2 * Definitions for the 'struct sk_buff' memory handlers.
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/kernel.h>
18 #include <linux/kmemcheck.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/bug.h>
22 #include <linux/cache.h>
23 #include <linux/rbtree.h>
24 #include <linux/socket.h>
26 #include <linux/atomic.h>
27 #include <asm/types.h>
28 #include <linux/spinlock.h>
29 #include <linux/net.h>
30 #include <linux/textsearch.h>
31 #include <net/checksum.h>
32 #include <linux/rcupdate.h>
33 #include <linux/hrtimer.h>
34 #include <linux/dma-mapping.h>
35 #include <linux/netdev_features.h>
36 #include <linux/sched.h>
37 #include <net/flow_dissector.h>
38 #include <linux/splice.h>
40 /* A. Checksumming of received packets by device.
44 * Device failed to checksum this packet e.g. due to lack of capabilities.
45 * The packet contains full (though not verified) checksum in packet but
46 * not in skb->csum. Thus, skb->csum is undefined in this case.
48 * CHECKSUM_UNNECESSARY:
50 * The hardware you're dealing with doesn't calculate the full checksum
51 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
52 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
53 * if their checksums are okay. skb->csum is still undefined in this case
54 * though. It is a bad option, but, unfortunately, nowadays most vendors do
55 * this. Apparently with the secret goal to sell you new devices, when you
56 * will add new protocol to your host, f.e. IPv6 8)
58 * CHECKSUM_UNNECESSARY is applicable to following protocols:
60 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
61 * zero UDP checksum for either IPv4 or IPv6, the networking stack
62 * may perform further validation in this case.
63 * GRE: only if the checksum is present in the header.
64 * SCTP: indicates the CRC in SCTP header has been validated.
66 * skb->csum_level indicates the number of consecutive checksums found in
67 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
68 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
69 * and a device is able to verify the checksums for UDP (possibly zero),
70 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
71 * two. If the device were only able to verify the UDP checksum and not
72 * GRE, either because it doesn't support GRE checksum of because GRE
73 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
74 * not considered in this case).
78 * This is the most generic way. The device supplied checksum of the _whole_
79 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
80 * hardware doesn't need to parse L3/L4 headers to implement this.
82 * Note: Even if device supports only some protocols, but is able to produce
83 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
87 * A checksum is set up to be offloaded to a device as described in the
88 * output description for CHECKSUM_PARTIAL. This may occur on a packet
89 * received directly from another Linux OS, e.g., a virtualized Linux kernel
90 * on the same host, or it may be set in the input path in GRO or remote
91 * checksum offload. For the purposes of checksum verification, the checksum
92 * referred to by skb->csum_start + skb->csum_offset and any preceding
93 * checksums in the packet are considered verified. Any checksums in the
94 * packet that are after the checksum being offloaded are not considered to
97 * B. Checksumming on output.
101 * The skb was already checksummed by the protocol, or a checksum is not
106 * The device is required to checksum the packet as seen by hard_start_xmit()
107 * from skb->csum_start up to the end, and to record/write the checksum at
108 * offset skb->csum_start + skb->csum_offset.
110 * The device must show its capabilities in dev->features, set up at device
111 * setup time, e.g. netdev_features.h:
113 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
114 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
115 * IPv4. Sigh. Vendors like this way for an unknown reason.
116 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
117 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
118 * NETIF_F_... - Well, you get the picture.
120 * CHECKSUM_UNNECESSARY:
122 * Normally, the device will do per protocol specific checksumming. Protocol
123 * implementations that do not want the NIC to perform the checksum
124 * calculation should use this flag in their outgoing skbs.
126 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
127 * offload. Correspondingly, the FCoE protocol driver
128 * stack should use CHECKSUM_UNNECESSARY.
130 * Any questions? No questions, good. --ANK
133 /* Don't change this without changing skb_csum_unnecessary! */
134 #define CHECKSUM_NONE 0
135 #define CHECKSUM_UNNECESSARY 1
136 #define CHECKSUM_COMPLETE 2
137 #define CHECKSUM_PARTIAL 3
139 /* Maximum value in skb->csum_level */
140 #define SKB_MAX_CSUM_LEVEL 3
142 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
143 #define SKB_WITH_OVERHEAD(X) \
144 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
145 #define SKB_MAX_ORDER(X, ORDER) \
146 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
147 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
148 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
150 /* return minimum truesize of one skb containing X bytes of data */
151 #define SKB_TRUESIZE(X) ((X) + \
152 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
153 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
157 struct pipe_inode_info
;
161 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
162 struct nf_conntrack
{
167 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
168 struct nf_bridge_info
{
171 BRNF_PROTO_UNCHANGED
,
177 struct net_device
*physindev
;
179 struct net_device
*physoutdev
;
180 char neigh_header
[8];
186 struct sk_buff_head
{
187 /* These two members must be first. */
188 struct sk_buff
*next
;
189 struct sk_buff
*prev
;
197 /* To allow 64K frame to be packed as single skb without frag_list we
198 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
199 * buffers which do not start on a page boundary.
201 * Since GRO uses frags we allocate at least 16 regardless of page
204 #if (65536/PAGE_SIZE + 1) < 16
205 #define MAX_SKB_FRAGS 16UL
207 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
210 typedef struct skb_frag_struct skb_frag_t
;
212 struct skb_frag_struct
{
216 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
225 static inline unsigned int skb_frag_size(const skb_frag_t
*frag
)
230 static inline void skb_frag_size_set(skb_frag_t
*frag
, unsigned int size
)
235 static inline void skb_frag_size_add(skb_frag_t
*frag
, int delta
)
240 static inline void skb_frag_size_sub(skb_frag_t
*frag
, int delta
)
245 #define HAVE_HW_TIME_STAMP
248 * struct skb_shared_hwtstamps - hardware time stamps
249 * @hwtstamp: hardware time stamp transformed into duration
250 * since arbitrary point in time
252 * Software time stamps generated by ktime_get_real() are stored in
255 * hwtstamps can only be compared against other hwtstamps from
258 * This structure is attached to packets as part of the
259 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
261 struct skb_shared_hwtstamps
{
265 /* Definitions for tx_flags in struct skb_shared_info */
267 /* generate hardware time stamp */
268 SKBTX_HW_TSTAMP
= 1 << 0,
270 /* generate software time stamp when queueing packet to NIC */
271 SKBTX_SW_TSTAMP
= 1 << 1,
273 /* device driver is going to provide hardware time stamp */
274 SKBTX_IN_PROGRESS
= 1 << 2,
276 /* device driver supports TX zero-copy buffers */
277 SKBTX_DEV_ZEROCOPY
= 1 << 3,
279 /* generate wifi status information (where possible) */
280 SKBTX_WIFI_STATUS
= 1 << 4,
282 /* This indicates at least one fragment might be overwritten
283 * (as in vmsplice(), sendfile() ...)
284 * If we need to compute a TX checksum, we'll need to copy
285 * all frags to avoid possible bad checksum
287 SKBTX_SHARED_FRAG
= 1 << 5,
289 /* generate software time stamp when entering packet scheduling */
290 SKBTX_SCHED_TSTAMP
= 1 << 6,
292 /* generate software timestamp on peer data acknowledgment */
293 SKBTX_ACK_TSTAMP
= 1 << 7,
296 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
297 SKBTX_SCHED_TSTAMP | \
299 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
302 * The callback notifies userspace to release buffers when skb DMA is done in
303 * lower device, the skb last reference should be 0 when calling this.
304 * The zerocopy_success argument is true if zero copy transmit occurred,
305 * false on data copy or out of memory error caused by data copy attempt.
306 * The ctx field is used to track device context.
307 * The desc field is used to track userspace buffer index.
310 void (*callback
)(struct ubuf_info
*, bool zerocopy_success
);
315 /* This data is invariant across clones and lives at
316 * the end of the header data, ie. at skb->end.
318 struct skb_shared_info
{
319 unsigned char nr_frags
;
321 unsigned short gso_size
;
322 /* Warning: this field is not always filled in (UFO)! */
323 unsigned short gso_segs
;
324 unsigned short gso_type
;
325 struct sk_buff
*frag_list
;
326 struct skb_shared_hwtstamps hwtstamps
;
331 * Warning : all fields before dataref are cleared in __alloc_skb()
335 /* Intermediate layers must ensure that destructor_arg
336 * remains valid until skb destructor */
337 void * destructor_arg
;
339 /* must be last field, see pskb_expand_head() */
340 skb_frag_t frags
[MAX_SKB_FRAGS
];
343 /* We divide dataref into two halves. The higher 16 bits hold references
344 * to the payload part of skb->data. The lower 16 bits hold references to
345 * the entire skb->data. A clone of a headerless skb holds the length of
346 * the header in skb->hdr_len.
348 * All users must obey the rule that the skb->data reference count must be
349 * greater than or equal to the payload reference count.
351 * Holding a reference to the payload part means that the user does not
352 * care about modifications to the header part of skb->data.
354 #define SKB_DATAREF_SHIFT 16
355 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
359 SKB_FCLONE_UNAVAILABLE
, /* skb has no fclone (from head_cache) */
360 SKB_FCLONE_ORIG
, /* orig skb (from fclone_cache) */
361 SKB_FCLONE_CLONE
, /* companion fclone skb (from fclone_cache) */
365 SKB_GSO_TCPV4
= 1 << 0,
366 SKB_GSO_UDP
= 1 << 1,
368 /* This indicates the skb is from an untrusted source. */
369 SKB_GSO_DODGY
= 1 << 2,
371 /* This indicates the tcp segment has CWR set. */
372 SKB_GSO_TCP_ECN
= 1 << 3,
374 SKB_GSO_TCPV6
= 1 << 4,
376 SKB_GSO_FCOE
= 1 << 5,
378 SKB_GSO_GRE
= 1 << 6,
380 SKB_GSO_GRE_CSUM
= 1 << 7,
382 SKB_GSO_IPIP
= 1 << 8,
384 SKB_GSO_SIT
= 1 << 9,
386 SKB_GSO_UDP_TUNNEL
= 1 << 10,
388 SKB_GSO_UDP_TUNNEL_CSUM
= 1 << 11,
390 SKB_GSO_TUNNEL_REMCSUM
= 1 << 12,
393 #if BITS_PER_LONG > 32
394 #define NET_SKBUFF_DATA_USES_OFFSET 1
397 #ifdef NET_SKBUFF_DATA_USES_OFFSET
398 typedef unsigned int sk_buff_data_t
;
400 typedef unsigned char *sk_buff_data_t
;
404 * struct skb_mstamp - multi resolution time stamps
405 * @stamp_us: timestamp in us resolution
406 * @stamp_jiffies: timestamp in jiffies
419 * skb_mstamp_get - get current timestamp
420 * @cl: place to store timestamps
422 static inline void skb_mstamp_get(struct skb_mstamp
*cl
)
424 u64 val
= local_clock();
426 do_div(val
, NSEC_PER_USEC
);
427 cl
->stamp_us
= (u32
)val
;
428 cl
->stamp_jiffies
= (u32
)jiffies
;
432 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
433 * @t1: pointer to newest sample
434 * @t0: pointer to oldest sample
436 static inline u32
skb_mstamp_us_delta(const struct skb_mstamp
*t1
,
437 const struct skb_mstamp
*t0
)
439 s32 delta_us
= t1
->stamp_us
- t0
->stamp_us
;
440 u32 delta_jiffies
= t1
->stamp_jiffies
- t0
->stamp_jiffies
;
442 /* If delta_us is negative, this might be because interval is too big,
443 * or local_clock() drift is too big : fallback using jiffies.
