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>
39 /* A. Checksumming of received packets by device.
43 * Device failed to checksum this packet e.g. due to lack of capabilities.
44 * The packet contains full (though not verified) checksum in packet but
45 * not in skb->csum. Thus, skb->csum is undefined in this case.
47 * CHECKSUM_UNNECESSARY:
49 * The hardware you're dealing with doesn't calculate the full checksum
50 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
51 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
52 * if their checksums are okay. skb->csum is still undefined in this case
53 * though. It is a bad option, but, unfortunately, nowadays most vendors do
54 * this. Apparently with the secret goal to sell you new devices, when you
55 * will add new protocol to your host, f.e. IPv6 8)
57 * CHECKSUM_UNNECESSARY is applicable to following protocols:
59 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
60 * zero UDP checksum for either IPv4 or IPv6, the networking stack
61 * may perform further validation in this case.
62 * GRE: only if the checksum is present in the header.
63 * SCTP: indicates the CRC in SCTP header has been validated.
65 * skb->csum_level indicates the number of consecutive checksums found in
66 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
67 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
68 * and a device is able to verify the checksums for UDP (possibly zero),
69 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
70 * two. If the device were only able to verify the UDP checksum and not
71 * GRE, either because it doesn't support GRE checksum of because GRE
72 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
73 * not considered in this case).
77 * This is the most generic way. The device supplied checksum of the _whole_
78 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
79 * hardware doesn't need to parse L3/L4 headers to implement this.
81 * Note: Even if device supports only some protocols, but is able to produce
82 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
86 * A checksum is set up to be offloaded to a device as described in the
87 * output description for CHECKSUM_PARTIAL. This may occur on a packet
88 * received directly from another Linux OS, e.g., a virtualized Linux kernel
89 * on the same host, or it may be set in the input path in GRO or remote
90 * checksum offload. For the purposes of checksum verification, the checksum
91 * referred to by skb->csum_start + skb->csum_offset and any preceding
92 * checksums in the packet are considered verified. Any checksums in the
93 * packet that are after the checksum being offloaded are not considered to
96 * B. Checksumming on output.
100 * The skb was already checksummed by the protocol, or a checksum is not
105 * The device is required to checksum the packet as seen by hard_start_xmit()
106 * from skb->csum_start up to the end, and to record/write the checksum at
107 * offset skb->csum_start + skb->csum_offset.
109 * The device must show its capabilities in dev->features, set up at device
110 * setup time, e.g. netdev_features.h:
112 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
113 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
114 * IPv4. Sigh. Vendors like this way for an unknown reason.
115 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
116 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
117 * NETIF_F_... - Well, you get the picture.
119 * CHECKSUM_UNNECESSARY:
121 * Normally, the device will do per protocol specific checksumming. Protocol
122 * implementations that do not want the NIC to perform the checksum
123 * calculation should use this flag in their outgoing skbs.
125 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
126 * offload. Correspondingly, the FCoE protocol driver
127 * stack should use CHECKSUM_UNNECESSARY.
129 * Any questions? No questions, good. --ANK
132 /* Don't change this without changing skb_csum_unnecessary! */
133 #define CHECKSUM_NONE 0
134 #define CHECKSUM_UNNECESSARY 1
135 #define CHECKSUM_COMPLETE 2
136 #define CHECKSUM_PARTIAL 3
138 /* Maximum value in skb->csum_level */
139 #define SKB_MAX_CSUM_LEVEL 3
141 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
142 #define SKB_WITH_OVERHEAD(X) \
143 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
144 #define SKB_MAX_ORDER(X, ORDER) \
145 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
146 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
147 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
149 /* return minimum truesize of one skb containing X bytes of data */
150 #define SKB_TRUESIZE(X) ((X) + \
151 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
152 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
156 struct pipe_inode_info
;
160 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
161 struct nf_conntrack
{
166 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
167 struct nf_bridge_info
{
170 BRNF_PROTO_UNCHANGED
,
176 struct net_device
*physindev
;
178 struct net_device
*physoutdev
;
179 char neigh_header
[8];
184 struct sk_buff_head
{
185 /* These two members must be first. */
186 struct sk_buff
*next
;
187 struct sk_buff
*prev
;
195 /* To allow 64K frame to be packed as single skb without frag_list we
196 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
197 * buffers which do not start on a page boundary.
