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_keys.h>
39 /* A. Checksumming of received packets by device.
43 * Device failed to checksum this packet e.g. due to lack of capabilities.
44 * The packet contains full (though not verified) checksum in packet but
45 * not in skb->csum. Thus, skb->csum is undefined in this case.
47 * CHECKSUM_UNNECESSARY:
49 * The hardware you're dealing with doesn't calculate the full checksum
50 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
51 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
52 * if their checksums are okay. skb->csum is still undefined in this case
53 * though. It is a bad option, but, unfortunately, nowadays most vendors do
54 * this. Apparently with the secret goal to sell you new devices, when you
55 * will add new protocol to your host, f.e. IPv6 8)
57 * CHECKSUM_UNNECESSARY is applicable to following protocols:
59 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
60 * zero UDP checksum for either IPv4 or IPv6, the networking stack
61 * may perform further validation in this case.
62 * GRE: only if the checksum is present in the header.
63 * SCTP: indicates the CRC in SCTP header has been validated.
65 * skb->csum_level indicates the number of consecutive checksums found in
66 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
67 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
68 * and a device is able to verify the checksums for UDP (possibly zero),
69 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
70 * two. If the device were only able to verify the UDP checksum and not
71 * GRE, either because it doesn't support GRE checksum of because GRE
72 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
73 * not considered in this case).
77 * This is the most generic way. The device supplied checksum of the _whole_
78 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
79 * hardware doesn't need to parse L3/L4 headers to implement this.
81 * Note: Even if device supports only some protocols, but is able to produce
82 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
86 * A checksum is set up to be offloaded to a device as described in the
87 * output description for CHECKSUM_PARTIAL. This may occur on a packet
88 * received directly from another Linux OS, e.g., a virtualized Linux kernel
89 * on the same host, or it may be set in the input path in GRO or remote
90 * checksum offload. For the purposes of checksum verification, the checksum
91 * referred to by skb->csum_start + skb->csum_offset and any preceding
92 * checksums in the packet are considered verified. Any checksums in the
93 * packet that are after the checksum being offloaded are not considered to
96 * B. Checksumming on output.
100 * The skb was already checksummed by the protocol, or a checksum is not
105 * The device is required to checksum the packet as seen by hard_start_xmit()
106 * from skb->csum_start up to the end, and to record/write the checksum at
107 * offset skb->csum_start + skb->csum_offset.
109 * The device must show its capabilities in dev->features, set up at device
110 * setup time, e.g. netdev_features.h:
112 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
113 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
114 * IPv4. Sigh. Vendors like this way for an unknown reason.
115 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
116 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
117 * NETIF_F_... - Well, you get the picture.
119 * CHECKSUM_UNNECESSARY:
121 * Normally, the device will do per protocol specific checksumming. Protocol
122 * implementations that do not want the NIC to perform the checksum
123 * calculation should use this flag in their outgoing skbs.
125 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
126 * offload. Correspondingly, the FCoE protocol driver
127 * stack should use CHECKSUM_UNNECESSARY.
129 * Any questions? No questions, good. --ANK
132 /* Don't change this without changing skb_csum_unnecessary! */
133 #define CHECKSUM_NONE 0
134 #define CHECKSUM_UNNECESSARY 1
135 #define CHECKSUM_COMPLETE 2
136 #define CHECKSUM_PARTIAL 3
138 /* Maximum value in skb->csum_level */
139 #define SKB_MAX_CSUM_LEVEL 3
141 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
142 #define SKB_WITH_OVERHEAD(X) \
143 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
144 #define SKB_MAX_ORDER(X, ORDER) \
145 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
146 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
147 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
149 /* return minimum truesize of one skb containing X bytes of data */
150 #define SKB_TRUESIZE(X) ((X) + \
151 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
152 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
156 struct pipe_inode_info
;
160 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
161 struct nf_conntrack
{
166 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
167 struct nf_bridge_info
{
170 struct net_device
*physindev
;
171 struct net_device
*physoutdev
;
172 unsigned long data
[32 / sizeof(unsigned long)];
176 struct sk_buff_head
{
177 /* These two members must be first. */
178 struct sk_buff
*next
;
179 struct sk_buff
*prev
;
187 /* To allow 64K frame to be packed as single skb without frag_list we
188 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
189 * buffers which do not start on a page boundary.
191 * Since GRO uses frags we allocate at least 16 regardless of page
194 #if (65536/PAGE_SIZE + 1) < 16
195 #define MAX_SKB_FRAGS 16UL
197 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
200 typedef struct skb_frag_struct skb_frag_t
;
202 struct skb_frag_struct
{
206 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
215 static inline unsigned int skb_frag_size(const skb_frag_t
*frag
)
220 static inline void skb_frag_size_set(skb_frag_t
*frag
, unsigned int size
)
225 static inline void skb_frag_size_add(skb_frag_t
*frag
, int delta
)
230 static inline void skb_frag_size_sub(skb_frag_t
*frag
, int delta
)
235 #define HAVE_HW_TIME_STAMP
238 * struct skb_shared_hwtstamps - hardware time stamps
239 * @hwtstamp: hardware time stamp transformed into duration
240 * since arbitrary point in time
242 * Software time stamps generated by ktime_get_real() are stored in
245 * hwtstamps can only be compared against other hwtstamps from
248 * This structure is attached to packets as part of the
249 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
251 struct skb_shared_hwtstamps
{
255 /* Definitions for tx_flags in struct skb_shared_info */
257 /* generate hardware time stamp */
258 SKBTX_HW_TSTAMP
= 1 << 0,
260 /* generate software time stamp when queueing packet to NIC */
261 SKBTX_SW_TSTAMP
= 1 << 1,
263 /* device driver is going to provide hardware time stamp */
264 SKBTX_IN_PROGRESS
= 1 << 2,
266 /* device driver supports TX zero-copy buffers */
267 SKBTX_DEV_ZEROCOPY
= 1 << 3,
269 /* generate wifi status information (where possible) */
270 SKBTX_WIFI_STATUS
= 1 << 4,
272 /* This indicates at least one fragment might be overwritten
273 * (as in vmsplice(), sendfile() ...)
274 * If we need to compute a TX checksum, we'll need to copy
275 * all frags to avoid possible bad checksum
277 SKBTX_SHARED_FRAG
= 1 << 5,
279 /* generate software time stamp when entering packet scheduling */
280 SKBTX_SCHED_TSTAMP
= 1 << 6,
282 /* generate software timestamp on peer data acknowledgment */
283 SKBTX_ACK_TSTAMP
= 1 << 7,
286 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
287 SKBTX_SCHED_TSTAMP | \
289 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
292 * The callback notifies userspace to release buffers when skb DMA is done in
293 * lower device, the skb last reference should be 0 when calling this.
294 * The zerocopy_success argument is true if zero copy transmit occurred,
295 * false on data copy or out of memory error caused by data copy attempt.
296 * The ctx field is used to track device context.
297 * The desc field is used to track userspace buffer index.
300 void (*callback
)(struct ubuf_info
*, bool zerocopy_success
);
305 /* This data is invariant across clones and lives at
306 * the end of the header data, ie. at skb->end.
308 struct skb_shared_info
{
309 unsigned char nr_frags
;
311 unsigned short gso_size
;
312 /* Warning: this field is not always filled in (UFO)! */
313 unsigned short gso_segs
;
314 unsigned short gso_type
;
315 struct sk_buff
*frag_list
;
316 struct skb_shared_hwtstamps hwtstamps
;
321 * Warning : all fields before dataref are cleared in __alloc_skb()
325 /* Intermediate layers must ensure that destructor_arg
326 * remains valid until skb destructor */
327 void * destructor_arg
;
329 /* must be last field, see pskb_expand_head() */
330 skb_frag_t frags
[MAX_SKB_FRAGS
];
333 /* We divide dataref into two halves. The higher 16 bits hold references
334 * to the payload part of skb->data. The lower 16 bits hold references to
335 * the entire skb->data. A clone of a headerless skb holds the length of
336 * the header in skb->hdr_len.
338 * All users must obey the rule that the skb->data reference count must be
339 * greater than or equal to the payload reference count.
341 * Holding a reference to the payload part means that the user does not
342 * care about modifications to the header part of skb->data.
344 #define SKB_DATAREF_SHIFT 16
345 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
349 SKB_FCLONE_UNAVAILABLE
, /* skb has no fclone (from head_cache) */
350 SKB_FCLONE_ORIG
, /* orig skb (from fclone_cache) */
351 SKB_FCLONE_CLONE
, /* companion fclone skb (from fclone_cache) */
355 SKB_GSO_TCPV4
= 1 << 0,
356 SKB_GSO_UDP
= 1 << 1,
358 /* This indicates the skb is from an untrusted source. */
359 SKB_GSO_DODGY
= 1 << 2,
361 /* This indicates the tcp segment has CWR set. */
362 SKB_GSO_TCP_ECN
= 1 << 3,
364 SKB_GSO_TCPV6
= 1 << 4,
366 SKB_GSO_FCOE
= 1 << 5,
368 SKB_GSO_GRE
= 1 << 6,
370 SKB_GSO_GRE_CSUM
= 1 << 7,
372 SKB_GSO_IPIP
= 1 << 8,
374 SKB_GSO_SIT
= 1 << 9,
376 SKB_GSO_UDP_TUNNEL
= 1 << 10,
378 SKB_GSO_UDP_TUNNEL_CSUM
= 1 << 11,
380 SKB_GSO_TUNNEL_REMCSUM
= 1 << 12,
383 #if BITS_PER_LONG > 32
384 #define NET_SKBUFF_DATA_USES_OFFSET 1
387 #ifdef NET_SKBUFF_DATA_USES_OFFSET
388 typedef unsigned int sk_buff_data_t
;
390 typedef unsigned char *sk_buff_data_t
;
394 * struct skb_mstamp - multi resolution time stamps
395 * @stamp_us: timestamp in us resolution
396 * @stamp_jiffies: timestamp in jiffies
409 * skb_mstamp_get - get current timestamp
410 * @cl: place to store timestamps
412 static inline void skb_mstamp_get(struct skb_mstamp
*cl
)
414 u64 val
= local_clock();
416 do_div(val
, NSEC_PER_USEC
);
417 cl
->stamp_us
= (u32
)val
;
418 cl
->stamp_jiffies
= (u32
)jiffies
;
422 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
423 * @t1: pointer to newest sample
424 * @t0: pointer to oldest sample
426 static inline u32
skb_mstamp_us_delta(const struct skb_mstamp
*t1
,
427 const struct skb_mstamp
*t0
)
429 s32 delta_us
= t1
->stamp_us
- t0
->stamp_us
;
430 u32 delta_jiffies
= t1
->stamp_jiffies
- t0
->stamp_jiffies
;
432 /* If delta_us is negative, this might be because interval is too big,
433 * or local_clock() drift is too big : fallback using jiffies.