446 delta_jiffies
>= (INT_MAX
/ (USEC_PER_SEC
/ HZ
)))
448 delta_us
= jiffies_to_usecs(delta_jiffies
);
455 * struct sk_buff - socket buffer
456 * @next: Next buffer in list
457 * @prev: Previous buffer in list
458 * @tstamp: Time we arrived/left
459 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
460 * @sk: Socket we are owned by
461 * @dev: Device we arrived on/are leaving by
462 * @cb: Control buffer. Free for use by every layer. Put private vars here
463 * @_skb_refdst: destination entry (with norefcount bit)
464 * @sp: the security path, used for xfrm
465 * @len: Length of actual data
466 * @data_len: Data length
467 * @mac_len: Length of link layer header
468 * @hdr_len: writable header length of cloned skb
469 * @csum: Checksum (must include start/offset pair)
470 * @csum_start: Offset from skb->head where checksumming should start
471 * @csum_offset: Offset from csum_start where checksum should be stored
472 * @priority: Packet queueing priority
473 * @ignore_df: allow local fragmentation
474 * @cloned: Head may be cloned (check refcnt to be sure)
475 * @ip_summed: Driver fed us an IP checksum
476 * @nohdr: Payload reference only, must not modify header
477 * @nfctinfo: Relationship of this skb to the connection
478 * @pkt_type: Packet class
479 * @fclone: skbuff clone status
480 * @ipvs_property: skbuff is owned by ipvs
481 * @peeked: this packet has been seen already, so stats have been
482 * done for it, don't do them again
483 * @nf_trace: netfilter packet trace flag
484 * @protocol: Packet protocol from driver
485 * @destructor: Destruct function
486 * @nfct: Associated connection, if any
487 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
488 * @skb_iif: ifindex of device we arrived on
489 * @tc_index: Traffic control index
490 * @tc_verd: traffic control verdict
491 * @hash: the packet hash
492 * @queue_mapping: Queue mapping for multiqueue devices
493 * @xmit_more: More SKBs are pending for this queue
494 * @ndisc_nodetype: router type (from link layer)
495 * @ooo_okay: allow the mapping of a socket to a queue to be changed
496 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
498 * @sw_hash: indicates hash was computed in software stack
499 * @wifi_acked_valid: wifi_acked was set
500 * @wifi_acked: whether frame was acked on wifi or not
501 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
502 * @napi_id: id of the NAPI struct this skb came from
503 * @secmark: security marking
504 * @mark: Generic packet mark
505 * @vlan_proto: vlan encapsulation protocol
506 * @vlan_tci: vlan tag control information
507 * @inner_protocol: Protocol (encapsulation)
508 * @inner_transport_header: Inner transport layer header (encapsulation)
509 * @inner_network_header: Network layer header (encapsulation)
510 * @inner_mac_header: Link layer header (encapsulation)
511 * @transport_header: Transport layer header
512 * @network_header: Network layer header
513 * @mac_header: Link layer header
514 * @tail: Tail pointer
516 * @head: Head of buffer
517 * @data: Data head pointer
518 * @truesize: Buffer size
519 * @users: User count - see {datagram,tcp}.c
525 /* These two members must be first. */
526 struct sk_buff
*next
;
527 struct sk_buff
*prev
;
531 struct skb_mstamp skb_mstamp
;
534 struct rb_node rbnode
; /* used in netem & tcp stack */
537 struct net_device
*dev
;
540 * This is the control buffer. It is free to use for every
541 * layer. Please put your private variables there. If you
542 * want to keep them across layers you have to do a skb_clone()
543 * first. This is owned by whoever has the skb queued ATM.
545 char cb
[48] __aligned(8);
547 unsigned long _skb_refdst
;
548 void (*destructor
)(struct sk_buff
*skb
);
552 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
553 struct nf_conntrack
*nfct
;
555 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
556 struct nf_bridge_info
*nf_bridge
;
563 /* Following fields are _not_ copied in __copy_skb_header()
564 * Note that queue_mapping is here mostly to fill a hole.
566 kmemcheck_bitfield_begin(flags1
);
575 kmemcheck_bitfield_end(flags1
);
577 /* fields enclosed in headers_start/headers_end are copied
578 * using a single memcpy() in __copy_skb_header()
581 __u32 headers_start
[0];
584 /* if you move pkt_type around you also must adapt those constants */
585 #ifdef __BIG_ENDIAN_BITFIELD
586 #define PKT_TYPE_MAX (7 << 5)
588 #define PKT_TYPE_MAX 7
590 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
592 __u8 __pkt_type_offset
[0];
603 __u8 wifi_acked_valid
:1;
607 /* Indicates the inner headers are valid in the skbuff. */
608 __u8 encapsulation
:1;
609 __u8 encap_hdr_csum
:1;
611 __u8 csum_complete_sw
:1;
615 #ifdef CONFIG_IPV6_NDISC_NODETYPE
616 __u8 ndisc_nodetype
:2;
618 __u8 ipvs_property
:1;
619 __u8 inner_protocol_type
:1;
620 __u8 remcsum_offload
:1;
621 /* 3 or 5 bit hole */
623 #ifdef CONFIG_NET_SCHED
624 __u16 tc_index
; /* traffic control index */
625 #ifdef CONFIG_NET_CLS_ACT
626 __u16 tc_verd
; /* traffic control verdict */
642 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
644 unsigned int napi_id
;
645 unsigned int sender_cpu
;
648 #ifdef CONFIG_NETWORK_SECMARK
653 __u32 reserved_tailroom
;
657 __be16 inner_protocol
;
661 __u16 inner_transport_header
;
662 __u16 inner_network_header
;
663 __u16 inner_mac_header
;
666 __u16 transport_header
;
667 __u16 network_header
;
671 __u32 headers_end
[0];
674 /* These elements must be at the end, see alloc_skb() for details. */
679 unsigned int truesize
;
685 * Handling routines are only of interest to the kernel
687 #include <linux/slab.h>
690 #define SKB_ALLOC_FCLONE 0x01
691 #define SKB_ALLOC_RX 0x02
692 #define SKB_ALLOC_NAPI 0x04
694 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
695 static inline bool skb_pfmemalloc(const struct sk_buff
*skb
)
697 return unlikely(skb
->pfmemalloc
);
701 * skb might have a dst pointer attached, refcounted or not.
702 * _skb_refdst low order bit is set if refcount was _not_ taken
704 #define SKB_DST_NOREF 1UL
705 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
708 * skb_dst - returns skb dst_entry
711 * Returns skb dst_entry, regardless of reference taken or not.
713 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
715 /* If refdst was not refcounted, check we still are in a
716 * rcu_read_lock section
718 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
719 !rcu_read_lock_held() &&
720 !rcu_read_lock_bh_held());
721 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
725 * skb_dst_set - sets skb dst
729 * Sets skb dst, assuming a reference was taken on dst and should
730 * be released by skb_dst_drop()
732 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
734 skb
->_skb_refdst
= (unsigned long)dst
;
738 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
742 * Sets skb dst, assuming a reference was not taken on dst.
743 * If dst entry is cached, we do not take reference and dst_release
744 * will be avoided by refdst_drop. If dst entry is not cached, we take
745 * reference, so that last dst_release can destroy the dst immediately.
747 static inline void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
)
749 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
750 skb
->_skb_refdst
= (unsigned long)dst
| SKB_DST_NOREF
;
754 * skb_dst_is_noref - Test if skb dst isn't refcounted
757 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
759 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
762 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
764 return (struct rtable
*)skb_dst(skb
);
767 void kfree_skb(struct sk_buff
*skb
);
768 void kfree_skb_list(struct sk_buff
*segs
);
769 void skb_tx_error(struct sk_buff
*skb
);
770 void consume_skb(struct sk_buff
*skb
);
771 void __kfree_skb(struct sk_buff
*skb
);
772 extern struct kmem_cache
*skbuff_head_cache
;
774 void kfree_skb_partial(struct sk_buff
*skb
, bool head_stolen
);
775 bool skb_try_coalesce(struct sk_buff
*to
, struct sk_buff
*from
,
776 bool *fragstolen
, int *delta_truesize
);
778 struct sk_buff
*__alloc_skb(unsigned int size
, gfp_t priority
, int flags
,
780 struct sk_buff
*__build_skb(void *data
, unsigned int frag_size
);
781 struct sk_buff
*build_skb(void *data
, unsigned int frag_size
);
782 static inline struct sk_buff
*alloc_skb(unsigned int size
,
785 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
788 struct sk_buff
*alloc_skb_with_frags(unsigned long header_len
,
789 unsigned long data_len
,
794 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
795 struct sk_buff_fclones
{
804 * skb_fclone_busy - check if fclone is busy
807 * Returns true is skb is a fast clone, and its clone is not freed.
808 * Some drivers call skb_orphan() in their ndo_start_xmit(),
809 * so we also check that this didnt happen.
811 static inline bool skb_fclone_busy(const struct sock
*sk
,
812 const struct sk_buff
*skb
)
814 const struct sk_buff_fclones
*fclones
;
816 fclones
= container_of(skb
, struct sk_buff_fclones
, skb1
);
818 return skb
->fclone
== SKB_FCLONE_ORIG
&&
819 atomic_read(&fclones
->fclone_ref
) > 1 &&
820 fclones
->skb2
.sk
== sk
;
823 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
826 return __alloc_skb(size
, priority
, SKB_ALLOC_FCLONE
, NUMA_NO_NODE
);
829 struct sk_buff
*__alloc_skb_head(gfp_t priority
, int node
);
830 static inline struct sk_buff
*alloc_skb_head(gfp_t priority
)
832 return __alloc_skb_head(priority
, -1);
835 struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
836 int skb_copy_ubufs(struct sk_buff
*skb
, gfp_t gfp_mask
);
837 struct sk_buff
*skb_clone(struct sk_buff
*skb
, gfp_t priority
);
838 struct sk_buff
*skb_copy(const struct sk_buff
*skb
, gfp_t priority
);
839 struct sk_buff
*__pskb_copy_fclone(struct sk_buff
*skb
, int headroom
,
840 gfp_t gfp_mask
, bool fclone
);
841 static inline struct sk_buff
*__pskb_copy(struct sk_buff
*skb
, int headroom
,
844 return __pskb_copy_fclone(skb
, headroom
, gfp_mask
, false);
847 int pskb_expand_head(struct sk_buff
*skb
, int nhead
, int ntail
, gfp_t gfp_mask
);
848 struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
849 unsigned int headroom
);
850 struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
, int newheadroom
,
851 int newtailroom
, gfp_t priority
);
852 int skb_to_sgvec_nomark(struct sk_buff
*skb
, struct scatterlist
*sg
,
853 int offset
, int len
);
854 int skb_to_sgvec(struct sk_buff
*skb
, struct scatterlist
*sg
, int offset
,
856 int skb_cow_data(struct sk_buff
*skb
, int tailbits
, struct sk_buff
**trailer
);
857 int skb_pad(struct sk_buff
*skb
, int pad
);
858 #define dev_kfree_skb(a) consume_skb(a)
860 int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
861 int getfrag(void *from
, char *to
, int offset
,
862 int len
, int odd
, struct sk_buff
*skb
),
863 void *from
, int length
);
865 int skb_append_pagefrags(struct sk_buff
*skb
, struct page
*page
,
866 int offset
, size_t size
);
868 struct skb_seq_state
{
872 __u32 stepped_offset
;
873 struct sk_buff
*root_skb
;
874 struct sk_buff
*cur_skb
;
878 void skb_prepare_seq_read(struct sk_buff
*skb
, unsigned int from
,
879 unsigned int to
, struct skb_seq_state
*st
);
880 unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
881 struct skb_seq_state
*st
);
882 void skb_abort_seq_read(struct skb_seq_state
*st
);
884 unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
885 unsigned int to
, struct ts_config
*config
);
888 * Packet hash types specify the type of hash in skb_set_hash.
