)
2716 return frag
->page
.p
;
2720 * __skb_frag_ref - take an addition reference on a paged fragment.
2721 * @frag: the paged fragment
2723 * Takes an additional reference on the paged fragment @frag.
2725 static inline void __skb_frag_ref(skb_frag_t
*frag
)
2727 get_page(skb_frag_page(frag
));
2731 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2733 * @f: the fragment offset.
2735 * Takes an additional reference on the @f'th paged fragment of @skb.
2737 static inline void skb_frag_ref(struct sk_buff
*skb
, int f
)
2739 __skb_frag_ref(&skb_shinfo(skb
)->frags
[f
]);
2743 * __skb_frag_unref - release a reference on a paged fragment.
2744 * @frag: the paged fragment
2746 * Releases a reference on the paged fragment @frag.
2748 static inline void __skb_frag_unref(skb_frag_t
*frag
)
2750 put_page(skb_frag_page(frag
));
2754 * skb_frag_unref - release a reference on a paged fragment of an skb.
2756 * @f: the fragment offset
2758 * Releases a reference on the @f'th paged fragment of @skb.
2760 static inline void skb_frag_unref(struct sk_buff
*skb
, int f
)
2762 __skb_frag_unref(&skb_shinfo(skb
)->frags
[f
]);
2766 * skb_frag_address - gets the address of the data contained in a paged fragment
2767 * @frag: the paged fragment buffer
2769 * Returns the address of the data within @frag. The page must already
2772 static inline void *skb_frag_address(const skb_frag_t
*frag
)
2774 return page_address(skb_frag_page(frag
)) + frag
->page_offset
;
2778 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2779 * @frag: the paged fragment buffer
2781 * Returns the address of the data within @frag. Checks that the page
2782 * is mapped and returns %NULL otherwise.
2784 static inline void *skb_frag_address_safe(const skb_frag_t
*frag
)
2786 void *ptr
= page_address(skb_frag_page(frag
));
2790 return ptr
+ frag
->page_offset
;
2794 * __skb_frag_set_page - sets the page contained in a paged fragment
2795 * @frag: the paged fragment
2796 * @page: the page to set
2798 * Sets the fragment @frag to contain @page.
2800 static inline void __skb_frag_set_page(skb_frag_t
*frag
, struct page
*page
)
2802 frag
->page
.p
= page
;
2806 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2808 * @f: the fragment offset
2809 * @page: the page to set
2811 * Sets the @f'th fragment of @skb to contain @page.
2813 static inline void skb_frag_set_page(struct sk_buff
*skb
, int f
,
2816 __skb_frag_set_page(&skb_shinfo(skb
)->frags
[f
], page
);
2819 bool skb_page_frag_refill(unsigned int sz
, struct page_frag
*pfrag
, gfp_t prio
);
2822 * skb_frag_dma_map - maps a paged fragment via the DMA API
2823 * @dev: the device to map the fragment to
2824 * @frag: the paged fragment to map
2825 * @offset: the offset within the fragment (starting at the
2826 * fragment's own offset)
2827 * @size: the number of bytes to map
2828 * @dir: the direction of the mapping (``PCI_DMA_*``)
2830 * Maps the page associated with @frag to @device.
2832 static inline dma_addr_t
skb_frag_dma_map(struct device
*dev
,
2833 const skb_frag_t
*frag
,
2834 size_t offset
, size_t size
,
2835 enum dma_data_direction dir
)
2837 return dma_map_page(dev
, skb_frag_page(frag
),
2838 frag
->page_offset
+ offset
, size
, dir
);
2841 static inline struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
2844 return __pskb_copy(skb
, skb_headroom(skb
), gfp_mask
);
2848 static inline struct sk_buff
*pskb_copy_for_clone(struct sk_buff
*skb
,
2851 return __pskb_copy_fclone(skb
, skb_headroom(skb
), gfp_mask
, true);
2856 * skb_clone_writable - is the header of a clone writable
2857 * @skb: buffer to check
2858 * @len: length up to which to write
2860 * Returns true if modifying the header part of the cloned buffer
2861 * does not requires the data to be copied.
2863 static inline int skb_clone_writable(const struct sk_buff
*skb
, unsigned int len
)
2865 return !skb_header_cloned(skb
) &&
2866 skb_headroom(skb
) + len
<= skb
->hdr_len
;
2869 static inline int skb_try_make_writable(struct sk_buff
*skb
,
2870 unsigned int write_len
)
2872 return skb_cloned(skb
) && !skb_clone_writable(skb
, write_len
) &&
2873 pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
);
2876 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
2881 if (headroom
> skb_headroom(skb
))
2882 delta
= headroom
- skb_headroom(skb
);
2884 if (delta
|| cloned
)
2885 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
2891 * skb_cow - copy header of skb when it is required
2892 * @skb: buffer to cow
2893 * @headroom: needed headroom
2895 * If the skb passed lacks sufficient headroom or its data part
2896 * is shared, data is reallocated. If reallocation fails, an error
2897 * is returned and original skb is not changed.
2899 * The result is skb with writable area skb->head...skb->tail
2900 * and at least @headroom of space at head.