436 delta_jiffies
>= (INT_MAX
/ (USEC_PER_SEC
/ HZ
)))
438 delta_us
= jiffies_to_usecs(delta_jiffies
);
445 * struct sk_buff - socket buffer
446 * @next: Next buffer in list
447 * @prev: Previous buffer in list
448 * @tstamp: Time we arrived/left
449 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
450 * @sk: Socket we are owned by
451 * @dev: Device we arrived on/are leaving by
452 * @cb: Control buffer. Free for use by every layer. Put private vars here
453 * @_skb_refdst: destination entry (with norefcount bit)
454 * @sp: the security path, used for xfrm
455 * @len: Length of actual data
456 * @data_len: Data length
457 * @mac_len: Length of link layer header
458 * @hdr_len: writable header length of cloned skb
459 * @csum: Checksum (must include start/offset pair)
460 * @csum_start: Offset from skb->head where checksumming should start
461 * @csum_offset: Offset from csum_start where checksum should be stored
462 * @priority: Packet queueing priority
463 * @ignore_df: allow local fragmentation
464 * @cloned: Head may be cloned (check refcnt to be sure)
465 * @ip_summed: Driver fed us an IP checksum
466 * @nohdr: Payload reference only, must not modify header
467 * @nfctinfo: Relationship of this skb to the connection
468 * @pkt_type: Packet class
469 * @fclone: skbuff clone status
470 * @ipvs_property: skbuff is owned by ipvs
471 * @peeked: this packet has been seen already, so stats have been
472 * done for it, don't do them again
473 * @nf_trace: netfilter packet trace flag
474 * @protocol: Packet protocol from driver
475 * @destructor: Destruct function
476 * @nfct: Associated connection, if any
477 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
478 * @skb_iif: ifindex of device we arrived on
479 * @tc_index: Traffic control index
480 * @tc_verd: traffic control verdict
481 * @hash: the packet hash
482 * @queue_mapping: Queue mapping for multiqueue devices
483 * @xmit_more: More SKBs are pending for this queue
484 * @ndisc_nodetype: router type (from link layer)
485 * @ooo_okay: allow the mapping of a socket to a queue to be changed
486 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
488 * @sw_hash: indicates hash was computed in software stack
489 * @wifi_acked_valid: wifi_acked was set
490 * @wifi_acked: whether frame was acked on wifi or not
491 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
492 * @napi_id: id of the NAPI struct this skb came from
493 * @secmark: security marking
494 * @mark: Generic packet mark
495 * @vlan_proto: vlan encapsulation protocol
496 * @vlan_tci: vlan tag control information
497 * @inner_protocol: Protocol (encapsulation)
498 * @inner_transport_header: Inner transport layer header (encapsulation)
499 * @inner_network_header: Network layer header (encapsulation)
500 * @inner_mac_header: Link layer header (encapsulation)
501 * @transport_header: Transport layer header
502 * @network_header: Network layer header
503 * @mac_header: Link layer header
504 * @tail: Tail pointer
506 * @head: Head of buffer
507 * @data: Data head pointer
508 * @truesize: Buffer size
509 * @users: User count - see {datagram,tcp}.c
515 /* These two members must be first. */
516 struct sk_buff
*next
;
517 struct sk_buff
*prev
;
521 struct skb_mstamp skb_mstamp
;
524 struct rb_node rbnode
; /* used in netem & tcp stack */
527 struct net_device
*dev
;
530 * This is the control buffer. It is free to use for every
531 * layer. Please put your private variables there. If you
532 * want to keep them across layers you have to do a skb_clone()
533 * first. This is owned by whoever has the skb queued ATM.
535 char cb
[48] __aligned(8);
537 unsigned long _skb_refdst
;
538 void (*destructor
)(struct sk_buff
*skb
);
542 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
543 struct nf_conntrack
*nfct
;
545 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
546 struct nf_bridge_info
*nf_bridge
;
553 /* Following fields are _not_ copied in __copy_skb_header()
554 * Note that queue_mapping is here mostly to fill a hole.
556 kmemcheck_bitfield_begin(flags1
);
565 kmemcheck_bitfield_end(flags1
);
567 /* fields enclosed in headers_start/headers_end are copied
568 * using a single memcpy() in __copy_skb_header()
571 __u32 headers_start
[0];
574 /* if you move pkt_type around you also must adapt those constants */
575 #ifdef __BIG_ENDIAN_BITFIELD
576 #define PKT_TYPE_MAX (7 << 5)
578 #define PKT_TYPE_MAX 7
580 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
582 __u8 __pkt_type_offset
[0];
593 __u8 wifi_acked_valid
:1;
597 /* Indicates the inner headers are valid in the skbuff. */
598 __u8 encapsulation
:1;
599 __u8 encap_hdr_csum
:1;
601 __u8 csum_complete_sw
:1;
605 #ifdef CONFIG_IPV6_NDISC_NODETYPE
606 __u8 ndisc_nodetype
:2;
608 __u8 ipvs_property
:1;
609 __u8 inner_protocol_type
:1;
610 __u8 remcsum_offload
:1;
611 /* 3 or 5 bit hole */
613 #ifdef CONFIG_NET_SCHED
614 __u16 tc_index
; /* traffic control index */
615 #ifdef CONFIG_NET_CLS_ACT
616 __u16 tc_verd
; /* traffic control verdict */
632 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
634 unsigned int napi_id
;
635 unsigned int sender_cpu
;
638 #ifdef CONFIG_NETWORK_SECMARK
643 __u32 reserved_tailroom
;
647 __be16 inner_protocol
;
651 __u16 inner_transport_header
;
652 __u16 inner_network_header
;
653 __u16 inner_mac_header
;
656 __u16 transport_header
;
657 __u16 network_header
;
661 __u32 headers_end
[0];
664 /* These elements must be at the end, see alloc_skb() for details. */
669 unsigned int truesize
;
675 * Handling routines are only of interest to the kernel
677 #include <linux/slab.h>
680 #define SKB_ALLOC_FCLONE 0x01
681 #define SKB_ALLOC_RX 0x02
682 #define SKB_ALLOC_NAPI 0x04
684 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
685 static inline bool skb_pfmemalloc(const struct sk_buff
*skb
)
687 return unlikely(skb
->pfmemalloc
);
691 * skb might have a dst pointer attached, refcounted or not.
692 * _skb_refdst low order bit is set if refcount was _not_ taken
694 #define SKB_DST_NOREF 1UL
695 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
698 * skb_dst - returns skb dst_entry
701 * Returns skb dst_entry, regardless of reference taken or not.
703 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
705 /* If refdst was not refcounted, check we still are in a
706 * rcu_read_lock section
708 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
709 !rcu_read_lock_held() &&
710 !rcu_read_lock_bh_held());
711 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
715 * skb_dst_set - sets skb dst
719 * Sets skb dst, assuming a reference was taken on dst and should
720 * be released by skb_dst_drop()
722 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
724 skb
->_skb_refdst
= (unsigned long)dst
;
728 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
732 * Sets skb dst, assuming a reference was not taken on dst.
733 * If dst entry is cached, we do not take reference and dst_release
734 * will be avoided by refdst_drop. If dst entry is not cached, we take
735 * reference, so that last dst_release can destroy the dst immediately.
737 static inline void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
)
739 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
740 skb
->_skb_refdst
= (unsigned long)dst
| SKB_DST_NOREF
;
744 * skb_dst_is_noref - Test if skb dst isn't refcounted
747 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
749 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
752 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
754 return (struct rtable
*)skb_dst(skb
);
757 void kfree_skb(struct sk_buff
*skb
);
758 void kfree_skb_list(struct sk_buff
*segs
);
759 void skb_tx_error(struct sk_buff
*skb
);
760 void consume_skb(struct sk_buff
*skb
);
761 void __kfree_skb(struct sk_buff
*skb
);
762 extern struct kmem_cache
*skbuff_head_cache
;
764 void kfree_skb_partial(struct sk_buff
*skb
, bool head_stolen
);
765 bool skb_try_coalesce(struct sk_buff
*to
, struct sk_buff
*from
,
766 bool *fragstolen
, int *delta_truesize
);
768 struct sk_buff
*__alloc_skb(unsigned int size
, gfp_t priority
, int flags
,
770 struct sk_buff
*build_skb(void *data
, unsigned int frag_size
);
771 static inline struct sk_buff
*alloc_skb(unsigned int size
,
774 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
777 struct sk_buff
*alloc_skb_with_frags(unsigned long header_len
,
778 unsigned long data_len
,
783 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
784 struct sk_buff_fclones
{
793 * skb_fclone_busy - check if fclone is busy
796 * Returns true is skb is a fast clone, and its clone is not freed.
797 * Some drivers call skb_orphan() in their ndo_start_xmit(),
798 * so we also check that this didnt happen.
800 static inline bool skb_fclone_busy(const struct sock
*sk
,
801 const struct sk_buff
*skb
)
803 const struct sk_buff_fclones
*fclones
;
805 fclones
= container_of(skb
, struct sk_buff_fclones
, skb1
);
807 return skb
->fclone
== SKB_FCLONE_ORIG
&&
808 atomic_read(&fclones
->fclone_ref
) > 1 &&
809 fclones
->skb2
.sk
== sk
;
812 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
815 return __alloc_skb(size
, priority
, SKB_ALLOC_FCLONE
, NUMA_NO_NODE
);
818 struct sk_buff
*__alloc_skb_head(gfp_t priority
, int node
);
819 static inline struct sk_buff
*alloc_skb_head(gfp_t priority
)
821 return __alloc_skb_head(priority
, -1);
824 struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
825 int skb_copy_ubufs(struct sk_buff
*skb
, gfp_t gfp_mask
);
826 struct sk_buff
*skb_clone(struct sk_buff
*skb
, gfp_t priority
);
827 struct sk_buff
*skb_copy(const struct sk_buff
*skb
, gfp_t priority
);
828 struct sk_buff
*__pskb_copy_fclone(struct sk_buff
*skb
, int headroom
,
829 gfp_t gfp_mask
, bool fclone
);
830 static inline struct sk_buff
*__pskb_copy(struct sk_buff
*skb
, int headroom
,
833 return __pskb_copy_fclone(skb
, headroom
, gfp_mask
, false);
836 int pskb_expand_head(struct sk_buff
*skb
, int nhead
, int ntail
, gfp_t gfp_mask
);
837 struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
838 unsigned int headroom
);
839 struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
, int newheadroom
,
840 int newtailroom
, gfp_t priority
);
841 int skb_to_sgvec_nomark(struct sk_buff
*skb
, struct scatterlist
*sg
,
842 int offset
, int len
);
843 int skb_to_sgvec(struct sk_buff
*skb
, struct scatterlist
*sg
, int offset
,
845 int skb_cow_data(struct sk_buff
*skb
, int tailbits
, struct sk_buff
**trailer
);
846 int skb_pad(struct sk_buff
*skb
, int pad
);
847 #define dev_kfree_skb(a) consume_skb(a)
849 int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
850 int getfrag(void *from
, char *to
, int offset
,
851 int len
, int odd
, struct sk_buff
*skb
),
852 void *from
, int length
);
854 struct skb_seq_state
{
858 __u32 stepped_offset
;
859 struct sk_buff
*root_skb
;
860 struct sk_buff
*cur_skb
;
864 void skb_prepare_seq_read(struct sk_buff
*skb
, unsigned int from
,
865 unsigned int to
, struct skb_seq_state
*st
);
866 unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
867 struct skb_seq_state
*st
);
868 void skb_abort_seq_read(struct skb_seq_state
*st
);
870 unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
871 unsigned int to
, struct ts_config
*config
);
874 * Packet hash types specify the type of hash in skb_set_hash.