890 * Hash types refer to the protocol layer addresses which are used to
891 * construct a packet's hash. The hashes are used to differentiate or identify
892 * flows of the protocol layer for the hash type. Hash types are either
893 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
895 * Properties of hashes:
897 * 1) Two packets in different flows have different hash values
898 * 2) Two packets in the same flow should have the same hash value
900 * A hash at a higher layer is considered to be more specific. A driver should
901 * set the most specific hash possible.
903 * A driver cannot indicate a more specific hash than the layer at which a hash
904 * was computed. For instance an L3 hash cannot be set as an L4 hash.
906 * A driver may indicate a hash level which is less specific than the
907 * actual layer the hash was computed on. For instance, a hash computed
908 * at L4 may be considered an L3 hash. This should only be done if the
909 * driver can't unambiguously determine that the HW computed the hash at
910 * the higher layer. Note that the "should" in the second property above
913 enum pkt_hash_types
{
914 PKT_HASH_TYPE_NONE
, /* Undefined type */
915 PKT_HASH_TYPE_L2
, /* Input: src_MAC, dest_MAC */
916 PKT_HASH_TYPE_L3
, /* Input: src_IP, dst_IP */
917 PKT_HASH_TYPE_L4
, /* Input: src_IP, dst_IP, src_port, dst_port */
921 skb_set_hash(struct sk_buff
*skb
, __u32 hash
, enum pkt_hash_types type
)
923 skb
->l4_hash
= (type
== PKT_HASH_TYPE_L4
);
928 static inline __u32
skb_get_hash(struct sk_buff
*skb
)
930 if (!skb
->l4_hash
&& !skb
->sw_hash
)
936 __u32
skb_get_hash_perturb(const struct sk_buff
*skb
, u32 perturb
);
938 static inline __u32
skb_get_hash_raw(const struct sk_buff
*skb
)
943 static inline void skb_clear_hash(struct sk_buff
*skb
)
950 static inline void skb_clear_hash_if_not_l4(struct sk_buff
*skb
)
956 static inline void skb_copy_hash(struct sk_buff
*to
, const struct sk_buff
*from
)
958 to
->hash
= from
->hash
;
959 to
->sw_hash
= from
->sw_hash
;
960 to
->l4_hash
= from
->l4_hash
;
963 static inline void skb_sender_cpu_clear(struct sk_buff
*skb
)
970 #ifdef NET_SKBUFF_DATA_USES_OFFSET
971 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
973 return skb
->head
+ skb
->end
;
976 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
981 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
986 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
988 return skb
->end
- skb
->head
;
993 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
995 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
997 return &skb_shinfo(skb
)->hwtstamps
;
1001 * skb_queue_empty - check if a queue is empty
1004 * Returns true if the queue is empty, false otherwise.
1006 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
1008 return list
->next
== (const struct sk_buff
*) list
;
1012 * skb_queue_is_last - check if skb is the last entry in the queue
1016 * Returns true if @skb is the last buffer on the list.
1018 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
1019 const struct sk_buff
*skb
)
1021 return skb
->next
== (const struct sk_buff
*) list
;
1025 * skb_queue_is_first - check if skb is the first entry in the queue
1029 * Returns true if @skb is the first buffer on the list.
1031 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
1032 const struct sk_buff
*skb
)
1034 return skb
->prev
== (const struct sk_buff
*) list
;
1038 * skb_queue_next - return the next packet in the queue
1040 * @skb: current buffer
1042 * Return the next packet in @list after @skb. It is only valid to
1043 * call this if skb_queue_is_last() evaluates to false.
1045 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
1046 const struct sk_buff
*skb
)
1048 /* This BUG_ON may seem severe, but if we just return then we
1049 * are going to dereference garbage.
1051 BUG_ON(skb_queue_is_last(list
, skb
));
1056 * skb_queue_prev - return the prev packet in the queue
1058 * @skb: current buffer
1060 * Return the prev packet in @list before @skb. It is only valid to
1061 * call this if skb_queue_is_first() evaluates to false.
1063 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
1064 const struct sk_buff
*skb
)
1066 /* This BUG_ON may seem severe, but if we just return then we
1067 * are going to dereference garbage.
1069 BUG_ON(skb_queue_is_first(list
, skb
));
1074 * skb_get - reference buffer
1075 * @skb: buffer to reference
1077 * Makes another reference to a socket buffer and returns a pointer
1080 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
1082 atomic_inc(&skb
->users
);
1087 * If users == 1, we are the only owner and are can avoid redundant
1092 * skb_cloned - is the buffer a clone
1093 * @skb: buffer to check
1095 * Returns true if the buffer was generated with skb_clone() and is
1096 * one of multiple shared copies of the buffer. Cloned buffers are
1097 * shared data so must not be written to under normal circumstances.
1099 static inline int skb_cloned(const struct sk_buff
*skb
)
1101 return skb
->cloned
&&
1102 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
1105 static inline int skb_unclone(struct sk_buff
*skb
, gfp_t pri
)
1107 might_sleep_if(pri
& __GFP_WAIT
);
1109 if (skb_cloned(skb
))
1110 return pskb_expand_head(skb
, 0, 0, pri
);
1116 * skb_header_cloned - is the header a clone
1117 * @skb: buffer to check
1119 * Returns true if modifying the header part of the buffer requires
1120 * the data to be copied.
1122 static inline int skb_header_cloned(const struct sk_buff
*skb
)
1129 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
1130 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
1131 return dataref
!= 1;
1135 * skb_header_release - release reference to header
1136 * @skb: buffer to operate on
1138 * Drop a reference to the header part of the buffer. This is done
1139 * by acquiring a payload reference. You must not read from the header
1140 * part of skb->data after this.
1141 * Note : Check if you can use __skb_header_release() instead.
1143 static inline void skb_header_release(struct sk_buff
*skb
)
1147 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
1151 * __skb_header_release - release reference to header
1152 * @skb: buffer to operate on
1154 * Variant of skb_header_release() assuming skb is private to caller.
1155 * We can avoid one atomic operation.
1157 static inline void __skb_header_release(struct sk_buff
*skb
)
1160 atomic_set(&skb_shinfo(skb
)->dataref
, 1 + (1 << SKB_DATAREF_SHIFT
));
1165 * skb_shared - is the buffer shared
1166 * @skb: buffer to check
1168 * Returns true if more than one person has a reference to this
1171 static inline int skb_shared(const struct sk_buff
*skb
)
1173 return atomic_read(&skb
->users
) != 1;
1177 * skb_share_check - check if buffer is shared and if so clone it
1178 * @skb: buffer to check
1179 * @pri: priority for memory allocation
1181 * If the buffer is shared the buffer is cloned and the old copy
1182 * drops a reference. A new clone with a single reference is returned.
1183 * If the buffer is not shared the original buffer is returned. When
1184 * being called from interrupt status or with spinlocks held pri must
1187 * NULL is returned on a memory allocation failure.
1189 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
, gfp_t pri
)
1191 might_sleep_if(pri
& __GFP_WAIT
);
1192 if (skb_shared(skb
)) {
1193 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
1205 * Copy shared buffers into a new sk_buff. We effectively do COW on
1206 * packets to handle cases where we have a local reader and forward
1207 * and a couple of other messy ones. The normal one is tcpdumping
1208 * a packet thats being forwarded.
1212 * skb_unshare - make a copy of a shared buffer
1213 * @skb: buffer to check
1214 * @pri: priority for memory allocation
1216 * If the socket buffer is a clone then this function creates a new
1217 * copy of the data, drops a reference count on the old copy and returns
1218 * the new copy with the reference count at 1. If the buffer is not a clone
1219 * the original buffer is returned. When called with a spinlock held or
1220 * from interrupt state @pri must be %GFP_ATOMIC
1222 * %NULL is returned on a memory allocation failure.
1224 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
1227 might_sleep_if(pri
& __GFP_WAIT
);
1228 if (skb_cloned(skb
)) {
1229 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
1231 /* Free our shared copy */
1242 * skb_peek - peek at the head of an &sk_buff_head
1243 * @list_: list to peek at
1245 * Peek an &sk_buff. Unlike most other operations you _MUST_
1246 * be careful with this one. A peek leaves the buffer on the
1247 * list and someone else may run off with it. You must hold
1248 * the appropriate locks or have a private queue to do this.
1250 * Returns %NULL for an empty list or a pointer to the head element.
1251 * The reference count is not incremented and the reference is therefore
1252 * volatile. Use with caution.
1254 static inline struct sk_buff
*skb_peek(const struct sk_buff_head
*list_
)
1256 struct sk_buff
*skb
= list_
->next
;
1258 if (skb
== (struct sk_buff
*)list_
)
1264 * skb_peek_next - peek skb following the given one from a queue
1265 * @skb: skb to start from
1266 * @list_: list to peek at
1268 * Returns %NULL when the end of the list is met or a pointer to the
1269 * next element. The reference count is not incremented and the
1270 * reference is therefore volatile. Use with caution.
1272 static inline struct sk_buff
*skb_peek_next(struct sk_buff
*skb
,
1273 const struct sk_buff_head
*list_
)
1275 struct sk_buff
*next
= skb
->next
;
1277 if (next
== (struct sk_buff
*)list_
)
1283 * skb_peek_tail - peek at the tail of an &sk_buff_head
1284 * @list_: list to peek at
1286 * Peek an &sk_buff. Unlike most other operations you _MUST_
1287 * be careful with this one. A peek leaves the buffer on the
1288 * list and someone else may run off with it. You must hold
1289 * the appropriate locks or have a private queue to do this.
1291 * Returns %NULL for an empty list or a pointer to the tail element.
1292 * The reference count is not incremented and the reference is therefore
1293 * volatile. Use with caution.