199 * Since GRO uses frags we allocate at least 16 regardless of page
202 #if (65536/PAGE_SIZE + 1) < 16
203 #define MAX_SKB_FRAGS 16UL
205 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
208 typedef struct skb_frag_struct skb_frag_t
;
210 struct skb_frag_struct
{
214 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
223 static inline unsigned int skb_frag_size(const skb_frag_t
*frag
)
228 static inline void skb_frag_size_set(skb_frag_t
*frag
, unsigned int size
)
233 static inline void skb_frag_size_add(skb_frag_t
*frag
, int delta
)
238 static inline void skb_frag_size_sub(skb_frag_t
*frag
, int delta
)
243 #define HAVE_HW_TIME_STAMP
246 * struct skb_shared_hwtstamps - hardware time stamps
247 * @hwtstamp: hardware time stamp transformed into duration
248 * since arbitrary point in time
250 * Software time stamps generated by ktime_get_real() are stored in
253 * hwtstamps can only be compared against other hwtstamps from
256 * This structure is attached to packets as part of the
257 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
259 struct skb_shared_hwtstamps
{
263 /* Definitions for tx_flags in struct skb_shared_info */
265 /* generate hardware time stamp */
266 SKBTX_HW_TSTAMP
= 1 << 0,
268 /* generate software time stamp when queueing packet to NIC */
269 SKBTX_SW_TSTAMP
= 1 << 1,
271 /* device driver is going to provide hardware time stamp */
272 SKBTX_IN_PROGRESS
= 1 << 2,
274 /* device driver supports TX zero-copy buffers */
275 SKBTX_DEV_ZEROCOPY
= 1 << 3,
277 /* generate wifi status information (where possible) */
278 SKBTX_WIFI_STATUS
= 1 << 4,
280 /* This indicates at least one fragment might be overwritten
281 * (as in vmsplice(), sendfile() ...)
282 * If we need to compute a TX checksum, we'll need to copy
283 * all frags to avoid possible bad checksum
285 SKBTX_SHARED_FRAG
= 1 << 5,
287 /* generate software time stamp when entering packet scheduling */
288 SKBTX_SCHED_TSTAMP
= 1 << 6,
290 /* generate software timestamp on peer data acknowledgment */
291 SKBTX_ACK_TSTAMP
= 1 << 7,
294 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
295 SKBTX_SCHED_TSTAMP | \
297 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
300 * The callback notifies userspace to release buffers when skb DMA is done in
301 * lower device, the skb last reference should be 0 when calling this.
302 * The zerocopy_success argument is true if zero copy transmit occurred,
303 * false on data copy or out of memory error caused by data copy attempt.
304 * The ctx field is used to track device context.
305 * The desc field is used to track userspace buffer index.
308 void (*callback
)(struct ubuf_info
*, bool zerocopy_success
);
313 /* This data is invariant across clones and lives at
314 * the end of the header data, ie. at skb->end.
316 struct skb_shared_info
{
317 unsigned char nr_frags
;
319 unsigned short gso_size
;
320 /* Warning: this field is not always filled in (UFO)! */
321 unsigned short gso_segs
;
322 unsigned short gso_type
;
323 struct sk_buff
*frag_list
;
324 struct skb_shared_hwtstamps hwtstamps
;
329 * Warning : all fields before dataref are cleared in __alloc_skb()
333 /* Intermediate layers must ensure that destructor_arg
334 * remains valid until skb destructor */
335 void * destructor_arg
;
337 /* must be last field, see pskb_expand_head() */
338 skb_frag_t frags
[MAX_SKB_FRAGS
];
341 /* We divide dataref into two halves. The higher 16 bits hold references
342 * to the payload part of skb->data. The lower 16 bits hold references to
343 * the entire skb->data. A clone of a headerless skb holds the length of
344 * the header in skb->hdr_len.
346 * All users must obey the rule that the skb->data reference count must be
347 * greater than or equal to the payload reference count.
349 * Holding a reference to the payload part means that the user does not
350 * care about modifications to the header part of skb->data.
352 #define SKB_DATAREF_SHIFT 16
353 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
357 SKB_FCLONE_UNAVAILABLE
, /* skb has no fclone (from head_cache) */
358 SKB_FCLONE_ORIG
, /* orig skb (from fclone_cache) */
359 SKB_FCLONE_CLONE
, /* companion fclone skb (from fclone_cache) */
363 SKB_GSO_TCPV4
= 1 << 0,
364 SKB_GSO_UDP
= 1 << 1,
366 /* This indicates the skb is from an untrusted source. */
367 SKB_GSO_DODGY
= 1 << 2,
369 /* This indicates the tcp segment has CWR set. */
370 SKB_GSO_TCP_ECN
= 1 << 3,
372 SKB_GSO_TCPV6
= 1 << 4,
374 SKB_GSO_FCOE
= 1 << 5,
376 SKB_GSO_GRE
= 1 << 6,
378 SKB_GSO_GRE_CSUM
= 1 << 7,
380 SKB_GSO_IPIP
= 1 << 8,
382 SKB_GSO_SIT
= 1 << 9,
384 SKB_GSO_UDP_TUNNEL
= 1 << 10,
386 SKB_GSO_UDP_TUNNEL_CSUM
= 1 << 11,
388 SKB_GSO_TUNNEL_REMCSUM
= 1 << 12,
391 #if BITS_PER_LONG > 32
392 #define NET_SKBUFF_DATA_USES_OFFSET 1
395 #ifdef NET_SKBUFF_DATA_USES_OFFSET
396 typedef unsigned int sk_buff_data_t
;
398 typedef unsigned char *sk_buff_data_t
;
402 * struct skb_mstamp - multi resolution time stamps
403 * @stamp_us: timestamp in us resolution
404 * @stamp_jiffies: timestamp in jiffies
417 * skb_mstamp_get - get current timestamp
418 * @cl: place to store timestamps
420 static inline void skb_mstamp_get(struct skb_mstamp
*cl
)
422 u64 val
= local_clock();
424 do_div(val
, NSEC_PER_USEC
);
425 cl
->stamp_us
= (u32
)val
;
426 cl
->stamp_jiffies
= (u32
)jiffies
;
430 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
431 * @t1: pointer to newest sample
432 * @t0: pointer to oldest sample
434 static inline u32
skb_mstamp_us_delta(const struct skb_mstamp
*t1
,
435 const struct skb_mstamp
*t0
)
437 s32 delta_us
= t1
->stamp_us
- t0
->stamp_us
;
438 u32 delta_jiffies
= t1
->stamp_jiffies
- t0
->stamp_jiffies
;
440 /* If delta_us is negative, this might be because interval is too big,
441 * or local_clock() drift is too big : fallback using jiffies.