876 * Hash types refer to the protocol layer addresses which are used to
877 * construct a packet's hash. The hashes are used to differentiate or identify
878 * flows of the protocol layer for the hash type. Hash types are either
879 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
881 * Properties of hashes:
883 * 1) Two packets in different flows have different hash values
884 * 2) Two packets in the same flow should have the same hash value
886 * A hash at a higher layer is considered to be more specific. A driver should
887 * set the most specific hash possible.
889 * A driver cannot indicate a more specific hash than the layer at which a hash
890 * was computed. For instance an L3 hash cannot be set as an L4 hash.
892 * A driver may indicate a hash level which is less specific than the
893 * actual layer the hash was computed on. For instance, a hash computed
894 * at L4 may be considered an L3 hash. This should only be done if the
895 * driver can't unambiguously determine that the HW computed the hash at
896 * the higher layer. Note that the "should" in the second property above
899 enum pkt_hash_types
{
900 PKT_HASH_TYPE_NONE
, /* Undefined type */
901 PKT_HASH_TYPE_L2
, /* Input: src_MAC, dest_MAC */
902 PKT_HASH_TYPE_L3
, /* Input: src_IP, dst_IP */
903 PKT_HASH_TYPE_L4
, /* Input: src_IP, dst_IP, src_port, dst_port */
907 skb_set_hash(struct sk_buff
*skb
, __u32 hash
, enum pkt_hash_types type
)
909 skb
->l4_hash
= (type
== PKT_HASH_TYPE_L4
);
914 void __skb_get_hash(struct sk_buff
*skb
);
915 static inline __u32
skb_get_hash(struct sk_buff
*skb
)
917 if (!skb
->l4_hash
&& !skb
->sw_hash
)
923 static inline __u32
skb_get_hash_raw(const struct sk_buff
*skb
)
928 static inline void skb_clear_hash(struct sk_buff
*skb
)
935 static inline void skb_clear_hash_if_not_l4(struct sk_buff
*skb
)
941 static inline void skb_copy_hash(struct sk_buff
*to
, const struct sk_buff
*from
)
943 to
->hash
= from
->hash
;
944 to
->sw_hash
= from
->sw_hash
;
945 to
->l4_hash
= from
->l4_hash
;
948 static inline void skb_sender_cpu_clear(struct sk_buff
*skb
)
955 #ifdef NET_SKBUFF_DATA_USES_OFFSET
956 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
958 return skb
->head
+ skb
->end
;
961 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
966 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
971 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
973 return skb
->end
- skb
->head
;
978 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
980 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
982 return &skb_shinfo(skb
)->hwtstamps
;
986 * skb_queue_empty - check if a queue is empty
989 * Returns true if the queue is empty, false otherwise.
991 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
993 return list
->next
== (const struct sk_buff
*) list
;
997 * skb_queue_is_last - check if skb is the last entry in the queue
1001 * Returns true if @skb is the last buffer on the list.
1003 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
1004 const struct sk_buff
*skb
)
1006 return skb
->next
== (const struct sk_buff
*) list
;
1010 * skb_queue_is_first - check if skb is the first entry in the queue
1014 * Returns true if @skb is the first buffer on the list.
1016 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
1017 const struct sk_buff
*skb
)
1019 return skb
->prev
== (const struct sk_buff
*) list
;
1023 * skb_queue_next - return the next packet in the queue
1025 * @skb: current buffer
1027 * Return the next packet in @list after @skb. It is only valid to
1028 * call this if skb_queue_is_last() evaluates to false.
1030 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
1031 const struct sk_buff
*skb
)
1033 /* This BUG_ON may seem severe, but if we just return then we
1034 * are going to dereference garbage.
1036 BUG_ON(skb_queue_is_last(list
, skb
));
1041 * skb_queue_prev - return the prev packet in the queue
1043 * @skb: current buffer
1045 * Return the prev packet in @list before @skb. It is only valid to
1046 * call this if skb_queue_is_first() evaluates to false.
1048 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
1049 const struct sk_buff
*skb
)
1051 /* This BUG_ON may seem severe, but if we just return then we
1052 * are going to dereference garbage.
1054 BUG_ON(skb_queue_is_first(list
, skb
));
1059 * skb_get - reference buffer
1060 * @skb: buffer to reference
1062 * Makes another reference to a socket buffer and returns a pointer
1065 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
1067 atomic_inc(&skb
->users
);
1072 * If users == 1, we are the only owner and are can avoid redundant
1077 * skb_cloned - is the buffer a clone
1078 * @skb: buffer to check
1080 * Returns true if the buffer was generated with skb_clone() and is
1081 * one of multiple shared copies of the buffer. Cloned buffers are
1082 * shared data so must not be written to under normal circumstances.
1084 static inline int skb_cloned(const struct sk_buff
*skb
)
1086 return skb
->cloned
&&
1087 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
1090 static inline int skb_unclone(struct sk_buff
*skb
, gfp_t pri
)
1092 might_sleep_if(pri
& __GFP_WAIT
);
1094 if (skb_cloned(skb
))
1095 return pskb_expand_head(skb
, 0, 0, pri
);
1101 * skb_header_cloned - is the header a clone
1102 * @skb: buffer to check
1104 * Returns true if modifying the header part of the buffer requires
1105 * the data to be copied.
1107 static inline int skb_header_cloned(const struct sk_buff
*skb
)
1114 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
1115 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
1116 return dataref
!= 1;
1120 * skb_header_release - release reference to header
1121 * @skb: buffer to operate on
1123 * Drop a reference to the header part of the buffer. This is done
1124 * by acquiring a payload reference. You must not read from the header
1125 * part of skb->data after this.
1126 * Note : Check if you can use __skb_header_release() instead.
1128 static inline void skb_header_release(struct sk_buff
*skb
)
1132 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
1136 * __skb_header_release - release reference to header
1137 * @skb: buffer to operate on
1139 * Variant of skb_header_release() assuming skb is private to caller.
1140 * We can avoid one atomic operation.
1142 static inline void __skb_header_release(struct sk_buff
*skb
)
1145 atomic_set(&skb_shinfo(skb
)->dataref
, 1 + (1 << SKB_DATAREF_SHIFT
));
1150 * skb_shared - is the buffer shared
1151 * @skb: buffer to check
1153 * Returns true if more than one person has a reference to this
1156 static inline int skb_shared(const struct sk_buff
*skb
)
1158 return atomic_read(&skb
->users
) != 1;
1162 * skb_share_check - check if buffer is shared and if so clone it
1163 * @skb: buffer to check
1164 * @pri: priority for memory allocation
1166 * If the buffer is shared the buffer is cloned and the old copy
1167 * drops a reference. A new clone with a single reference is returned.
1168 * If the buffer is not shared the original buffer is returned. When
1169 * being called from interrupt status or with spinlocks held pri must
1172 * NULL is returned on a memory allocation failure.
1174 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
, gfp_t pri
)
1176 might_sleep_if(pri
& __GFP_WAIT
);
1177 if (skb_shared(skb
)) {
1178 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
1190 * Copy shared buffers into a new sk_buff. We effectively do COW on
1191 * packets to handle cases where we have a local reader and forward
1192 * and a couple of other messy ones. The normal one is tcpdumping
1193 * a packet thats being forwarded.
1197 * skb_unshare - make a copy of a shared buffer
1198 * @skb: buffer to check
1199 * @pri: priority for memory allocation
1201 * If the socket buffer is a clone then this function creates a new
1202 * copy of the data, drops a reference count on the old copy and returns
1203 * the new copy with the reference count at 1. If the buffer is not a clone
1204 * the original buffer is returned. When called with a spinlock held or
1205 * from interrupt state @pri must be %GFP_ATOMIC
1207 * %NULL is returned on a memory allocation failure.
1209 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
1212 might_sleep_if(pri
& __GFP_WAIT
);
1213 if (skb_cloned(skb
)) {
1214 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
1216 /* Free our shared copy */
1227 * skb_peek - peek at the head of an &sk_buff_head
1228 * @list_: list to peek at
1230 * Peek an &sk_buff. Unlike most other operations you _MUST_
1231 * be careful with this one. A peek leaves the buffer on the
1232 * list and someone else may run off with it. You must hold
1233 * the appropriate locks or have a private queue to do this.
1235 * Returns %NULL for an empty list or a pointer to the head element.
1236 * The reference count is not incremented and the reference is therefore
1237 * volatile. Use with caution.
1239 static inline struct sk_buff
*skb_peek(const struct sk_buff_head
*list_
)
1241 struct sk_buff
*skb
= list_
->next
;
1243 if (skb
== (struct sk_buff
*)list_
)
1249 * skb_peek_next - peek skb following the given one from a queue
1250 * @skb: skb to start from
1251 * @list_: list to peek at
1253 * Returns %NULL when the end of the list is met or a pointer to the
1254 * next element. The reference count is not incremented and the
1255 * reference is therefore volatile. Use with caution.
1257 static inline struct sk_buff
*skb_peek_next(struct sk_buff
*skb
,
1258 const struct sk_buff_head
*list_
)
1260 struct sk_buff
*next
= skb
->next
;
1262 if (next
== (struct sk_buff
*)list_
)
1268 * skb_peek_tail - peek at the tail of an &sk_buff_head
1269 * @list_: list to peek at
1271 * Peek an &sk_buff. Unlike most other operations you _MUST_
1272 * be careful with this one. A peek leaves the buffer on the
1273 * list and someone else may run off with it. You must hold
1274 * the appropriate locks or have a private queue to do this.
1276 * Returns %NULL for an empty list or a pointer to the tail element.
1277 * The reference count is not incremented and the reference is therefore
1278 * volatile. Use with caution.
1280 static inline struct sk_buff
*skb_peek_tail(const struct sk_buff_head
*list_
)
1282 struct sk_buff
*skb
= list_
->prev
;
1284 if (skb
== (struct sk_buff
*)list_
)
1291 * skb_queue_len - get queue length
1292 * @list_: list to measure
1294 * Return the length of an &sk_buff queue.