2902 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
2904 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
2908 * skb_cow_head - skb_cow but only making the head writable
2909 * @skb: buffer to cow
2910 * @headroom: needed headroom
2912 * This function is identical to skb_cow except that we replace the
2913 * skb_cloned check by skb_header_cloned. It should be used when
2914 * you only need to push on some header and do not need to modify
2917 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
2919 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
2923 * skb_padto - pad an skbuff up to a minimal size
2924 * @skb: buffer to pad
2925 * @len: minimal length
2927 * Pads up a buffer to ensure the trailing bytes exist and are
2928 * blanked. If the buffer already contains sufficient data it
2929 * is untouched. Otherwise it is extended. Returns zero on
2930 * success. The skb is freed on error.
2932 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
2934 unsigned int size
= skb
->len
;
2935 if (likely(size
>= len
))
2937 return skb_pad(skb
, len
- size
);
2941 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2942 * @skb: buffer to pad
2943 * @len: minimal length
2944 * @free_on_error: free buffer on error
2946 * Pads up a buffer to ensure the trailing bytes exist and are
2947 * blanked. If the buffer already contains sufficient data it
2948 * is untouched. Otherwise it is extended. Returns zero on
2949 * success. The skb is freed on error if @free_on_error is true.
2951 static inline int __skb_put_padto(struct sk_buff
*skb
, unsigned int len
,
2954 unsigned int size
= skb
->len
;
2956 if (unlikely(size
< len
)) {
2958 if (__skb_pad(skb
, len
, free_on_error
))
2960 __skb_put(skb
, len
);
2966 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2967 * @skb: buffer to pad
2968 * @len: minimal length
2970 * Pads up a buffer to ensure the trailing bytes exist and are
2971 * blanked. If the buffer already contains sufficient data it
2972 * is untouched. Otherwise it is extended. Returns zero on
2973 * success. The skb is freed on error.
2975 static inline int skb_put_padto(struct sk_buff
*skb
, unsigned int len
)
2977 return __skb_put_padto(skb
, len
, true);
2980 static inline int skb_add_data(struct sk_buff
*skb
,
2981 struct iov_iter
*from
, int copy
)
2983 const int off
= skb
->len
;
2985 if (skb
->ip_summed
== CHECKSUM_NONE
) {
2987 if (csum_and_copy_from_iter_full(skb_put(skb
, copy
), copy
,
2989 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
2992 } else if (copy_from_iter_full(skb_put(skb
, copy
), copy
, from
))
2995 __skb_trim(skb
, off
);
2999 static inline bool skb_can_coalesce(struct sk_buff
*skb
, int i
,
3000 const struct page
*page
, int off
)
3005 const struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
3007 return page
== skb_frag_page(frag
) &&
3008 off
== frag
->page_offset
+ skb_frag_size(frag
);
3013 static inline int __skb_linearize(struct sk_buff
*skb
)
3015 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
3019 * skb_linearize - convert paged skb to linear one
3020 * @skb: buffer to linarize
3022 * If there is no free memory -ENOMEM is returned, otherwise zero
3023 * is returned and the old skb data released.
3025 static inline int skb_linearize(struct sk_buff
*skb
)
3027 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
3031 * skb_has_shared_frag - can any frag be overwritten
3032 * @skb: buffer to test
3034 * Return true if the skb has at least one frag that might be modified
3035 * by an external entity (as in vmsplice()/sendfile())
3037 static inline bool skb_has_shared_frag(const struct sk_buff
*skb
)
3039 return skb_is_nonlinear(skb
) &&
3040 skb_shinfo(skb
)->tx_flags
& SKBTX_SHARED_FRAG
;
3044 * skb_linearize_cow - make sure skb is linear and writable
3045 * @skb: buffer to process
3047 * If there is no free memory -ENOMEM is returned, otherwise zero
3048 * is returned and the old skb data released.
3050 static inline int skb_linearize_cow(struct sk_buff
*skb
)
3052 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
3053 __skb_linearize(skb
) : 0;
3056 static __always_inline
void
3057 __skb_postpull_rcsum(struct sk_buff
*skb
, const void *start
, unsigned int len
,
3060 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3061 skb
->csum
= csum_block_sub(skb
->csum
,
3062 csum_partial(start
, len
, 0), off
);
3063 else if (skb
->ip_summed
== CHECKSUM_PARTIAL
&&
3064 skb_checksum_start_offset(skb
) < 0)
3065 skb
->ip_summed
= CHECKSUM_NONE
;
3069 * skb_postpull_rcsum - update checksum for received skb after pull
3070 * @skb: buffer to update
3071 * @start: start of data before pull
3072 * @len: length of data pulled
3074 * After doing a pull on a received packet, you need to call this to
3075 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3076 * CHECKSUM_NONE so that it can be recomputed from scratch.
3078 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
3079 const void *start
, unsigned int len
)
3081 __skb_postpull_rcsum(skb
, start
, len
, 0);
3084 static __always_inline
void
3085 __skb_postpush_rcsum(struct sk_buff
*skb
, const void *start
, unsigned int len
,
3088 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3089 skb
->csum
= csum_block_add(skb
->csum
,
3090 csum_partial(start
, len
, 0), off
);
3094 * skb_postpush_rcsum - update checksum for received skb after push
3095 * @skb: buffer to update
3096 * @start: start of data after push
3097 * @len: length of data pushed
3099 * After doing a push on a received packet, you need to call this to
3100 * update the CHECKSUM_COMPLETE checksum.