1295 static inline struct sk_buff
*skb_peek_tail(const struct sk_buff_head
*list_
)
1297 struct sk_buff
*skb
= list_
->prev
;
1299 if (skb
== (struct sk_buff
*)list_
)
1306 * skb_queue_len - get queue length
1307 * @list_: list to measure
1309 * Return the length of an &sk_buff queue.
1311 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
1317 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1318 * @list: queue to initialize
1320 * This initializes only the list and queue length aspects of
1321 * an sk_buff_head object. This allows to initialize the list
1322 * aspects of an sk_buff_head without reinitializing things like
1323 * the spinlock. It can also be used for on-stack sk_buff_head
1324 * objects where the spinlock is known to not be used.
1326 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
1328 list
->prev
= list
->next
= (struct sk_buff
*)list
;
1333 * This function creates a split out lock class for each invocation;
1334 * this is needed for now since a whole lot of users of the skb-queue
1335 * infrastructure in drivers have different locking usage (in hardirq)
1336 * than the networking core (in softirq only). In the long run either the
1337 * network layer or drivers should need annotation to consolidate the
1338 * main types of usage into 3 classes.
1340 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
1342 spin_lock_init(&list
->lock
);
1343 __skb_queue_head_init(list
);
1346 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
1347 struct lock_class_key
*class)
1349 skb_queue_head_init(list
);
1350 lockdep_set_class(&list
->lock
, class);
1354 * Insert an sk_buff on a list.
1356 * The "__skb_xxxx()" functions are the non-atomic ones that
1357 * can only be called with interrupts disabled.
1359 void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
,
1360 struct sk_buff_head
*list
);
1361 static inline void __skb_insert(struct sk_buff
*newsk
,
1362 struct sk_buff
*prev
, struct sk_buff
*next
,
1363 struct sk_buff_head
*list
)
1367 next
->prev
= prev
->next
= newsk
;
1371 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
1372 struct sk_buff
*prev
,
1373 struct sk_buff
*next
)
1375 struct sk_buff
*first
= list
->next
;
1376 struct sk_buff
*last
= list
->prev
;
1386 * skb_queue_splice - join two skb lists, this is designed for stacks
1387 * @list: the new list to add
1388 * @head: the place to add it in the first list
1390 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
1391 struct sk_buff_head
*head
)
1393 if (!skb_queue_empty(list
)) {
1394 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1395 head
->qlen
+= list
->qlen
;
1400 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1401 * @list: the new list to add
1402 * @head: the place to add it in the first list
1404 * The list at @list is reinitialised
1406 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
1407 struct sk_buff_head
*head
)
1409 if (!skb_queue_empty(list
)) {
1410 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1411 head
->qlen
+= list
->qlen
;
1412 __skb_queue_head_init(list
);
1417 * skb_queue_splice_tail - join two skb lists, each list being a queue
1418 * @list: the new list to add
1419 * @head: the place to add it in the first list
1421 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
1422 struct sk_buff_head
*head
)
1424 if (!skb_queue_empty(list
)) {
1425 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1426 head
->qlen
+= list
->qlen
;
1431 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1432 * @list: the new list to add
1433 * @head: the place to add it in the first list
1435 * Each of the lists is a queue.
1436 * The list at @list is reinitialised
1438 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
1439 struct sk_buff_head
*head
)
1441 if (!skb_queue_empty(list
)) {
1442 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1443 head
->qlen
+= list
->qlen
;
1444 __skb_queue_head_init(list
);
1449 * __skb_queue_after - queue a buffer at the list head
1450 * @list: list to use
1451 * @prev: place after this buffer
1452 * @newsk: buffer to queue
1454 * Queue a buffer int the middle of a list. This function takes no locks
1455 * and you must therefore hold required locks before calling it.
1457 * A buffer cannot be placed on two lists at the same time.
1459 static inline void __skb_queue_after(struct sk_buff_head
*list
,
1460 struct sk_buff
*prev
,
1461 struct sk_buff
*newsk
)
1463 __skb_insert(newsk
, prev
, prev
->next
, list
);
1466 void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
1467 struct sk_buff_head
*list
);
1469 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1470 struct sk_buff
*next
,
1471 struct sk_buff
*newsk
)
1473 __skb_insert(newsk
, next
->prev
, next
, list
);
1477 * __skb_queue_head - queue a buffer at the list head
1478 * @list: list to use
1479 * @newsk: buffer to queue
1481 * Queue a buffer at the start of a list. This function takes no locks
1482 * and you must therefore hold required locks before calling it.
1484 * A buffer cannot be placed on two lists at the same time.
1486 void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1487 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1488 struct sk_buff
*newsk
)
1490 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1494 * __skb_queue_tail - queue a buffer at the list tail
1495 * @list: list to use
1496 * @newsk: buffer to queue
1498 * Queue a buffer at the end of a list. This function takes no locks
1499 * and you must therefore hold required locks before calling it.
1501 * A buffer cannot be placed on two lists at the same time.
1503 void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1504 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1505 struct sk_buff
*newsk
)
1507 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1511 * remove sk_buff from list. _Must_ be called atomically, and with
1514 void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1515 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1517 struct sk_buff
*next
, *prev
;
1522 skb
->next
= skb
->prev
= NULL
;
1528 * __skb_dequeue - remove from the head of the queue
1529 * @list: list to dequeue from
1531 * Remove the head of the list. This function does not take any locks
1532 * so must be used with appropriate locks held only. The head item is
1533 * returned or %NULL if the list is empty.
1535 struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1536 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1538 struct sk_buff
*skb
= skb_peek(list
);
1540 __skb_unlink(skb
, list
);
1545 * __skb_dequeue_tail - remove from the tail of the queue
1546 * @list: list to dequeue from
1548 * Remove the tail of the list. This function does not take any locks
1549 * so must be used with appropriate locks held only. The tail item is
1550 * returned or %NULL if the list is empty.
1552 struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1553 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1555 struct sk_buff
*skb
= skb_peek_tail(list
);
1557 __skb_unlink(skb
, list
);
1562 static inline bool skb_is_nonlinear(const struct sk_buff
*skb
)
1564 return skb
->data_len
;
1567 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1569 return skb
->len
- skb
->data_len
;
1572 static inline int skb_pagelen(const struct sk_buff
*skb
)
1576 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1577 len
+= skb_frag_size(&skb_shinfo(skb
)->frags
[i
]);
1578 return len
+ skb_headlen(skb
);
1582 * __skb_fill_page_desc - initialise a paged fragment in an skb
1583 * @skb: buffer containing fragment to be initialised
1584 * @i: paged fragment index to initialise
1585 * @page: the page to use for this fragment
1586 * @off: the offset to the data with @page
1587 * @size: the length of the data
1589 * Initialises the @i'th fragment of @skb to point to &size bytes at
1590 * offset @off within @page.
1592 * Does not take any additional reference on the fragment.
1594 static inline void __skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1595 struct page
*page
, int off
, int size
)
1597 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1600 * Propagate page->pfmemalloc to the skb if we can. The problem is
1601 * that not all callers have unique ownership of the page. If
1602 * pfmemalloc is set, we check the mapping as a mapping implies
1603 * page->index is set (index and pfmemalloc share space).
1604 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1605 * do not lose pfmemalloc information as the pages would not be
1606 * allocated using __GFP_MEMALLOC.
1608 frag
->page
.p
= page
;
1609 frag
->page_offset
= off
;
1610 skb_frag_size_set(frag
, size
);
1612 page
= compound_head(page
);
1613 if (page
->pfmemalloc
&& !page
->mapping
)
1614 skb
->pfmemalloc
= true;
1618 * skb_fill_page_desc - initialise a paged fragment in an skb
1619 * @skb: buffer containing fragment to be initialised
1620 * @i: paged fragment index to initialise
1621 * @page: the page to use for this fragment
1622 * @off: the offset to the data with @page
1623 * @size: the length of the data
1625 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1626 * @skb to point to @size bytes at offset @off within @page. In
1627 * addition updates @skb such that @i is the last fragment.
1629 * Does not take any additional reference on the fragment.
1631 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1632 struct page
*page
, int off
, int size
)
1634 __skb_fill_page_desc(skb
, i
, page
, off
, size
);
1635 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1638 void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
, int off
,
1639 int size
, unsigned int truesize
);
1641 void skb_coalesce_rx_frag(struct sk_buff
*skb
, int i
, int size
,
1642 unsigned int truesize
);
1644 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1645 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1646 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1648 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1649 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1651 return skb
->head
+ skb
->tail
;
1654 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1656 skb
->tail
= skb
->data
- skb
->head
;
1659 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1661 skb_reset_tail_pointer(skb
);
1662 skb
->tail
+= offset
;
1665 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1666 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1671 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1673 skb
->tail
= skb
->data
;
1676 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1678 skb
->tail
= skb
->data
+ offset
;
1681 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1684 * Add data to an sk_buff
1686 unsigned char *pskb_put(struct sk_buff
*skb
, struct sk_buff
*tail
, int len
);
1687 unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1688 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1690 unsigned char *tmp
= skb_tail_pointer(skb
);
1691 SKB_LINEAR_ASSERT(skb
);
1697 unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1698 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1705 unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1706 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1709 BUG_ON(skb
->len
< skb
->data_len
);
1710 return skb
->data
+= len
;
1713 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1715 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1718 unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1720 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1722 if (len
> skb_headlen(skb
) &&
1723 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1726 return skb
->data
+= len
;
1729 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1731 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1734 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1736 if (likely(len
<= skb_headlen(skb
)))
1738 if (unlikely(len
> skb
->len
))
1740 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1744 * skb_headroom - bytes at buffer head
1745 * @skb: buffer to check
1747 * Return the number of bytes of free space at the head of an &sk_buff.
1749 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1751 return skb
->data
- skb
->head
;
1755 * skb_tailroom - bytes at buffer end
1756 * @skb: buffer to check
1758 * Return the number of bytes of free space at the tail of an sk_buff
1760 static inline int skb_tailroom(const struct sk_buff
*skb
)
1762 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1766 * skb_availroom - bytes at buffer end
1767 * @skb: buffer to check
1769 * Return the number of bytes of free space at the tail of an sk_buff
1770 * allocated by sk_stream_alloc()
1772 static inline int skb_availroom(const struct sk_buff
*skb
)
1774 if (skb_is_nonlinear(skb
))
1777 return skb
->end
- skb
->tail
- skb
->reserved_tailroom
;
1781 * skb_reserve - adjust headroom
1782 * @skb: buffer to alter
1783 * @len: bytes to move
1785 * Increase the headroom of an empty &sk_buff by reducing the tail
1786 * room. This is only allowed for an empty buffer.