1296 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
1302 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1303 * @list: queue to initialize
1305 * This initializes only the list and queue length aspects of
1306 * an sk_buff_head object. This allows to initialize the list
1307 * aspects of an sk_buff_head without reinitializing things like
1308 * the spinlock. It can also be used for on-stack sk_buff_head
1309 * objects where the spinlock is known to not be used.
1311 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
1313 list
->prev
= list
->next
= (struct sk_buff
*)list
;
1318 * This function creates a split out lock class for each invocation;
1319 * this is needed for now since a whole lot of users of the skb-queue
1320 * infrastructure in drivers have different locking usage (in hardirq)
1321 * than the networking core (in softirq only). In the long run either the
1322 * network layer or drivers should need annotation to consolidate the
1323 * main types of usage into 3 classes.
1325 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
1327 spin_lock_init(&list
->lock
);
1328 __skb_queue_head_init(list
);
1331 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
1332 struct lock_class_key
*class)
1334 skb_queue_head_init(list
);
1335 lockdep_set_class(&list
->lock
, class);
1339 * Insert an sk_buff on a list.
1341 * The "__skb_xxxx()" functions are the non-atomic ones that
1342 * can only be called with interrupts disabled.
1344 void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
,
1345 struct sk_buff_head
*list
);
1346 static inline void __skb_insert(struct sk_buff
*newsk
,
1347 struct sk_buff
*prev
, struct sk_buff
*next
,
1348 struct sk_buff_head
*list
)
1352 next
->prev
= prev
->next
= newsk
;
1356 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
1357 struct sk_buff
*prev
,
1358 struct sk_buff
*next
)
1360 struct sk_buff
*first
= list
->next
;
1361 struct sk_buff
*last
= list
->prev
;
1371 * skb_queue_splice - join two skb lists, this is designed for stacks
1372 * @list: the new list to add
1373 * @head: the place to add it in the first list
1375 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
1376 struct sk_buff_head
*head
)
1378 if (!skb_queue_empty(list
)) {
1379 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1380 head
->qlen
+= list
->qlen
;
1385 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1386 * @list: the new list to add
1387 * @head: the place to add it in the first list
1389 * The list at @list is reinitialised
1391 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
1392 struct sk_buff_head
*head
)
1394 if (!skb_queue_empty(list
)) {
1395 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1396 head
->qlen
+= list
->qlen
;
1397 __skb_queue_head_init(list
);
1402 * skb_queue_splice_tail - join two skb lists, each list being a queue
1403 * @list: the new list to add
1404 * @head: the place to add it in the first list
1406 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
1407 struct sk_buff_head
*head
)
1409 if (!skb_queue_empty(list
)) {
1410 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1411 head
->qlen
+= list
->qlen
;
1416 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1417 * @list: the new list to add
1418 * @head: the place to add it in the first list
1420 * Each of the lists is a queue.
1421 * The list at @list is reinitialised
1423 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
1424 struct sk_buff_head
*head
)
1426 if (!skb_queue_empty(list
)) {
1427 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1428 head
->qlen
+= list
->qlen
;
1429 __skb_queue_head_init(list
);
1434 * __skb_queue_after - queue a buffer at the list head
1435 * @list: list to use
1436 * @prev: place after this buffer
1437 * @newsk: buffer to queue
1439 * Queue a buffer int the middle of a list. This function takes no locks
1440 * and you must therefore hold required locks before calling it.
1442 * A buffer cannot be placed on two lists at the same time.
1444 static inline void __skb_queue_after(struct sk_buff_head
*list
,
1445 struct sk_buff
*prev
,
1446 struct sk_buff
*newsk
)
1448 __skb_insert(newsk
, prev
, prev
->next
, list
);
1451 void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
1452 struct sk_buff_head
*list
);
1454 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1455 struct sk_buff
*next
,
1456 struct sk_buff
*newsk
)
1458 __skb_insert(newsk
, next
->prev
, next
, list
);
1462 * __skb_queue_head - queue a buffer at the list head
1463 * @list: list to use
1464 * @newsk: buffer to queue
1466 * Queue a buffer at the start of a list. This function takes no locks
1467 * and you must therefore hold required locks before calling it.
1469 * A buffer cannot be placed on two lists at the same time.
1471 void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1472 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1473 struct sk_buff
*newsk
)
1475 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1479 * __skb_queue_tail - queue a buffer at the list tail
1480 * @list: list to use
1481 * @newsk: buffer to queue
1483 * Queue a buffer at the end of a list. This function takes no locks
1484 * and you must therefore hold required locks before calling it.
1486 * A buffer cannot be placed on two lists at the same time.
1488 void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1489 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1490 struct sk_buff
*newsk
)
1492 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1496 * remove sk_buff from list. _Must_ be called atomically, and with
1499 void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1500 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1502 struct sk_buff
*next
, *prev
;
1507 skb
->next
= skb
->prev
= NULL
;
1513 * __skb_dequeue - remove from the head of the queue
1514 * @list: list to dequeue from
1516 * Remove the head of the list. This function does not take any locks
1517 * so must be used with appropriate locks held only. The head item is
1518 * returned or %NULL if the list is empty.
1520 struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1521 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1523 struct sk_buff
*skb
= skb_peek(list
);
1525 __skb_unlink(skb
, list
);
1530 * __skb_dequeue_tail - remove from the tail of the queue
1531 * @list: list to dequeue from
1533 * Remove the tail of the list. This function does not take any locks
1534 * so must be used with appropriate locks held only. The tail item is
1535 * returned or %NULL if the list is empty.
1537 struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1538 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1540 struct sk_buff
*skb
= skb_peek_tail(list
);
1542 __skb_unlink(skb
, list
);
1547 static inline bool skb_is_nonlinear(const struct sk_buff
*skb
)
1549 return skb
->data_len
;
1552 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1554 return skb
->len
- skb
->data_len
;
1557 static inline int skb_pagelen(const struct sk_buff
*skb
)
1561 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1562 len
+= skb_frag_size(&skb_shinfo(skb
)->frags
[i
]);
1563 return len
+ skb_headlen(skb
);
1567 * __skb_fill_page_desc - initialise a paged fragment in an skb
1568 * @skb: buffer containing fragment to be initialised
1569 * @i: paged fragment index to initialise
1570 * @page: the page to use for this fragment
1571 * @off: the offset to the data with @page
1572 * @size: the length of the data
1574 * Initialises the @i'th fragment of @skb to point to &size bytes at
1575 * offset @off within @page.
1577 * Does not take any additional reference on the fragment.
1579 static inline void __skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1580 struct page
*page
, int off
, int size
)
1582 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1585 * Propagate page->pfmemalloc to the skb if we can. The problem is
1586 * that not all callers have unique ownership of the page. If
1587 * pfmemalloc is set, we check the mapping as a mapping implies
1588 * page->index is set (index and pfmemalloc share space).
1589 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1590 * do not lose pfmemalloc information as the pages would not be
1591 * allocated using __GFP_MEMALLOC.
1593 frag
->page
.p
= page
;
1594 frag
->page_offset
= off
;
1595 skb_frag_size_set(frag
, size
);
1597 page
= compound_head(page
);
1598 if (page
->pfmemalloc
&& !page
->mapping
)
1599 skb
->pfmemalloc
= true;
1603 * skb_fill_page_desc - initialise a paged fragment in an skb
1604 * @skb: buffer containing fragment to be initialised
1605 * @i: paged fragment index to initialise
1606 * @page: the page to use for this fragment
1607 * @off: the offset to the data with @page
1608 * @size: the length of the data
1610 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1611 * @skb to point to @size bytes at offset @off within @page. In
1612 * addition updates @skb such that @i is the last fragment.
1614 * Does not take any additional reference on the fragment.
1616 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1617 struct page
*page
, int off
, int size
)
1619 __skb_fill_page_desc(skb
, i
, page
, off
, size
);
1620 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1623 void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
, int off
,
1624 int size
, unsigned int truesize
);
1626 void skb_coalesce_rx_frag(struct sk_buff
*skb
, int i
, int size
,
1627 unsigned int truesize
);
1629 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1630 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1631 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1633 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1634 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1636 return skb
->head
+ skb
->tail
;
1639 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1641 skb
->tail
= skb
->data
- skb
->head
;
1644 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1646 skb_reset_tail_pointer(skb
);
1647 skb
->tail
+= offset
;
1650 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1651 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1656 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1658 skb
->tail
= skb
->data
;
1661 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1663 skb
->tail
= skb
->data
+ offset
;
1666 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1669 * Add data to an sk_buff
1671 unsigned char *pskb_put(struct sk_buff
*skb
, struct sk_buff
*tail
, int len
);
1672 unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1673 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1675 unsigned char *tmp
= skb_tail_pointer(skb
);
1676 SKB_LINEAR_ASSERT(skb
);
1682 unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1683 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1690 unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1691 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1694 BUG_ON(skb
->len
< skb
->data_len
);
1695 return skb
->data
+= len
;
1698 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1700 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1703 unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1705 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1707 if (len
> skb_headlen(skb
) &&
1708 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1711 return skb
->data
+= len
;
1714 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1716 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1719 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1721 if (likely(len
<= skb_headlen(skb
)))
1723 if (unlikely(len
> skb
->len
))
1725 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1729 * skb_headroom - bytes at buffer head
1730 * @skb: buffer to check
1732 * Return the number of bytes of free space at the head of an &sk_buff.
1734 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1736 return skb
->data
- skb
->head
;
1740 * skb_tailroom - bytes at buffer end
1741 * @skb: buffer to check
1743 * Return the number of bytes of free space at the tail of an sk_buff
1745 static inline int skb_tailroom(const struct sk_buff
*skb
)
1747 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1751 * skb_availroom - bytes at buffer end
1752 * @skb: buffer to check
1754 * Return the number of bytes of free space at the tail of an sk_buff
1755 * allocated by sk_stream_alloc()
1757 static inline int skb_availroom(const struct sk_buff
*skb
)
1759 if (skb_is_nonlinear(skb
))
1762 return skb
->end
- skb
->tail
- skb
->reserved_tailroom
;
1766 * skb_reserve - adjust headroom
1767 * @skb: buffer to alter
1768 * @len: bytes to move
1770 * Increase the headroom of an empty &sk_buff by reducing the tail
1771 * room. This is only allowed for an empty buffer.