3102 static inline void skb_postpush_rcsum(struct sk_buff
*skb
,
3103 const void *start
, unsigned int len
)
3105 __skb_postpush_rcsum(skb
, start
, len
, 0);
3108 void *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
3111 * skb_push_rcsum - push skb and update receive checksum
3112 * @skb: buffer to update
3113 * @len: length of data pulled
3115 * This function performs an skb_push on the packet and updates
3116 * the CHECKSUM_COMPLETE checksum. It should be used on
3117 * receive path processing instead of skb_push unless you know
3118 * that the checksum difference is zero (e.g., a valid IP header)
3119 * or you are setting ip_summed to CHECKSUM_NONE.
3121 static inline void *skb_push_rcsum(struct sk_buff
*skb
, unsigned int len
)
3124 skb_postpush_rcsum(skb
, skb
->data
, len
);
3129 * pskb_trim_rcsum - trim received skb and update checksum
3130 * @skb: buffer to trim
3133 * This is exactly the same as pskb_trim except that it ensures the
3134 * checksum of received packets are still valid after the operation.
3137 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
3139 if (likely(len
>= skb
->len
))
3141 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3142 skb
->ip_summed
= CHECKSUM_NONE
;
3143 return __pskb_trim(skb
, len
);
3146 static inline int __skb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
3148 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3149 skb
->ip_summed
= CHECKSUM_NONE
;
3150 __skb_trim(skb
, len
);
3154 static inline int __skb_grow_rcsum(struct sk_buff
*skb
, unsigned int len
)
3156 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3157 skb
->ip_summed
= CHECKSUM_NONE
;
3158 return __skb_grow(skb
, len
);
3161 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3162 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3163 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3164 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3165 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3167 #define skb_queue_walk(queue, skb) \
3168 for (skb = (queue)->next; \
3169 skb != (struct sk_buff *)(queue); \
3172 #define skb_queue_walk_safe(queue, skb, tmp) \
3173 for (skb = (queue)->next, tmp = skb->next; \
3174 skb != (struct sk_buff *)(queue); \
3175 skb = tmp, tmp = skb->next)
3177 #define skb_queue_walk_from(queue, skb) \
3178 for (; skb != (struct sk_buff *)(queue); \
3181 #define skb_rbtree_walk(skb, root) \
3182 for (skb = skb_rb_first(root); skb != NULL; \
3183 skb = skb_rb_next(skb))
3185 #define skb_rbtree_walk_from(skb) \
3186 for (; skb != NULL; \
3187 skb = skb_rb_next(skb))
3189 #define skb_rbtree_walk_from_safe(skb, tmp) \
3190 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3193 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3194 for (tmp = skb->next; \
3195 skb != (struct sk_buff *)(queue); \
3196 skb = tmp, tmp = skb->next)
3198 #define skb_queue_reverse_walk(queue, skb) \
3199 for (skb = (queue)->prev; \
3200 skb != (struct sk_buff *)(queue); \
3203 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3204 for (skb = (queue)->prev, tmp = skb->prev; \
3205 skb != (struct sk_buff *)(queue); \
3206 skb = tmp, tmp = skb->prev)
3208 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3209 for (tmp = skb->prev; \
3210 skb != (struct sk_buff *)(queue); \
3211 skb = tmp, tmp = skb->prev)
3213 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
3215 return skb_shinfo(skb
)->frag_list
!= NULL
;
3218 static inline void skb_frag_list_init(struct sk_buff
*skb
)
3220 skb_shinfo(skb
)->frag_list
= NULL
;
3223 #define skb_walk_frags(skb, iter) \
3224 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3227 int __skb_wait_for_more_packets(struct sock
*sk
, int *err
, long *timeo_p
,
3228 const struct sk_buff
*skb
);
3229 struct sk_buff
*__skb_try_recv_from_queue(struct sock
*sk
,
3230 struct sk_buff_head
*queue
,
3232 void (*destructor
)(struct sock
*sk
,
3233 struct sk_buff
*skb
),
3234 int *peeked
, int *off
, int *err
,
3235 struct sk_buff
**last
);
3236 struct sk_buff
*__skb_try_recv_datagram(struct sock
*sk
, unsigned flags
,
3237 void (*destructor
)(struct sock
*sk
,
3238 struct sk_buff
*skb
),
3239 int *peeked
, int *off
, int *err
,
3240 struct sk_buff
**last
);
3241 struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
3242 void (*destructor
)(struct sock
*sk
,
3243 struct sk_buff
*skb
),
3244 int *peeked
, int *off
, int *err
);
3245 struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