1788 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
1794 #define ENCAP_TYPE_ETHER 0
1795 #define ENCAP_TYPE_IPPROTO 1
1797 static inline void skb_set_inner_protocol(struct sk_buff
*skb
,
1800 skb
->inner_protocol
= protocol
;
1801 skb
->inner_protocol_type
= ENCAP_TYPE_ETHER
;
1804 static inline void skb_set_inner_ipproto(struct sk_buff
*skb
,
1807 skb
->inner_ipproto
= ipproto
;
1808 skb
->inner_protocol_type
= ENCAP_TYPE_IPPROTO
;
1811 static inline void skb_reset_inner_headers(struct sk_buff
*skb
)
1813 skb
->inner_mac_header
= skb
->mac_header
;
1814 skb
->inner_network_header
= skb
->network_header
;
1815 skb
->inner_transport_header
= skb
->transport_header
;
1818 static inline void skb_reset_mac_len(struct sk_buff
*skb
)
1820 skb
->mac_len
= skb
->network_header
- skb
->mac_header
;
1823 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1826 return skb
->head
+ skb
->inner_transport_header
;
1829 static inline void skb_reset_inner_transport_header(struct sk_buff
*skb
)
1831 skb
->inner_transport_header
= skb
->data
- skb
->head
;
1834 static inline void skb_set_inner_transport_header(struct sk_buff
*skb
,
1837 skb_reset_inner_transport_header(skb
);
1838 skb
->inner_transport_header
+= offset
;
1841 static inline unsigned char *skb_inner_network_header(const struct sk_buff
*skb
)
1843 return skb
->head
+ skb
->inner_network_header
;
1846 static inline void skb_reset_inner_network_header(struct sk_buff
*skb
)
1848 skb
->inner_network_header
= skb
->data
- skb
->head
;
1851 static inline void skb_set_inner_network_header(struct sk_buff
*skb
,
1854 skb_reset_inner_network_header(skb
);
1855 skb
->inner_network_header
+= offset
;
1858 static inline unsigned char *skb_inner_mac_header(const struct sk_buff
*skb
)
1860 return skb
->head
+ skb
->inner_mac_header
;
1863 static inline void skb_reset_inner_mac_header(struct sk_buff
*skb
)
1865 skb
->inner_mac_header
= skb
->data
- skb
->head
;
1868 static inline void skb_set_inner_mac_header(struct sk_buff
*skb
,
1871 skb_reset_inner_mac_header(skb
);
1872 skb
->inner_mac_header
+= offset
;
1874 static inline bool skb_transport_header_was_set(const struct sk_buff
*skb
)
1876 return skb
->transport_header
!= (typeof(skb
->transport_header
))~0U;
1879 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1881 return skb
->head
+ skb
->transport_header
;
1884 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1886 skb
->transport_header
= skb
->data
- skb
->head
;
1889 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1892 skb_reset_transport_header(skb
);
1893 skb
->transport_header
+= offset
;
1896 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1898 return skb
->head
+ skb
->network_header
;
1901 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1903 skb
->network_header
= skb
->data
- skb
->head
;
1906 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1908 skb_reset_network_header(skb
);
1909 skb
->network_header
+= offset
;
1912 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1914 return skb
->head
+ skb
->mac_header
;
1917 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1919 return skb
->mac_header
!= (typeof(skb
->mac_header
))~0U;
1922 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1924 skb
->mac_header
= skb
->data
- skb
->head
;
1927 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1929 skb_reset_mac_header(skb
);
1930 skb
->mac_header
+= offset
;
1933 static inline void skb_pop_mac_header(struct sk_buff
*skb
)
1935 skb
->mac_header
= skb
->network_header
;
1938 static inline void skb_probe_transport_header(struct sk_buff
*skb
,
1939 const int offset_hint
)
1941 struct flow_keys keys
;
1943 if (skb_transport_header_was_set(skb
))
1945 else if (skb_flow_dissect_flow_keys(skb
, &keys
))
1946 skb_set_transport_header(skb
, keys
.control
.thoff
);
1948 skb_set_transport_header(skb
, offset_hint
);
1951 static inline void skb_mac_header_rebuild(struct sk_buff
*skb
)
1953 if (skb_mac_header_was_set(skb
)) {
1954 const unsigned char *old_mac
= skb_mac_header(skb
);
1956 skb_set_mac_header(skb
, -skb
->mac_len
);
1957 memmove(skb_mac_header(skb
), old_mac
, skb
->mac_len
);
1961 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
1963 return skb
->csum_start
- skb_headroom(skb
);
1966 static inline int skb_transport_offset(const struct sk_buff
*skb
)
1968 return skb_transport_header(skb
) - skb
->data
;
1971 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
1973 return skb
->transport_header
- skb
->network_header
;
1976 static inline u32
skb_inner_network_header_len(const struct sk_buff
*skb
)
1978 return skb
->inner_transport_header
- skb
->inner_network_header
;
1981 static inline int skb_network_offset(const struct sk_buff
*skb
)
1983 return skb_network_header(skb
) - skb
->data
;
1986 static inline int skb_inner_network_offset(const struct sk_buff
*skb
)
1988 return skb_inner_network_header(skb
) - skb
->data
;
1991 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
1993 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
1997 * CPUs often take a performance hit when accessing unaligned memory
1998 * locations. The actual performance hit varies, it can be small if the
1999 * hardware handles it or large if we have to take an exception and fix it
2002 * Since an ethernet header is 14 bytes network drivers often end up with
2003 * the IP header at an unaligned offset. The IP header can be aligned by
2004 * shifting the start of the packet by 2 bytes. Drivers should do this
2007 * skb_reserve(skb, NET_IP_ALIGN);
2009 * The downside to this alignment of the IP header is that the DMA is now
2010 * unaligned. On some architectures the cost of an unaligned DMA is high
2011 * and this cost outweighs the gains made by aligning the IP header.
2013 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2016 #ifndef NET_IP_ALIGN
2017 #define NET_IP_ALIGN 2
2021 * The networking layer reserves some headroom in skb data (via
2022 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2023 * the header has to grow. In the default case, if the header has to grow
2024 * 32 bytes or less we avoid the reallocation.
2026 * Unfortunately this headroom changes the DMA alignment of the resulting
2027 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2028 * on some architectures. An architecture can override this value,
2029 * perhaps setting it to a cacheline in size (since that will maintain
2030 * cacheline alignment of the DMA). It must be a power of 2.
2032 * Various parts of the networking layer expect at least 32 bytes of
2033 * headroom, you should not reduce this.
2035 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2036 * to reduce average number of cache lines per packet.
2037 * get_rps_cpus() for example only access one 64 bytes aligned block :
2038 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2041 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2044 int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
2046 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
2048 if (unlikely(skb_is_nonlinear(skb
))) {
2053 skb_set_tail_pointer(skb
, len
);
2056 void skb_trim(struct sk_buff
*skb
, unsigned int len
);
2058 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2061 return ___pskb_trim(skb
, len
);
2062 __skb_trim(skb
, len
);
2066 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2068 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
2072 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2073 * @skb: buffer to alter
2076 * This is identical to pskb_trim except that the caller knows that
2077 * the skb is not cloned so we should never get an error due to out-
2080 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
2082 int err
= pskb_trim(skb
, len
);
2087 * skb_orphan - orphan a buffer
2088 * @skb: buffer to orphan
2090 * If a buffer currently has an owner then we call the owner's
2091 * destructor function and make the @skb unowned. The buffer continues
2092 * to exist but is no longer charged to its former owner.
2094 static inline void skb_orphan(struct sk_buff
*skb
)
2096 if (skb
->destructor
) {
2097 skb
->destructor(skb
);
2098 skb
->destructor
= NULL
;
2106 * skb_orphan_frags - orphan the frags contained in a buffer
2107 * @skb: buffer to orphan frags from
2108 * @gfp_mask: allocation mask for replacement pages
2110 * For each frag in the SKB which needs a destructor (i.e. has an
2111 * owner) create a copy of that frag and release the original
2112 * page by calling the destructor.
2114 static inline int skb_orphan_frags(struct sk_buff
*skb
, gfp_t gfp_mask
)
2116 if (likely(!(skb_shinfo(skb
)->tx_flags
& SKBTX_DEV_ZEROCOPY
)))
2118 return skb_copy_ubufs(skb
, gfp_mask
);
2122 * __skb_queue_purge - empty a list
2123 * @list: list to empty
2125 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2126 * the list and one reference dropped. This function does not take the
2127 * list lock and the caller must hold the relevant locks to use it.
2129 void skb_queue_purge(struct sk_buff_head
*list
);
2130 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
2132 struct sk_buff
*skb
;
2133 while ((skb
= __skb_dequeue(list
)) != NULL
)
2137 void *netdev_alloc_frag(unsigned int fragsz
);
2139 struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
, unsigned int length
,
2143 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2144 * @dev: network device to receive on
2145 * @length: length to allocate
2147 * Allocate a new &sk_buff and assign it a usage count of one. The
2148 * buffer has unspecified headroom built in. Users should allocate
2149 * the headroom they think they need without accounting for the
2150 * built in space. The built in space is used for optimisations.
2152 * %NULL is returned if there is no free memory. Although this function
2153 * allocates memory it can be called from an interrupt.
2155 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
2156 unsigned int length
)
2158 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
2161 /* legacy helper around __netdev_alloc_skb() */
2162 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
2165 return __netdev_alloc_skb(NULL
, length
, gfp_mask
);
2168 /* legacy helper around netdev_alloc_skb() */
2169 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
2171 return netdev_alloc_skb(NULL
, length
);
2175 static inline struct sk_buff
*__netdev_alloc_skb_ip_align(struct net_device
*dev
,
2176 unsigned int length
, gfp_t gfp
)
2178 struct sk_buff
*skb
= __netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
, gfp
);
2180 if (NET_IP_ALIGN
&& skb
)
2181 skb_reserve(skb
, NET_IP_ALIGN
);
2185 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
2186 unsigned int length
)
2188 return __netdev_alloc_skb_ip_align(dev
, length
, GFP_ATOMIC
);
2191 static inline void skb_free_frag(void *addr
)
2193 __free_page_frag(addr
);
2196 void *napi_alloc_frag(unsigned int fragsz
);
2197 struct sk_buff
*__napi_alloc_skb(struct napi_struct
*napi
,
2198 unsigned int length
, gfp_t gfp_mask
);
2199 static inline struct sk_buff
*napi_alloc_skb(struct napi_struct
*napi
,
2200 unsigned int length
)
2202 return __napi_alloc_skb(napi
, length
, GFP_ATOMIC
);
2206 * __dev_alloc_pages - allocate page for network Rx
2207 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2208 * @order: size of the allocation
2210 * Allocate a new page.
2212 * %NULL is returned if there is no free memory.
2214 static inline struct page
*__dev_alloc_pages(gfp_t gfp_mask
,
2217 /* This piece of code contains several assumptions.
2218 * 1. This is for device Rx, therefor a cold page is preferred.
2219 * 2. The expectation is the user wants a compound page.
2220 * 3. If requesting a order 0 page it will not be compound
2221 * due to the check to see if order has a value in prep_new_page
2222 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2223 * code in gfp_to_alloc_flags that should be enforcing this.
2225 gfp_mask
|= __GFP_COLD
| __GFP_COMP
| __GFP_MEMALLOC
;
2227 return alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, order
);
2230 static inline struct page
*dev_alloc_pages(unsigned int order
)
2232 return __dev_alloc_pages(GFP_ATOMIC
, order
);
2236 * __dev_alloc_page - allocate a page for network Rx
2237 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2239 * Allocate a new page.