1773 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
1779 #define ENCAP_TYPE_ETHER 0
1780 #define ENCAP_TYPE_IPPROTO 1
1782 static inline void skb_set_inner_protocol(struct sk_buff
*skb
,
1785 skb
->inner_protocol
= protocol
;
1786 skb
->inner_protocol_type
= ENCAP_TYPE_ETHER
;
1789 static inline void skb_set_inner_ipproto(struct sk_buff
*skb
,
1792 skb
->inner_ipproto
= ipproto
;
1793 skb
->inner_protocol_type
= ENCAP_TYPE_IPPROTO
;
1796 static inline void skb_reset_inner_headers(struct sk_buff
*skb
)
1798 skb
->inner_mac_header
= skb
->mac_header
;
1799 skb
->inner_network_header
= skb
->network_header
;
1800 skb
->inner_transport_header
= skb
->transport_header
;
1803 static inline void skb_reset_mac_len(struct sk_buff
*skb
)
1805 skb
->mac_len
= skb
->network_header
- skb
->mac_header
;
1808 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1811 return skb
->head
+ skb
->inner_transport_header
;
1814 static inline void skb_reset_inner_transport_header(struct sk_buff
*skb
)
1816 skb
->inner_transport_header
= skb
->data
- skb
->head
;
1819 static inline void skb_set_inner_transport_header(struct sk_buff
*skb
,
1822 skb_reset_inner_transport_header(skb
);
1823 skb
->inner_transport_header
+= offset
;
1826 static inline unsigned char *skb_inner_network_header(const struct sk_buff
*skb
)
1828 return skb
->head
+ skb
->inner_network_header
;
1831 static inline void skb_reset_inner_network_header(struct sk_buff
*skb
)
1833 skb
->inner_network_header
= skb
->data
- skb
->head
;
1836 static inline void skb_set_inner_network_header(struct sk_buff
*skb
,
1839 skb_reset_inner_network_header(skb
);
1840 skb
->inner_network_header
+= offset
;
1843 static inline unsigned char *skb_inner_mac_header(const struct sk_buff
*skb
)
1845 return skb
->head
+ skb
->inner_mac_header
;
1848 static inline void skb_reset_inner_mac_header(struct sk_buff
*skb
)
1850 skb
->inner_mac_header
= skb
->data
- skb
->head
;
1853 static inline void skb_set_inner_mac_header(struct sk_buff
*skb
,
1856 skb_reset_inner_mac_header(skb
);
1857 skb
->inner_mac_header
+= offset
;
1859 static inline bool skb_transport_header_was_set(const struct sk_buff
*skb
)
1861 return skb
->transport_header
!= (typeof(skb
->transport_header
))~0U;
1864 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1866 return skb
->head
+ skb
->transport_header
;
1869 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1871 skb
->transport_header
= skb
->data
- skb
->head
;
1874 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1877 skb_reset_transport_header(skb
);
1878 skb
->transport_header
+= offset
;
1881 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1883 return skb
->head
+ skb
->network_header
;
1886 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1888 skb
->network_header
= skb
->data
- skb
->head
;
1891 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1893 skb_reset_network_header(skb
);
1894 skb
->network_header
+= offset
;
1897 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1899 return skb
->head
+ skb
->mac_header
;
1902 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1904 return skb
->mac_header
!= (typeof(skb
->mac_header
))~0U;
1907 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1909 skb
->mac_header
= skb
->data
- skb
->head
;
1912 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1914 skb_reset_mac_header(skb
);
1915 skb
->mac_header
+= offset
;
1918 static inline void skb_pop_mac_header(struct sk_buff
*skb
)
1920 skb
->mac_header
= skb
->network_header
;
1923 static inline void skb_probe_transport_header(struct sk_buff
*skb
,
1924 const int offset_hint
)
1926 struct flow_keys keys
;
1928 if (skb_transport_header_was_set(skb
))
1930 else if (skb_flow_dissect(skb
, &keys
))
1931 skb_set_transport_header(skb
, keys
.thoff
);
1933 skb_set_transport_header(skb
, offset_hint
);
1936 static inline void skb_mac_header_rebuild(struct sk_buff
*skb
)
1938 if (skb_mac_header_was_set(skb
)) {
1939 const unsigned char *old_mac
= skb_mac_header(skb
);
1941 skb_set_mac_header(skb
, -skb
->mac_len
);
1942 memmove(skb_mac_header(skb
), old_mac
, skb
->mac_len
);
1946 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
1948 return skb
->csum_start
- skb_headroom(skb
);
1951 static inline int skb_transport_offset(const struct sk_buff
*skb
)
1953 return skb_transport_header(skb
) - skb
->data
;
1956 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
1958 return skb
->transport_header
- skb
->network_header
;
1961 static inline u32
skb_inner_network_header_len(const struct sk_buff
*skb
)
1963 return skb
->inner_transport_header
- skb
->inner_network_header
;
1966 static inline int skb_network_offset(const struct sk_buff
*skb
)
1968 return skb_network_header(skb
) - skb
->data
;
1971 static inline int skb_inner_network_offset(const struct sk_buff
*skb
)
1973 return skb_inner_network_header(skb
) - skb
->data
;
1976 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
1978 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
1982 * CPUs often take a performance hit when accessing unaligned memory
1983 * locations. The actual performance hit varies, it can be small if the
1984 * hardware handles it or large if we have to take an exception and fix it
1987 * Since an ethernet header is 14 bytes network drivers often end up with
1988 * the IP header at an unaligned offset. The IP header can be aligned by
1989 * shifting the start of the packet by 2 bytes. Drivers should do this
1992 * skb_reserve(skb, NET_IP_ALIGN);
1994 * The downside to this alignment of the IP header is that the DMA is now
1995 * unaligned. On some architectures the cost of an unaligned DMA is high
1996 * and this cost outweighs the gains made by aligning the IP header.
1998 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2001 #ifndef NET_IP_ALIGN
2002 #define NET_IP_ALIGN 2
2006 * The networking layer reserves some headroom in skb data (via
2007 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2008 * the header has to grow. In the default case, if the header has to grow
2009 * 32 bytes or less we avoid the reallocation.
2011 * Unfortunately this headroom changes the DMA alignment of the resulting
2012 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2013 * on some architectures. An architecture can override this value,
2014 * perhaps setting it to a cacheline in size (since that will maintain
2015 * cacheline alignment of the DMA). It must be a power of 2.
2017 * Various parts of the networking layer expect at least 32 bytes of
2018 * headroom, you should not reduce this.
2020 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2021 * to reduce average number of cache lines per packet.
2022 * get_rps_cpus() for example only access one 64 bytes aligned block :
2023 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2026 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2029 int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
2031 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
2033 if (unlikely(skb_is_nonlinear(skb
))) {
2038 skb_set_tail_pointer(skb
, len
);
2041 void skb_trim(struct sk_buff
*skb
, unsigned int len
);
2043 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2046 return ___pskb_trim(skb
, len
);
2047 __skb_trim(skb
, len
);
2051 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2053 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
2057 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2058 * @skb: buffer to alter
2061 * This is identical to pskb_trim except that the caller knows that
2062 * the skb is not cloned so we should never get an error due to out-
2065 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
2067 int err
= pskb_trim(skb
, len
);
2072 * skb_orphan - orphan a buffer
2073 * @skb: buffer to orphan
2075 * If a buffer currently has an owner then we call the owner's
2076 * destructor function and make the @skb unowned. The buffer continues
2077 * to exist but is no longer charged to its former owner.
2079 static inline void skb_orphan(struct sk_buff
*skb
)
2081 if (skb
->destructor
) {
2082 skb
->destructor(skb
);
2083 skb
->destructor
= NULL
;
2091 * skb_orphan_frags - orphan the frags contained in a buffer
2092 * @skb: buffer to orphan frags from
2093 * @gfp_mask: allocation mask for replacement pages
2095 * For each frag in the SKB which needs a destructor (i.e. has an
2096 * owner) create a copy of that frag and release the original
2097 * page by calling the destructor.
2099 static inline int skb_orphan_frags(struct sk_buff
*skb
, gfp_t gfp_mask
)
2101 if (likely(!(skb_shinfo(skb
)->tx_flags
& SKBTX_DEV_ZEROCOPY
)))
2103 return skb_copy_ubufs(skb
, gfp_mask
);
2107 * __skb_queue_purge - empty a list
2108 * @list: list to empty
2110 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2111 * the list and one reference dropped. This function does not take the
2112 * list lock and the caller must hold the relevant locks to use it.
2114 void skb_queue_purge(struct sk_buff_head
*list
);
2115 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
2117 struct sk_buff
*skb
;
2118 while ((skb
= __skb_dequeue(list
)) != NULL
)
2122 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2123 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2124 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2126 void *netdev_alloc_frag(unsigned int fragsz
);
2128 struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
, unsigned int length
,
2132 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2133 * @dev: network device to receive on
2134 * @length: length to allocate
2136 * Allocate a new &sk_buff and assign it a usage count of one. The
2137 * buffer has unspecified headroom built in. Users should allocate
2138 * the headroom they think they need without accounting for the
2139 * built in space. The built in space is used for optimisations.
2141 * %NULL is returned if there is no free memory. Although this function
2142 * allocates memory it can be called from an interrupt.
2144 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
2145 unsigned int length
)
2147 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
2150 /* legacy helper around __netdev_alloc_skb() */
2151 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
2154 return __netdev_alloc_skb(NULL
, length
, gfp_mask
);
2157 /* legacy helper around netdev_alloc_skb() */
2158 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
2160 return netdev_alloc_skb(NULL
, length
);
2164 static inline struct sk_buff
*__netdev_alloc_skb_ip_align(struct net_device
*dev
,
2165 unsigned int length
, gfp_t gfp
)
2167 struct sk_buff
*skb
= __netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
, gfp
);
2169 if (NET_IP_ALIGN
&& skb
)
2170 skb_reserve(skb
, NET_IP_ALIGN
);
2174 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
2175 unsigned int length
)
2177 return __netdev_alloc_skb_ip_align(dev
, length
, GFP_ATOMIC
);
2180 void *napi_alloc_frag(unsigned int fragsz
);
2181 struct sk_buff
*__napi_alloc_skb(struct napi_struct
*napi
,
2182 unsigned int length
, gfp_t gfp_mask
);
2183 static inline struct sk_buff
*napi_alloc_skb(struct napi_struct
*napi
,
2184 unsigned int length
)
2186 return __napi_alloc_skb(napi
, length
, GFP_ATOMIC
);
2190 * __dev_alloc_pages - allocate page for network Rx
2191 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2192 * @order: size of the allocation
2194 * Allocate a new page.
2196 * %NULL is returned if there is no free memory.
2198 static inline struct page
*__dev_alloc_pages(gfp_t gfp_mask
,
2201 /* This piece of code contains several assumptions.
2202 * 1. This is for device Rx, therefor a cold page is preferred.
2203 * 2. The expectation is the user wants a compound page.
2204 * 3. If requesting a order 0 page it will not be compound
2205 * due to the check to see if order has a value in prep_new_page
2206 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2207 * code in gfp_to_alloc_flags that should be enforcing this.
2209 gfp_mask
|= __GFP_COLD
| __GFP_COMP
| __GFP_MEMALLOC
;
2211 return alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, order
);
2214 static inline struct page
*dev_alloc_pages(unsigned int order
)
2216 return __dev_alloc_pages(GFP_ATOMIC
, order
);
2220 * __dev_alloc_page - allocate a page for network Rx
2221 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2223 * Allocate a new page.
