2 * Definitions for the 'struct sk_buff' memory handlers.
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/kernel.h>
18 #include <linux/kmemcheck.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/bug.h>
22 #include <linux/cache.h>
23 #include <linux/rbtree.h>
24 #include <linux/socket.h>
26 #include <linux/atomic.h>
27 #include <asm/types.h>
28 #include <linux/spinlock.h>
29 #include <linux/net.h>
30 #include <linux/textsearch.h>
31 #include <net/checksum.h>
32 #include <linux/rcupdate.h>
33 #include <linux/hrtimer.h>
34 #include <linux/dma-mapping.h>
35 #include <linux/netdev_features.h>
36 #include <linux/sched.h>
37 #include <net/flow_dissector.h>
38 #include <linux/splice.h>
39 #include <linux/in6.h>
40 #include <linux/if_packet.h>
43 /* The interface for checksum offload between the stack and networking drivers
46 * A. IP checksum related features
48 * Drivers advertise checksum offload capabilities in the features of a device.
49 * From the stack's point of view these are capabilities offered by the driver,
50 * a driver typically only advertises features that it is capable of offloading
53 * The checksum related features are:
55 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
56 * IP (one's complement) checksum for any combination
57 * of protocols or protocol layering. The checksum is
58 * computed and set in a packet per the CHECKSUM_PARTIAL
59 * interface (see below).
61 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
62 * TCP or UDP packets over IPv4. These are specifically
63 * unencapsulated packets of the form IPv4|TCP or
64 * IPv4|UDP where the Protocol field in the IPv4 header
65 * is TCP or UDP. The IPv4 header may contain IP options
66 * This feature cannot be set in features for a device
67 * with NETIF_F_HW_CSUM also set. This feature is being
68 * DEPRECATED (see below).
70 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
71 * TCP or UDP packets over IPv6. These are specifically
72 * unencapsulated packets of the form IPv6|TCP or
73 * IPv4|UDP where the Next Header field in the IPv6
74 * header is either TCP or UDP. IPv6 extension headers
75 * are not supported with this feature. This feature
76 * cannot be set in features for a device with
77 * NETIF_F_HW_CSUM also set. This feature is being
78 * DEPRECATED (see below).
80 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
81 * This flag is used only used to disable the RX checksum
82 * feature for a device. The stack will accept receive
83 * checksum indication in packets received on a device
84 * regardless of whether NETIF_F_RXCSUM is set.
86 * B. Checksumming of received packets by device. Indication of checksum
87 * verification is in set skb->ip_summed. Possible values are:
91 * Device did not checksum this packet e.g. due to lack of capabilities.
92 * The packet contains full (though not verified) checksum in packet but
93 * not in skb->csum. Thus, skb->csum is undefined in this case.
95 * CHECKSUM_UNNECESSARY:
97 * The hardware you're dealing with doesn't calculate the full checksum
98 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
99 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
100 * if their checksums are okay. skb->csum is still undefined in this case
101 * though. A driver or device must never modify the checksum field in the
102 * packet even if checksum is verified.
104 * CHECKSUM_UNNECESSARY is applicable to following protocols:
105 * TCP: IPv6 and IPv4.
106 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
107 * zero UDP checksum for either IPv4 or IPv6, the networking stack
108 * may perform further validation in this case.
109 * GRE: only if the checksum is present in the header.
110 * SCTP: indicates the CRC in SCTP header has been validated.
112 * skb->csum_level indicates the number of consecutive checksums found in
113 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
114 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
115 * and a device is able to verify the checksums for UDP (possibly zero),
116 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
117 * two. If the device were only able to verify the UDP checksum and not
118 * GRE, either because it doesn't support GRE checksum of because GRE
119 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
120 * not considered in this case).
124 * This is the most generic way. The device supplied checksum of the _whole_
125 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
126 * hardware doesn't need to parse L3/L4 headers to implement this.
128 * Note: Even if device supports only some protocols, but is able to produce
129 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133 * A checksum is set up to be offloaded to a device as described in the
134 * output description for CHECKSUM_PARTIAL. This may occur on a packet
135 * received directly from another Linux OS, e.g., a virtualized Linux kernel
136 * on the same host, or it may be set in the input path in GRO or remote
137 * checksum offload. For the purposes of checksum verification, the checksum
138 * referred to by skb->csum_start + skb->csum_offset and any preceding
139 * checksums in the packet are considered verified. Any checksums in the
140 * packet that are after the checksum being offloaded are not considered to
143 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
144 * in the skb->ip_summed for a packet. Values are:
148 * The driver is required to checksum the packet as seen by hard_start_xmit()
149 * from skb->csum_start up to the end, and to record/write the checksum at
150 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
151 * csum_start and csum_offset values are valid values given the length and
152 * offset of the packet, however they should not attempt to validate that the
153 * checksum refers to a legitimate transport layer checksum-- it is the
154 * purview of the stack to validate that csum_start and csum_offset are set
157 * When the stack requests checksum offload for a packet, the driver MUST
158 * ensure that the checksum is set correctly. A driver can either offload the
159 * checksum calculation to the device, or call skb_checksum_help (in the case
160 * that the device does not support offload for a particular checksum).
162 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
163 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
164 * checksum offload capability. If a device has limited checksum capabilities
165 * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
166 * described above) a helper function can be called to resolve
167 * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
168 * function takes a spec argument that describes the protocol layer that is
169 * supported for checksum offload and can be called for each packet. If a
170 * packet does not match the specification for offload, skb_checksum_help
171 * is called to resolve the checksum.
175 * The skb was already checksummed by the protocol, or a checksum is not
178 * CHECKSUM_UNNECESSARY:
180 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
184 * Not used in checksum output. If a driver observes a packet with this value
185 * set in skbuff, if should treat as CHECKSUM_NONE being set.
187 * D. Non-IP checksum (CRC) offloads
189 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
190 * offloading the SCTP CRC in a packet. To perform this offload the stack
191 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
192 * accordingly. Note the there is no indication in the skbuff that the
193 * CHECKSUM_PARTIAL refers to an SCTP checksum, a driver that supports
194 * both IP checksum offload and SCTP CRC offload must verify which offload
195 * is configured for a packet presumably by inspecting packet headers.
197 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
198 * offloading the FCOE CRC in a packet. To perform this offload the stack
199 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
200 * accordingly. Note the there is no indication in the skbuff that the
201 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
202 * both IP checksum offload and FCOE CRC offload must verify which offload
203 * is configured for a packet presumably by inspecting packet headers.
205 * E. Checksumming on output with GSO.
207 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
208 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
209 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
210 * part of the GSO operation is implied. If a checksum is being offloaded
211 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
212 * are set to refer to the outermost checksum being offload (two offloaded
213 * checksums are possible with UDP encapsulation).
216 /* Don't change this without changing skb_csum_unnecessary! */
217 #define CHECKSUM_NONE 0
218 #define CHECKSUM_UNNECESSARY 1
219 #define CHECKSUM_COMPLETE 2
220 #define CHECKSUM_PARTIAL 3
222 /* Maximum value in skb->csum_level */
223 #define SKB_MAX_CSUM_LEVEL 3
225 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
226 #define SKB_WITH_OVERHEAD(X) \
227 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
228 #define SKB_MAX_ORDER(X, ORDER) \
229 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
230 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
231 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
233 /* return minimum truesize of one skb containing X bytes of data */
234 #define SKB_TRUESIZE(X) ((X) + \
235 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
236 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240 struct pipe_inode_info
;
244 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
245 struct nf_conntrack
{
250 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
251 struct nf_bridge_info
{
254 BRNF_PROTO_UNCHANGED
,
262 struct net_device
*physindev
;
264 /* always valid & non-NULL from FORWARD on, for physdev match */
265 struct net_device
*physoutdev
;
267 /* prerouting: detect dnat in orig/reply direction */
269 struct in6_addr ipv6_daddr
;
271 /* after prerouting + nat detected: store original source
272 * mac since neigh resolution overwrites it, only used while
273 * skb is out in neigh layer.
275 char neigh_header
[8];
280 struct sk_buff_head
{
281 /* These two members must be first. */
282 struct sk_buff
*next
;
283 struct sk_buff
*prev
;
291 /* To allow 64K frame to be packed as single skb without frag_list we
292 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
293 * buffers which do not start on a page boundary.
295 * Since GRO uses frags we allocate at least 16 regardless of page
298 #if (65536/PAGE_SIZE + 1) < 16
299 #define MAX_SKB_FRAGS 16UL
301 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
303 extern int sysctl_max_skb_frags
;
305 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
306 * segment using its current segmentation instead.
308 #define GSO_BY_FRAGS 0xFFFF
310 typedef struct skb_frag_struct skb_frag_t
;
312 struct skb_frag_struct
{
316 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
325 static inline unsigned int skb_frag_size(const skb_frag_t
*frag
)
330 static inline void skb_frag_size_set(skb_frag_t
*frag
, unsigned int size
)
335 static inline void skb_frag_size_add(skb_frag_t
*frag
, int delta
)
340 static inline void skb_frag_size_sub(skb_frag_t
*frag
, int delta
)
345 #define HAVE_HW_TIME_STAMP
348 * struct skb_shared_hwtstamps - hardware time stamps
349 * @hwtstamp: hardware time stamp transformed into duration
350 * since arbitrary point in time
352 * Software time stamps generated by ktime_get_real() are stored in
355 * hwtstamps can only be compared against other hwtstamps from
358 * This structure is attached to packets as part of the
359 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
361 struct skb_shared_hwtstamps
{
365 /* Definitions for tx_flags in struct skb_shared_info */
367 /* generate hardware time stamp */
368 SKBTX_HW_TSTAMP
= 1 << 0,
370 /* generate software time stamp when queueing packet to NIC */
371 SKBTX_SW_TSTAMP
= 1 << 1,
373 /* device driver is going to provide hardware time stamp */
374 SKBTX_IN_PROGRESS
= 1 << 2,
376 /* device driver supports TX zero-copy buffers */
377 SKBTX_DEV_ZEROCOPY
= 1 << 3,
379 /* generate wifi status information (where possible) */
380 SKBTX_WIFI_STATUS
= 1 << 4,
382 /* This indicates at least one fragment might be overwritten
383 * (as in vmsplice(), sendfile() ...)
384 * If we need to compute a TX checksum, we'll need to copy
385 * all frags to avoid possible bad checksum
387 SKBTX_SHARED_FRAG
= 1 << 5,
389 /* generate software time stamp when entering packet scheduling */
390 SKBTX_SCHED_TSTAMP
= 1 << 6,
393 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
395 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
398 * The callback notifies userspace to release buffers when skb DMA is done in
399 * lower device, the skb last reference should be 0 when calling this.
400 * The zerocopy_success argument is true if zero copy transmit occurred,
401 * false on data copy or out of memory error caused by data copy attempt.
402 * The ctx field is used to track device context.
403 * The desc field is used to track userspace buffer index.
406 void (*callback
)(struct ubuf_info
*, bool zerocopy_success
);
411 /* This data is invariant across clones and lives at
412 * the end of the header data, ie. at skb->end.
414 struct skb_shared_info
{
415 unsigned char nr_frags
;
417 unsigned short gso_size
;
418 /* Warning: this field is not always filled in (UFO)! */
419 unsigned short gso_segs
;
420 unsigned short gso_type
;
421 struct sk_buff
*frag_list
;
422 struct skb_shared_hwtstamps hwtstamps
;
427 * Warning : all fields before dataref are cleared in __alloc_skb()
431 /* Intermediate layers must ensure that destructor_arg
432 * remains valid until skb destructor */
433 void * destructor_arg
;
435 /* must be last field, see pskb_expand_head() */
436 skb_frag_t frags
[MAX_SKB_FRAGS
];
439 /* We divide dataref into two halves. The higher 16 bits hold references
440 * to the payload part of skb->data. The lower 16 bits hold references to
441 * the entire skb->data. A clone of a headerless skb holds the length of
442 * the header in skb->hdr_len.
444 * All users must obey the rule that the skb->data reference count must be
445 * greater than or equal to the payload reference count.
447 * Holding a reference to the payload part means that the user does not
448 * care about modifications to the header part of skb->data.