2241 * %NULL is returned if there is no free memory.
2243 static inline struct page
*__dev_alloc_page(gfp_t gfp_mask
)
2245 return __dev_alloc_pages(gfp_mask
, 0);
2248 static inline struct page
*dev_alloc_page(void)
2250 return __dev_alloc_page(GFP_ATOMIC
);
2254 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2255 * @page: The page that was allocated from skb_alloc_page
2256 * @skb: The skb that may need pfmemalloc set
2258 static inline void skb_propagate_pfmemalloc(struct page
*page
,
2259 struct sk_buff
*skb
)
2261 if (page
&& page
->pfmemalloc
)
2262 skb
->pfmemalloc
= true;
2266 * skb_frag_page - retrieve the page referred to by a paged fragment
2267 * @frag: the paged fragment
2269 * Returns the &struct page associated with @frag.
2271 static inline struct page
*skb_frag_page(const skb_frag_t
*frag
)
2273 return frag
->page
.p
;
2277 * __skb_frag_ref - take an addition reference on a paged fragment.
2278 * @frag: the paged fragment
2280 * Takes an additional reference on the paged fragment @frag.
2282 static inline void __skb_frag_ref(skb_frag_t
*frag
)
2284 get_page(skb_frag_page(frag
));
2288 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2290 * @f: the fragment offset.
2292 * Takes an additional reference on the @f'th paged fragment of @skb.
2294 static inline void skb_frag_ref(struct sk_buff
*skb
, int f
)
2296 __skb_frag_ref(&skb_shinfo(skb
)->frags
[f
]);
2300 * __skb_frag_unref - release a reference on a paged fragment.
2301 * @frag: the paged fragment
2303 * Releases a reference on the paged fragment @frag.
2305 static inline void __skb_frag_unref(skb_frag_t
*frag
)
2307 put_page(skb_frag_page(frag
));
2311 * skb_frag_unref - release a reference on a paged fragment of an skb.
2313 * @f: the fragment offset
2315 * Releases a reference on the @f'th paged fragment of @skb.
2317 static inline void skb_frag_unref(struct sk_buff
*skb
, int f
)
2319 __skb_frag_unref(&skb_shinfo(skb
)->frags
[f
]);
2323 * skb_frag_address - gets the address of the data contained in a paged fragment
2324 * @frag: the paged fragment buffer
2326 * Returns the address of the data within @frag. The page must already
2329 static inline void *skb_frag_address(const skb_frag_t
*frag
)
2331 return page_address(skb_frag_page(frag
)) + frag
->page_offset
;
2335 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2336 * @frag: the paged fragment buffer
2338 * Returns the address of the data within @frag. Checks that the page
2339 * is mapped and returns %NULL otherwise.
2341 static inline void *skb_frag_address_safe(const skb_frag_t
*frag
)
2343 void *ptr
= page_address(skb_frag_page(frag
));
2347 return ptr
+ frag
->page_offset
;
2351 * __skb_frag_set_page - sets the page contained in a paged fragment
2352 * @frag: the paged fragment
2353 * @page: the page to set
2355 * Sets the fragment @frag to contain @page.
2357 static inline void __skb_frag_set_page(skb_frag_t
*frag
, struct page
*page
)
2359 frag
->page
.p
= page
;
2363 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2365 * @f: the fragment offset
2366 * @page: the page to set
2368 * Sets the @f'th fragment of @skb to contain @page.
2370 static inline void skb_frag_set_page(struct sk_buff
*skb
, int f
,
2373 __skb_frag_set_page(&skb_shinfo(skb
)->frags
[f
], page
);
2376 bool skb_page_frag_refill(unsigned int sz
, struct page_frag
*pfrag
, gfp_t prio
);
2379 * skb_frag_dma_map - maps a paged fragment via the DMA API
2380 * @dev: the device to map the fragment to
2381 * @frag: the paged fragment to map
2382 * @offset: the offset within the fragment (starting at the
2383 * fragment's own offset)
2384 * @size: the number of bytes to map
2385 * @dir: the direction of the mapping (%PCI_DMA_*)
2387 * Maps the page associated with @frag to @device.
2389 static inline dma_addr_t
skb_frag_dma_map(struct device
*dev
,
2390 const skb_frag_t
*frag
,
2391 size_t offset
, size_t size
,
2392 enum dma_data_direction dir
)
2394 return dma_map_page(dev
, skb_frag_page(frag
),
2395 frag
->page_offset
+ offset
, size
, dir
);
2398 static inline struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
2401 return __pskb_copy(skb
, skb_headroom(skb
), gfp_mask
);
2405 static inline struct sk_buff
*pskb_copy_for_clone(struct sk_buff
*skb
,
2408 return __pskb_copy_fclone(skb
, skb_headroom(skb
), gfp_mask
, true);
2413 * skb_clone_writable - is the header of a clone writable
2414 * @skb: buffer to check
2415 * @len: length up to which to write
2417 * Returns true if modifying the header part of the cloned buffer
2418 * does not requires the data to be copied.
2420 static inline int skb_clone_writable(const struct sk_buff
*skb
, unsigned int len
)
2422 return !skb_header_cloned(skb
) &&
2423 skb_headroom(skb
) + len
<= skb
->hdr_len
;
2426 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
2431 if (headroom
> skb_headroom(skb
))
2432 delta
= headroom
- skb_headroom(skb
);
2434 if (delta
|| cloned
)
2435 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
2441 * skb_cow - copy header of skb when it is required
2442 * @skb: buffer to cow
2443 * @headroom: needed headroom
2445 * If the skb passed lacks sufficient headroom or its data part
2446 * is shared, data is reallocated. If reallocation fails, an error
2447 * is returned and original skb is not changed.
2449 * The result is skb with writable area skb->head...skb->tail
2450 * and at least @headroom of space at head.
2452 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
2454 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
2458 * skb_cow_head - skb_cow but only making the head writable
2459 * @skb: buffer to cow
2460 * @headroom: needed headroom
2462 * This function is identical to skb_cow except that we replace the
2463 * skb_cloned check by skb_header_cloned. It should be used when
2464 * you only need to push on some header and do not need to modify
2467 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
2469 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
2473 * skb_padto - pad an skbuff up to a minimal size
2474 * @skb: buffer to pad
2475 * @len: minimal length
2477 * Pads up a buffer to ensure the trailing bytes exist and are
2478 * blanked. If the buffer already contains sufficient data it
2479 * is untouched. Otherwise it is extended. Returns zero on
2480 * success. The skb is freed on error.
2482 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
2484 unsigned int size
= skb
->len
;
2485 if (likely(size
>= len
))
2487 return skb_pad(skb
, len
- size
);
2491 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2492 * @skb: buffer to pad
2493 * @len: minimal length
2495 * Pads up a buffer to ensure the trailing bytes exist and are
2496 * blanked. If the buffer already contains sufficient data it
2497 * is untouched. Otherwise it is extended. Returns zero on
2498 * success. The skb is freed on error.
2500 static inline int skb_put_padto(struct sk_buff
*skb
, unsigned int len
)
2502 unsigned int size
= skb
->len
;
2504 if (unlikely(size
< len
)) {
2506 if (skb_pad(skb
, len
))
2508 __skb_put(skb
, len
);
2513 static inline int skb_add_data(struct sk_buff
*skb
,
2514 struct iov_iter
*from
, int copy
)
2516 const int off
= skb
->len
;
2518 if (skb
->ip_summed
== CHECKSUM_NONE
) {
2520 if (csum_and_copy_from_iter(skb_put(skb
, copy
), copy
,
2521 &csum
, from
) == copy
) {
2522 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
2525 } else if (copy_from_iter(skb_put(skb
, copy
), copy
, from
) == copy
)
2528 __skb_trim(skb
, off
);
2532 static inline bool skb_can_coalesce(struct sk_buff
*skb
, int i
,
2533 const struct page
*page
, int off
)
2536 const struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
2538 return page
== skb_frag_page(frag
) &&
2539 off
== frag
->page_offset
+ skb_frag_size(frag
);
2544 static inline int __skb_linearize(struct sk_buff
*skb
)
2546 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
2550 * skb_linearize - convert paged skb to linear one
2551 * @skb: buffer to linarize
2553 * If there is no free memory -ENOMEM is returned, otherwise zero
2554 * is returned and the old skb data released.
2556 static inline int skb_linearize(struct sk_buff
*skb
)
2558 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
2562 * skb_has_shared_frag - can any frag be overwritten
2563 * @skb: buffer to test
2565 * Return true if the skb has at least one frag that might be modified
2566 * by an external entity (as in vmsplice()/sendfile())
2568 static inline bool skb_has_shared_frag(const struct sk_buff
*skb
)
2570 return skb_is_nonlinear(skb
) &&
2571 skb_shinfo(skb
)->tx_flags
& SKBTX_SHARED_FRAG
;
2575 * skb_linearize_cow - make sure skb is linear and writable
2576 * @skb: buffer to process
2578 * If there is no free memory -ENOMEM is returned, otherwise zero
2579 * is returned and the old skb data released.
2581 static inline int skb_linearize_cow(struct sk_buff
*skb
)
2583 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
2584 __skb_linearize(skb
) : 0;
2588 * skb_postpull_rcsum - update checksum for received skb after pull
2589 * @skb: buffer to update
2590 * @start: start of data before pull
2591 * @len: length of data pulled
2593 * After doing a pull on a received packet, you need to call this to
2594 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2595 * CHECKSUM_NONE so that it can be recomputed from scratch.
2598 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
2599 const void *start
, unsigned int len
)
2601 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2602 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
2605 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
2608 * pskb_trim_rcsum - trim received skb and update checksum
2609 * @skb: buffer to trim
2612 * This is exactly the same as pskb_trim except that it ensures the
2613 * checksum of received packets are still valid after the operation.