444 delta_jiffies
>= (INT_MAX
/ (USEC_PER_SEC
/ HZ
)))
446 delta_us
= jiffies_to_usecs(delta_jiffies
);
453 * struct sk_buff - socket buffer
454 * @next: Next buffer in list
455 * @prev: Previous buffer in list
456 * @tstamp: Time we arrived/left
457 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
458 * @sk: Socket we are owned by
459 * @dev: Device we arrived on/are leaving by
460 * @cb: Control buffer. Free for use by every layer. Put private vars here
461 * @_skb_refdst: destination entry (with norefcount bit)
462 * @sp: the security path, used for xfrm
463 * @len: Length of actual data
464 * @data_len: Data length
465 * @mac_len: Length of link layer header
466 * @hdr_len: writable header length of cloned skb
467 * @csum: Checksum (must include start/offset pair)
468 * @csum_start: Offset from skb->head where checksumming should start
469 * @csum_offset: Offset from csum_start where checksum should be stored
470 * @priority: Packet queueing priority
471 * @ignore_df: allow local fragmentation
472 * @cloned: Head may be cloned (check refcnt to be sure)
473 * @ip_summed: Driver fed us an IP checksum
474 * @nohdr: Payload reference only, must not modify header
475 * @nfctinfo: Relationship of this skb to the connection
476 * @pkt_type: Packet class
477 * @fclone: skbuff clone status
478 * @ipvs_property: skbuff is owned by ipvs
479 * @peeked: this packet has been seen already, so stats have been
480 * done for it, don't do them again
481 * @nf_trace: netfilter packet trace flag
482 * @protocol: Packet protocol from driver
483 * @destructor: Destruct function
484 * @nfct: Associated connection, if any
485 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
486 * @skb_iif: ifindex of device we arrived on
487 * @tc_index: Traffic control index
488 * @tc_verd: traffic control verdict
489 * @hash: the packet hash
490 * @queue_mapping: Queue mapping for multiqueue devices
491 * @xmit_more: More SKBs are pending for this queue
492 * @ndisc_nodetype: router type (from link layer)
493 * @ooo_okay: allow the mapping of a socket to a queue to be changed
494 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
496 * @sw_hash: indicates hash was computed in software stack
497 * @wifi_acked_valid: wifi_acked was set
498 * @wifi_acked: whether frame was acked on wifi or not
499 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
500 * @napi_id: id of the NAPI struct this skb came from
501 * @secmark: security marking
502 * @mark: Generic packet mark
503 * @vlan_proto: vlan encapsulation protocol
504 * @vlan_tci: vlan tag control information
505 * @inner_protocol: Protocol (encapsulation)
506 * @inner_transport_header: Inner transport layer header (encapsulation)
507 * @inner_network_header: Network layer header (encapsulation)
508 * @inner_mac_header: Link layer header (encapsulation)
509 * @transport_header: Transport layer header
510 * @network_header: Network layer header
511 * @mac_header: Link layer header
512 * @tail: Tail pointer
514 * @head: Head of buffer
515 * @data: Data head pointer
516 * @truesize: Buffer size
517 * @users: User count - see {datagram,tcp}.c
523 /* These two members must be first. */
524 struct sk_buff
*next
;
525 struct sk_buff
*prev
;
529 struct skb_mstamp skb_mstamp
;
532 struct rb_node rbnode
; /* used in netem & tcp stack */
535 struct net_device
*dev
;
538 * This is the control buffer. It is free to use for every
539 * layer. Please put your private variables there. If you
540 * want to keep them across layers you have to do a skb_clone()
541 * first. This is owned by whoever has the skb queued ATM.
543 char cb
[48] __aligned(8);
545 unsigned long _skb_refdst
;
546 void (*destructor
)(struct sk_buff
*skb
);
550 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
551 struct nf_conntrack
*nfct
;
553 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
554 struct nf_bridge_info
*nf_bridge
;
561 /* Following fields are _not_ copied in __copy_skb_header()
562 * Note that queue_mapping is here mostly to fill a hole.