, int noblock
,
3247 unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
3248 struct poll_table_struct
*wait
);
3249 int skb_copy_datagram_iter(const struct sk_buff
*from
, int offset
,
3250 struct iov_iter
*to
, int size
);
3251 static inline int skb_copy_datagram_msg(const struct sk_buff
*from
, int offset
,
3252 struct msghdr
*msg
, int size
)
3254 return skb_copy_datagram_iter(from
, offset
, &msg
->msg_iter
, size
);
3256 int skb_copy_and_csum_datagram_msg(struct sk_buff
*skb
, int hlen
,
3257 struct msghdr
*msg
);
3258 int skb_copy_datagram_from_iter(struct sk_buff
*skb
, int offset
,
3259 struct iov_iter
*from
, int len
);
3260 int zerocopy_sg_from_iter(struct sk_buff
*skb
, struct iov_iter
*frm
);
3261 void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
3262 void __skb_free_datagram_locked(struct sock
*sk
, struct sk_buff
*skb
, int len
);
3263 static inline void skb_free_datagram_locked(struct sock
*sk
,
3264 struct sk_buff
*skb
)
3266 __skb_free_datagram_locked(sk
, skb
, 0);
3268 int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
, unsigned int flags
);
3269 int skb_copy_bits(const struct sk_buff
*skb
, int offset
, void *to
, int len
);
3270 int skb_store_bits(struct sk_buff
*skb
, int offset
, const void *from
, int len
);
3271 __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
, int offset
, u8
*to
,
3272 int len
, __wsum csum
);
3273 int skb_splice_bits(struct sk_buff
*skb
, struct sock
*sk
, unsigned int offset
,
3274 struct pipe_inode_info
*pipe
, unsigned int len
,
3275 unsigned int flags
);
3276 int skb_send_sock_locked(struct sock
*sk
, struct sk_buff
*skb
, int offset
,
3278 int skb_send_sock(struct sock
*sk
, struct sk_buff
*skb
, int offset
, int len
);
3279 void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
3280 unsigned int skb_zerocopy_headlen(const struct sk_buff
*from
);
3281 int skb_zerocopy(struct sk_buff
*to
, struct sk_buff
*from
,
3283 void skb_split(struct sk_buff
*skb
, struct sk_buff
*skb1
, const u32 len
);
3284 int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
, int shiftlen
);
3285 void skb_scrub_packet(struct sk_buff
*skb
, bool xnet
);
3286 unsigned int skb_gso_transport_seglen(const struct sk_buff
*skb
);
3287 bool skb_gso_validate_mtu(const struct sk_buff
*skb
, unsigned int mtu
);
3288 struct sk_buff
*skb_segment(struct sk_buff
*skb
, netdev_features_t features
);
3289 struct sk_buff
*skb_vlan_untag(struct sk_buff
*skb
);
3290 int skb_ensure_writable(struct sk_buff
*skb
, int write_len
);
3291 int __skb_vlan_pop(struct sk_buff
*skb
, u16
*vlan_tci
);
3292 int skb_vlan_pop(struct sk_buff
*skb
);
3293 int skb_vlan_push(struct sk_buff
*skb
, __be16 vlan_proto
, u16 vlan_tci
);
3294 struct sk_buff
*pskb_extract(struct sk_buff
*skb
, int off
, int to_copy
,
3297 static inline int memcpy_from_msg(void *data
, struct msghdr
*msg
, int len
)
3299 return copy_from_iter_full(data
, len
, &msg
->msg_iter
) ? 0 : -EFAULT
;
3302 static inline int memcpy_to_msg(struct msghdr
*msg
, void *data
, int len
)
3304 return copy_to_iter(data
, len
, &msg
->msg_iter
) == len
? 0 : -EFAULT
;
3307 struct skb_checksum_ops
{
3308 __wsum (*update
)(const void *mem
, int len
, __wsum wsum
);
3309 __wsum (*combine
)(__wsum csum
, __wsum csum2
, int offset
, int len
);
3312 extern const struct skb_checksum_ops
*crc32c_csum_stub __read_mostly
;
3314 __wsum
__skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
3315 __wsum csum
, const struct skb_checksum_ops
*ops
);
3316 __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
3319 static inline void * __must_check
3320 __skb_header_pointer(const struct sk_buff
*skb
, int offset
,
3321 int len
, void *data
, int hlen
, void *buffer
)
3323 if (hlen
- offset
>= len
)
3324 return data
+ offset
;
3327 skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
3333 static inline void * __must_check
3334 skb_header_pointer(const struct sk_buff
*skb
, int offset
, int len
, void *buffer
)
3336 return __skb_header_pointer(skb
, offset
, len
, skb
->data
,
3337 skb_headlen(skb
), buffer
);
3341 * skb_needs_linearize - check if we need to linearize a given skb
3342 * depending on the given device features.
3343 * @skb: socket buffer to check
3344 * @features: net device features
3346 * Returns true if either:
3347 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3348 * 2. skb is fragmented and the device does not support SG.