2616 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
2618 if (likely(len
>= skb
->len
))
2620 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2621 skb
->ip_summed
= CHECKSUM_NONE
;
2622 return __pskb_trim(skb
, len
);
2625 #define skb_queue_walk(queue, skb) \
2626 for (skb = (queue)->next; \
2627 skb != (struct sk_buff *)(queue); \
2630 #define skb_queue_walk_safe(queue, skb, tmp) \
2631 for (skb = (queue)->next, tmp = skb->next; \
2632 skb != (struct sk_buff *)(queue); \
2633 skb = tmp, tmp = skb->next)
2635 #define skb_queue_walk_from(queue, skb) \
2636 for (; skb != (struct sk_buff *)(queue); \
2639 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2640 for (tmp = skb->next; \
2641 skb != (struct sk_buff *)(queue); \
2642 skb = tmp, tmp = skb->next)
2644 #define skb_queue_reverse_walk(queue, skb) \
2645 for (skb = (queue)->prev; \
2646 skb != (struct sk_buff *)(queue); \
2649 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2650 for (skb = (queue)->prev, tmp = skb->prev; \
2651 skb != (struct sk_buff *)(queue); \
2652 skb = tmp, tmp = skb->prev)
2654 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2655 for (tmp = skb->prev; \
2656 skb != (struct sk_buff *)(queue); \
2657 skb = tmp, tmp = skb->prev)
2659 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
2661 return skb_shinfo(skb
)->frag_list
!= NULL
;
2664 static inline void skb_frag_list_init(struct sk_buff
*skb
)
2666 skb_shinfo(skb
)->frag_list
= NULL
;
2669 static inline void skb_frag_add_head(struct sk_buff
*skb
, struct sk_buff
*frag
)
2671 frag
->next
= skb_shinfo(skb
)->frag_list
;
2672 skb_shinfo(skb
)->frag_list
= frag
;
2675 #define skb_walk_frags(skb, iter) \
2676 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2678 struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
2679 int *peeked
, int *off
, int *err
);
2680 struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
, int noblock
,
2682 unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
2683 struct poll_table_struct
*wait
);
2684 int skb_copy_datagram_iter(const struct sk_buff
*from
, int offset
,
2685 struct iov_iter
*to
, int size
);
2686 static inline int skb_copy_datagram_msg(const struct sk_buff
*from
, int offset
,
2687 struct msghdr
*msg
, int size
)
2689 return skb_copy_datagram_iter(from
, offset
, &msg
->msg_iter
, size
);
2691 int skb_copy_and_csum_datagram_msg(struct sk_buff
*skb
, int hlen
,
2692 struct msghdr
*msg
);
2693 int skb_copy_datagram_from_iter(struct sk_buff
*skb
, int offset
,
2694 struct iov_iter
*from
, int len
);
2695 int zerocopy_sg_from_iter(struct sk_buff
*skb
, struct iov_iter
*frm
);
2696 void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
2697 void skb_free_datagram_locked(struct sock
*sk
, struct sk_buff
*skb
);
2698 int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
, unsigned int flags
);
2699 int skb_copy_bits(const struct sk_buff
*skb
, int offset
, void *to
, int len
);
2700 int skb_store_bits(struct sk_buff
*skb
, int offset
, const void *from
, int len
);
2701 __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
, int offset
, u8
*to
,
2702 int len
, __wsum csum
);
2703 ssize_t
skb_socket_splice(struct sock
*sk
,
2704 struct pipe_inode_info
*pipe
,
2705 struct splice_pipe_desc
*spd
);
2706 int skb_splice_bits(struct sk_buff
*skb
, struct sock
*sk
, unsigned int offset
,
2707 struct pipe_inode_info
*pipe
, unsigned int len
,
2709 ssize_t (*splice_cb
)(struct sock
*,
2710 struct pipe_inode_info
*,
2711 struct splice_pipe_desc
*));
2712 void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
2713 unsigned int skb_zerocopy_headlen(const struct sk_buff
*from
);
2714 int skb_zerocopy(struct sk_buff
*to
, struct sk_buff
*from
,
2716 void skb_split(struct sk_buff
*skb
, struct sk_buff
*skb1
, const u32 len
);
2717 int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
, int shiftlen
);
2718 void skb_scrub_packet(struct sk_buff
*skb
, bool xnet
);
2719 unsigned int skb_gso_transport_seglen(const struct sk_buff
*skb
);
2720 struct sk_buff
*skb_segment(struct sk_buff
*skb
, netdev_features_t features
);
2721 struct sk_buff
*skb_vlan_untag(struct sk_buff
*skb
);
2722 int skb_ensure_writable(struct sk_buff
*skb
, int write_len
);
2723 int skb_vlan_pop(struct sk_buff
*skb
);
2724 int skb_vlan_push(struct sk_buff
*skb
, __be16 vlan_proto
, u16 vlan_tci
);
2726 static inline int memcpy_from_msg(void *data
, struct msghdr
*msg
, int len
)
2728 return copy_from_iter(data
, len
, &msg
->msg_iter
) == len
? 0 : -EFAULT
;
2731 static inline int memcpy_to_msg(struct msghdr
*msg
, void *data
, int len
)
2733 return copy_to_iter(data
, len
, &msg
->msg_iter
) == len
? 0 : -EFAULT
;
2736 struct skb_checksum_ops
{
2737 __wsum (*update
)(const void *mem
, int len
, __wsum wsum
);
2738 __wsum (*combine
)(__wsum csum
, __wsum csum2
, int offset
, int len
);
2741 __wsum
__skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
2742 __wsum csum
, const struct skb_checksum_ops
*ops
);
2743 __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
2746 static inline void *__skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2747 int len
, void *data
, int hlen
, void *buffer
)
2749 if (hlen
- offset
>= len
)
2750 return data
+ offset
;
2753 skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
2759 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2760 int len
, void *buffer
)
2762 return __skb_header_pointer(skb
, offset
, len
, skb
->data
,
2763 skb_headlen(skb
), buffer
);
2767 * skb_needs_linearize - check if we need to linearize a given skb
2768 * depending on the given device features.
2769 * @skb: socket buffer to check
2770 * @features: net device features
2772 * Returns true if either:
2773 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2774 * 2. skb is fragmented and the device does not support SG.
2776 static inline bool skb_needs_linearize(struct sk_buff
*skb
,
2777 netdev_features_t features
)
2779 return skb_is_nonlinear(skb
) &&
2780 ((skb_has_frag_list(skb
) && !(features
& NETIF_F_FRAGLIST
)) ||
2781 (skb_shinfo(skb
)->nr_frags
&& !(features
& NETIF_F_SG
)));
2784 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
2786 const unsigned int len
)
2788 memcpy(to
, skb
->data
, len
);
2791 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
2792 const int offset
, void *to
,
2793 const unsigned int len
)
2795 memcpy(to
, skb
->data
+ offset
, len
);
2798 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
2800 const unsigned int len
)
2802 memcpy(skb
->data
, from
, len
);
2805 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
2808 const unsigned int len
)
2810 memcpy(skb
->data
+ offset
, from
, len
);
2813 void skb_init(void);
2815 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
2821 * skb_get_timestamp - get timestamp from a skb
2822 * @skb: skb to get stamp from
2823 * @stamp: pointer to struct timeval to store stamp in
2825 * Timestamps are stored in the skb as offsets to a base timestamp.
2826 * This function converts the offset back to a struct timeval and stores
2829 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
2830 struct timeval
*stamp
)
2832 *stamp
= ktime_to_timeval(skb
->tstamp
);
2835 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
2836 struct timespec
*stamp
)
2838 *stamp
= ktime_to_timespec(skb
->tstamp
);
2841 static inline void __net_timestamp(struct sk_buff
*skb
)
2843 skb
->tstamp
= ktime_get_real();
2846 static inline ktime_t
net_timedelta(ktime_t t
)
2848 return ktime_sub(ktime_get_real(), t
);
2851 static inline ktime_t
net_invalid_timestamp(void)
2853 return ktime_set(0, 0);
2856 struct sk_buff
*skb_clone_sk(struct sk_buff
*skb
);
2858 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2860 void skb_clone_tx_timestamp(struct sk_buff
*skb
);
2861 bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
2863 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2865 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
2869 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
2874 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2877 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2879 * PHY drivers may accept clones of transmitted packets for
2880 * timestamping via their phy_driver.txtstamp method. These drivers
2881 * must call this function to return the skb back to the stack, with
2882 * or without a timestamp.
2884 * @skb: clone of the the original outgoing packet
2885 * @hwtstamps: hardware time stamps, may be NULL if not available
2888 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
2889 struct skb_shared_hwtstamps
*hwtstamps
);
2891 void __skb_tstamp_tx(struct sk_buff
*orig_skb
,
2892 struct skb_shared_hwtstamps
*hwtstamps
,
2893 struct sock
*sk
, int tstype
);
2896 * skb_tstamp_tx - queue clone of skb with send time stamps
2897 * @orig_skb: the original outgoing packet
2898 * @hwtstamps: hardware time stamps, may be NULL if not available
2900 * If the skb has a socket associated, then this function clones the
2901 * skb (thus sharing the actual data and optional structures), stores
2902 * the optional hardware time stamping information (if non NULL) or
2903 * generates a software time stamp (otherwise), then queues the clone
2904 * to the error queue of the socket. Errors are silently ignored.
2906 void skb_tstamp_tx(struct sk_buff
*orig_skb
,
2907 struct skb_shared_hwtstamps
*hwtstamps
);
2909 static inline void sw_tx_timestamp(struct sk_buff
*skb
)
2911 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
&&
2912 !(skb_shinfo(skb
)->tx_flags
& SKBTX_IN_PROGRESS
))
2913 skb_tstamp_tx(skb
, NULL
);
2917 * skb_tx_timestamp() - Driver hook for transmit timestamping
2919 * Ethernet MAC Drivers should call this function in their hard_xmit()
2920 * function immediately before giving the sk_buff to the MAC hardware.
2922 * Specifically, one should make absolutely sure that this function is
2923 * called before TX completion of this packet can trigger. Otherwise
2924 * the packet could potentially already be freed.
2926 * @skb: A socket buffer.
2928 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
2930 skb_clone_tx_timestamp(skb
);
2931 sw_tx_timestamp(skb
);
2935 * skb_complete_wifi_ack - deliver skb with wifi status
2937 * @skb: the original outgoing packet
2938 * @acked: ack status
2941 void skb_complete_wifi_ack(struct sk_buff
*skb
, bool acked
);
2943 __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
2944 __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
2946 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
2948 return ((skb
->ip_summed
== CHECKSUM_UNNECESSARY
) ||
2950 (skb
->ip_summed
== CHECKSUM_PARTIAL
&&
2951 skb_checksum_start_offset(skb
) >= 0));
2955 * skb_checksum_complete - Calculate checksum of an entire packet
2956 * @skb: packet to process
2958 * This function calculates the checksum over the entire packet plus
2959 * the value of skb->csum. The latter can be used to supply the
2960 * checksum of a pseudo header as used by TCP/UDP. It returns the
2963 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2964 * this function can be used to verify that checksum on received
2965 * packets. In that case the function should return zero if the
2966 * checksum is correct. In particular, this function will return zero
2967 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2968 * hardware has already verified the correctness of the checksum.
2970 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
2972 return skb_csum_unnecessary(skb
) ?
2973 0 : __skb_checksum_complete(skb
);
2976 static inline void __skb_decr_checksum_unnecessary(struct sk_buff
*skb
)
2978 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
2979 if (skb
->csum_level
== 0)
2980 skb
->ip_summed
= CHECKSUM_NONE
;
2986 static inline void __skb_incr_checksum_unnecessary(struct sk_buff
*skb
)
2988 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
2989 if (skb
->csum_level
< SKB_MAX_CSUM_LEVEL
)
2991 } else if (skb
->ip_summed
== CHECKSUM_NONE
) {
2992 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
2993 skb
->csum_level
= 0;
2997 static inline void __skb_mark_checksum_bad(struct sk_buff
*skb
)
2999 /* Mark current checksum as bad (typically called from GRO
3000 * path). In the case that ip_summed is CHECKSUM_NONE
3001 * this must be the first checksum encountered in the packet.
3002 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3003 * checksum after the last one validated. For UDP, a zero
3004 * checksum can not be marked as bad.