564 kmemcheck_bitfield_begin(flags1
);
573 kmemcheck_bitfield_end(flags1
);
575 /* fields enclosed in headers_start/headers_end are copied
576 * using a single memcpy() in __copy_skb_header()
579 __u32 headers_start
[0];
582 /* if you move pkt_type around you also must adapt those constants */
583 #ifdef __BIG_ENDIAN_BITFIELD
584 #define PKT_TYPE_MAX (7 << 5)
586 #define PKT_TYPE_MAX 7
588 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
590 __u8 __pkt_type_offset
[0];
601 __u8 wifi_acked_valid
:1;
605 /* Indicates the inner headers are valid in the skbuff. */
606 __u8 encapsulation
:1;
607 __u8 encap_hdr_csum
:1;
609 __u8 csum_complete_sw
:1;
613 #ifdef CONFIG_IPV6_NDISC_NODETYPE
614 __u8 ndisc_nodetype
:2;
616 __u8 ipvs_property
:1;
617 __u8 inner_protocol_type
:1;
618 __u8 remcsum_offload
:1;
619 /* 3 or 5 bit hole */
621 #ifdef CONFIG_NET_SCHED
622 __u16 tc_index
; /* traffic control index */
623 #ifdef CONFIG_NET_CLS_ACT
624 __u16 tc_verd
; /* traffic control verdict */
640 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
642 unsigned int napi_id
;
643 unsigned int sender_cpu
;
646 #ifdef CONFIG_NETWORK_SECMARK
651 __u32 reserved_tailroom
;
655 __be16 inner_protocol
;
659 __u16 inner_transport_header
;
660 __u16 inner_network_header
;
661 __u16 inner_mac_header
;
664 __u16 transport_header
;
665 __u16 network_header
;
669 __u32 headers_end
[0];
672 /* These elements must be at the end, see alloc_skb() for details. */
677 unsigned int truesize
;
683 * Handling routines are only of interest to the kernel
685 #include <linux/slab.h>
688 #define SKB_ALLOC_FCLONE 0x01
689 #define SKB_ALLOC_RX 0x02
690 #define SKB_ALLOC_NAPI 0x04
692 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
693 static inline bool skb_pfmemalloc(const struct sk_buff
*skb
)
695 return unlikely(skb
->pfmemalloc
);
699 * skb might have a dst pointer attached, refcounted or not.
700 * _skb_refdst low order bit is set if refcount was _not_ taken
702 #define SKB_DST_NOREF 1UL
703 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
706 * skb_dst - returns skb dst_entry
709 * Returns skb dst_entry, regardless of reference taken or not.
711 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
713 /* If refdst was not refcounted, check we still are in a
714 * rcu_read_lock section
716 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
717 !rcu_read_lock_held() &&
718 !rcu_read_lock_bh_held());
719 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
723 * skb_dst_set - sets skb dst
727 * Sets skb dst, assuming a reference was taken on dst and should
728 * be released by skb_dst_drop()
730 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
732 skb
->_skb_refdst
= (unsigned long)dst
;
736 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
740 * Sets skb dst, assuming a reference was not taken on dst.
741 * If dst entry is cached, we do not take reference and dst_release
742 * will be avoided by refdst_drop. If dst entry is not cached, we take
743 * reference, so that last dst_release can destroy the dst immediately.
745 static inline void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
)
747 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
748 skb
->_skb_refdst
= (unsigned long)dst
| SKB_DST_NOREF
;
752 * skb_dst_is_noref - Test if skb dst isn't refcounted
755 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
757 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
760 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
762 return (struct rtable
*)skb_dst(skb
);
765 void kfree_skb(struct sk_buff
*skb
);
766 void kfree_skb_list(struct sk_buff
*segs
);
767 void skb_tx_error(struct sk_buff
*skb
);
768 void consume_skb(struct sk_buff
*skb
);
769 void __kfree_skb(struct sk_buff
*skb
);
770 extern struct kmem_cache
*skbuff_head_cache
;
772 void kfree_skb_partial(struct sk_buff
*skb
, bool head_stolen
);
773 bool skb_try_coalesce(struct sk_buff
*to
, struct sk_buff
*from
,
774 bool *fragstolen
, int *delta_truesize
);
776 struct sk_buff
*__alloc_skb(unsigned int size
, gfp_t priority
, int flags
,
778 struct sk_buff
*__build_skb(void *data
, unsigned int frag_size
);
779 struct sk_buff
*build_skb(void *data
, unsigned int frag_size
);
780 static inline struct sk_buff
*alloc_skb(unsigned int size
,
783 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
786 struct sk_buff
*alloc_skb_with_frags(unsigned long header_len
,
787 unsigned long data_len
,
792 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
793 struct sk_buff_fclones
{
802 * skb_fclone_busy - check if fclone is busy
805 * Returns true is skb is a fast clone, and its clone is not freed.
806 * Some drivers call skb_orphan() in their ndo_start_xmit(),
807 * so we also check that this didnt happen.