3350 static inline bool skb_needs_linearize(struct sk_buff
*skb
,
3351 netdev_features_t features
)
3353 return skb_is_nonlinear(skb
) &&
3354 ((skb_has_frag_list(skb
) && !(features
& NETIF_F_FRAGLIST
)) ||
3355 (skb_shinfo(skb
)->nr_frags
&& !(features
& NETIF_F_SG
)));
3358 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
3360 const unsigned int len
)
3362 memcpy(to
, skb
->data
, len
);
3365 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
3366 const int offset
, void *to
,
3367 const unsigned int len
)
3369 memcpy(to
, skb
->data
+ offset
, len
);
3372 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
3374 const unsigned int len
)
3376 memcpy(skb
->data
, from
, len
);
3379 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
3382 const unsigned int len
)
3384 memcpy(skb
->data
+ offset
, from
, len
);
3387 void skb_init(void);
3389 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
3395 * skb_get_timestamp - get timestamp from a skb
3396 * @skb: skb to get stamp from
3397 * @stamp: pointer to struct timeval to store stamp in
3399 * Timestamps are stored in the skb as offsets to a base timestamp.
3400 * This function converts the offset back to a struct timeval and stores
3403 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
3404 struct timeval
*stamp
)
3406 *stamp
= ktime_to_timeval(skb
->tstamp
);
3409 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
3410 struct timespec
*stamp
)
3412 *stamp
= ktime_to_timespec(skb
->tstamp
);
3415 static inline void __net_timestamp(struct sk_buff
*skb
)
3417 skb
->tstamp
= ktime_get_real();
3420 static inline ktime_t
net_timedelta(ktime_t t
)
3422 return ktime_sub(ktime_get_real(), t
);
3425 static inline ktime_t
net_invalid_timestamp(void)
3430 static inline u8
skb_metadata_len(const struct sk_buff
*skb
)
3432 return skb_shinfo(skb
)->meta_len
;
3435 static inline void *skb_metadata_end(const struct sk_buff
*skb
)
3437 return skb_mac_header(skb
);
3440 static inline bool __skb_metadata_differs(const struct sk_buff
*skb_a
,
3441 const struct sk_buff
*skb_b
,
3444 const void *a
= skb_metadata_end(skb_a
);
3445 const void *b
= skb_metadata_end(skb_b
);
3446 /* Using more efficient varaiant than plain call to memcmp(). */
3447 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3451 #define __it(x, op) (x -= sizeof(u##op))
3452 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3453 case 32: diffs
|= __it_diff(a
, b
, 64);
3454 case 24: diffs
|= __it_diff(a
, b
, 64);
3455 case 16: diffs
|= __it_diff(a
, b
, 64);
3456 case 8: diffs
|= __it_diff(a
, b
, 64);
3458 case 28: diffs
|= __it_diff(a
, b
, 64);
3459 case 20: diffs
|= __it_diff(a
, b
, 64);
3460 case 12: diffs
|= __it_diff(a
, b
, 64);
3461 case 4: diffs
|= __it_diff(a
, b
, 32);
3466 return memcmp(a
- meta_len
, b
- meta_len
, meta_len
);
3470 static inline bool skb_metadata_differs(const struct sk_buff
*skb_a
,
3471 const struct sk_buff
*skb_b
)
3473 u8 len_a
= skb_metadata_len(skb_a
);
3474 u8 len_b
= skb_metadata_len(skb_b
);
3476 if (!(len_a
| len_b
))
3479 return len_a
!= len_b
?
3480 true : __skb_metadata_differs(skb_a
, skb_b
, len_a
);
3483 static inline void skb_metadata_set(struct sk_buff
*skb
, u8 meta_len
)
3485 skb_shinfo(skb
)->meta_len
= meta_len
;
3488 static inline void skb_metadata_clear(struct sk_buff
*skb
)
3490 skb_metadata_set(skb
, 0);
3493 struct sk_buff
*skb_clone_sk(struct sk_buff
*skb
);
3495 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3497 void skb_clone_tx_timestamp(struct sk_buff
*skb
);
3498 bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
3500 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3502 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
3506 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
3511 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3514 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3516 * PHY drivers may accept clones of transmitted packets for
3517 * timestamping via their phy_driver.txtstamp method. These drivers
3518 * must call this function to return the skb back to the stack with a
3521 * @skb: clone of the the original outgoing packet
3522 * @hwtstamps: hardware time stamps
3525 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
3526 struct skb_shared_hwtstamps
*hwtstamps
);
3528 void __skb_tstamp_tx(struct sk_buff
*orig_skb
,
3529 struct skb_shared_hwtstamps
*hwtstamps
,
3530 struct sock
*sk
, int tstype
);
3533 * skb_tstamp_tx - queue clone of skb with send time stamps
3534 * @orig_skb: the original outgoing packet
3535 * @hwtstamps: hardware time stamps, may be NULL if not available
3537 * If the skb has a socket associated, then this function clones the
3538 * skb (thus sharing the actual data and optional structures), stores
3539 * the optional hardware time stamping information (if non NULL) or
3540 * generates a software time stamp (otherwise), then queues the clone
3541 * to the error queue of the socket. Errors are silently ignored.
3543 void skb_tstamp_tx(struct sk_buff
*orig_skb
,
3544 struct skb_shared_hwtstamps
*hwtstamps
);
3547 * skb_tx_timestamp() - Driver hook for transmit timestamping
3549 * Ethernet MAC Drivers should call this function in their hard_xmit()
3550 * function immediately before giving the sk_buff to the MAC hardware.