2225 * %NULL is returned if there is no free memory.
2227 static inline struct page
*__dev_alloc_page(gfp_t gfp_mask
)
2229 return __dev_alloc_pages(gfp_mask
, 0);
2232 static inline struct page
*dev_alloc_page(void)
2234 return __dev_alloc_page(GFP_ATOMIC
);
2238 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2239 * @page: The page that was allocated from skb_alloc_page
2240 * @skb: The skb that may need pfmemalloc set
2242 static inline void skb_propagate_pfmemalloc(struct page
*page
,
2243 struct sk_buff
*skb
)
2245 if (page
&& page
->pfmemalloc
)
2246 skb
->pfmemalloc
= true;
2250 * skb_frag_page - retrieve the page referred to by a paged fragment
2251 * @frag: the paged fragment
2253 * Returns the &struct page associated with @frag.
2255 static inline struct page
*skb_frag_page(const skb_frag_t
*frag
)
2257 return frag
->page
.p
;
2261 * __skb_frag_ref - take an addition reference on a paged fragment.
2262 * @frag: the paged fragment
2264 * Takes an additional reference on the paged fragment @frag.
2266 static inline void __skb_frag_ref(skb_frag_t
*frag
)
2268 get_page(skb_frag_page(frag
));
2272 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2274 * @f: the fragment offset.
2276 * Takes an additional reference on the @f'th paged fragment of @skb.
2278 static inline void skb_frag_ref(struct sk_buff
*skb
, int f
)
2280 __skb_frag_ref(&skb_shinfo(skb
)->frags
[f
]);
2284 * __skb_frag_unref - release a reference on a paged fragment.
2285 * @frag: the paged fragment
2287 * Releases a reference on the paged fragment @frag.
2289 static inline void __skb_frag_unref(skb_frag_t
*frag
)
2291 put_page(skb_frag_page(frag
));
2295 * skb_frag_unref - release a reference on a paged fragment of an skb.
2297 * @f: the fragment offset
2299 * Releases a reference on the @f'th paged fragment of @skb.
2301 static inline void skb_frag_unref(struct sk_buff
*skb
, int f
)
2303 __skb_frag_unref(&skb_shinfo(skb
)->frags
[f
]);
2307 * skb_frag_address - gets the address of the data contained in a paged fragment
2308 * @frag: the paged fragment buffer
2310 * Returns the address of the data within @frag. The page must already
2313 static inline void *skb_frag_address(const skb_frag_t
*frag
)
2315 return page_address(skb_frag_page(frag
)) + frag
->page_offset
;
2319 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2320 * @frag: the paged fragment buffer
2322 * Returns the address of the data within @frag. Checks that the page
2323 * is mapped and returns %NULL otherwise.
2325 static inline void *skb_frag_address_safe(const skb_frag_t
*frag
)
2327 void *ptr
= page_address(skb_frag_page(frag
));
2331 return ptr
+ frag
->page_offset
;
2335 * __skb_frag_set_page - sets the page contained in a paged fragment
2336 * @frag: the paged fragment
2337 * @page: the page to set
2339 * Sets the fragment @frag to contain @page.
2341 static inline void __skb_frag_set_page(skb_frag_t
*frag
, struct page
*page
)
2343 frag
->page
.p
= page
;
2347 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2349 * @f: the fragment offset
2350 * @page: the page to set
2352 * Sets the @f'th fragment of @skb to contain @page.
2354 static inline void skb_frag_set_page(struct sk_buff
*skb
, int f
,
2357 __skb_frag_set_page(&skb_shinfo(skb
)->frags
[f
], page
);
2360 bool skb_page_frag_refill(unsigned int sz
, struct page_frag
*pfrag
, gfp_t prio
);
2363 * skb_frag_dma_map - maps a paged fragment via the DMA API
2364 * @dev: the device to map the fragment to
2365 * @frag: the paged fragment to map
2366 * @offset: the offset within the fragment (starting at the
2367 * fragment's own offset)
2368 * @size: the number of bytes to map
2369 * @dir: the direction of the mapping (%PCI_DMA_*)
2371 * Maps the page associated with @frag to @device.
2373 static inline dma_addr_t
skb_frag_dma_map(struct device
*dev
,
2374 const skb_frag_t
*frag
,
2375 size_t offset
, size_t size
,
2376 enum dma_data_direction dir
)
2378 return dma_map_page(dev
, skb_frag_page(frag
),
2379 frag
->page_offset
+ offset
, size
, dir
);
2382 static inline struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
2385 return __pskb_copy(skb
, skb_headroom(skb
), gfp_mask
);
2389 static inline struct sk_buff
*pskb_copy_for_clone(struct sk_buff
*skb
,
2392 return __pskb_copy_fclone(skb
, skb_headroom(skb
), gfp_mask
, true);
2397 * skb_clone_writable - is the header of a clone writable
2398 * @skb: buffer to check
2399 * @len: length up to which to write
2401 * Returns true if modifying the header part of the cloned buffer
2402 * does not requires the data to be copied.
2404 static inline int skb_clone_writable(const struct sk_buff
*skb
, unsigned int len
)
2406 return !skb_header_cloned(skb
) &&
2407 skb_headroom(skb
) + len
<= skb
->hdr_len
;
2410 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
2415 if (headroom
> skb_headroom(skb
))
2416 delta
= headroom
- skb_headroom(skb
);
2418 if (delta
|| cloned
)
2419 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
2425 * skb_cow - copy header of skb when it is required
2426 * @skb: buffer to cow
2427 * @headroom: needed headroom
2429 * If the skb passed lacks sufficient headroom or its data part
2430 * is shared, data is reallocated. If reallocation fails, an error
2431 * is returned and original skb is not changed.
2433 * The result is skb with writable area skb->head...skb->tail
2434 * and at least @headroom of space at head.
2436 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
2438 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
2442 * skb_cow_head - skb_cow but only making the head writable
2443 * @skb: buffer to cow
2444 * @headroom: needed headroom
2446 * This function is identical to skb_cow except that we replace the
2447 * skb_cloned check by skb_header_cloned. It should be used when
2448 * you only need to push on some header and do not need to modify
2451 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
2453 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
2457 * skb_padto - pad an skbuff up to a minimal size
2458 * @skb: buffer to pad
2459 * @len: minimal length
2461 * Pads up a buffer to ensure the trailing bytes exist and are
2462 * blanked. If the buffer already contains sufficient data it
2463 * is untouched. Otherwise it is extended. Returns zero on
2464 * success. The skb is freed on error.
2466 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
2468 unsigned int size
= skb
->len
;
2469 if (likely(size
>= len
))
2471 return skb_pad(skb
, len
- size
);
2475 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2476 * @skb: buffer to pad
2477 * @len: minimal length
2479 * Pads up a buffer to ensure the trailing bytes exist and are
2480 * blanked. If the buffer already contains sufficient data it
2481 * is untouched. Otherwise it is extended. Returns zero on
2482 * success. The skb is freed on error.
2484 static inline int skb_put_padto(struct sk_buff
*skb
, unsigned int len
)
2486 unsigned int size
= skb
->len
;
2488 if (unlikely(size
< len
)) {
2490 if (skb_pad(skb
, len
))
2492 __skb_put(skb
, len
);
2497 static inline int skb_add_data(struct sk_buff
*skb
,
2498 struct iov_iter
*from
, int copy
)
2500 const int off
= skb
->len
;
2502 if (skb
->ip_summed
== CHECKSUM_NONE
) {
2504 if (csum_and_copy_from_iter(skb_put(skb
, copy
), copy
,
2505 &csum
, from
) == copy
) {
2506 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
2509 } else if (copy_from_iter(skb_put(skb
, copy
), copy
, from
) == copy
)
2512 __skb_trim(skb
, off
);
2516 static inline bool skb_can_coalesce(struct sk_buff
*skb
, int i
,
2517 const struct page
*page
, int off
)
2520 const struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
2522 return page
== skb_frag_page(frag
) &&
2523 off
== frag
->page_offset
+ skb_frag_size(frag
);
2528 static inline int __skb_linearize(struct sk_buff
*skb
)
2530 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
2534 * skb_linearize - convert paged skb to linear one
2535 * @skb: buffer to linarize
2537 * If there is no free memory -ENOMEM is returned, otherwise zero
2538 * is returned and the old skb data released.
2540 static inline int skb_linearize(struct sk_buff
*skb
)
2542 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
2546 * skb_has_shared_frag - can any frag be overwritten
2547 * @skb: buffer to test
2549 * Return true if the skb has at least one frag that might be modified
2550 * by an external entity (as in vmsplice()/sendfile())
2552 static inline bool skb_has_shared_frag(const struct sk_buff
*skb
)
2554 return skb_is_nonlinear(skb
) &&
2555 skb_shinfo(skb
)->tx_flags
& SKBTX_SHARED_FRAG
;
2559 * skb_linearize_cow - make sure skb is linear and writable
2560 * @skb: buffer to process
2562 * If there is no free memory -ENOMEM is returned, otherwise zero
2563 * is returned and the old skb data released.
2565 static inline int skb_linearize_cow(struct sk_buff
*skb
)
2567 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
2568 __skb_linearize(skb
) : 0;
2572 * skb_postpull_rcsum - update checksum for received skb after pull
2573 * @skb: buffer to update
2574 * @start: start of data before pull
2575 * @len: length of data pulled
2577 * After doing a pull on a received packet, you need to call this to
2578 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2579 * CHECKSUM_NONE so that it can be recomputed from scratch.
2582 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
2583 const void *start
, unsigned int len
)
2585 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2586 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
2589 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
2592 * pskb_trim_rcsum - trim received skb and update checksum
2593 * @skb: buffer to trim
2596 * This is exactly the same as pskb_trim except that it ensures the
2597 * checksum of received packets are still valid after the operation.