809 static inline bool skb_fclone_busy(const struct sock
*sk
,
810 const struct sk_buff
*skb
)
812 const struct sk_buff_fclones
*fclones
;
814 fclones
= container_of(skb
, struct sk_buff_fclones
, skb1
);
816 return skb
->fclone
== SKB_FCLONE_ORIG
&&
817 atomic_read(&fclones
->fclone_ref
) > 1 &&
818 fclones
->skb2
.sk
== sk
;
821 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
824 return __alloc_skb(size
, priority
, SKB_ALLOC_FCLONE
, NUMA_NO_NODE
);
827 struct sk_buff
*__alloc_skb_head(gfp_t priority
, int node
);
828 static inline struct sk_buff
*alloc_skb_head(gfp_t priority
)
830 return __alloc_skb_head(priority
, -1);
833 struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
834 int skb_copy_ubufs(struct sk_buff
*skb
, gfp_t gfp_mask
);
835 struct sk_buff
*skb_clone(struct sk_buff
*skb
, gfp_t priority
);
836 struct sk_buff
*skb_copy(const struct sk_buff
*skb
, gfp_t priority
);
837 struct sk_buff
*__pskb_copy_fclone(struct sk_buff
*skb
, int headroom
,
838 gfp_t gfp_mask
, bool fclone
);
839 static inline struct sk_buff
*__pskb_copy(struct sk_buff
*skb
, int headroom
,
842 return __pskb_copy_fclone(skb
, headroom
, gfp_mask
, false);
845 int pskb_expand_head(struct sk_buff
*skb
, int nhead
, int ntail
, gfp_t gfp_mask
);
846 struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
847 unsigned int headroom
);
848 struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
, int newheadroom
,
849 int newtailroom
, gfp_t priority
);
850 int skb_to_sgvec_nomark(struct sk_buff
*skb
, struct scatterlist
*sg
,
851 int offset
, int len
);
852 int skb_to_sgvec(struct sk_buff
*skb
, struct scatterlist
*sg
, int offset
,
854 int skb_cow_data(struct sk_buff
*skb
, int tailbits
, struct sk_buff
**trailer
);
855 int skb_pad(struct sk_buff
*skb
, int pad
);
856 #define dev_kfree_skb(a) consume_skb(a)
858 int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
859 int getfrag(void *from
, char *to
, int offset
,
860 int len
, int odd
, struct sk_buff
*skb
),
861 void *from
, int length
);
863 struct skb_seq_state
{
867 __u32 stepped_offset
;
868 struct sk_buff
*root_skb
;
869 struct sk_buff
*cur_skb
;
873 void skb_prepare_seq_read(struct sk_buff
*skb
, unsigned int from
,
874 unsigned int to
, struct skb_seq_state
*st
);
875 unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
876 struct skb_seq_state
*st
);
877 void skb_abort_seq_read(struct skb_seq_state
*st
);
879 unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
880 unsigned int to
, struct ts_config
*config
);
883 * Packet hash types specify the type of hash in skb_set_hash.
885 * Hash types refer to the protocol layer addresses which are used to
886 * construct a packet's hash. The hashes are used to differentiate or identify
887 * flows of the protocol layer for the hash type. Hash types are either
888 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
890 * Properties of hashes:
892 * 1) Two packets in different flows have different hash values
893 * 2) Two packets in the same flow should have the same hash value
895 * A hash at a higher layer is considered to be more specific. A driver should
896 * set the most specific hash possible.
898 * A driver cannot indicate a more specific hash than the layer at which a hash
899 * was computed. For instance an L3 hash cannot be set as an L4 hash.
901 * A driver may indicate a hash level which is less specific than the
902 * actual layer the hash was computed on. For instance, a hash computed
903 * at L4 may be considered an L3 hash. This should only be done if the
904 * driver can't unambiguously determine that the HW computed the hash at
905 * the higher layer. Note that the "should" in the second property above
908 enum pkt_hash_types
{
909 PKT_HASH_TYPE_NONE
, /* Undefined type */
910 PKT_HASH_TYPE_L2
, /* Input: src_MAC, dest_MAC */
911 PKT_HASH_TYPE_L3
, /* Input: src_IP, dst_IP */
912 PKT_HASH_TYPE_L4
, /* Input: src_IP, dst_IP, src_port, dst_port */
916 skb_set_hash(struct sk_buff
*skb
, __u32 hash
, enum pkt_hash_types type
)
918 skb
->l4_hash
= (type
== PKT_HASH_TYPE_L4
);
923 static inline __u32
skb_get_hash(struct sk_buff
*skb
)
925 if (!skb
->l4_hash
&& !skb
->sw_hash
)
931 __u32
skb_get_hash_perturb(const struct sk_buff
*skb
, u32 perturb
);
933 static inline __u32
skb_get_hash_raw(const struct sk_buff
*skb
)
938 static inline void skb_clear_hash(struct sk_buff
*skb
)
945 static inline void skb_clear_hash_if_not_l4(struct sk_buff
*skb
)
951 static inline void skb_copy_hash(struct sk_buff
*to
, const struct sk_buff
*from
)
953 to
->hash
= from
->hash
;
954 to
->sw_hash
= from
->sw_hash
;
955 to
->l4_hash
= from
->l4_hash
;
958 static inline void skb_sender_cpu_clear(struct sk_buff
*skb
)
965 #ifdef NET_SKBUFF_DATA_USES_OFFSET
966 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
968 return skb
->head
+ skb
->end
;
971 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
976 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
981 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
983 return skb
->end
- skb
->head
;
988 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
990 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
992 return &skb_shinfo(skb
)->hwtstamps
;
996 * skb_queue_empty - check if a queue is empty
999 * Returns true if the queue is empty, false otherwise.
1001 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
1003 return list
->next
== (const struct sk_buff
*) list
;
1007 * skb_queue_is_last - check if skb is the last entry in the queue
1011 * Returns true if @skb is the last buffer on the list.