3552 * Specifically, one should make absolutely sure that this function is
3553 * called before TX completion of this packet can trigger. Otherwise
3554 * the packet could potentially already be freed.
3556 * @skb: A socket buffer.
3558 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
3560 skb_clone_tx_timestamp(skb
);
3561 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
)
3562 skb_tstamp_tx(skb
, NULL
);
3566 * skb_complete_wifi_ack - deliver skb with wifi status
3568 * @skb: the original outgoing packet
3569 * @acked: ack status
3572 void skb_complete_wifi_ack(struct sk_buff
*skb
, bool acked
);
3574 __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
3575 __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
3577 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
3579 return ((skb
->ip_summed
== CHECKSUM_UNNECESSARY
) ||
3581 (skb
->ip_summed
== CHECKSUM_PARTIAL
&&
3582 skb_checksum_start_offset(skb
) >= 0));
3586 * skb_checksum_complete - Calculate checksum of an entire packet
3587 * @skb: packet to process
3589 * This function calculates the checksum over the entire packet plus
3590 * the value of skb->csum. The latter can be used to supply the
3591 * checksum of a pseudo header as used by TCP/UDP. It returns the
3594 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3595 * this function can be used to verify that checksum on received
3596 * packets. In that case the function should return zero if the
3597 * checksum is correct. In particular, this function will return zero
3598 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3599 * hardware has already verified the correctness of the checksum.
3601 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
3603 return skb_csum_unnecessary(skb
) ?
3604 0 : __skb_checksum_complete(skb
);
3607 static inline void __skb_decr_checksum_unnecessary(struct sk_buff
*skb
)
3609 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
3610 if (skb
->csum_level
== 0)
3611 skb
->ip_summed
= CHECKSUM_NONE
;
3617 static inline void __skb_incr_checksum_unnecessary(struct sk_buff
*skb
)
3619 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
3620 if (skb
->csum_level
< SKB_MAX_CSUM_LEVEL
)
3622 } else if (skb
->ip_summed
== CHECKSUM_NONE
) {
3623 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
3624 skb
->csum_level
= 0;
3628 /* Check if we need to perform checksum complete validation.
3630 * Returns true if checksum complete is needed, false otherwise
3631 * (either checksum is unnecessary or zero checksum is allowed).
3633 static inline bool __skb_checksum_validate_needed(struct sk_buff
*skb
,
3637 if (skb_csum_unnecessary(skb
) || (zero_okay
&& !check
)) {
3638 skb
->csum_valid
= 1;
3639 __skb_decr_checksum_unnecessary(skb
);
3646 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3649 #define CHECKSUM_BREAK 76
3651 /* Unset checksum-complete
3653 * Unset checksum complete can be done when packet is being modified
3654 * (uncompressed for instance) and checksum-complete value is
3657 static inline void skb_checksum_complete_unset(struct sk_buff
*skb
)
3659 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3660 skb
->ip_summed
= CHECKSUM_NONE
;
3663 /* Validate (init) checksum based on checksum complete.
3666 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3667 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3668 * checksum is stored in skb->csum for use in __skb_checksum_complete
3669 * non-zero: value of invalid checksum
3672 static inline __sum16
__skb_checksum_validate_complete(struct sk_buff
*skb
,
3676 if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
3677 if (!csum_fold(csum_add(psum
, skb
->csum
))) {
3678 skb
->csum_valid
= 1;
3685 if (complete
|| skb
->len
<= CHECKSUM_BREAK
) {
3688 csum
= __skb_checksum_complete(skb
);
3689 skb
->csum_valid
= !csum
;
3696 static inline __wsum
null_compute_pseudo(struct sk_buff
*skb
, int proto
)
3701 /* Perform checksum validate (init). Note that this is a macro since we only
3702 * want to calculate the pseudo header which is an input function if necessary.
3703 * First we try to validate without any computation (checksum unnecessary) and
3704 * then calculate based on checksum complete calling the function to compute
3708 * 0: checksum is validated or try to in skb_checksum_complete
3709 * non-zero: value of invalid checksum
3711 #define __skb_checksum_validate(skb, proto, complete, \
3712 zero_okay, check, compute_pseudo) \
3714 __sum16 __ret = 0; \
3715 skb->csum_valid = 0; \
3716 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3717 __ret = __skb_checksum_validate_complete(skb, \
3718 complete, compute_pseudo(skb, proto)); \
3722 #define skb_checksum_init(skb, proto, compute_pseudo) \
3723 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3725 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3726 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3728 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3729 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3731 #define skb_checksum_validate_zero_check(skb, proto, check, \
3733 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3735 #define skb_checksum_simple_validate(skb) \
3736 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3738 static inline bool __skb_checksum_convert_check(struct sk_buff
*skb
)
3740 return (skb
->ip_summed
== CHECKSUM_NONE
&& skb
->csum_valid
);
3743 static inline void __skb_checksum_convert(struct sk_buff
*skb
,
3744 __sum16 check
, __wsum pseudo
)
3746 skb
->csum
= ~pseudo
;
3747 skb
->ip_summed
= CHECKSUM_COMPLETE
;
3750 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3752 if (__skb_checksum_convert_check(skb)) \
3753 __skb_checksum_convert(skb, check, \
3754 compute_pseudo(skb, proto)); \
3757 static inline void skb_remcsum_adjust_partial(struct sk_buff
*skb
, void *ptr
,
3758 u16 start
, u16 offset
)
3760 skb
->ip_summed
= CHECKSUM_PARTIAL
;
3761 skb
->csum_start
= ((unsigned char *)ptr
+ start
) - skb
->head
;
3762 skb
->csum_offset
= offset
- start
;
3765 /* Update skbuf and packet to reflect the remote checksum offload operation.