450 #define SKB_DATAREF_SHIFT 16
451 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
455 SKB_FCLONE_UNAVAILABLE
, /* skb has no fclone (from head_cache) */
456 SKB_FCLONE_ORIG
, /* orig skb (from fclone_cache) */
457 SKB_FCLONE_CLONE
, /* companion fclone skb (from fclone_cache) */
461 SKB_GSO_TCPV4
= 1 << 0,
462 SKB_GSO_UDP
= 1 << 1,
464 /* This indicates the skb is from an untrusted source. */
465 SKB_GSO_DODGY
= 1 << 2,
467 /* This indicates the tcp segment has CWR set. */
468 SKB_GSO_TCP_ECN
= 1 << 3,
470 SKB_GSO_TCP_FIXEDID
= 1 << 4,
472 SKB_GSO_TCPV6
= 1 << 5,
474 SKB_GSO_FCOE
= 1 << 6,
476 SKB_GSO_GRE
= 1 << 7,
478 SKB_GSO_GRE_CSUM
= 1 << 8,
480 SKB_GSO_IPXIP4
= 1 << 9,
482 SKB_GSO_IPXIP6
= 1 << 10,
484 SKB_GSO_UDP_TUNNEL
= 1 << 11,
486 SKB_GSO_UDP_TUNNEL_CSUM
= 1 << 12,
488 SKB_GSO_PARTIAL
= 1 << 13,
490 SKB_GSO_TUNNEL_REMCSUM
= 1 << 14,
492 SKB_GSO_SCTP
= 1 << 15,
495 #if BITS_PER_LONG > 32
496 #define NET_SKBUFF_DATA_USES_OFFSET 1
499 #ifdef NET_SKBUFF_DATA_USES_OFFSET
500 typedef unsigned int sk_buff_data_t
;
502 typedef unsigned char *sk_buff_data_t
;
506 * struct skb_mstamp - multi resolution time stamps
507 * @stamp_us: timestamp in us resolution
508 * @stamp_jiffies: timestamp in jiffies
521 * skb_mstamp_get - get current timestamp
522 * @cl: place to store timestamps
524 static inline void skb_mstamp_get(struct skb_mstamp
*cl
)
526 u64 val
= local_clock();
528 do_div(val
, NSEC_PER_USEC
);
529 cl
->stamp_us
= (u32
)val
;
530 cl
->stamp_jiffies
= (u32
)jiffies
;
534 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
535 * @t1: pointer to newest sample
536 * @t0: pointer to oldest sample
538 static inline u32
skb_mstamp_us_delta(const struct skb_mstamp
*t1
,
539 const struct skb_mstamp
*t0
)
541 s32 delta_us
= t1
->stamp_us
- t0
->stamp_us
;
542 u32 delta_jiffies
= t1
->stamp_jiffies
- t0
->stamp_jiffies
;
544 /* If delta_us is negative, this might be because interval is too big,
545 * or local_clock() drift is too big : fallback using jiffies.
548 delta_jiffies
>= (INT_MAX
/ (USEC_PER_SEC
/ HZ
)))
550 delta_us
= jiffies_to_usecs(delta_jiffies
);
555 static inline bool skb_mstamp_after(const struct skb_mstamp
*t1
,
556 const struct skb_mstamp
*t0
)
558 s32 diff
= t1
->stamp_jiffies
- t0
->stamp_jiffies
;
561 diff
= t1
->stamp_us
- t0
->stamp_us
;
566 * struct sk_buff - socket buffer
567 * @next: Next buffer in list
568 * @prev: Previous buffer in list
569 * @tstamp: Time we arrived/left
570 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
571 * @sk: Socket we are owned by
572 * @dev: Device we arrived on/are leaving by
573 * @cb: Control buffer. Free for use by every layer. Put private vars here
574 * @_skb_refdst: destination entry (with norefcount bit)
575 * @sp: the security path, used for xfrm
576 * @len: Length of actual data
577 * @data_len: Data length
578 * @mac_len: Length of link layer header
579 * @hdr_len: writable header length of cloned skb
580 * @csum: Checksum (must include start/offset pair)
581 * @csum_start: Offset from skb->head where checksumming should start
582 * @csum_offset: Offset from csum_start where checksum should be stored
583 * @priority: Packet queueing priority
584 * @ignore_df: allow local fragmentation
585 * @cloned: Head may be cloned (check refcnt to be sure)
586 * @ip_summed: Driver fed us an IP checksum
587 * @nohdr: Payload reference only, must not modify header
588 * @nfctinfo: Relationship of this skb to the connection
589 * @pkt_type: Packet class
590 * @fclone: skbuff clone status
591 * @ipvs_property: skbuff is owned by ipvs
592 * @peeked: this packet has been seen already, so stats have been
593 * done for it, don't do them again
594 * @nf_trace: netfilter packet trace flag
595 * @protocol: Packet protocol from driver
596 * @destructor: Destruct function
597 * @nfct: Associated connection, if any
598 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
599 * @skb_iif: ifindex of device we arrived on
600 * @tc_index: Traffic control index
601 * @tc_verd: traffic control verdict
602 * @hash: the packet hash
603 * @queue_mapping: Queue mapping for multiqueue devices
604 * @xmit_more: More SKBs are pending for this queue
605 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
606 * @ndisc_nodetype: router type (from link layer)
607 * @ooo_okay: allow the mapping of a socket to a queue to be changed
608 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
610 * @sw_hash: indicates hash was computed in software stack
611 * @wifi_acked_valid: wifi_acked was set
612 * @wifi_acked: whether frame was acked on wifi or not
613 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
614 * @napi_id: id of the NAPI struct this skb came from
615 * @secmark: security marking
616 * @mark: Generic packet mark
617 * @vlan_proto: vlan encapsulation protocol
618 * @vlan_tci: vlan tag control information
619 * @inner_protocol: Protocol (encapsulation)
620 * @inner_transport_header: Inner transport layer header (encapsulation)
621 * @inner_network_header: Network layer header (encapsulation)
622 * @inner_mac_header: Link layer header (encapsulation)
623 * @transport_header: Transport layer header
624 * @network_header: Network layer header
625 * @mac_header: Link layer header
626 * @tail: Tail pointer
628 * @head: Head of buffer
629 * @data: Data head pointer
630 * @truesize: Buffer size
631 * @users: User count - see {datagram,tcp}.c
637 /* These two members must be first. */
638 struct sk_buff
*next
;
639 struct sk_buff
*prev
;
643 struct skb_mstamp skb_mstamp
;
646 struct rb_node rbnode
; /* used in netem, ip4 defrag, and tcp stack */
651 int ip_defrag_offset
;
654 struct net_device
*dev
;
657 * This is the control buffer. It is free to use for every
658 * layer. Please put your private variables there. If you
659 * want to keep them across layers you have to do a skb_clone()
660 * first. This is owned by whoever has the skb queued ATM.
662 char cb
[48] __aligned(8);
664 unsigned long _skb_refdst
;
665 void (*destructor
)(struct sk_buff
*skb
);
669 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
670 struct nf_conntrack
*nfct
;
672 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
673 struct nf_bridge_info
*nf_bridge
;
680 /* Following fields are _not_ copied in __copy_skb_header()
681 * Note that queue_mapping is here mostly to fill a hole.
683 kmemcheck_bitfield_begin(flags1
);
686 /* if you move cloned around you also must adapt those constants */
687 #ifdef __BIG_ENDIAN_BITFIELD
688 #define CLONED_MASK (1 << 7)
690 #define CLONED_MASK 1
692 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
694 __u8 __cloned_offset
[0];
702 kmemcheck_bitfield_end(flags1
);
704 /* fields enclosed in headers_start/headers_end are copied
705 * using a single memcpy() in __copy_skb_header()
708 __u32 headers_start
[0];
711 /* if you move pkt_type around you also must adapt those constants */
712 #ifdef __BIG_ENDIAN_BITFIELD
713 #define PKT_TYPE_MAX (7 << 5)
715 #define PKT_TYPE_MAX 7
717 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
719 __u8 __pkt_type_offset
[0];
729 __u8 wifi_acked_valid
:1;
733 /* Indicates the inner headers are valid in the skbuff. */
734 __u8 encapsulation
:1;
735 __u8 encap_hdr_csum
:1;
737 __u8 csum_complete_sw
:1;
740 #ifdef CONFIG_IPV6_NDISC_NODETYPE
741 __u8 ndisc_nodetype
:2;
743 __u8 ipvs_property
:1;
745 __u8 inner_protocol_type
:1;
746 __u8 remcsum_offload
:1;
747 #ifdef CONFIG_NET_SWITCHDEV
748 __u8 offload_fwd_mark
:1;
750 /* 2, 4 or 5 bit hole */
752 #ifdef CONFIG_NET_SCHED
753 __u16 tc_index
; /* traffic control index */
754 #ifdef CONFIG_NET_CLS_ACT
755 __u16 tc_verd
; /* traffic control verdict */
771 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
773 unsigned int napi_id
;
774 unsigned int sender_cpu
;
777 #ifdef CONFIG_NETWORK_SECMARK
783 __u32 reserved_tailroom
;
787 __be16 inner_protocol
;
791 __u16 inner_transport_header
;
792 __u16 inner_network_header
;
793 __u16 inner_mac_header
;
796 __u16 transport_header
;
797 __u16 network_header
;
801 __u32 headers_end
[0];
804 /* These elements must be at the end, see alloc_skb() for details. */
809 unsigned int truesize
;
815 * Handling routines are only of interest to the kernel
817 #include <linux/slab.h>
820 #define SKB_ALLOC_FCLONE 0x01
821 #define SKB_ALLOC_RX 0x02
822 #define SKB_ALLOC_NAPI 0x04
824 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
825 static inline bool skb_pfmemalloc(const struct sk_buff
*skb
)
827 return unlikely(skb
->pfmemalloc
);
831 * skb might have a dst pointer attached, refcounted or not.
832 * _skb_refdst low order bit is set if refcount was _not_ taken
834 #define SKB_DST_NOREF 1UL
835 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
838 * skb_dst - returns skb dst_entry
841 * Returns skb dst_entry, regardless of reference taken or not.
843 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
845 /* If refdst was not refcounted, check we still are in a
846 * rcu_read_lock section
848 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
849 !rcu_read_lock_held() &&
850 !rcu_read_lock_bh_held());
851 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
855 * skb_dst_set - sets skb dst
859 * Sets skb dst, assuming a reference was taken on dst and should
860 * be released by skb_dst_drop()
862 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
864 skb
->_skb_refdst
= (unsigned long)dst
;
868 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
872 * Sets skb dst, assuming a reference was not taken on dst.
873 * If dst entry is cached, we do not take reference and dst_release
874 * will be avoided by refdst_drop. If dst entry is not cached, we take
875 * reference, so that last dst_release can destroy the dst immediately.
877 static inline void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
)
879 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
880 skb
->_skb_refdst
= (unsigned long)dst
| SKB_DST_NOREF
;
884 * skb_dst_is_noref - Test if skb dst isn't refcounted
887 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
889 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
892 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
894 return (struct rtable
*)skb_dst(skb
);
897 /* For mangling skb->pkt_type from user space side from applications
898 * such as nft, tc, etc, we only allow a conservative subset of
899 * possible pkt_types to be set.
901 static inline bool skb_pkt_type_ok(u32 ptype
)
903 return ptype
<= PACKET_OTHERHOST
;
906 void kfree_skb(struct sk_buff
*skb
);
907 void kfree_skb_list(struct sk_buff
*segs
);
908 void skb_tx_error(struct sk_buff
*skb
);
909 void consume_skb(struct sk_buff
*skb
);
910 void __kfree_skb(struct sk_buff
*skb
);
911 extern struct kmem_cache
*skbuff_head_cache
;
913 void kfree_skb_partial(struct sk_buff
*skb
, bool head_stolen
);
914 bool skb_try_coalesce(struct sk_buff
*to
, struct sk_buff
*from
,
915 bool *fragstolen
, int *delta_truesize
);
917 struct sk_buff
*__alloc_skb(unsigned int size
, gfp_t priority
, int flags
,
919 struct sk_buff
*__build_skb(void *data
, unsigned int frag_size
);
920 struct sk_buff
*build_skb(void *data
, unsigned int frag_size
);
921 static inline struct sk_buff
*alloc_skb(unsigned int size
,
924 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
927 struct sk_buff
*alloc_skb_with_frags(unsigned long header_len
,
928 unsigned long data_len
,
933 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
934 struct sk_buff_fclones
{
943 * skb_fclone_busy - check if fclone is busy
947 * Returns true if skb is a fast clone, and its clone is not freed.
948 * Some drivers call skb_orphan() in their ndo_start_xmit(),
949 * so we also check that this didnt happen.