1013 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
1014 const struct sk_buff
*skb
)
1016 return skb
->next
== (const struct sk_buff
*) list
;
1020 * skb_queue_is_first - check if skb is the first entry in the queue
1024 * Returns true if @skb is the first buffer on the list.
1026 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
1027 const struct sk_buff
*skb
)
1029 return skb
->prev
== (const struct sk_buff
*) list
;
1033 * skb_queue_next - return the next packet in the queue
1035 * @skb: current buffer
1037 * Return the next packet in @list after @skb. It is only valid to
1038 * call this if skb_queue_is_last() evaluates to false.
1040 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
1041 const struct sk_buff
*skb
)
1043 /* This BUG_ON may seem severe, but if we just return then we
1044 * are going to dereference garbage.
1046 BUG_ON(skb_queue_is_last(list
, skb
));
1051 * skb_queue_prev - return the prev packet in the queue
1053 * @skb: current buffer
1055 * Return the prev packet in @list before @skb. It is only valid to
1056 * call this if skb_queue_is_first() evaluates to false.
1058 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
1059 const struct sk_buff
*skb
)
1061 /* This BUG_ON may seem severe, but if we just return then we
1062 * are going to dereference garbage.
1064 BUG_ON(skb_queue_is_first(list
, skb
));
1069 * skb_get - reference buffer
1070 * @skb: buffer to reference
1072 * Makes another reference to a socket buffer and returns a pointer
1075 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
1077 atomic_inc(&skb
->users
);
1082 * If users == 1, we are the only owner and are can avoid redundant
1087 * skb_cloned - is the buffer a clone
1088 * @skb: buffer to check
1090 * Returns true if the buffer was generated with skb_clone() and is
1091 * one of multiple shared copies of the buffer. Cloned buffers are
1092 * shared data so must not be written to under normal circumstances.
1094 static inline int skb_cloned(const struct sk_buff
*skb
)
1096 return skb
->cloned
&&
1097 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
1100 static inline int skb_unclone(struct sk_buff
*skb
, gfp_t pri
)
1102 might_sleep_if(pri
& __GFP_WAIT
);
1104 if (skb_cloned(skb
))
1105 return pskb_expand_head(skb
, 0, 0, pri
);
1111 * skb_header_cloned - is the header a clone
1112 * @skb: buffer to check
1114 * Returns true if modifying the header part of the buffer requires
1115 * the data to be copied.
1117 static inline int skb_header_cloned(const struct sk_buff
*skb
)
1124 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
1125 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
1126 return dataref
!= 1;
1130 * skb_header_release - release reference to header
1131 * @skb: buffer to operate on
1133 * Drop a reference to the header part of the buffer. This is done
1134 * by acquiring a payload reference. You must not read from the header
1135 * part of skb->data after this.
1136 * Note : Check if you can use __skb_header_release() instead.
1138 static inline void skb_header_release(struct sk_buff
*skb
)
1142 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
1146 * __skb_header_release - release reference to header
1147 * @skb: buffer to operate on
1149 * Variant of skb_header_release() assuming skb is private to caller.
1150 * We can avoid one atomic operation.
1152 static inline void __skb_header_release(struct sk_buff
*skb
)
1155 atomic_set(&skb_shinfo(skb
)->dataref
, 1 + (1 << SKB_DATAREF_SHIFT
));
1160 * skb_shared - is the buffer shared
1161 * @skb: buffer to check
1163 * Returns true if more than one person has a reference to this
1166 static inline int skb_shared(const struct sk_buff
*skb
)
1168 return atomic_read(&skb
->users
) != 1;
1172 * skb_share_check - check if buffer is shared and if so clone it
1173 * @skb: buffer to check
1174 * @pri: priority for memory allocation
1176 * If the buffer is shared the buffer is cloned and the old copy
1177 * drops a reference. A new clone with a single reference is returned.
1178 * If the buffer is not shared the original buffer is returned. When
1179 * being called from interrupt status or with spinlocks held pri must
1182 * NULL is returned on a memory allocation failure.
1184 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
, gfp_t pri
)
1186 might_sleep_if(pri
& __GFP_WAIT
);
1187 if (skb_shared(skb
)) {
1188 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
1200 * Copy shared buffers into a new sk_buff. We effectively do COW on
1201 * packets to handle cases where we have a local reader and forward
1202 * and a couple of other messy ones. The normal one is tcpdumping
1203 * a packet thats being forwarded.
1207 * skb_unshare - make a copy of a shared buffer
1208 * @skb: buffer to check
1209 * @pri: priority for memory allocation
1211 * If the socket buffer is a clone then this function creates a new
1212 * copy of the data, drops a reference count on the old copy and returns
1213 * the new copy with the reference count at 1. If the buffer is not a clone
1214 * the original buffer is returned. When called with a spinlock held or
1215 * from interrupt state @pri must be %GFP_ATOMIC
1217 * %NULL is returned on a memory allocation failure.
1219 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
1222 might_sleep_if(pri
& __GFP_WAIT
);
1223 if (skb_cloned(skb
)) {
1224 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
1226 /* Free our shared copy */
1237 * skb_peek - peek at the head of an &sk_buff_head
1238 * @list_: list to peek at
1240 * Peek an &sk_buff. Unlike most other operations you _MUST_
1241 * be careful with this one. A peek leaves the buffer on the
1242 * list and someone else may run off with it. You must hold
1243 * the appropriate locks or have a private queue to do this.