3766 * When called, ptr indicates the starting point for skb->csum when
3767 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3768 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3770 static inline void skb_remcsum_process(struct sk_buff
*skb
, void *ptr
,
3771 int start
, int offset
, bool nopartial
)
3776 skb_remcsum_adjust_partial(skb
, ptr
, start
, offset
);
3780 if (unlikely(skb
->ip_summed
!= CHECKSUM_COMPLETE
)) {
3781 __skb_checksum_complete(skb
);
3782 skb_postpull_rcsum(skb
, skb
->data
, ptr
- (void *)skb
->data
);
3785 delta
= remcsum_adjust(ptr
, skb
->csum
, start
, offset
);
3787 /* Adjust skb->csum since we changed the packet */
3788 skb
->csum
= csum_add(skb
->csum
, delta
);
3791 static inline struct nf_conntrack
*skb_nfct(const struct sk_buff
*skb
)
3793 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3794 return (void *)(skb
->_nfct
& SKB_NFCT_PTRMASK
);
3800 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3801 void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
3802 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
3804 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
3805 nf_conntrack_destroy(nfct
);
3807 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
3810 atomic_inc(&nfct
->use
);
3813 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3814 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
3816 if (nf_bridge
&& refcount_dec_and_test(&nf_bridge
->use
))
3819 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
3822 refcount_inc(&nf_bridge
->use
);
3824 #endif /* CONFIG_BRIDGE_NETFILTER */
3825 static inline void nf_reset(struct sk_buff
*skb
)
3827 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3828 nf_conntrack_put(skb_nfct(skb
));
3831 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3832 nf_bridge_put(skb
->nf_bridge
);
3833 skb
->nf_bridge
= NULL
;
3837 static inline void nf_reset_trace(struct sk_buff
*skb
)
3839 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3844 /* Note: This doesn't put any conntrack and bridge info in dst. */
3845 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
,
3848 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3849 dst
->_nfct
= src
->_nfct
;
3850 nf_conntrack_get(skb_nfct(src
));
3852 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3853 dst
->nf_bridge
= src
->nf_bridge
;
3854 nf_bridge_get(src
->nf_bridge
);
3856 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3858 dst
->nf_trace
= src
->nf_trace
;
3862 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
3864 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3865 nf_conntrack_put(skb_nfct(dst
));
3867 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3868 nf_bridge_put(dst
->nf_bridge
);
3870 __nf_copy(dst
, src
, true);
3873 #ifdef CONFIG_NETWORK_SECMARK
3874 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3876 to
->secmark
= from
->secmark
;
3879 static inline void skb_init_secmark(struct sk_buff
*skb
)
3884 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3887 static inline void skb_init_secmark(struct sk_buff
*skb
)
3891 static inline bool skb_irq_freeable(const struct sk_buff
*skb
)
3893 return !skb
->destructor
&&
3894 #if IS_ENABLED(CONFIG_XFRM)
3898 !skb
->_skb_refdst
&&
3899 !skb_has_frag_list(skb
);
3902 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
3904 skb
->queue_mapping
= queue_mapping
;
3907 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
3909 return skb
->queue_mapping
;
3912 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
3914 to
->queue_mapping
= from
->queue_mapping
;
3917 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
3919 skb
->queue_mapping
= rx_queue
+ 1;
3922 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
3924 return skb
->queue_mapping
- 1;
3927 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
3929 return skb
->queue_mapping
!= 0;
3932 static inline void skb_set_dst_pending_confirm(struct sk_buff
*skb
, u32 val
)
3934 skb
->dst_pending_confirm
= val
;
3937 static inline bool skb_get_dst_pending_confirm(const struct sk_buff
*skb
)
3939 return skb
->dst_pending_confirm
!= 0;
3942 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
3951 /* Keeps track of mac header offset relative to skb->head.
3952 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3953 * For non-tunnel skb it points to skb_mac_header() and for
3954 * tunnel skb it points to outer mac header.
3955 * Keeps track of level of encapsulation of network headers.