951 static inline bool skb_fclone_busy(const struct sock
*sk
,
952 const struct sk_buff
*skb
)
954 const struct sk_buff_fclones
*fclones
;
956 fclones
= container_of(skb
, struct sk_buff_fclones
, skb1
);
958 return skb
->fclone
== SKB_FCLONE_ORIG
&&
959 atomic_read(&fclones
->fclone_ref
) > 1 &&
960 fclones
->skb2
.sk
== sk
;
963 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
966 return __alloc_skb(size
, priority
, SKB_ALLOC_FCLONE
, NUMA_NO_NODE
);
969 struct sk_buff
*__alloc_skb_head(gfp_t priority
, int node
);
970 static inline struct sk_buff
*alloc_skb_head(gfp_t priority
)
972 return __alloc_skb_head(priority
, -1);
975 struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
976 int skb_copy_ubufs(struct sk_buff
*skb
, gfp_t gfp_mask
);
977 struct sk_buff
*skb_clone(struct sk_buff
*skb
, gfp_t priority
);
978 struct sk_buff
*skb_copy(const struct sk_buff
*skb
, gfp_t priority
);
979 struct sk_buff
*__pskb_copy_fclone(struct sk_buff
*skb
, int headroom
,
980 gfp_t gfp_mask
, bool fclone
);
981 static inline struct sk_buff
*__pskb_copy(struct sk_buff
*skb
, int headroom
,
984 return __pskb_copy_fclone(skb
, headroom
, gfp_mask
, false);
987 int pskb_expand_head(struct sk_buff
*skb
, int nhead
, int ntail
, gfp_t gfp_mask
);
988 struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
989 unsigned int headroom
);
990 struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
, int newheadroom
,
991 int newtailroom
, gfp_t priority
);
992 int __must_check
skb_to_sgvec_nomark(struct sk_buff
*skb
, struct scatterlist
*sg
,
993 int offset
, int len
);
994 int __must_check
skb_to_sgvec(struct sk_buff
*skb
, struct scatterlist
*sg
,
995 int offset
, int len
);
996 int skb_cow_data(struct sk_buff
*skb
, int tailbits
, struct sk_buff
**trailer
);
997 int skb_pad(struct sk_buff
*skb
, int pad
);
998 #define dev_kfree_skb(a) consume_skb(a)
1000 int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
1001 int getfrag(void *from
, char *to
, int offset
,
1002 int len
, int odd
, struct sk_buff
*skb
),
1003 void *from
, int length
);
1005 int skb_append_pagefrags(struct sk_buff
*skb
, struct page
*page
,
1006 int offset
, size_t size
);
1008 struct skb_seq_state
{
1012 __u32 stepped_offset
;
1013 struct sk_buff
*root_skb
;
1014 struct sk_buff
*cur_skb
;
1018 void skb_prepare_seq_read(struct sk_buff
*skb
, unsigned int from
,
1019 unsigned int to
, struct skb_seq_state
*st
);
1020 unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
1021 struct skb_seq_state
*st
);
1022 void skb_abort_seq_read(struct skb_seq_state
*st
);
1024 unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
1025 unsigned int to
, struct ts_config
*config
);
1028 * Packet hash types specify the type of hash in skb_set_hash.
1030 * Hash types refer to the protocol layer addresses which are used to
1031 * construct a packet's hash. The hashes are used to differentiate or identify
1032 * flows of the protocol layer for the hash type. Hash types are either
1033 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1035 * Properties of hashes:
1037 * 1) Two packets in different flows have different hash values
1038 * 2) Two packets in the same flow should have the same hash value
1040 * A hash at a higher layer is considered to be more specific. A driver should
1041 * set the most specific hash possible.
1043 * A driver cannot indicate a more specific hash than the layer at which a hash
1044 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1046 * A driver may indicate a hash level which is less specific than the
1047 * actual layer the hash was computed on. For instance, a hash computed
1048 * at L4 may be considered an L3 hash. This should only be done if the
1049 * driver can't unambiguously determine that the HW computed the hash at
1050 * the higher layer. Note that the "should" in the second property above
1053 enum pkt_hash_types
{
1054 PKT_HASH_TYPE_NONE
, /* Undefined type */
1055 PKT_HASH_TYPE_L2
, /* Input: src_MAC, dest_MAC */
1056 PKT_HASH_TYPE_L3
, /* Input: src_IP, dst_IP */
1057 PKT_HASH_TYPE_L4
, /* Input: src_IP, dst_IP, src_port, dst_port */
1060 static inline void skb_clear_hash(struct sk_buff
*skb
)
1067 static inline void skb_clear_hash_if_not_l4(struct sk_buff
*skb
)
1070 skb_clear_hash(skb
);
1074 __skb_set_hash(struct sk_buff
*skb
, __u32 hash
, bool is_sw
, bool is_l4
)
1076 skb
->l4_hash
= is_l4
;
1077 skb
->sw_hash
= is_sw
;
1082 skb_set_hash(struct sk_buff
*skb
, __u32 hash
, enum pkt_hash_types type
)
1084 /* Used by drivers to set hash from HW */
1085 __skb_set_hash(skb
, hash
, false, type
== PKT_HASH_TYPE_L4
);
1089 __skb_set_sw_hash(struct sk_buff
*skb
, __u32 hash
, bool is_l4
)
1091 __skb_set_hash(skb
, hash
, true, is_l4
);
1094 void __skb_get_hash(struct sk_buff
*skb
);
1095 u32
__skb_get_hash_symmetric(struct sk_buff
*skb
);
1096 u32
skb_get_poff(const struct sk_buff
*skb
);
1097 u32
__skb_get_poff(const struct sk_buff
*skb
, void *data
,
1098 const struct flow_keys
*keys
, int hlen
);
1099 __be32
__skb_flow_get_ports(const struct sk_buff
*skb
, int thoff
, u8 ip_proto
,
1100 void *data
, int hlen_proto
);
1102 static inline __be32
skb_flow_get_ports(const struct sk_buff
*skb
,
1103 int thoff
, u8 ip_proto
)
1105 return __skb_flow_get_ports(skb
, thoff
, ip_proto
, NULL
, 0);
1108 void skb_flow_dissector_init(struct flow_dissector
*flow_dissector
,
1109 const struct flow_dissector_key
*key
,
1110 unsigned int key_count
);
1112 bool __skb_flow_dissect(const struct sk_buff
*skb
,
1113 struct flow_dissector
*flow_dissector
,
1114 void *target_container
,
1115 void *data
, __be16 proto
, int nhoff
, int hlen
,
1116 unsigned int flags
);
1118 static inline bool skb_flow_dissect(const struct sk_buff
*skb
,
1119 struct flow_dissector
*flow_dissector
,
1120 void *target_container
, unsigned int flags
)
1122 return __skb_flow_dissect(skb
, flow_dissector
, target_container
,
1123 NULL
, 0, 0, 0, flags
);
1126 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff
*skb
,
1127 struct flow_keys
*flow
,
1130 memset(flow
, 0, sizeof(*flow
));
1131 return __skb_flow_dissect(skb
, &flow_keys_dissector
, flow
,
1132 NULL
, 0, 0, 0, flags
);
1135 static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys
*flow
,
1136 void *data
, __be16 proto
,
1137 int nhoff
, int hlen
,
1140 memset(flow
, 0, sizeof(*flow
));
1141 return __skb_flow_dissect(NULL
, &flow_keys_buf_dissector
, flow
,
1142 data
, proto
, nhoff
, hlen
, flags
);
1145 static inline __u32
skb_get_hash(struct sk_buff
*skb
)
1147 if (!skb
->l4_hash
&& !skb
->sw_hash
)
1148 __skb_get_hash(skb
);
1153 __u32
__skb_get_hash_flowi6(struct sk_buff
*skb
, const struct flowi6
*fl6
);
1155 static inline __u32
skb_get_hash_flowi6(struct sk_buff
*skb
, const struct flowi6
*fl6
)
1157 if (!skb
->l4_hash
&& !skb
->sw_hash
) {
1158 struct flow_keys keys
;
1159 __u32 hash
= __get_hash_from_flowi6(fl6
, &keys
);
1161 __skb_set_sw_hash(skb
, hash
, flow_keys_have_l4(&keys
));
1167 __u32
__skb_get_hash_flowi4(struct sk_buff
*skb
, const struct flowi4
*fl
);
1169 static inline __u32
skb_get_hash_flowi4(struct sk_buff
*skb
, const struct flowi4
*fl4
)
1171 if (!skb
->l4_hash
&& !skb
->sw_hash
) {
1172 struct flow_keys keys
;
1173 __u32 hash
= __get_hash_from_flowi4(fl4
, &keys
);
1175 __skb_set_sw_hash(skb
, hash
, flow_keys_have_l4(&keys
));
1181 __u32
skb_get_hash_perturb(const struct sk_buff
*skb
, u32 perturb
);
1183 static inline __u32
skb_get_hash_raw(const struct sk_buff
*skb
)
1188 static inline void skb_copy_hash(struct sk_buff
*to
, const struct sk_buff
*from
)
1190 to
->hash
= from
->hash
;
1191 to
->sw_hash
= from
->sw_hash
;
1192 to
->l4_hash
= from
->l4_hash
;
1195 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1196 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
1198 return skb
->head
+ skb
->end
;
1201 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
1206 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
1211 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
1213 return skb
->end
- skb
->head
;
1218 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1220 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
1222 return &skb_shinfo(skb
)->hwtstamps
;
1226 * skb_queue_empty - check if a queue is empty
1229 * Returns true if the queue is empty, false otherwise.
1231 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
1233 return list
->next
== (const struct sk_buff
*) list
;
1237 * skb_queue_is_last - check if skb is the last entry in the queue
1241 * Returns true if @skb is the last buffer on the list.
1243 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
1244 const struct sk_buff
*skb
)
1246 return skb
->next
== (const struct sk_buff
*) list
;
1250 * skb_queue_is_first - check if skb is the first entry in the queue
1254 * Returns true if @skb is the first buffer on the list.
1256 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
1257 const struct sk_buff
*skb
)
1259 return skb
->prev
== (const struct sk_buff
*) list
;
1263 * skb_queue_next - return the next packet in the queue
1265 * @skb: current buffer
1267 * Return the next packet in @list after @skb. It is only valid to
1268 * call this if skb_queue_is_last() evaluates to false.
1270 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
1271 const struct sk_buff
*skb
)
1273 /* This BUG_ON may seem severe, but if we just return then we
1274 * are going to dereference garbage.
1276 BUG_ON(skb_queue_is_last(list
, skb
));
1281 * skb_queue_prev - return the prev packet in the queue
1283 * @skb: current buffer
1285 * Return the prev packet in @list before @skb. It is only valid to
1286 * call this if skb_queue_is_first() evaluates to false.
1288 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
1289 const struct sk_buff
*skb
)
1291 /* This BUG_ON may seem severe, but if we just return then we
1292 * are going to dereference garbage.
1294 BUG_ON(skb_queue_is_first(list
, skb
));
1299 * skb_get - reference buffer
1300 * @skb: buffer to reference
1302 * Makes another reference to a socket buffer and returns a pointer
1305 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
1307 atomic_inc(&skb
->users
);
1312 * If users == 1, we are the only owner and are can avoid redundant
1317 * skb_cloned - is the buffer a clone
1318 * @skb: buffer to check
1320 * Returns true if the buffer was generated with skb_clone() and is
1321 * one of multiple shared copies of the buffer. Cloned buffers are
1322 * shared data so must not be written to under normal circumstances.
1324 static inline int skb_cloned(const struct sk_buff
*skb
)
1326 return skb
->cloned
&&
1327 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
1330 static inline int skb_unclone(struct sk_buff
*skb
, gfp_t pri
)
1332 might_sleep_if(gfpflags_allow_blocking(pri
));
1334 if (skb_cloned(skb
))
1335 return pskb_expand_head(skb
, 0, 0, pri
);
1341 * skb_header_cloned - is the header a clone
1342 * @skb: buffer to check
1344 * Returns true if modifying the header part of the buffer requires
1345 * the data to be copied.
1347 static inline int skb_header_cloned(const struct sk_buff
*skb
)
1354 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
1355 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
1356 return dataref
!= 1;
1359 static inline int skb_header_unclone(struct sk_buff
*skb
, gfp_t pri
)
1361 might_sleep_if(gfpflags_allow_blocking(pri
));
1363 if (skb_header_cloned(skb
))
1364 return pskb_expand_head(skb
, 0, 0, pri
);
1370 * skb_header_release - release reference to header
1371 * @skb: buffer to operate on
1373 * Drop a reference to the header part of the buffer. This is done
1374 * by acquiring a payload reference. You must not read from the header
1375 * part of skb->data after this.
1376 * Note : Check if you can use __skb_header_release() instead.
1378 static inline void skb_header_release(struct sk_buff
*skb
)
1382 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
1386 * __skb_header_release - release reference to header
1387 * @skb: buffer to operate on
1389 * Variant of skb_header_release() assuming skb is private to caller.
1390 * We can avoid one atomic operation.
1392 static inline void __skb_header_release(struct sk_buff
*skb
)
1395 atomic_set(&skb_shinfo(skb
)->dataref
, 1 + (1 << SKB_DATAREF_SHIFT
));
1400 * skb_shared - is the buffer shared
1401 * @skb: buffer to check
1403 * Returns true if more than one person has a reference to this
1406 static inline int skb_shared(const struct sk_buff
*skb
)
1408 return atomic_read(&skb
->users
) != 1;
1412 * skb_share_check - check if buffer is shared and if so clone it
1413 * @skb: buffer to check
1414 * @pri: priority for memory allocation
1416 * If the buffer is shared the buffer is cloned and the old copy
1417 * drops a reference. A new clone with a single reference is returned.
1418 * If the buffer is not shared the original buffer is returned. When
1419 * being called from interrupt status or with spinlocks held pri must
1422 * NULL is returned on a memory allocation failure.
1424 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
, gfp_t pri
)
1426 might_sleep_if(gfpflags_allow_blocking(pri
));
1427 if (skb_shared(skb
)) {
1428 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
1440 * Copy shared buffers into a new sk_buff. We effectively do COW on
1441 * packets to handle cases where we have a local reader and forward
1442 * and a couple of other messy ones. The normal one is tcpdumping
1443 * a packet thats being forwarded.