2600 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
2602 if (likely(len
>= skb
->len
))
2604 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2605 skb
->ip_summed
= CHECKSUM_NONE
;
2606 return __pskb_trim(skb
, len
);
2609 #define skb_queue_walk(queue, skb) \
2610 for (skb = (queue)->next; \
2611 skb != (struct sk_buff *)(queue); \
2614 #define skb_queue_walk_safe(queue, skb, tmp) \
2615 for (skb = (queue)->next, tmp = skb->next; \
2616 skb != (struct sk_buff *)(queue); \
2617 skb = tmp, tmp = skb->next)
2619 #define skb_queue_walk_from(queue, skb) \
2620 for (; skb != (struct sk_buff *)(queue); \
2623 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2624 for (tmp = skb->next; \
2625 skb != (struct sk_buff *)(queue); \
2626 skb = tmp, tmp = skb->next)
2628 #define skb_queue_reverse_walk(queue, skb) \
2629 for (skb = (queue)->prev; \
2630 skb != (struct sk_buff *)(queue); \
2633 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2634 for (skb = (queue)->prev, tmp = skb->prev; \
2635 skb != (struct sk_buff *)(queue); \
2636 skb = tmp, tmp = skb->prev)
2638 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2639 for (tmp = skb->prev; \
2640 skb != (struct sk_buff *)(queue); \
2641 skb = tmp, tmp = skb->prev)
2643 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
2645 return skb_shinfo(skb
)->frag_list
!= NULL
;
2648 static inline void skb_frag_list_init(struct sk_buff
*skb
)
2650 skb_shinfo(skb
)->frag_list
= NULL
;
2653 static inline void skb_frag_add_head(struct sk_buff
*skb
, struct sk_buff
*frag
)
2655 frag
->next
= skb_shinfo(skb
)->frag_list
;
2656 skb_shinfo(skb
)->frag_list
= frag
;
2659 #define skb_walk_frags(skb, iter) \
2660 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2662 struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
2663 int *peeked
, int *off
, int *err
);
2664 struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
, int noblock
,
2666 unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
2667 struct poll_table_struct
*wait
);
2668 int skb_copy_datagram_iter(const struct sk_buff
*from
, int offset
,
2669 struct iov_iter
*to
, int size
);
2670 static inline int skb_copy_datagram_msg(const struct sk_buff
*from
, int offset
,
2671 struct msghdr
*msg
, int size
)
2673 return skb_copy_datagram_iter(from
, offset
, &msg
->msg_iter
, size
);
2675 int skb_copy_and_csum_datagram_msg(struct sk_buff
*skb
, int hlen
,
2676 struct msghdr
*msg
);
2677 int skb_copy_datagram_from_iter(struct sk_buff
*skb
, int offset
,
2678 struct iov_iter
*from
, int len
);
2679 int zerocopy_sg_from_iter(struct sk_buff
*skb
, struct iov_iter
*frm
);
2680 void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
2681 void skb_free_datagram_locked(struct sock
*sk
, struct sk_buff
*skb
);
2682 int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
, unsigned int flags
);
2683 int skb_copy_bits(const struct sk_buff
*skb
, int offset
, void *to
, int len
);
2684 int skb_store_bits(struct sk_buff
*skb
, int offset
, const void *from
, int len
);
2685 __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
, int offset
, u8
*to
,
2686 int len
, __wsum csum
);
2687 int skb_splice_bits(struct sk_buff
*skb
, unsigned int offset
,
2688 struct pipe_inode_info
*pipe
, unsigned int len
,
2689 unsigned int flags
);
2690 void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
2691 unsigned int skb_zerocopy_headlen(const struct sk_buff
*from
);
2692 int skb_zerocopy(struct sk_buff
*to
, struct sk_buff
*from
,
2694 void skb_split(struct sk_buff
*skb
, struct sk_buff
*skb1
, const u32 len
);
2695 int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
, int shiftlen
);
2696 void skb_scrub_packet(struct sk_buff
*skb
, bool xnet
);
2697 unsigned int skb_gso_transport_seglen(const struct sk_buff
*skb
);
2698 struct sk_buff
*skb_segment(struct sk_buff
*skb
, netdev_features_t features
);
2699 struct sk_buff
*skb_vlan_untag(struct sk_buff
*skb
);
2700 int skb_ensure_writable(struct sk_buff
*skb
, int write_len
);
2701 int skb_vlan_pop(struct sk_buff
*skb
);
2702 int skb_vlan_push(struct sk_buff
*skb
, __be16 vlan_proto
, u16 vlan_tci
);
2704 static inline int memcpy_from_msg(void *data
, struct msghdr
*msg
, int len
)
2706 return copy_from_iter(data
, len
, &msg
->msg_iter
) == len
? 0 : -EFAULT
;
2709 static inline int memcpy_to_msg(struct msghdr
*msg
, void *data
, int len
)
2711 return copy_to_iter(data
, len
, &msg
->msg_iter
) == len
? 0 : -EFAULT
;
2714 struct skb_checksum_ops
{
2715 __wsum (*update
)(const void *mem
, int len
, __wsum wsum
);
2716 __wsum (*combine
)(__wsum csum
, __wsum csum2
, int offset
, int len
);
2719 __wsum
__skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
2720 __wsum csum
, const struct skb_checksum_ops
*ops
);
2721 __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
2724 static inline void *__skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2725 int len
, void *data
, int hlen
, void *buffer
)
2727 if (hlen
- offset
>= len
)
2728 return data
+ offset
;
2731 skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
2737 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2738 int len
, void *buffer
)
2740 return __skb_header_pointer(skb
, offset
, len
, skb
->data
,
2741 skb_headlen(skb
), buffer
);
2745 * skb_needs_linearize - check if we need to linearize a given skb
2746 * depending on the given device features.
2747 * @skb: socket buffer to check
2748 * @features: net device features
2750 * Returns true if either:
2751 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2752 * 2. skb is fragmented and the device does not support SG.
2754 static inline bool skb_needs_linearize(struct sk_buff
*skb
,
2755 netdev_features_t features
)
2757 return skb_is_nonlinear(skb
) &&
2758 ((skb_has_frag_list(skb
) && !(features
& NETIF_F_FRAGLIST
)) ||
2759 (skb_shinfo(skb
)->nr_frags
&& !(features
& NETIF_F_SG
)));
2762 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
2764 const unsigned int len
)
2766 memcpy(to
, skb
->data
, len
);
2769 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
2770 const int offset
, void *to
,
2771 const unsigned int len
)
2773 memcpy(to
, skb
->data
+ offset
, len
);
2776 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
2778 const unsigned int len
)
2780 memcpy(skb
->data
, from
, len
);
2783 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
2786 const unsigned int len
)
2788 memcpy(skb
->data
+ offset
, from
, len
);
2791 void skb_init(void);
2793 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
2799 * skb_get_timestamp - get timestamp from a skb
2800 * @skb: skb to get stamp from
2801 * @stamp: pointer to struct timeval to store stamp in
2803 * Timestamps are stored in the skb as offsets to a base timestamp.
2804 * This function converts the offset back to a struct timeval and stores
2807 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
2808 struct timeval
*stamp
)
2810 *stamp
= ktime_to_timeval(skb
->tstamp
);
2813 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
2814 struct timespec
*stamp
)
2816 *stamp
= ktime_to_timespec(skb
->tstamp
);
2819 static inline void __net_timestamp(struct sk_buff
*skb
)
2821 skb
->tstamp
= ktime_get_real();
2824 static inline ktime_t
net_timedelta(ktime_t t
)
2826 return ktime_sub(ktime_get_real(), t
);
2829 static inline ktime_t
net_invalid_timestamp(void)
2831 return ktime_set(0, 0);
2834 struct sk_buff
*skb_clone_sk(struct sk_buff
*skb
);
2836 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2838 void skb_clone_tx_timestamp(struct sk_buff
*skb
);
2839 bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
2841 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2843 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
2847 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
2852 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2855 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2857 * PHY drivers may accept clones of transmitted packets for
2858 * timestamping via their phy_driver.txtstamp method. These drivers
2859 * must call this function to return the skb back to the stack, with
2860 * or without a timestamp.
2862 * @skb: clone of the the original outgoing packet
2863 * @hwtstamps: hardware time stamps, may be NULL if not available
2866 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
2867 struct skb_shared_hwtstamps
*hwtstamps
);
2869 void __skb_tstamp_tx(struct sk_buff
*orig_skb
,
2870 struct skb_shared_hwtstamps
*hwtstamps
,
2871 struct sock
*sk
, int tstype
);
2874 * skb_tstamp_tx - queue clone of skb with send time stamps
2875 * @orig_skb: the original outgoing packet
2876 * @hwtstamps: hardware time stamps, may be NULL if not available
2878 * If the skb has a socket associated, then this function clones the
2879 * skb (thus sharing the actual data and optional structures), stores
2880 * the optional hardware time stamping information (if non NULL) or
2881 * generates a software time stamp (otherwise), then queues the clone
2882 * to the error queue of the socket. Errors are silently ignored.
2884 void skb_tstamp_tx(struct sk_buff
*orig_skb
,
2885 struct skb_shared_hwtstamps
*hwtstamps
);
2887 static inline void sw_tx_timestamp(struct sk_buff
*skb
)
2889 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
&&
2890 !(skb_shinfo(skb
)->tx_flags
& SKBTX_IN_PROGRESS
))
2891 skb_tstamp_tx(skb
, NULL
);
2895 * skb_tx_timestamp() - Driver hook for transmit timestamping
2897 * Ethernet MAC Drivers should call this function in their hard_xmit()
2898 * function immediately before giving the sk_buff to the MAC hardware.
2900 * Specifically, one should make absolutely sure that this function is
2901 * called before TX completion of this packet can trigger. Otherwise
2902 * the packet could potentially already be freed.
2904 * @skb: A socket buffer.
2906 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
2908 skb_clone_tx_timestamp(skb
);
2909 sw_tx_timestamp(skb
);
2913 * skb_complete_wifi_ack - deliver skb with wifi status
2915 * @skb: the original outgoing packet
2916 * @acked: ack status
2919 void skb_complete_wifi_ack(struct sk_buff
*skb
, bool acked
);
2921 __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
2922 __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
2924 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
2926 return ((skb
->ip_summed
== CHECKSUM_UNNECESSARY
) ||
2928 (skb
->ip_summed
== CHECKSUM_PARTIAL
&&
2929 skb_checksum_start_offset(skb
) >= 0));
2933 * skb_checksum_complete - Calculate checksum of an entire packet
2934 * @skb: packet to process
2936 * This function calculates the checksum over the entire packet plus
2937 * the value of skb->csum. The latter can be used to supply the
2938 * checksum of a pseudo header as used by TCP/UDP. It returns the
2941 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2942 * this function can be used to verify that checksum on received
2943 * packets. In that case the function should return zero if the
2944 * checksum is correct. In particular, this function will return zero
2945 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2946 * hardware has already verified the correctness of the checksum.
2948 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
2950 return skb_csum_unnecessary(skb
) ?
2951 0 : __skb_checksum_complete(skb
);
2954 static inline void __skb_decr_checksum_unnecessary(struct sk_buff
*skb
)
2956 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
2957 if (skb
->csum_level
== 0)
2958 skb
->ip_summed
= CHECKSUM_NONE
;
2964 static inline void __skb_incr_checksum_unnecessary(struct sk_buff
*skb
)
2966 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
2967 if (skb
->csum_level
< SKB_MAX_CSUM_LEVEL
)
2969 } else if (skb
->ip_summed
== CHECKSUM_NONE
) {
2970 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
2971 skb
->csum_level
= 0;
2975 static inline void __skb_mark_checksum_bad(struct sk_buff
*skb
)
2977 /* Mark current checksum as bad (typically called from GRO
2978 * path). In the case that ip_summed is CHECKSUM_NONE
2979 * this must be the first checksum encountered in the packet.