1245 * Returns %NULL for an empty list or a pointer to the head element.
1246 * The reference count is not incremented and the reference is therefore
1247 * volatile. Use with caution.
1249 static inline struct sk_buff
*skb_peek(const struct sk_buff_head
*list_
)
1251 struct sk_buff
*skb
= list_
->next
;
1253 if (skb
== (struct sk_buff
*)list_
)
1259 * skb_peek_next - peek skb following the given one from a queue
1260 * @skb: skb to start from
1261 * @list_: list to peek at
1263 * Returns %NULL when the end of the list is met or a pointer to the
1264 * next element. The reference count is not incremented and the
1265 * reference is therefore volatile. Use with caution.
1267 static inline struct sk_buff
*skb_peek_next(struct sk_buff
*skb
,
1268 const struct sk_buff_head
*list_
)
1270 struct sk_buff
*next
= skb
->next
;
1272 if (next
== (struct sk_buff
*)list_
)
1278 * skb_peek_tail - peek at the tail of an &sk_buff_head
1279 * @list_: list to peek at
1281 * Peek an &sk_buff. Unlike most other operations you _MUST_
1282 * be careful with this one. A peek leaves the buffer on the
1283 * list and someone else may run off with it. You must hold
1284 * the appropriate locks or have a private queue to do this.
1286 * Returns %NULL for an empty list or a pointer to the tail element.
1287 * The reference count is not incremented and the reference is therefore
1288 * volatile. Use with caution.
1290 static inline struct sk_buff
*skb_peek_tail(const struct sk_buff_head
*list_
)
1292 struct sk_buff
*skb
= list_
->prev
;
1294 if (skb
== (struct sk_buff
*)list_
)
1301 * skb_queue_len - get queue length
1302 * @list_: list to measure
1304 * Return the length of an &sk_buff queue.
1306 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
1312 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1313 * @list: queue to initialize
1315 * This initializes only the list and queue length aspects of
1316 * an sk_buff_head object. This allows to initialize the list
1317 * aspects of an sk_buff_head without reinitializing things like
1318 * the spinlock. It can also be used for on-stack sk_buff_head
1319 * objects where the spinlock is known to not be used.
1321 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
1323 list
->prev
= list
->next
= (struct sk_buff
*)list
;
1328 * This function creates a split out lock class for each invocation;
1329 * this is needed for now since a whole lot of users of the skb-queue
1330 * infrastructure in drivers have different locking usage (in hardirq)
1331 * than the networking core (in softirq only). In the long run either the
1332 * network layer or drivers should need annotation to consolidate the
1333 * main types of usage into 3 classes.
1335 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
1337 spin_lock_init(&list
->lock
);
1338 __skb_queue_head_init(list
);
1341 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
1342 struct lock_class_key
*class)
1344 skb_queue_head_init(list
);
1345 lockdep_set_class(&list
->lock
, class);
1349 * Insert an sk_buff on a list.
1351 * The "__skb_xxxx()" functions are the non-atomic ones that
1352 * can only be called with interrupts disabled.
1354 void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
,
1355 struct sk_buff_head
*list
);
1356 static inline void __skb_insert(struct sk_buff
*newsk
,
1357 struct sk_buff
*prev
, struct sk_buff
*next
,
1358 struct sk_buff_head
*list
)
1362 next
->prev
= prev
->next
= newsk
;
1366 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
1367 struct sk_buff
*prev
,
1368 struct sk_buff
*next
)
1370 struct sk_buff
*first
= list
->next
;
1371 struct sk_buff
*last
= list
->prev
;
1381 * skb_queue_splice - join two skb lists, this is designed for stacks
1382 * @list: the new list to add
1383 * @head: the place to add it in the first list
1385 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
1386 struct sk_buff_head
*head
)
1388 if (!skb_queue_empty(list
)) {
1389 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1390 head
->qlen
+= list
->qlen
;
1395 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1396 * @list: the new list to add
1397 * @head: the place to add it in the first list
1399 * The list at @list is reinitialised
1401 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
1402 struct sk_buff_head
*head
)
1404 if (!skb_queue_empty(list
)) {
1405 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1406 head
->qlen
+= list
->qlen
;
1407 __skb_queue_head_init(list
);
1412 * skb_queue_splice_tail - join two skb lists, each list being a queue
1413 * @list: the new list to add
1414 * @head: the place to add it in the first list
1416 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
1417 struct sk_buff_head
*head
)
1419 if (!skb_queue_empty(list
)) {
1420 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1421 head
->qlen
+= list
->qlen
;
1426 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1427 * @list: the new list to add
1428 * @head: the place to add it in the first list
1430 * Each of the lists is a queue.
1431 * The list at @list is reinitialised
1433 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
1434 struct sk_buff_head
*head
)
1436 if (!skb_queue_empty(list
)) {
1437 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1438 head
->qlen
+= list
->qlen
;
1439 __skb_queue_head_init(list
);
1444 * __skb_queue_after - queue a buffer at the list head
1445 * @list: list to use
1446 * @prev: place after this buffer
1447 * @newsk: buffer to queue
1449 * Queue a buffer int the middle of a list. This function takes no locks
1450 * and you must therefore hold required locks before calling it.