1447 * skb_unshare - make a copy of a shared buffer
1448 * @skb: buffer to check
1449 * @pri: priority for memory allocation
1451 * If the socket buffer is a clone then this function creates a new
1452 * copy of the data, drops a reference count on the old copy and returns
1453 * the new copy with the reference count at 1. If the buffer is not a clone
1454 * the original buffer is returned. When called with a spinlock held or
1455 * from interrupt state @pri must be %GFP_ATOMIC
1457 * %NULL is returned on a memory allocation failure.
1459 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
1462 might_sleep_if(gfpflags_allow_blocking(pri
));
1463 if (skb_cloned(skb
)) {
1464 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
1466 /* Free our shared copy */
1477 * skb_peek - peek at the head of an &sk_buff_head
1478 * @list_: list to peek at
1480 * Peek an &sk_buff. Unlike most other operations you _MUST_
1481 * be careful with this one. A peek leaves the buffer on the
1482 * list and someone else may run off with it. You must hold
1483 * the appropriate locks or have a private queue to do this.
1485 * Returns %NULL for an empty list or a pointer to the head element.
1486 * The reference count is not incremented and the reference is therefore
1487 * volatile. Use with caution.
1489 static inline struct sk_buff
*skb_peek(const struct sk_buff_head
*list_
)
1491 struct sk_buff
*skb
= list_
->next
;
1493 if (skb
== (struct sk_buff
*)list_
)
1499 * skb_peek_next - peek skb following the given one from a queue
1500 * @skb: skb to start from
1501 * @list_: list to peek at
1503 * Returns %NULL when the end of the list is met or a pointer to the
1504 * next element. The reference count is not incremented and the
1505 * reference is therefore volatile. Use with caution.
1507 static inline struct sk_buff
*skb_peek_next(struct sk_buff
*skb
,
1508 const struct sk_buff_head
*list_
)
1510 struct sk_buff
*next
= skb
->next
;
1512 if (next
== (struct sk_buff
*)list_
)
1518 * skb_peek_tail - peek at the tail of an &sk_buff_head
1519 * @list_: list to peek at
1521 * Peek an &sk_buff. Unlike most other operations you _MUST_
1522 * be careful with this one. A peek leaves the buffer on the
1523 * list and someone else may run off with it. You must hold
1524 * the appropriate locks or have a private queue to do this.
1526 * Returns %NULL for an empty list or a pointer to the tail element.
1527 * The reference count is not incremented and the reference is therefore
1528 * volatile. Use with caution.
1530 static inline struct sk_buff
*skb_peek_tail(const struct sk_buff_head
*list_
)
1532 struct sk_buff
*skb
= list_
->prev
;
1534 if (skb
== (struct sk_buff
*)list_
)
1541 * skb_queue_len - get queue length
1542 * @list_: list to measure
1544 * Return the length of an &sk_buff queue.
1546 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
1552 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1553 * @list: queue to initialize
1555 * This initializes only the list and queue length aspects of
1556 * an sk_buff_head object. This allows to initialize the list
1557 * aspects of an sk_buff_head without reinitializing things like
1558 * the spinlock. It can also be used for on-stack sk_buff_head
1559 * objects where the spinlock is known to not be used.
1561 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
1563 list
->prev
= list
->next
= (struct sk_buff
*)list
;
1568 * This function creates a split out lock class for each invocation;
1569 * this is needed for now since a whole lot of users of the skb-queue
1570 * infrastructure in drivers have different locking usage (in hardirq)
1571 * than the networking core (in softirq only). In the long run either the
1572 * network layer or drivers should need annotation to consolidate the
1573 * main types of usage into 3 classes.
1575 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
1577 spin_lock_init(&list
->lock
);
1578 __skb_queue_head_init(list
);
1581 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
1582 struct lock_class_key
*class)
1584 skb_queue_head_init(list
);
1585 lockdep_set_class(&list
->lock
, class);
1589 * Insert an sk_buff on a list.
1591 * The "__skb_xxxx()" functions are the non-atomic ones that
1592 * can only be called with interrupts disabled.
1594 void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
,
1595 struct sk_buff_head
*list
);
1596 static inline void __skb_insert(struct sk_buff
*newsk
,
1597 struct sk_buff
*prev
, struct sk_buff
*next
,
1598 struct sk_buff_head
*list
)
1602 next
->prev
= prev
->next
= newsk
;
1606 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
1607 struct sk_buff
*prev
,
1608 struct sk_buff
*next
)
1610 struct sk_buff
*first
= list
->next
;
1611 struct sk_buff
*last
= list
->prev
;
1621 * skb_queue_splice - join two skb lists, this is designed for stacks
1622 * @list: the new list to add
1623 * @head: the place to add it in the first list
1625 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
1626 struct sk_buff_head
*head
)
1628 if (!skb_queue_empty(list
)) {
1629 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1630 head
->qlen
+= list
->qlen
;
1635 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1636 * @list: the new list to add
1637 * @head: the place to add it in the first list
1639 * The list at @list is reinitialised
1641 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
1642 struct sk_buff_head
*head
)
1644 if (!skb_queue_empty(list
)) {
1645 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1646 head
->qlen
+= list
->qlen
;
1647 __skb_queue_head_init(list
);
1652 * skb_queue_splice_tail - join two skb lists, each list being a queue
1653 * @list: the new list to add
1654 * @head: the place to add it in the first list
1656 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
1657 struct sk_buff_head
*head
)
1659 if (!skb_queue_empty(list
)) {
1660 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1661 head
->qlen
+= list
->qlen
;
1666 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1667 * @list: the new list to add
1668 * @head: the place to add it in the first list
1670 * Each of the lists is a queue.
1671 * The list at @list is reinitialised
1673 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
1674 struct sk_buff_head
*head
)
1676 if (!skb_queue_empty(list
)) {
1677 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1678 head
->qlen
+= list
->qlen
;
1679 __skb_queue_head_init(list
);
1684 * __skb_queue_after - queue a buffer at the list head
1685 * @list: list to use
1686 * @prev: place after this buffer
1687 * @newsk: buffer to queue
1689 * Queue a buffer int the middle of a list. This function takes no locks
1690 * and you must therefore hold required locks before calling it.
1692 * A buffer cannot be placed on two lists at the same time.
1694 static inline void __skb_queue_after(struct sk_buff_head
*list
,
1695 struct sk_buff
*prev
,
1696 struct sk_buff
*newsk
)
1698 __skb_insert(newsk
, prev
, prev
->next
, list
);
1701 void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
1702 struct sk_buff_head
*list
);
1704 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1705 struct sk_buff
*next
,
1706 struct sk_buff
*newsk
)
1708 __skb_insert(newsk
, next
->prev
, next
, list
);
1712 * __skb_queue_head - queue a buffer at the list head
1713 * @list: list to use
1714 * @newsk: buffer to queue
1716 * Queue a buffer at the start of a list. This function takes no locks
1717 * and you must therefore hold required locks before calling it.
1719 * A buffer cannot be placed on two lists at the same time.
1721 void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1722 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1723 struct sk_buff
*newsk
)
1725 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1729 * __skb_queue_tail - queue a buffer at the list tail
1730 * @list: list to use
1731 * @newsk: buffer to queue
1733 * Queue a buffer at the end of a list. This function takes no locks
1734 * and you must therefore hold required locks before calling it.
1736 * A buffer cannot be placed on two lists at the same time.
1738 void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1739 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1740 struct sk_buff
*newsk
)
1742 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1746 * remove sk_buff from list. _Must_ be called atomically, and with
1749 void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1750 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1752 struct sk_buff
*next
, *prev
;
1757 skb
->next
= skb
->prev
= NULL
;
1763 * __skb_dequeue - remove from the head of the queue
1764 * @list: list to dequeue from
1766 * Remove the head of the list. This function does not take any locks
1767 * so must be used with appropriate locks held only. The head item is
1768 * returned or %NULL if the list is empty.
1770 struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1771 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1773 struct sk_buff
*skb
= skb_peek(list
);
1775 __skb_unlink(skb
, list
);
1780 * __skb_dequeue_tail - remove from the tail of the queue
1781 * @list: list to dequeue from
1783 * Remove the tail of the list. This function does not take any locks
1784 * so must be used with appropriate locks held only. The tail item is
1785 * returned or %NULL if the list is empty.
1787 struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1788 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1790 struct sk_buff
*skb
= skb_peek_tail(list
);
1792 __skb_unlink(skb
, list
);
1797 static inline bool skb_is_nonlinear(const struct sk_buff
*skb
)
1799 return skb
->data_len
;
1802 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1804 return skb
->len
- skb
->data_len
;
1807 static inline int skb_pagelen(const struct sk_buff
*skb
)
1811 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1812 len
+= skb_frag_size(&skb_shinfo(skb
)->frags
[i
]);
1813 return len
+ skb_headlen(skb
);
1817 * __skb_fill_page_desc - initialise a paged fragment in an skb
1818 * @skb: buffer containing fragment to be initialised
1819 * @i: paged fragment index to initialise
1820 * @page: the page to use for this fragment
1821 * @off: the offset to the data with @page
1822 * @size: the length of the data
1824 * Initialises the @i'th fragment of @skb to point to &size bytes at
1825 * offset @off within @page.
1827 * Does not take any additional reference on the fragment.
1829 static inline void __skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1830 struct page
*page
, int off
, int size
)
1832 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1835 * Propagate page pfmemalloc to the skb if we can. The problem is
1836 * that not all callers have unique ownership of the page but rely
1837 * on page_is_pfmemalloc doing the right thing(tm).
1839 frag
->page
.p
= page
;
1840 frag
->page_offset
= off
;
1841 skb_frag_size_set(frag
, size
);
1843 page
= compound_head(page
);
1844 if (page_is_pfmemalloc(page
))
1845 skb
->pfmemalloc
= true;
1849 * skb_fill_page_desc - initialise a paged fragment in an skb
1850 * @skb: buffer containing fragment to be initialised
1851 * @i: paged fragment index to initialise
1852 * @page: the page to use for this fragment
1853 * @off: the offset to the data with @page
1854 * @size: the length of the data
1856 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1857 * @skb to point to @size bytes at offset @off within @page. In
1858 * addition updates @skb such that @i is the last fragment.
1860 * Does not take any additional reference on the fragment.
1862 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1863 struct page
*page
, int off
, int size
)
1865 __skb_fill_page_desc(skb
, i
, page
, off
, size
);
1866 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1869 void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
, int off
,
1870 int size
, unsigned int truesize
);
1872 void skb_coalesce_rx_frag(struct sk_buff
*skb
, int i
, int size
,
1873 unsigned int truesize
);
1875 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1876 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1877 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1879 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1880 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1882 return skb
->head
+ skb
->tail
;
1885 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1887 skb
->tail
= skb
->data
- skb
->head
;
1890 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1892 skb_reset_tail_pointer(skb
);
1893 skb
->tail
+= offset
;
1896 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1897 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1902 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1904 skb
->tail
= skb
->data
;
1907 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1909 skb
->tail
= skb
->data
+ offset
;
1912 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1915 * Add data to an sk_buff
1917 unsigned char *pskb_put(struct sk_buff
*skb
, struct sk_buff
*tail
, int len
);
1918 unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1919 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1921 unsigned char *tmp
= skb_tail_pointer(skb
);
1922 SKB_LINEAR_ASSERT(skb
);
1928 unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1929 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1936 unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1937 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1940 BUG_ON(skb
->len
< skb
->data_len
);
1941 return skb
->data
+= len
;
1944 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1946 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1949 unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1951 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1953 if (len
> skb_headlen(skb
) &&
1954 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1957 return skb
->data
+= len
;
1960 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1962 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1965 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1967 if (likely(len
<= skb_headlen(skb
)))
1969 if (unlikely(len
> skb
->len
))
1971 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1975 * skb_headroom - bytes at buffer head
1976 * @skb: buffer to check
1978 * Return the number of bytes of free space at the head of an &sk_buff.
1980 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1982 return skb
->data
- skb
->head
;
1986 * skb_tailroom - bytes at buffer end
1987 * @skb: buffer to check
1989 * Return the number of bytes of free space at the tail of an sk_buff
1991 static inline int skb_tailroom(const struct sk_buff
*skb
)
1993 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1997 * skb_availroom - bytes at buffer end
1998 * @skb: buffer to check
2000 * Return the number of bytes of free space at the tail of an sk_buff
2001 * allocated by sk_stream_alloc()
2003 static inline int skb_availroom(const struct sk_buff
*skb
)
2005 if (skb_is_nonlinear(skb
))
2008 return skb
->end
- skb
->tail
- skb
->reserved_tailroom
;
2012 * skb_reserve - adjust headroom
2013 * @skb: buffer to alter
2014 * @len: bytes to move
2016 * Increase the headroom of an empty &sk_buff by reducing the tail
2017 * room. This is only allowed for an empty buffer.