3966 #define SKB_SGO_CB_OFFSET 32
3967 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
3969 static inline int skb_tnl_header_len(const struct sk_buff
*inner_skb
)
3971 return (skb_mac_header(inner_skb
) - inner_skb
->head
) -
3972 SKB_GSO_CB(inner_skb
)->mac_offset
;
3975 static inline int gso_pskb_expand_head(struct sk_buff
*skb
, int extra
)
3977 int new_headroom
, headroom
;
3980 headroom
= skb_headroom(skb
);
3981 ret
= pskb_expand_head(skb
, extra
, 0, GFP_ATOMIC
);
3985 new_headroom
= skb_headroom(skb
);
3986 SKB_GSO_CB(skb
)->mac_offset
+= (new_headroom
- headroom
);
3990 static inline void gso_reset_checksum(struct sk_buff
*skb
, __wsum res
)
3992 /* Do not update partial checksums if remote checksum is enabled. */
3993 if (skb
->remcsum_offload
)
3996 SKB_GSO_CB(skb
)->csum
= res
;
3997 SKB_GSO_CB(skb
)->csum_start
= skb_checksum_start(skb
) - skb
->head
;
4000 /* Compute the checksum for a gso segment. First compute the checksum value
4001 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4002 * then add in skb->csum (checksum from csum_start to end of packet).
4003 * skb->csum and csum_start are then updated to reflect the checksum of the
4004 * resultant packet starting from the transport header-- the resultant checksum
4005 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4008 static inline __sum16
gso_make_checksum(struct sk_buff
*skb
, __wsum res
)
4010 unsigned char *csum_start
= skb_transport_header(skb
);
4011 int plen
= (skb
->head
+ SKB_GSO_CB(skb
)->csum_start
) - csum_start
;
4012 __wsum partial
= SKB_GSO_CB(skb
)->csum
;
4014 SKB_GSO_CB(skb
)->csum
= res
;
4015 SKB_GSO_CB(skb
)->csum_start
= csum_start
- skb
->head
;
4017 return csum_fold(csum_partial(csum_start
, plen
, partial
));
4020 static inline bool skb_is_gso(const struct sk_buff
*skb
)
4022 return skb_shinfo(skb
)->gso_size
;
4025 /* Note: Should be called only if skb_is_gso(skb) is true */
4026 static inline bool skb_is_gso_v6(const struct sk_buff
*skb
)
4028 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
4031 static inline void skb_gso_reset(struct sk_buff
*skb
)
4033 skb_shinfo(skb
)->gso_size
= 0;
4034 skb_shinfo(skb
)->gso_segs
= 0;
4035 skb_shinfo(skb
)->gso_type
= 0;
4038 void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
4040 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
4042 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4043 * wanted then gso_type will be set. */
4044 const struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
4046 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
4047 unlikely(shinfo
->gso_type
== 0)) {
4048 __skb_warn_lro_forwarding(skb
);
4054 static inline void skb_forward_csum(struct sk_buff
*skb
)
4056 /* Unfortunately we don't support this one. Any brave souls? */
4057 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
4058 skb
->ip_summed
= CHECKSUM_NONE
;
4062 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4063 * @skb: skb to check
4065 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4066 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4067 * use this helper, to document places where we make this assertion.
4069 static inline void skb_checksum_none_assert(const struct sk_buff
*skb
)
4072 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
4076 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
4078 int skb_checksum_setup(struct sk_buff
*skb
, bool recalculate
);
4079 struct sk_buff
*skb_checksum_trimmed(struct sk_buff
*skb
,
4080 unsigned int transport_len
,
4081 __sum16(*skb_chkf
)(struct sk_buff
*skb
));
4084 * skb_head_is_locked - Determine if the skb->head is locked down
4085 * @skb: skb to check
4087 * The head on skbs build around a head frag can be removed if they are
4088 * not cloned. This function returns true if the skb head is locked down
4089 * due to either being allocated via kmalloc, or by being a clone with
4090 * multiple references to the head.
4092 static inline bool skb_head_is_locked(const struct sk_buff
*skb
)
4094 return !skb
->head_frag
|| skb_cloned(skb
);
4098 * skb_gso_network_seglen - Return length of individual segments of a gso packet
4102 * skb_gso_network_seglen is used to determine the real size of the
4103 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
4105 * The MAC/L2 header is not accounted for.
4107 static inline unsigned int skb_gso_network_seglen(const struct sk_buff
*skb
)
4109 unsigned int hdr_len
= skb_transport_header(skb
) -
4110 skb_network_header(skb
);
4111 return hdr_len
+ skb_gso_transport_seglen(skb
);
4114 /* Local Checksum Offload.
4115 * Compute outer checksum based on the assumption that the
4116 * inner checksum will be offloaded later.
4117 * See Documentation/networking/checksum-offloads.txt for
4118 * explanation of how this works.
4119 * Fill in outer checksum adjustment (e.g. with sum of outer
4120 * pseudo-header) before calling.
4121 * Also ensure that inner checksum is in linear data area.
4123 static inline __wsum
lco_csum(struct sk_buff
*skb
)
4125 unsigned char *csum_start
= skb_checksum_start(skb
);
4126 unsigned char *l4_hdr
= skb_transport_header(skb
);
4129 /* Start with complement of inner checksum adjustment */
4130 partial
= ~csum_unfold(*(__force __sum16
*)(csum_start
+
4133 /* Add in checksum of our headers (incl. outer checksum
4134 * adjustment filled in by caller) and return result.
4136 return csum_partial(l4_hdr
, csum_start
- l4_hdr
, partial
);
4139 #endif /* __KERNEL__ */
4140 #endif /* _LINUX_SKBUFF_H */