2980 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2981 * checksum after the last one validated. For UDP, a zero
2982 * checksum can not be marked as bad.
2985 if (skb
->ip_summed
== CHECKSUM_NONE
||
2986 skb
->ip_summed
== CHECKSUM_UNNECESSARY
)
2990 /* Check if we need to perform checksum complete validation.
2992 * Returns true if checksum complete is needed, false otherwise
2993 * (either checksum is unnecessary or zero checksum is allowed).
2995 static inline bool __skb_checksum_validate_needed(struct sk_buff
*skb
,
2999 if (skb_csum_unnecessary(skb
) || (zero_okay
&& !check
)) {
3000 skb
->csum_valid
= 1;
3001 __skb_decr_checksum_unnecessary(skb
);
3008 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3011 #define CHECKSUM_BREAK 76
3013 /* Validate (init) checksum based on checksum complete.
3016 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3017 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3018 * checksum is stored in skb->csum for use in __skb_checksum_complete
3019 * non-zero: value of invalid checksum
3022 static inline __sum16
__skb_checksum_validate_complete(struct sk_buff
*skb
,
3026 if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
3027 if (!csum_fold(csum_add(psum
, skb
->csum
))) {
3028 skb
->csum_valid
= 1;
3031 } else if (skb
->csum_bad
) {
3032 /* ip_summed == CHECKSUM_NONE in this case */
3038 if (complete
|| skb
->len
<= CHECKSUM_BREAK
) {
3041 csum
= __skb_checksum_complete(skb
);
3042 skb
->csum_valid
= !csum
;
3049 static inline __wsum
null_compute_pseudo(struct sk_buff
*skb
, int proto
)
3054 /* Perform checksum validate (init). Note that this is a macro since we only
3055 * want to calculate the pseudo header which is an input function if necessary.
3056 * First we try to validate without any computation (checksum unnecessary) and
3057 * then calculate based on checksum complete calling the function to compute
3061 * 0: checksum is validated or try to in skb_checksum_complete
3062 * non-zero: value of invalid checksum
3064 #define __skb_checksum_validate(skb, proto, complete, \
3065 zero_okay, check, compute_pseudo) \
3067 __sum16 __ret = 0; \
3068 skb->csum_valid = 0; \
3069 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3070 __ret = __skb_checksum_validate_complete(skb, \
3071 complete, compute_pseudo(skb, proto)); \
3075 #define skb_checksum_init(skb, proto, compute_pseudo) \
3076 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3078 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3079 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3081 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3082 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3084 #define skb_checksum_validate_zero_check(skb, proto, check, \
3086 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3088 #define skb_checksum_simple_validate(skb) \
3089 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3091 static inline bool __skb_checksum_convert_check(struct sk_buff
*skb
)
3093 return (skb
->ip_summed
== CHECKSUM_NONE
&&
3094 skb
->csum_valid
&& !skb
->csum_bad
);
3097 static inline void __skb_checksum_convert(struct sk_buff
*skb
,
3098 __sum16 check
, __wsum pseudo
)
3100 skb
->csum
= ~pseudo
;
3101 skb
->ip_summed
= CHECKSUM_COMPLETE
;
3104 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3106 if (__skb_checksum_convert_check(skb)) \
3107 __skb_checksum_convert(skb, check, \
3108 compute_pseudo(skb, proto)); \
3111 static inline void skb_remcsum_adjust_partial(struct sk_buff
*skb
, void *ptr
,
3112 u16 start
, u16 offset
)
3114 skb
->ip_summed
= CHECKSUM_PARTIAL
;
3115 skb
->csum_start
= ((unsigned char *)ptr
+ start
) - skb
->head
;
3116 skb
->csum_offset
= offset
- start
;
3119 /* Update skbuf and packet to reflect the remote checksum offload operation.
3120 * When called, ptr indicates the starting point for skb->csum when
3121 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3122 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3124 static inline void skb_remcsum_process(struct sk_buff
*skb
, void *ptr
,
3125 int start
, int offset
, bool nopartial
)
3130 skb_remcsum_adjust_partial(skb
, ptr
, start
, offset
);
3134 if (unlikely(skb
->ip_summed
!= CHECKSUM_COMPLETE
)) {
3135 __skb_checksum_complete(skb
);
3136 skb_postpull_rcsum(skb
, skb
->data
, ptr
- (void *)skb
->data
);
3139 delta
= remcsum_adjust(ptr
, skb
->csum
, start
, offset
);
3141 /* Adjust skb->csum since we changed the packet */
3142 skb
->csum
= csum_add(skb
->csum
, delta
);
3145 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3146 void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
3147 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
3149 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
3150 nf_conntrack_destroy(nfct
);
3152 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
3155 atomic_inc(&nfct
->use
);
3158 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3159 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
3161 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
3164 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
3167 atomic_inc(&nf_bridge
->use
);
3169 #endif /* CONFIG_BRIDGE_NETFILTER */
3170 static inline void nf_reset(struct sk_buff
*skb
)
3172 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3173 nf_conntrack_put(skb
->nfct
);
3176 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3177 nf_bridge_put(skb
->nf_bridge
);
3178 skb
->nf_bridge
= NULL
;
3182 static inline void nf_reset_trace(struct sk_buff
*skb
)
3184 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3189 /* Note: This doesn't put any conntrack and bridge info in dst. */
3190 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
,
3193 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3194 dst
->nfct
= src
->nfct
;
3195 nf_conntrack_get(src
->nfct
);
3197 dst
->nfctinfo
= src
->nfctinfo
;
3199 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3200 dst
->nf_bridge
= src
->nf_bridge
;
3201 nf_bridge_get(src
->nf_bridge
);
3203 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3205 dst
->nf_trace
= src
->nf_trace
;
3209 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
3211 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3212 nf_conntrack_put(dst
->nfct
);
3214 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3215 nf_bridge_put(dst
->nf_bridge
);
3217 __nf_copy(dst
, src
, true);
3220 #ifdef CONFIG_NETWORK_SECMARK
3221 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3223 to
->secmark
= from
->secmark
;
3226 static inline void skb_init_secmark(struct sk_buff
*skb
)
3231 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3234 static inline void skb_init_secmark(struct sk_buff
*skb
)
3238 static inline bool skb_irq_freeable(const struct sk_buff
*skb
)
3240 return !skb
->destructor
&&
3241 #if IS_ENABLED(CONFIG_XFRM)
3244 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3247 !skb
->_skb_refdst
&&
3248 !skb_has_frag_list(skb
);
3251 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
3253 skb
->queue_mapping
= queue_mapping
;
3256 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
3258 return skb
->queue_mapping
;
3261 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
3263 to
->queue_mapping
= from
->queue_mapping
;
3266 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
3268 skb
->queue_mapping
= rx_queue
+ 1;
3271 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
3273 return skb
->queue_mapping
- 1;
3276 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
3278 return skb
->queue_mapping
!= 0;
3281 u16
__skb_tx_hash(const struct net_device
*dev
, struct sk_buff
*skb
,
3282 unsigned int num_tx_queues
);
3284 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
3293 /* Keeps track of mac header offset relative to skb->head.
3294 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3295 * For non-tunnel skb it points to skb_mac_header() and for
3296 * tunnel skb it points to outer mac header.
3297 * Keeps track of level of encapsulation of network headers.
3304 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3306 static inline int skb_tnl_header_len(const struct sk_buff
*inner_skb
)
3308 return (skb_mac_header(inner_skb
) - inner_skb
->head
) -
3309 SKB_GSO_CB(inner_skb
)->mac_offset
;
3312 static inline int gso_pskb_expand_head(struct sk_buff
*skb
, int extra
)
3314 int new_headroom
, headroom
;
3317 headroom
= skb_headroom(skb
);
3318 ret
= pskb_expand_head(skb
, extra
, 0, GFP_ATOMIC
);
3322 new_headroom
= skb_headroom(skb
);
3323 SKB_GSO_CB(skb
)->mac_offset
+= (new_headroom
- headroom
);
3327 /* Compute the checksum for a gso segment. First compute the checksum value
3328 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3329 * then add in skb->csum (checksum from csum_start to end of packet).
3330 * skb->csum and csum_start are then updated to reflect the checksum of the
3331 * resultant packet starting from the transport header-- the resultant checksum
3332 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3335 static inline __sum16
gso_make_checksum(struct sk_buff
*skb
, __wsum res
)
3337 int plen
= SKB_GSO_CB(skb
)->csum_start
- skb_headroom(skb
) -
3338 skb_transport_offset(skb
);
3341 csum
= csum_fold(csum_partial(skb_transport_header(skb
),
3344 SKB_GSO_CB(skb
)->csum_start
-= plen
;
3349 static inline bool skb_is_gso(const struct sk_buff
*skb
)
3351 return skb_shinfo(skb
)->gso_size
;
3354 /* Note: Should be called only if skb_is_gso(skb) is true */
3355 static inline bool skb_is_gso_v6(const struct sk_buff
*skb
)
3357 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
3360 void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
3362 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
3364 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3365 * wanted then gso_type will be set. */
3366 const struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
3368 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
3369 unlikely(shinfo
->gso_type
== 0)) {
3370 __skb_warn_lro_forwarding(skb
);
3376 static inline void skb_forward_csum(struct sk_buff
*skb
)
3378 /* Unfortunately we don't support this one. Any brave souls? */
3379 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3380 skb
->ip_summed
= CHECKSUM_NONE
;
3384 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3385 * @skb: skb to check
3387 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3388 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3389 * use this helper, to document places where we make this assertion.
3391 static inline void skb_checksum_none_assert(const struct sk_buff
*skb
)
3394 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
3398 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
3400 int skb_checksum_setup(struct sk_buff
*skb
, bool recalculate
);
3402 u32
skb_get_poff(const struct sk_buff
*skb
);
3403 u32
__skb_get_poff(const struct sk_buff
*skb
, void *data
,
3404 const struct flow_keys
*keys
, int hlen
);
3407 * skb_head_is_locked - Determine if the skb->head is locked down
3408 * @skb: skb to check
3410 * The head on skbs build around a head frag can be removed if they are
3411 * not cloned. This function returns true if the skb head is locked down
3412 * due to either being allocated via kmalloc, or by being a clone with
3413 * multiple references to the head.
3415 static inline bool skb_head_is_locked(const struct sk_buff
*skb
)
3417 return !skb
->head_frag
|| skb_cloned(skb
);
3421 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3425 * skb_gso_network_seglen is used to determine the real size of the
3426 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3428 * The MAC/L2 header is not accounted for.
3430 static inline unsigned int skb_gso_network_seglen(const struct sk_buff
*skb
)
3432 unsigned int hdr_len
= skb_transport_header(skb
) -
3433 skb_network_header(skb
);
3434 return hdr_len
+ skb_gso_transport_seglen(skb
);
3436 #endif /* __KERNEL__ */
3437 #endif /* _LINUX_SKBUFF_H */