2019 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
2026 * skb_tailroom_reserve - adjust reserved_tailroom
2027 * @skb: buffer to alter
2028 * @mtu: maximum amount of headlen permitted
2029 * @needed_tailroom: minimum amount of reserved_tailroom
2031 * Set reserved_tailroom so that headlen can be as large as possible but
2032 * not larger than mtu and tailroom cannot be smaller than
2034 * The required headroom should already have been reserved before using
2037 static inline void skb_tailroom_reserve(struct sk_buff
*skb
, unsigned int mtu
,
2038 unsigned int needed_tailroom
)
2040 SKB_LINEAR_ASSERT(skb
);
2041 if (mtu
< skb_tailroom(skb
) - needed_tailroom
)
2042 /* use at most mtu */
2043 skb
->reserved_tailroom
= skb_tailroom(skb
) - mtu
;
2045 /* use up to all available space */
2046 skb
->reserved_tailroom
= needed_tailroom
;
2049 #define ENCAP_TYPE_ETHER 0
2050 #define ENCAP_TYPE_IPPROTO 1
2052 static inline void skb_set_inner_protocol(struct sk_buff
*skb
,
2055 skb
->inner_protocol
= protocol
;
2056 skb
->inner_protocol_type
= ENCAP_TYPE_ETHER
;
2059 static inline void skb_set_inner_ipproto(struct sk_buff
*skb
,
2062 skb
->inner_ipproto
= ipproto
;
2063 skb
->inner_protocol_type
= ENCAP_TYPE_IPPROTO
;
2066 static inline void skb_reset_inner_headers(struct sk_buff
*skb
)
2068 skb
->inner_mac_header
= skb
->mac_header
;
2069 skb
->inner_network_header
= skb
->network_header
;
2070 skb
->inner_transport_header
= skb
->transport_header
;
2073 static inline void skb_reset_mac_len(struct sk_buff
*skb
)
2075 skb
->mac_len
= skb
->network_header
- skb
->mac_header
;
2078 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2081 return skb
->head
+ skb
->inner_transport_header
;
2084 static inline int skb_inner_transport_offset(const struct sk_buff
*skb
)
2086 return skb_inner_transport_header(skb
) - skb
->data
;
2089 static inline void skb_reset_inner_transport_header(struct sk_buff
*skb
)
2091 skb
->inner_transport_header
= skb
->data
- skb
->head
;
2094 static inline void skb_set_inner_transport_header(struct sk_buff
*skb
,
2097 skb_reset_inner_transport_header(skb
);
2098 skb
->inner_transport_header
+= offset
;
2101 static inline unsigned char *skb_inner_network_header(const struct sk_buff
*skb
)
2103 return skb
->head
+ skb
->inner_network_header
;
2106 static inline void skb_reset_inner_network_header(struct sk_buff
*skb
)
2108 skb
->inner_network_header
= skb
->data
- skb
->head
;
2111 static inline void skb_set_inner_network_header(struct sk_buff
*skb
,
2114 skb_reset_inner_network_header(skb
);
2115 skb
->inner_network_header
+= offset
;
2118 static inline unsigned char *skb_inner_mac_header(const struct sk_buff
*skb
)
2120 return skb
->head
+ skb
->inner_mac_header
;
2123 static inline void skb_reset_inner_mac_header(struct sk_buff
*skb
)
2125 skb
->inner_mac_header
= skb
->data
- skb
->head
;
2128 static inline void skb_set_inner_mac_header(struct sk_buff
*skb
,
2131 skb_reset_inner_mac_header(skb
);
2132 skb
->inner_mac_header
+= offset
;
2134 static inline bool skb_transport_header_was_set(const struct sk_buff
*skb
)
2136 return skb
->transport_header
!= (typeof(skb
->transport_header
))~0U;
2139 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
2141 return skb
->head
+ skb
->transport_header
;
2144 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
2146 skb
->transport_header
= skb
->data
- skb
->head
;
2149 static inline void skb_set_transport_header(struct sk_buff
*skb
,
2152 skb_reset_transport_header(skb
);
2153 skb
->transport_header
+= offset
;
2156 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
2158 return skb
->head
+ skb
->network_header
;
2161 static inline void skb_reset_network_header(struct sk_buff
*skb
)
2163 skb
->network_header
= skb
->data
- skb
->head
;
2166 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
2168 skb_reset_network_header(skb
);
2169 skb
->network_header
+= offset
;
2172 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
2174 return skb
->head
+ skb
->mac_header
;
2177 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
2179 return skb
->mac_header
!= (typeof(skb
->mac_header
))~0U;
2182 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
2184 skb
->mac_header
= skb
->data
- skb
->head
;
2187 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
2189 skb_reset_mac_header(skb
);
2190 skb
->mac_header
+= offset
;
2193 static inline void skb_pop_mac_header(struct sk_buff
*skb
)
2195 skb
->mac_header
= skb
->network_header
;
2198 static inline void skb_probe_transport_header(struct sk_buff
*skb
,
2199 const int offset_hint
)
2201 struct flow_keys keys
;
2203 if (skb_transport_header_was_set(skb
))
2205 else if (skb_flow_dissect_flow_keys(skb
, &keys
, 0))
2206 skb_set_transport_header(skb
, keys
.control
.thoff
);
2208 skb_set_transport_header(skb
, offset_hint
);
2211 static inline void skb_mac_header_rebuild(struct sk_buff
*skb
)
2213 if (skb_mac_header_was_set(skb
)) {
2214 const unsigned char *old_mac
= skb_mac_header(skb
);
2216 skb_set_mac_header(skb
, -skb
->mac_len
);
2217 memmove(skb_mac_header(skb
), old_mac
, skb
->mac_len
);
2221 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
2223 return skb
->csum_start
- skb_headroom(skb
);
2226 static inline unsigned char *skb_checksum_start(const struct sk_buff
*skb
)
2228 return skb
->head
+ skb
->csum_start
;
2231 static inline int skb_transport_offset(const struct sk_buff
*skb
)
2233 return skb_transport_header(skb
) - skb
->data
;
2236 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
2238 return skb
->transport_header
- skb
->network_header
;
2241 static inline u32
skb_inner_network_header_len(const struct sk_buff
*skb
)
2243 return skb
->inner_transport_header
- skb
->inner_network_header
;
2246 static inline int skb_network_offset(const struct sk_buff
*skb
)
2248 return skb_network_header(skb
) - skb
->data
;
2251 static inline int skb_inner_network_offset(const struct sk_buff
*skb
)
2253 return skb_inner_network_header(skb
) - skb
->data
;
2256 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
2258 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
2262 * CPUs often take a performance hit when accessing unaligned memory
2263 * locations. The actual performance hit varies, it can be small if the
2264 * hardware handles it or large if we have to take an exception and fix it
2267 * Since an ethernet header is 14 bytes network drivers often end up with
2268 * the IP header at an unaligned offset. The IP header can be aligned by
2269 * shifting the start of the packet by 2 bytes. Drivers should do this
2272 * skb_reserve(skb, NET_IP_ALIGN);
2274 * The downside to this alignment of the IP header is that the DMA is now
2275 * unaligned. On some architectures the cost of an unaligned DMA is high
2276 * and this cost outweighs the gains made by aligning the IP header.
2278 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2281 #ifndef NET_IP_ALIGN
2282 #define NET_IP_ALIGN 2
2286 * The networking layer reserves some headroom in skb data (via
2287 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2288 * the header has to grow. In the default case, if the header has to grow
2289 * 32 bytes or less we avoid the reallocation.
2291 * Unfortunately this headroom changes the DMA alignment of the resulting
2292 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2293 * on some architectures. An architecture can override this value,
2294 * perhaps setting it to a cacheline in size (since that will maintain
2295 * cacheline alignment of the DMA). It must be a power of 2.
2297 * Various parts of the networking layer expect at least 32 bytes of
2298 * headroom, you should not reduce this.
2300 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2301 * to reduce average number of cache lines per packet.
2302 * get_rps_cpus() for example only access one 64 bytes aligned block :
2303 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2306 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2309 int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
2311 static inline void __skb_set_length(struct sk_buff
*skb
, unsigned int len
)
2313 if (unlikely(skb_is_nonlinear(skb
))) {
2318 skb_set_tail_pointer(skb
, len
);
2321 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
2323 __skb_set_length(skb
, len
);
2326 void skb_trim(struct sk_buff
*skb
, unsigned int len
);
2328 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2331 return ___pskb_trim(skb
, len
);
2332 __skb_trim(skb
, len
);
2336 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2338 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
2342 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2343 * @skb: buffer to alter
2346 * This is identical to pskb_trim except that the caller knows that
2347 * the skb is not cloned so we should never get an error due to out-
2350 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
2352 int err
= pskb_trim(skb
, len
);
2356 static inline int __skb_grow(struct sk_buff
*skb
, unsigned int len
)
2358 unsigned int diff
= len
- skb
->len
;
2360 if (skb_tailroom(skb
) < diff
) {
2361 int ret
= pskb_expand_head(skb
, 0, diff
- skb_tailroom(skb
),
2366 __skb_set_length(skb
, len
);
2371 * skb_orphan - orphan a buffer
2372 * @skb: buffer to orphan
2374 * If a buffer currently has an owner then we call the owner's
2375 * destructor function and make the @skb unowned. The buffer continues
2376 * to exist but is no longer charged to its former owner.
2378 static inline void skb_orphan(struct sk_buff
*skb
)
2380 if (skb
->destructor
) {
2381 skb
->destructor(skb
);
2382 skb
->destructor
= NULL
;
2390 * skb_orphan_frags - orphan the frags contained in a buffer
2391 * @skb: buffer to orphan frags from
2392 * @gfp_mask: allocation mask for replacement pages
2394 * For each frag in the SKB which needs a destructor (i.e. has an
2395 * owner) create a copy of that frag and release the original
2396 * page by calling the destructor.
2398 static inline int skb_orphan_frags(struct sk_buff
*skb
, gfp_t gfp_mask
)
2400 if (likely(!(skb_shinfo(skb
)->tx_flags
& SKBTX_DEV_ZEROCOPY
)))
2402 return skb_copy_ubufs(skb
, gfp_mask
);
2406 * __skb_queue_purge - empty a list
2407 * @list: list to empty
2409 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2410 * the list and one reference dropped. This function does not take the
2411 * list lock and the caller must hold the relevant locks to use it.
2413 void skb_queue_purge(struct sk_buff_head
*list
);
2414 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
2416 struct sk_buff
*skb
;
2417 while ((skb
= __skb_dequeue(list
)) != NULL
)
2421 unsigned int skb_rbtree_purge(struct rb_root
*root
);
2423 void *netdev_alloc_frag(unsigned int fragsz
);
2425 struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
, unsigned int length
,
2429 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2430 * @dev: network device to receive on
2431 * @length: length to allocate
2433 * Allocate a new &sk_buff and assign it a usage count of one. The
2434 * buffer has unspecified headroom built in. Users should allocate
2435 * the headroom they think they need without accounting for the
2436 * built in space. The built in space is used for optimisations.
2438 * %NULL is returned if there is no free memory. Although this function
2439 * allocates memory it can be called from an interrupt.
2441 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
2442 unsigned int length
)
2444 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
2447 /* legacy helper around __netdev_alloc_skb() */
2448 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
2451 return __netdev_alloc_skb(NULL
, length
, gfp_mask
);
2454 /* legacy helper around netdev_alloc_skb() */
2455 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
2457 return netdev_alloc_skb(NULL
, length
);
2461 static inline struct sk_buff
*__netdev_alloc_skb_ip_align(struct net_device
*dev
,
2462 unsigned int length
, gfp_t gfp
)
2464 struct sk_buff
*skb
= __netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
, gfp
);
2466 if (NET_IP_ALIGN
&& skb
)
2467 skb_reserve(skb
, NET_IP_ALIGN
);
2471 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
2472 unsigned int length
)
2474 return __netdev_alloc_skb_ip_align(dev
, length
, GFP_ATOMIC
);
2477 static inline void skb_free_frag(void *addr
)
2479 __free_page_frag(addr
);
2482 void *napi_alloc_frag(unsigned int fragsz
);
2483 struct sk_buff
*__napi_alloc_skb(struct napi_struct
*napi
,
2484 unsigned int length
, gfp_t gfp_mask
);
2485 static inline struct sk_buff
*napi_alloc_skb(struct napi_struct
*napi
,
2486 unsigned int length
)
2488 return __napi_alloc_skb(napi
, length
, GFP_ATOMIC
);
2490 void napi_consume_skb(struct sk_buff
*skb
, int budget
);
2492 void __kfree_skb_flush(void);
2493 void __kfree_skb_defer(struct sk_buff
*skb
);
2496 * __dev_alloc_pages - allocate page for network Rx
2497 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2498 * @order: size of the allocation
2500 * Allocate a new page.
2502 * %NULL is returned if there is no free memory.
2504 static inline struct page
*__dev_alloc_pages(gfp_t gfp_mask
,
2507 /* This piece of code contains several assumptions.
2508 * 1. This is for device Rx, therefor a cold page is preferred.
2509 * 2. The expectation is the user wants a compound page.
2510 * 3. If requesting a order 0 page it will not be compound
2511 * due to the check to see if order has a value in prep_new_page
2512 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2513 * code in gfp_to_alloc_flags that should be enforcing this.
2515 gfp_mask
|= __GFP_COLD
| __GFP_COMP
| __GFP_MEMALLOC
;
2517 return alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, order
);
2520 static inline struct page
*dev_alloc_pages(unsigned int order
)
2522 return __dev_alloc_pages(GFP_ATOMIC
| __GFP_NOWARN
, order
);
2526 * __dev_alloc_page - allocate a page for network Rx
2527 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2529 * Allocate a new page.