3007 if (skb
->ip_summed
== CHECKSUM_NONE
||
3008 skb
->ip_summed
== CHECKSUM_UNNECESSARY
)
3012 /* Check if we need to perform checksum complete validation.
3014 * Returns true if checksum complete is needed, false otherwise
3015 * (either checksum is unnecessary or zero checksum is allowed).
3017 static inline bool __skb_checksum_validate_needed(struct sk_buff
*skb
,
3021 if (skb_csum_unnecessary(skb
) || (zero_okay
&& !check
)) {
3022 skb
->csum_valid
= 1;
3023 __skb_decr_checksum_unnecessary(skb
);
3030 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3033 #define CHECKSUM_BREAK 76
3035 /* Unset checksum-complete
3037 * Unset checksum complete can be done when packet is being modified
3038 * (uncompressed for instance) and checksum-complete value is
3041 static inline void skb_checksum_complete_unset(struct sk_buff
*skb
)
3043 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3044 skb
->ip_summed
= CHECKSUM_NONE
;
3047 /* Validate (init) checksum based on checksum complete.
3050 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3051 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3052 * checksum is stored in skb->csum for use in __skb_checksum_complete
3053 * non-zero: value of invalid checksum
3056 static inline __sum16
__skb_checksum_validate_complete(struct sk_buff
*skb
,
3060 if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
3061 if (!csum_fold(csum_add(psum
, skb
->csum
))) {
3062 skb
->csum_valid
= 1;
3065 } else if (skb
->csum_bad
) {
3066 /* ip_summed == CHECKSUM_NONE in this case */
3067 return (__force __sum16
)1;
3072 if (complete
|| skb
->len
<= CHECKSUM_BREAK
) {
3075 csum
= __skb_checksum_complete(skb
);
3076 skb
->csum_valid
= !csum
;
3083 static inline __wsum
null_compute_pseudo(struct sk_buff
*skb
, int proto
)
3088 /* Perform checksum validate (init). Note that this is a macro since we only
3089 * want to calculate the pseudo header which is an input function if necessary.
3090 * First we try to validate without any computation (checksum unnecessary) and
3091 * then calculate based on checksum complete calling the function to compute
3095 * 0: checksum is validated or try to in skb_checksum_complete
3096 * non-zero: value of invalid checksum
3098 #define __skb_checksum_validate(skb, proto, complete, \
3099 zero_okay, check, compute_pseudo) \
3101 __sum16 __ret = 0; \
3102 skb->csum_valid = 0; \
3103 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3104 __ret = __skb_checksum_validate_complete(skb, \
3105 complete, compute_pseudo(skb, proto)); \
3109 #define skb_checksum_init(skb, proto, compute_pseudo) \
3110 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3112 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3113 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3115 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3116 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3118 #define skb_checksum_validate_zero_check(skb, proto, check, \
3120 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3122 #define skb_checksum_simple_validate(skb) \
3123 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3125 static inline bool __skb_checksum_convert_check(struct sk_buff
*skb
)
3127 return (skb
->ip_summed
== CHECKSUM_NONE
&&
3128 skb
->csum_valid
&& !skb
->csum_bad
);
3131 static inline void __skb_checksum_convert(struct sk_buff
*skb
,
3132 __sum16 check
, __wsum pseudo
)
3134 skb
->csum
= ~pseudo
;
3135 skb
->ip_summed
= CHECKSUM_COMPLETE
;
3138 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3140 if (__skb_checksum_convert_check(skb)) \
3141 __skb_checksum_convert(skb, check, \
3142 compute_pseudo(skb, proto)); \
3145 static inline void skb_remcsum_adjust_partial(struct sk_buff
*skb
, void *ptr
,
3146 u16 start
, u16 offset
)
3148 skb
->ip_summed
= CHECKSUM_PARTIAL
;
3149 skb
->csum_start
= ((unsigned char *)ptr
+ start
) - skb
->head
;
3150 skb
->csum_offset
= offset
- start
;
3153 /* Update skbuf and packet to reflect the remote checksum offload operation.
3154 * When called, ptr indicates the starting point for skb->csum when
3155 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3156 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3158 static inline void skb_remcsum_process(struct sk_buff
*skb
, void *ptr
,
3159 int start
, int offset
, bool nopartial
)
3164 skb_remcsum_adjust_partial(skb
, ptr
, start
, offset
);
3168 if (unlikely(skb
->ip_summed
!= CHECKSUM_COMPLETE
)) {
3169 __skb_checksum_complete(skb
);
3170 skb_postpull_rcsum(skb
, skb
->data
, ptr
- (void *)skb
->data
);
3173 delta
= remcsum_adjust(ptr
, skb
->csum
, start
, offset
);
3175 /* Adjust skb->csum since we changed the packet */
3176 skb
->csum
= csum_add(skb
->csum
, delta
);
3179 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3180 void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
3181 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
3183 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
3184 nf_conntrack_destroy(nfct
);
3186 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
3189 atomic_inc(&nfct
->use
);
3192 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3193 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
3195 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
3198 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
3201 atomic_inc(&nf_bridge
->use
);
3203 #endif /* CONFIG_BRIDGE_NETFILTER */
3204 static inline void nf_reset(struct sk_buff
*skb
)
3206 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3207 nf_conntrack_put(skb
->nfct
);
3210 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3211 nf_bridge_put(skb
->nf_bridge
);
3212 skb
->nf_bridge
= NULL
;
3216 static inline void nf_reset_trace(struct sk_buff
*skb
)
3218 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3223 /* Note: This doesn't put any conntrack and bridge info in dst. */
3224 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
,
3227 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3228 dst
->nfct
= src
->nfct
;
3229 nf_conntrack_get(src
->nfct
);
3231 dst
->nfctinfo
= src
->nfctinfo
;
3233 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3234 dst
->nf_bridge
= src
->nf_bridge
;
3235 nf_bridge_get(src
->nf_bridge
);
3237 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3239 dst
->nf_trace
= src
->nf_trace
;
3243 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
3245 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3246 nf_conntrack_put(dst
->nfct
);
3248 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3249 nf_bridge_put(dst
->nf_bridge
);
3251 __nf_copy(dst
, src
, true);
3254 #ifdef CONFIG_NETWORK_SECMARK
3255 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3257 to
->secmark
= from
->secmark
;
3260 static inline void skb_init_secmark(struct sk_buff
*skb
)
3265 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3268 static inline void skb_init_secmark(struct sk_buff
*skb
)
3272 static inline bool skb_irq_freeable(const struct sk_buff
*skb
)
3274 return !skb
->destructor
&&
3275 #if IS_ENABLED(CONFIG_XFRM)
3278 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3281 !skb
->_skb_refdst
&&
3282 !skb_has_frag_list(skb
);
3285 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
3287 skb
->queue_mapping
= queue_mapping
;
3290 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
3292 return skb
->queue_mapping
;
3295 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
3297 to
->queue_mapping
= from
->queue_mapping
;
3300 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
3302 skb
->queue_mapping
= rx_queue
+ 1;
3305 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
3307 return skb
->queue_mapping
- 1;
3310 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
3312 return skb
->queue_mapping
!= 0;
3315 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
3324 /* Keeps track of mac header offset relative to skb->head.
3325 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3326 * For non-tunnel skb it points to skb_mac_header() and for
3327 * tunnel skb it points to outer mac header.
3328 * Keeps track of level of encapsulation of network headers.
3335 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3337 static inline int skb_tnl_header_len(const struct sk_buff
*inner_skb
)
3339 return (skb_mac_header(inner_skb
) - inner_skb
->head
) -
3340 SKB_GSO_CB(inner_skb
)->mac_offset
;
3343 static inline int gso_pskb_expand_head(struct sk_buff
*skb
, int extra
)
3345 int new_headroom
, headroom
;
3348 headroom
= skb_headroom(skb
);
3349 ret
= pskb_expand_head(skb
, extra
, 0, GFP_ATOMIC
);
3353 new_headroom
= skb_headroom(skb
);
3354 SKB_GSO_CB(skb
)->mac_offset
+= (new_headroom
- headroom
);
3358 /* Compute the checksum for a gso segment. First compute the checksum value
3359 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3360 * then add in skb->csum (checksum from csum_start to end of packet).
3361 * skb->csum and csum_start are then updated to reflect the checksum of the
3362 * resultant packet starting from the transport header-- the resultant checksum
3363 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3366 static inline __sum16
gso_make_checksum(struct sk_buff
*skb
, __wsum res
)
3368 int plen
= SKB_GSO_CB(skb
)->csum_start
- skb_headroom(skb
) -
3369 skb_transport_offset(skb
);
3372 partial
= csum_partial(skb_transport_header(skb
), plen
, skb
->csum
);
3374 SKB_GSO_CB(skb
)->csum_start
-= plen
;
3376 return csum_fold(partial
);
3379 static inline bool skb_is_gso(const struct sk_buff
*skb
)
3381 return skb_shinfo(skb
)->gso_size
;
3384 /* Note: Should be called only if skb_is_gso(skb) is true */
3385 static inline bool skb_is_gso_v6(const struct sk_buff
*skb
)
3387 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
3390 void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
3392 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
3394 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3395 * wanted then gso_type will be set. */
3396 const struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
3398 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
3399 unlikely(shinfo
->gso_type
== 0)) {
3400 __skb_warn_lro_forwarding(skb
);
3406 static inline void skb_forward_csum(struct sk_buff
*skb
)
3408 /* Unfortunately we don't support this one. Any brave souls? */
3409 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3410 skb
->ip_summed
= CHECKSUM_NONE
;
3414 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3415 * @skb: skb to check
3417 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3418 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3419 * use this helper, to document places where we make this assertion.
3421 static inline void skb_checksum_none_assert(const struct sk_buff
*skb
)
3424 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
3428 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
3430 int skb_checksum_setup(struct sk_buff
*skb
, bool recalculate
);
3431 struct sk_buff
*skb_checksum_trimmed(struct sk_buff
*skb
,
3432 unsigned int transport_len
,
3433 __sum16(*skb_chkf
)(struct sk_buff
*skb
));
3436 * skb_head_is_locked - Determine if the skb->head is locked down
3437 * @skb: skb to check
3439 * The head on skbs build around a head frag can be removed if they are
3440 * not cloned. This function returns true if the skb head is locked down
3441 * due to either being allocated via kmalloc, or by being a clone with
3442 * multiple references to the head.
3444 static inline bool skb_head_is_locked(const struct sk_buff
*skb
)
3446 return !skb
->head_frag
|| skb_cloned(skb
);
3450 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3454 * skb_gso_network_seglen is used to determine the real size of the
3455 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3457 * The MAC/L2 header is not accounted for.
3459 static inline unsigned int skb_gso_network_seglen(const struct sk_buff
*skb
)
3461 unsigned int hdr_len
= skb_transport_header(skb
) -
3462 skb_network_header(skb
);
3463 return hdr_len
+ skb_gso_transport_seglen(skb
);
3465 #endif /* __KERNEL__ */
3466 #endif /* _LINUX_SKBUFF_H */