2531 * %NULL is returned if there is no free memory.
2533 static inline struct page
*__dev_alloc_page(gfp_t gfp_mask
)
2535 return __dev_alloc_pages(gfp_mask
, 0);
2538 static inline struct page
*dev_alloc_page(void)
2540 return dev_alloc_pages(0);
2544 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2545 * @page: The page that was allocated from skb_alloc_page
2546 * @skb: The skb that may need pfmemalloc set
2548 static inline void skb_propagate_pfmemalloc(struct page
*page
,
2549 struct sk_buff
*skb
)
2551 if (page_is_pfmemalloc(page
))
2552 skb
->pfmemalloc
= true;
2556 * skb_frag_page - retrieve the page referred to by a paged fragment
2557 * @frag: the paged fragment
2559 * Returns the &struct page associated with @frag.
2561 static inline struct page
*skb_frag_page(const skb_frag_t
*frag
)
2563 return frag
->page
.p
;
2567 * __skb_frag_ref - take an addition reference on a paged fragment.
2568 * @frag: the paged fragment
2570 * Takes an additional reference on the paged fragment @frag.
2572 static inline void __skb_frag_ref(skb_frag_t
*frag
)
2574 get_page(skb_frag_page(frag
));
2578 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2580 * @f: the fragment offset.
2582 * Takes an additional reference on the @f'th paged fragment of @skb.
2584 static inline void skb_frag_ref(struct sk_buff
*skb
, int f
)
2586 __skb_frag_ref(&skb_shinfo(skb
)->frags
[f
]);
2590 * __skb_frag_unref - release a reference on a paged fragment.
2591 * @frag: the paged fragment
2593 * Releases a reference on the paged fragment @frag.
2595 static inline void __skb_frag_unref(skb_frag_t
*frag
)
2597 put_page(skb_frag_page(frag
));
2601 * skb_frag_unref - release a reference on a paged fragment of an skb.
2603 * @f: the fragment offset
2605 * Releases a reference on the @f'th paged fragment of @skb.
2607 static inline void skb_frag_unref(struct sk_buff
*skb
, int f
)
2609 __skb_frag_unref(&skb_shinfo(skb
)->frags
[f
]);
2613 * skb_frag_address - gets the address of the data contained in a paged fragment
2614 * @frag: the paged fragment buffer
2616 * Returns the address of the data within @frag. The page must already
2619 static inline void *skb_frag_address(const skb_frag_t
*frag
)
2621 return page_address(skb_frag_page(frag
)) + frag
->page_offset
;
2625 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2626 * @frag: the paged fragment buffer
2628 * Returns the address of the data within @frag. Checks that the page
2629 * is mapped and returns %NULL otherwise.
2631 static inline void *skb_frag_address_safe(const skb_frag_t
*frag
)
2633 void *ptr
= page_address(skb_frag_page(frag
));
2637 return ptr
+ frag
->page_offset
;
2641 * __skb_frag_set_page - sets the page contained in a paged fragment
2642 * @frag: the paged fragment
2643 * @page: the page to set
2645 * Sets the fragment @frag to contain @page.
2647 static inline void __skb_frag_set_page(skb_frag_t
*frag
, struct page
*page
)
2649 frag
->page
.p
= page
;
2653 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2655 * @f: the fragment offset
2656 * @page: the page to set
2658 * Sets the @f'th fragment of @skb to contain @page.
2660 static inline void skb_frag_set_page(struct sk_buff
*skb
, int f
,
2663 __skb_frag_set_page(&skb_shinfo(skb
)->frags
[f
], page
);
2666 bool skb_page_frag_refill(unsigned int sz
, struct page_frag
*pfrag
, gfp_t prio
);
2669 * skb_frag_dma_map - maps a paged fragment via the DMA API
2670 * @dev: the device to map the fragment to
2671 * @frag: the paged fragment to map
2672 * @offset: the offset within the fragment (starting at the
2673 * fragment's own offset)
2674 * @size: the number of bytes to map
2675 * @dir: the direction of the mapping (%PCI_DMA_*)
2677 * Maps the page associated with @frag to @device.
2679 static inline dma_addr_t
skb_frag_dma_map(struct device
*dev
,
2680 const skb_frag_t
*frag
,
2681 size_t offset
, size_t size
,
2682 enum dma_data_direction dir
)
2684 return dma_map_page(dev
, skb_frag_page(frag
),
2685 frag
->page_offset
+ offset
, size
, dir
);
2688 static inline struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
2691 return __pskb_copy(skb
, skb_headroom(skb
), gfp_mask
);
2695 static inline struct sk_buff
*pskb_copy_for_clone(struct sk_buff
*skb
,
2698 return __pskb_copy_fclone(skb
, skb_headroom(skb
), gfp_mask
, true);
2703 * skb_clone_writable - is the header of a clone writable
2704 * @skb: buffer to check
2705 * @len: length up to which to write
2707 * Returns true if modifying the header part of the cloned buffer
2708 * does not requires the data to be copied.
2710 static inline int skb_clone_writable(const struct sk_buff
*skb
, unsigned int len
)
2712 return !skb_header_cloned(skb
) &&
2713 skb_headroom(skb
) + len
<= skb
->hdr_len
;
2716 static inline int skb_try_make_writable(struct sk_buff
*skb
,
2717 unsigned int write_len
)
2719 return skb_cloned(skb
) && !skb_clone_writable(skb
, write_len
) &&
2720 pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
);
2723 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
2728 if (headroom
> skb_headroom(skb
))
2729 delta
= headroom
- skb_headroom(skb
);
2731 if (delta
|| cloned
)
2732 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
2738 * skb_cow - copy header of skb when it is required
2739 * @skb: buffer to cow
2740 * @headroom: needed headroom
2742 * If the skb passed lacks sufficient headroom or its data part
2743 * is shared, data is reallocated. If reallocation fails, an error
2744 * is returned and original skb is not changed.
2746 * The result is skb with writable area skb->head...skb->tail
2747 * and at least @headroom of space at head.
2749 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
2751 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
2755 * skb_cow_head - skb_cow but only making the head writable
2756 * @skb: buffer to cow
2757 * @headroom: needed headroom
2759 * This function is identical to skb_cow except that we replace the
2760 * skb_cloned check by skb_header_cloned. It should be used when
2761 * you only need to push on some header and do not need to modify
2764 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
2766 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
2770 * skb_padto - pad an skbuff up to a minimal size
2771 * @skb: buffer to pad
2772 * @len: minimal length
2774 * Pads up a buffer to ensure the trailing bytes exist and are
2775 * blanked. If the buffer already contains sufficient data it
2776 * is untouched. Otherwise it is extended. Returns zero on
2777 * success. The skb is freed on error.
2779 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
2781 unsigned int size
= skb
->len
;
2782 if (likely(size
>= len
))
2784 return skb_pad(skb
, len
- size
);
2788 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2789 * @skb: buffer to pad
2790 * @len: minimal length
2792 * Pads up a buffer to ensure the trailing bytes exist and are
2793 * blanked. If the buffer already contains sufficient data it
2794 * is untouched. Otherwise it is extended. Returns zero on
2795 * success. The skb is freed on error.
2797 static inline int skb_put_padto(struct sk_buff
*skb
, unsigned int len
)
2799 unsigned int size
= skb
->len
;
2801 if (unlikely(size
< len
)) {
2803 if (skb_pad(skb
, len
))
2805 __skb_put(skb
, len
);
2810 static inline int skb_add_data(struct sk_buff
*skb
,
2811 struct iov_iter
*from
, int copy
)
2813 const int off
= skb
->len
;
2815 if (skb
->ip_summed
== CHECKSUM_NONE
) {
2817 if (csum_and_copy_from_iter(skb_put(skb
, copy
), copy
,
2818 &csum
, from
) == copy
) {
2819 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
2822 } else if (copy_from_iter(skb_put(skb
, copy
), copy
, from
) == copy
)
2825 __skb_trim(skb
, off
);
2829 static inline bool skb_can_coalesce(struct sk_buff
*skb
, int i
,
2830 const struct page
*page
, int off
)
2833 const struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
2835 return page
== skb_frag_page(frag
) &&
2836 off
== frag
->page_offset
+ skb_frag_size(frag
);
2841 static inline int __skb_linearize(struct sk_buff
*skb
)
2843 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
2847 * skb_linearize - convert paged skb to linear one
2848 * @skb: buffer to linarize
2850 * If there is no free memory -ENOMEM is returned, otherwise zero
2851 * is returned and the old skb data released.
2853 static inline int skb_linearize(struct sk_buff
*skb
)
2855 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
2859 * skb_has_shared_frag - can any frag be overwritten
2860 * @skb: buffer to test
2862 * Return true if the skb has at least one frag that might be modified
2863 * by an external entity (as in vmsplice()/sendfile())
2865 static inline bool skb_has_shared_frag(const struct sk_buff
*skb
)
2867 return skb_is_nonlinear(skb
) &&
2868 skb_shinfo(skb
)->tx_flags
& SKBTX_SHARED_FRAG
;
2872 * skb_linearize_cow - make sure skb is linear and writable
2873 * @skb: buffer to process
2875 * If there is no free memory -ENOMEM is returned, otherwise zero
2876 * is returned and the old skb data released.
2878 static inline int skb_linearize_cow(struct sk_buff
*skb
)
2880 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
2881 __skb_linearize(skb
) : 0;
2884 static __always_inline
void
2885 __skb_postpull_rcsum(struct sk_buff
*skb
, const void *start
, unsigned int len
,
2888 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2889 skb
->csum
= csum_block_sub(skb
->csum
,
2890 csum_partial(start
, len
, 0), off
);
2891 else if (skb
->ip_summed
== CHECKSUM_PARTIAL
&&
2892 skb_checksum_start_offset(skb
) < 0)
2893 skb
->ip_summed
= CHECKSUM_NONE
;
2897 * skb_postpull_rcsum - update checksum for received skb after pull
2898 * @skb: buffer to update
2899 * @start: start of data before pull
2900 * @len: length of data pulled
2902 * After doing a pull on a received packet, you need to call this to
2903 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2904 * CHECKSUM_NONE so that it can be recomputed from scratch.
2906 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
2907 const void *start
, unsigned int len
)
2909 __skb_postpull_rcsum(skb
, start
, len
, 0);
2912 static __always_inline
void
2913 __skb_postpush_rcsum(struct sk_buff
*skb
, const void *start
, unsigned int len
,
2916 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2917 skb
->csum
= csum_block_add(skb
->csum
,
2918 csum_partial(start
, len
, 0), off
);
2922 * skb_postpush_rcsum - update checksum for received skb after push
2923 * @skb: buffer to update
2924 * @start: start of data after push
2925 * @len: length of data pushed
2927 * After doing a push on a received packet, you need to call this to
2928 * update the CHECKSUM_COMPLETE checksum.
2930 static inline void skb_postpush_rcsum(struct sk_buff
*skb
,
2931 const void *start
, unsigned int len
)
2933 __skb_postpush_rcsum(skb
, start
, len
, 0);
2936 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
2939 * skb_push_rcsum - push skb and update receive checksum
2940 * @skb: buffer to update
2941 * @len: length of data pulled
2943 * This function performs an skb_push on the packet and updates
2944 * the CHECKSUM_COMPLETE checksum. It should be used on
2945 * receive path processing instead of skb_push unless you know
2946 * that the checksum difference is zero (e.g., a valid IP header)
2947 * or you are setting ip_summed to CHECKSUM_NONE.
2949 static inline unsigned char *skb_push_rcsum(struct sk_buff
*skb
,
2953 skb_postpush_rcsum(skb
, skb
->data
, len
);
2957 int pskb_trim_rcsum_slow(struct sk_buff
*skb
, unsigned int len
);
2959 * pskb_trim_rcsum - trim received skb and update checksum
2960 * @skb: buffer to trim
2963 * This is exactly the same as pskb_trim except that it ensures the
2964 * checksum of received packets are still valid after the operation.
2965 * It can change skb pointers.
2968 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
2970 if (likely(len
>= skb
->len
))
2972 return pskb_trim_rcsum_slow(skb
, len
);
2975 static inline int __skb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
2977 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2978 skb
->ip_summed
= CHECKSUM_NONE
;
2979 __skb_trim(skb
, len
);
2983 static inline int __skb_grow_rcsum(struct sk_buff
*skb
, unsigned int len
)
2985 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2986 skb
->ip_summed
= CHECKSUM_NONE
;
2987 return __skb_grow(skb
, len
);
2990 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
2992 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
2993 #define skb_rb_first(root) rb_to_skb(rb_first(root))
2994 #define skb_rb_last(root) rb_to_skb(rb_last(root))
2995 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
2996 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
2998 #define skb_queue_walk(queue, skb) \
2999 for (skb = (queue)->next; \
3000 skb != (struct sk_buff *)(queue); \
3003 #define skb_queue_walk_safe(queue, skb, tmp) \
3004 for (skb = (queue)->next, tmp = skb->next; \
3005 skb != (struct sk_buff *)(queue); \
3006 skb = tmp, tmp = skb->next)
3008 #define skb_queue_walk_from(queue, skb) \
3009 for (; skb != (struct sk_buff *)(queue); \
3012 #define skb_rbtree_walk(skb, root) \
3013 for (skb = skb_rb_first(root); skb != NULL; \
3014 skb = skb_rb_next(skb))
3016 #define skb_rbtree_walk_from(skb) \
3017 for (; skb != NULL; \
3018 skb = skb_rb_next(skb))
3020 #define skb_rbtree_walk_from_safe(skb, tmp) \
3021 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3024 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3025 for (tmp = skb->next; \
3026 skb != (struct sk_buff *)(queue); \
3027 skb = tmp, tmp = skb->next)
3029 #define skb_queue_reverse_walk(queue, skb) \
3030 for (skb = (queue)->prev; \
3031 skb != (struct sk_buff *)(queue); \
3034 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3035 for (skb = (queue)->prev, tmp = skb->prev; \
3036 skb != (struct sk_buff *)(queue); \
3037 skb = tmp, tmp = skb->prev)
3039 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3040 for (tmp = skb->prev; \
3041 skb != (struct sk_buff *)(queue); \
3042 skb = tmp, tmp = skb->prev)
3044 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
3046 return skb_shinfo(skb
)->frag_list
!= NULL
;
3049 static inline void skb_frag_list_init(struct sk_buff
*skb
)
3051 skb_shinfo(skb
)->frag_list
= NULL
;
3054 #define skb_walk_frags(skb, iter) \
3055 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3058 int __skb_wait_for_more_packets(struct sock
*sk
, int *err
, long *timeo_p
,
3059 const struct sk_buff
*skb
);
3060 struct sk_buff
*__skb_try_recv_datagram(struct sock
*sk
, unsigned flags
,
3061 int *peeked
, int *off
, int *err
,
3062 struct sk_buff
**last
);
3063 struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
3064 int *peeked
, int *off
, int *err
);
3065 struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
, int noblock
,
3067 unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
3068 struct poll_table_struct
*wait
);
3069 int skb_copy_datagram_iter(const struct sk_buff
*from
, int offset
,
3070 struct iov_iter
*to
, int size
);
3071 static inline int skb_copy_datagram_msg(const struct sk_buff
*from
, int offset
,
3072 struct msghdr
*msg
, int size
)
3074 return skb_copy_datagram_iter(from
, offset
, &msg
->msg_iter
, size
);
3076 int skb_copy_and_csum_datagram_msg(struct sk_buff
*skb
, int hlen
,
3077 struct msghdr
*msg
);
3078 int skb_copy_datagram_from_iter(struct sk_buff
*skb
, int offset
,
3079 struct iov_iter
*from
, int len
);
3080 int zerocopy_sg_from_iter(struct sk_buff
*skb
, struct iov_iter
*frm
);
3081 void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
3082 void __skb_free_datagram_locked(struct sock
*sk
, struct sk_buff
*skb
, int len
);
3083 static inline void skb_free_datagram_locked(struct sock
*sk
,
3084 struct sk_buff
*skb
)
3086 __skb_free_datagram_locked(sk
, skb
, 0);
3088 int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
, unsigned int flags
);
3089 int skb_copy_bits(const struct sk_buff
*skb
, int offset
, void *to
, int len
);
3090 int skb_store_bits(struct sk_buff
*skb
, int offset
, const void *from
, int len
);
3091 __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
, int offset
, u8
*to
,
3092 int len
, __wsum csum
);
3093 int skb_splice_bits(struct sk_buff
*skb
, struct sock
*sk
, unsigned int offset
,
3094 struct pipe_inode_info
*pipe
, unsigned int len
,
3095 unsigned int flags
);
3096 void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
3097 unsigned int skb_zerocopy_headlen(const struct sk_buff
*from
);
3098 int skb_zerocopy(struct sk_buff
*to
, struct sk_buff
*from
,
3100 void skb_split(struct sk_buff
*skb
, struct sk_buff
*skb1
, const u32 len
);
3101 int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
, int shiftlen
);
3102 void skb_scrub_packet(struct sk_buff
*skb
, bool xnet
);
3103 unsigned int skb_gso_transport_seglen(const struct sk_buff
*skb
);
3104 bool skb_gso_validate_mtu(const struct sk_buff
*skb
, unsigned int mtu
);
3105 bool skb_gso_validate_mac_len(const struct sk_buff
*skb
, unsigned int len
);
3106 struct sk_buff
*skb_segment(struct sk_buff
*skb
, netdev_features_t features
);
3107 struct sk_buff
*skb_vlan_untag(struct sk_buff
*skb
);
3108 int skb_ensure_writable(struct sk_buff
*skb
, int write_len
);
3109 int __skb_vlan_pop(struct sk_buff
*skb
, u16
*vlan_tci
);
3110 int skb_vlan_pop(struct sk_buff
*skb
);
3111 int skb_vlan_push(struct sk_buff
*skb
, __be16 vlan_proto
, u16 vlan_tci
);
3112 struct sk_buff
*pskb_extract(struct sk_buff
*skb
, int off
, int to_copy
,
3115 static inline int memcpy_from_msg(void *data
, struct msghdr
*msg
, int len
)
3117 return copy_from_iter(data
, len
, &msg
->msg_iter
) == len
? 0 : -EFAULT
;
3120 static inline int memcpy_to_msg(struct msghdr
*msg
, void *data
, int len
)
3122 return copy_to_iter(data
, len
, &msg
->msg_iter
) == len
? 0 : -EFAULT
;
3125 struct skb_checksum_ops
{
3126 __wsum (*update
)(const void *mem
, int len
, __wsum wsum
);
3127 __wsum (*combine
)(__wsum csum
, __wsum csum2
, int offset
, int len
);
3130 __wsum
__skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
3131 __wsum csum
, const struct skb_checksum_ops
*ops
);
3132 __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
3135 static inline void * __must_check
3136 __skb_header_pointer(const struct sk_buff
*skb
, int offset
,
3137 int len
, void *data
, int hlen
, void *buffer
)
3139 if (hlen
- offset
>= len
)
3140 return data
+ offset
;
3143 skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
3149 static inline void * __must_check
3150 skb_header_pointer(const struct sk_buff
*skb
, int offset
, int len
, void *buffer
)
3152 return __skb_header_pointer(skb
, offset
, len
, skb
->data
,
3153 skb_headlen(skb
), buffer
);
3157 * skb_needs_linearize - check if we need to linearize a given skb
3158 * depending on the given device features.
3159 * @skb: socket buffer to check
3160 * @features: net device features
3162 * Returns true if either:
3163 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3164 * 2. skb is fragmented and the device does not support SG.
3166 static inline bool skb_needs_linearize(struct sk_buff
*skb
,
3167 netdev_features_t features
)
3169 return skb_is_nonlinear(skb
) &&
3170 ((skb_has_frag_list(skb
) && !(features
& NETIF_F_FRAGLIST
)) ||
3171 (skb_shinfo(skb
)->nr_frags
&& !(features
& NETIF_F_SG
)));
3174 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
3176 const unsigned int len
)
3178 memcpy(to
, skb
->data
, len
);
3181 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
3182 const int offset
, void *to
,
3183 const unsigned int len
)
3185 memcpy(to
, skb
->data
+ offset
, len
);
3188 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
3190 const unsigned int len
)
3192 memcpy(skb
->data
, from
, len
);
3195 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
3198 const unsigned int len
)
3200 memcpy(skb
->data
+ offset
, from
, len
);
3203 void skb_init(void);
3205 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
3211 * skb_get_timestamp - get timestamp from a skb
3212 * @skb: skb to get stamp from
3213 * @stamp: pointer to struct timeval to store stamp in
3215 * Timestamps are stored in the skb as offsets to a base timestamp.
3216 * This function converts the offset back to a struct timeval and stores
3219 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
3220 struct timeval
*stamp
)
3222 *stamp
= ktime_to_timeval(skb
->tstamp
);
3225 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
3226 struct timespec
*stamp
)
3228 *stamp
= ktime_to_timespec(skb
->tstamp
);
3231 static inline void __net_timestamp(struct sk_buff
*skb
)
3233 skb
->tstamp
= ktime_get_real();
3236 static inline ktime_t
net_timedelta(ktime_t t
)
3238 return ktime_sub(ktime_get_real(), t
);
3241 static inline ktime_t
net_invalid_timestamp(void)
3243 return ktime_set(0, 0);
3246 struct sk_buff
*skb_clone_sk(struct sk_buff
*skb
);
3248 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3250 void skb_clone_tx_timestamp(struct sk_buff
*skb
);
3251 bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
3253 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3255 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
3259 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
3264 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3267 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3269 * PHY drivers may accept clones of transmitted packets for
3270 * timestamping via their phy_driver.txtstamp method. These drivers
3271 * must call this function to return the skb back to the stack with a
3274 * @skb: clone of the the original outgoing packet
3275 * @hwtstamps: hardware time stamps
3278 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
3279 struct skb_shared_hwtstamps
*hwtstamps
);
3281 void __skb_tstamp_tx(struct sk_buff
*orig_skb
,
3282 struct skb_shared_hwtstamps
*hwtstamps
,
3283 struct sock
*sk
, int tstype
);
3286 * skb_tstamp_tx - queue clone of skb with send time stamps
3287 * @orig_skb: the original outgoing packet
3288 * @hwtstamps: hardware time stamps, may be NULL if not available
3290 * If the skb has a socket associated, then this function clones the
3291 * skb (thus sharing the actual data and optional structures), stores
3292 * the optional hardware time stamping information (if non NULL) or
3293 * generates a software time stamp (otherwise), then queues the clone
3294 * to the error queue of the socket. Errors are silently ignored.
3296 void skb_tstamp_tx(struct sk_buff
*orig_skb
,
3297 struct skb_shared_hwtstamps
*hwtstamps
);
3299 static inline void sw_tx_timestamp(struct sk_buff
*skb
)
3301 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
&&
3302 !(skb_shinfo(skb
)->tx_flags
& SKBTX_IN_PROGRESS
))
3303 skb_tstamp_tx(skb
, NULL
);
3307 * skb_tx_timestamp() - Driver hook for transmit timestamping
3309 * Ethernet MAC Drivers should call this function in their hard_xmit()
3310 * function immediately before giving the sk_buff to the MAC hardware.
3312 * Specifically, one should make absolutely sure that this function is
3313 * called before TX completion of this packet can trigger. Otherwise
3314 * the packet could potentially already be freed.
3316 * @skb: A socket buffer.
3318 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
3320 skb_clone_tx_timestamp(skb
);
3321 sw_tx_timestamp(skb
);
3325 * skb_complete_wifi_ack - deliver skb with wifi status
3327 * @skb: the original outgoing packet
3328 * @acked: ack status
3331 void skb_complete_wifi_ack(struct sk_buff
*skb
, bool acked
);
3333 __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
3334 __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
3336 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
3338 return ((skb
->ip_summed
== CHECKSUM_UNNECESSARY
) ||
3340 (skb
->ip_summed
== CHECKSUM_PARTIAL
&&
3341 skb_checksum_start_offset(skb
) >= 0));
3345 * skb_checksum_complete - Calculate checksum of an entire packet
3346 * @skb: packet to process
3348 * This function calculates the checksum over the entire packet plus
3349 * the value of skb->csum. The latter can be used to supply the
3350 * checksum of a pseudo header as used by TCP/UDP. It returns the
3353 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3354 * this function can be used to verify that checksum on received
3355 * packets. In that case the function should return zero if the
3356 * checksum is correct. In particular, this function will return zero
3357 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3358 * hardware has already verified the correctness of the checksum.
3360 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
3362 return skb_csum_unnecessary(skb
) ?
3363 0 : __skb_checksum_complete(skb
);
3366 static inline void __skb_decr_checksum_unnecessary(struct sk_buff
*skb
)
3368 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
3369 if (skb
->csum_level
== 0)
3370 skb
->ip_summed
= CHECKSUM_NONE
;
3376 static inline void __skb_incr_checksum_unnecessary(struct sk_buff
*skb
)
3378 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
3379 if (skb
->csum_level
< SKB_MAX_CSUM_LEVEL
)
3381 } else if (skb
->ip_summed
== CHECKSUM_NONE
) {
3382 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
3383 skb
->csum_level
= 0;
3387 static inline void __skb_mark_checksum_bad(struct sk_buff
*skb
)
3389 /* Mark current checksum as bad (typically called from GRO
3390 * path). In the case that ip_summed is CHECKSUM_NONE
3391 * this must be the first checksum encountered in the packet.
3392 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3393 * checksum after the last one validated. For UDP, a zero
3394 * checksum can not be marked as bad.
3397 if (skb
->ip_summed
== CHECKSUM_NONE
||
3398 skb
->ip_summed
== CHECKSUM_UNNECESSARY
)
3402 /* Check if we need to perform checksum complete validation.
3404 * Returns true if checksum complete is needed, false otherwise
3405 * (either checksum is unnecessary or zero checksum is allowed).
3407 static inline bool __skb_checksum_validate_needed(struct sk_buff
*skb
,
3411 if (skb_csum_unnecessary(skb
) || (zero_okay
&& !check
)) {
3412 skb
->csum_valid
= 1;
3413 __skb_decr_checksum_unnecessary(skb
);
3420 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3423 #define CHECKSUM_BREAK 76
3425 /* Unset checksum-complete
3427 * Unset checksum complete can be done when packet is being modified
3428 * (uncompressed for instance) and checksum-complete value is
3431 static inline void skb_checksum_complete_unset(struct sk_buff
*skb
)
3433 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3434 skb
->ip_summed
= CHECKSUM_NONE
;
3437 /* Validate (init) checksum based on checksum complete.
3440 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3441 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3442 * checksum is stored in skb->csum for use in __skb_checksum_complete
3443 * non-zero: value of invalid checksum
3446 static inline __sum16
__skb_checksum_validate_complete(struct sk_buff
*skb
,
3450 if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
3451 if (!csum_fold(csum_add(psum
, skb
->csum
))) {
3452 skb
->csum_valid
= 1;
3455 } else if (skb
->csum_bad
) {
3456 /* ip_summed == CHECKSUM_NONE in this case */
3457 return (__force __sum16
)1;
3462 if (complete
|| skb
->len
<= CHECKSUM_BREAK
) {
3465 csum
= __skb_checksum_complete(skb
);
3466 skb
->csum_valid
= !csum
;
3473 static inline __wsum
null_compute_pseudo(struct sk_buff
*skb
, int proto
)
3478 /* Perform checksum validate (init). Note that this is a macro since we only
3479 * want to calculate the pseudo header which is an input function if necessary.
3480 * First we try to validate without any computation (checksum unnecessary) and
3481 * then calculate based on checksum complete calling the function to compute
3485 * 0: checksum is validated or try to in skb_checksum_complete
3486 * non-zero: value of invalid checksum
3488 #define __skb_checksum_validate(skb, proto, complete, \
3489 zero_okay, check, compute_pseudo) \
3491 __sum16 __ret = 0; \
3492 skb->csum_valid = 0; \
3493 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3494 __ret = __skb_checksum_validate_complete(skb, \
3495 complete, compute_pseudo(skb, proto)); \
3499 #define skb_checksum_init(skb, proto, compute_pseudo) \
3500 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3502 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3503 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3505 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3506 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3508 #define skb_checksum_validate_zero_check(skb, proto, check, \
3510 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3512 #define skb_checksum_simple_validate(skb) \
3513 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3515 static inline bool __skb_checksum_convert_check(struct sk_buff
*skb
)
3517 return (skb
->ip_summed
== CHECKSUM_NONE
&&
3518 skb
->csum_valid
&& !skb
->csum_bad
);
3521 static inline void __skb_checksum_convert(struct sk_buff
*skb
,
3522 __sum16 check
, __wsum pseudo
)
3524 skb
->csum
= ~pseudo
;
3525 skb
->ip_summed
= CHECKSUM_COMPLETE
;
3528 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3530 if (__skb_checksum_convert_check(skb)) \
3531 __skb_checksum_convert(skb, check, \
3532 compute_pseudo(skb, proto)); \
3535 static inline void skb_remcsum_adjust_partial(struct sk_buff
*skb
, void *ptr
,
3536 u16 start
, u16 offset
)
3538 skb
->ip_summed
= CHECKSUM_PARTIAL
;
3539 skb
->csum_start
= ((unsigned char *)ptr
+ start
) - skb
->head
;
3540 skb
->csum_offset
= offset
- start
;
3543 /* Update skbuf and packet to reflect the remote checksum offload operation.
3544 * When called, ptr indicates the starting point for skb->csum when
3545 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3546 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3548 static inline void skb_remcsum_process(struct sk_buff
*skb
, void *ptr
,
3549 int start
, int offset
, bool nopartial
)
3554 skb_remcsum_adjust_partial(skb
, ptr
, start
, offset
);
3558 if (unlikely(skb
->ip_summed
!= CHECKSUM_COMPLETE
)) {
3559 __skb_checksum_complete(skb
);
3560 skb_postpull_rcsum(skb
, skb
->data
, ptr
- (void *)skb
->data
);
3563 delta
= remcsum_adjust(ptr
, skb
->csum
, start
, offset
);
3565 /* Adjust skb->csum since we changed the packet */
3566 skb
->csum
= csum_add(skb
->csum
, delta
);
3569 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3570 void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
3571 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
3573 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
3574 nf_conntrack_destroy(nfct
);
3576 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
3579 atomic_inc(&nfct
->use
);
3582 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3583 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
3585 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
3588 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
3591 atomic_inc(&nf_bridge
->use
);
3593 #endif /* CONFIG_BRIDGE_NETFILTER */
3594 static inline void nf_reset(struct sk_buff
*skb
)
3596 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3597 nf_conntrack_put(skb
->nfct
);
3600 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3601 nf_bridge_put(skb
->nf_bridge
);
3602 skb
->nf_bridge
= NULL
;
3606 static inline void nf_reset_trace(struct sk_buff
*skb
)
3608 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3613 static inline void ipvs_reset(struct sk_buff
*skb
)
3615 #if IS_ENABLED(CONFIG_IP_VS)
3616 skb
->ipvs_property
= 0;
3620 /* Note: This doesn't put any conntrack and bridge info in dst. */
3621 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
,
3624 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3625 dst
->nfct
= src
->nfct
;
3626 nf_conntrack_get(src
->nfct
);
3628 dst
->nfctinfo
= src
->nfctinfo
;
3630 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3631 dst
->nf_bridge
= src
->nf_bridge
;
3632 nf_bridge_get(src
->nf_bridge
);
3634 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3636 dst
->nf_trace
= src
->nf_trace
;
3640 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
3642 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3643 nf_conntrack_put(dst
->nfct
);
3645 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3646 nf_bridge_put(dst
->nf_bridge
);
3648 __nf_copy(dst
, src
, true);
3651 #ifdef CONFIG_NETWORK_SECMARK
3652 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3654 to
->secmark
= from
->secmark
;
3657 static inline void skb_init_secmark(struct sk_buff
*skb
)
3662 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3665 static inline void skb_init_secmark(struct sk_buff
*skb
)
3669 static inline bool skb_irq_freeable(const struct sk_buff
*skb
)
3671 return !skb
->destructor
&&
3672 #if IS_ENABLED(CONFIG_XFRM)
3675 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3678 !skb
->_skb_refdst
&&
3679 !skb_has_frag_list(skb
);
3682 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
3684 skb
->queue_mapping
= queue_mapping
;
3687 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
3689 return skb
->queue_mapping
;
3692 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
3694 to
->queue_mapping
= from
->queue_mapping
;
3697 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
3699 skb
->queue_mapping
= rx_queue
+ 1;
3702 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
3704 return skb
->queue_mapping
- 1;
3707 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
3709 return skb
->queue_mapping
!= 0;
3712 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
3721 /* Keeps track of mac header offset relative to skb->head.
3722 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3723 * For non-tunnel skb it points to skb_mac_header() and for
3724 * tunnel skb it points to outer mac header.
3725 * Keeps track of level of encapsulation of network headers.
3736 #define SKB_SGO_CB_OFFSET 32
3737 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
3739 static inline int skb_tnl_header_len(const struct sk_buff
*inner_skb
)
3741 return (skb_mac_header(inner_skb
) - inner_skb
->head
) -
3742 SKB_GSO_CB(inner_skb
)->mac_offset
;
3745 static inline int gso_pskb_expand_head(struct sk_buff
*skb
, int extra
)
3747 int new_headroom
, headroom
;
3750 headroom
= skb_headroom(skb
);
3751 ret
= pskb_expand_head(skb
, extra
, 0, GFP_ATOMIC
);
3755 new_headroom
= skb_headroom(skb
);
3756 SKB_GSO_CB(skb
)->mac_offset
+= (new_headroom
- headroom
);
3760 static inline void gso_reset_checksum(struct sk_buff
*skb
, __wsum res
)
3762 /* Do not update partial checksums if remote checksum is enabled. */
3763 if (skb
->remcsum_offload
)
3766 SKB_GSO_CB(skb
)->csum
= res
;
3767 SKB_GSO_CB(skb
)->csum_start
= skb_checksum_start(skb
) - skb
->head
;
3770 /* Compute the checksum for a gso segment. First compute the checksum value
3771 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3772 * then add in skb->csum (checksum from csum_start to end of packet).
3773 * skb->csum and csum_start are then updated to reflect the checksum of the
3774 * resultant packet starting from the transport header-- the resultant checksum
3775 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3778 static inline __sum16
gso_make_checksum(struct sk_buff
*skb
, __wsum res
)
3780 unsigned char *csum_start
= skb_transport_header(skb
);
3781 int plen
= (skb
->head
+ SKB_GSO_CB(skb
)->csum_start
) - csum_start
;
3782 __wsum partial
= SKB_GSO_CB(skb
)->csum
;
3784 SKB_GSO_CB(skb
)->csum
= res
;
3785 SKB_GSO_CB(skb
)->csum_start
= csum_start
- skb
->head
;
3787 return csum_fold(csum_partial(csum_start
, plen
, partial
));
3790 static inline bool skb_is_gso(const struct sk_buff
*skb
)
3792 return skb_shinfo(skb
)->gso_size
;
3795 /* Note: Should be called only if skb_is_gso(skb) is true */
3796 static inline bool skb_is_gso_v6(const struct sk_buff
*skb
)
3798 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
3801 static inline void skb_gso_reset(struct sk_buff
*skb
)
3803 skb_shinfo(skb
)->gso_size
= 0;
3804 skb_shinfo(skb
)->gso_segs
= 0;
3805 skb_shinfo(skb
)->gso_type
= 0;
3808 void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
3810 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
3812 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3813 * wanted then gso_type will be set. */
3814 const struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
3816 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
3817 unlikely(shinfo
->gso_type
== 0)) {
3818 __skb_warn_lro_forwarding(skb
);
3824 static inline void skb_forward_csum(struct sk_buff
*skb
)
3826 /* Unfortunately we don't support this one. Any brave souls? */
3827 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3828 skb
->ip_summed
= CHECKSUM_NONE
;
3832 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3833 * @skb: skb to check
3835 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3836 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3837 * use this helper, to document places where we make this assertion.
3839 static inline void skb_checksum_none_assert(const struct sk_buff
*skb
)
3842 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
3846 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
3848 int skb_checksum_setup(struct sk_buff
*skb
, bool recalculate
);
3849 struct sk_buff
*skb_checksum_trimmed(struct sk_buff
*skb
,
3850 unsigned int transport_len
,
3851 __sum16(*skb_chkf
)(struct sk_buff
*skb
));
3854 * skb_head_is_locked - Determine if the skb->head is locked down
3855 * @skb: skb to check
3857 * The head on skbs build around a head frag can be removed if they are
3858 * not cloned. This function returns true if the skb head is locked down
3859 * due to either being allocated via kmalloc, or by being a clone with
3860 * multiple references to the head.
3862 static inline bool skb_head_is_locked(const struct sk_buff
*skb
)
3864 return !skb
->head_frag
|| skb_cloned(skb
);
3868 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3872 * skb_gso_network_seglen is used to determine the real size of the
3873 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3875 * The MAC/L2 header is not accounted for.
3877 static inline unsigned int skb_gso_network_seglen(const struct sk_buff
*skb
)
3879 unsigned int hdr_len
= skb_transport_header(skb
) -
3880 skb_network_header(skb
);
3881 return hdr_len
+ skb_gso_transport_seglen(skb
);
3885 * skb_gso_mac_seglen - Return length of individual segments of a gso packet
3889 * skb_gso_mac_seglen is used to determine the real size of the
3890 * individual segments, including MAC/L2, Layer3 (IP, IPv6) and L4
3891 * headers (TCP/UDP).
3893 static inline unsigned int skb_gso_mac_seglen(const struct sk_buff
*skb
)
3895 unsigned int hdr_len
= skb_transport_header(skb
) - skb_mac_header(skb
);
3896 return hdr_len
+ skb_gso_transport_seglen(skb
);
3899 /* Local Checksum Offload.
3900 * Compute outer checksum based on the assumption that the
3901 * inner checksum will be offloaded later.
3902 * See Documentation/networking/checksum-offloads.txt for
3903 * explanation of how this works.
3904 * Fill in outer checksum adjustment (e.g. with sum of outer
3905 * pseudo-header) before calling.
3906 * Also ensure that inner checksum is in linear data area.
3908 static inline __wsum
lco_csum(struct sk_buff
*skb
)
3910 unsigned char *csum_start
= skb_checksum_start(skb
);
3911 unsigned char *l4_hdr
= skb_transport_header(skb
);
3914 /* Start with complement of inner checksum adjustment */
3915 partial
= ~csum_unfold(*(__force __sum16
*)(csum_start
+
3918 /* Add in checksum of our headers (incl. outer checksum
3919 * adjustment filled in by caller) and return result.
3921 return csum_partial(l4_hdr
, csum_start
- l4_hdr
, partial
);
3924 #endif /* __KERNEL__ */
3925 #endif /* _LINUX_SKBUFF_H */