1452 * A buffer cannot be placed on two lists at the same time.
1454 static inline void __skb_queue_after(struct sk_buff_head
*list
,
1455 struct sk_buff
*prev
,
1456 struct sk_buff
*newsk
)
1458 __skb_insert(newsk
, prev
, prev
->next
, list
);
1461 void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
1462 struct sk_buff_head
*list
);
1464 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1465 struct sk_buff
*next
,
1466 struct sk_buff
*newsk
)
1468 __skb_insert(newsk
, next
->prev
, next
, list
);
1472 * __skb_queue_head - queue a buffer at the list head
1473 * @list: list to use
1474 * @newsk: buffer to queue
1476 * Queue a buffer at the start of a list. This function takes no locks
1477 * and you must therefore hold required locks before calling it.
1479 * A buffer cannot be placed on two lists at the same time.
1481 void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1482 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1483 struct sk_buff
*newsk
)
1485 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1489 * __skb_queue_tail - queue a buffer at the list tail
1490 * @list: list to use
1491 * @newsk: buffer to queue
1493 * Queue a buffer at the end of a list. This function takes no locks
1494 * and you must therefore hold required locks before calling it.
1496 * A buffer cannot be placed on two lists at the same time.
1498 void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1499 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1500 struct sk_buff
*newsk
)
1502 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1506 * remove sk_buff from list. _Must_ be called atomically, and with
1509 void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1510 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1512 struct sk_buff
*next
, *prev
;
1517 skb
->next
= skb
->prev
= NULL
;
1523 * __skb_dequeue - remove from the head of the queue
1524 * @list: list to dequeue from
1526 * Remove the head of the list. This function does not take any locks
1527 * so must be used with appropriate locks held only. The head item is
1528 * returned or %NULL if the list is empty.
1530 struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1531 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1533 struct sk_buff
*skb
= skb_peek(list
);
1535 __skb_unlink(skb
, list
);
1540 * __skb_dequeue_tail - remove from the tail of the queue
1541 * @list: list to dequeue from
1543 * Remove the tail of the list. This function does not take any locks
1544 * so must be used with appropriate locks held only. The tail item is
1545 * returned or %NULL if the list is empty.
1547 struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1548 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1550 struct sk_buff
*skb
= skb_peek_tail(list
);
1552 __skb_unlink(skb
, list
);
1557 static inline bool skb_is_nonlinear(const struct sk_buff
*skb
)
1559 return skb
->data_len
;
1562 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1564 return skb
->len
- skb
->data_len
;
1567 static inline int skb_pagelen(const struct sk_buff
*skb
)
1571 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1572 len
+= skb_frag_size(&skb_shinfo(skb
)->frags
[i
]);
1573 return len
+ skb_headlen(skb
);
1577 * __skb_fill_page_desc - initialise a paged fragment in an skb
1578 * @skb: buffer containing fragment to be initialised
1579 * @i: paged fragment index to initialise
1580 * @page: the page to use for this fragment
1581 * @off: the offset to the data with @page
1582 * @size: the length of the data
1584 * Initialises the @i'th fragment of @skb to point to &size bytes at
1585 * offset @off within @page.
1587 * Does not take any additional reference on the fragment.
1589 static inline void __skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1590 struct page
*page
, int off
, int size
)
1592 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1595 * Propagate page->pfmemalloc to the skb if we can. The problem is
1596 * that not all callers have unique ownership of the page. If
1597 * pfmemalloc is set, we check the mapping as a mapping implies
1598 * page->index is set (index and pfmemalloc share space).
1599 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1600 * do not lose pfmemalloc information as the pages would not be
1601 * allocated using __GFP_MEMALLOC.
1603 frag
->page
.p
= page
;
1604 frag
->page_offset
= off
;
1605 skb_frag_size_set(frag
, size
);
1607 page
= compound_head(page
);
1608 if (page
->pfmemalloc
&& !page
->mapping
)
1609 skb
->pfmemalloc
= true;
1613 * skb_fill_page_desc - initialise a paged fragment in an skb
1614 * @skb: buffer containing fragment to be initialised
1615 * @i: paged fragment index to initialise
1616 * @page: the page to use for this fragment
1617 * @off: the offset to the data with @page
1618 * @size: the length of the data
1620 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1621 * @skb to point to @size bytes at offset @off within @page. In
1622 * addition updates @skb such that @i is the last fragment.
1624 * Does not take any additional reference on the fragment.
1626 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1627 struct page
*page
, int off
, int size
)
1629 __skb_fill_page_desc(skb
, i
, page
, off
, size
);
1630 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1633 void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
, int off
,
1634 int size
, unsigned int truesize
);
1636 void skb_coalesce_rx_frag(struct sk_buff
*skb
, int i
, int size
,
1637 unsigned int truesize
);
1639 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1640 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1641 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1643 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1644 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1646 return skb
->head
+ skb
->tail
;
1649 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1651 skb
->tail
= skb
->data
- skb
->head
;
1654 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1656 skb_reset_tail_pointer(skb
);
1657 skb
->tail
+= offset
;
1660 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1661 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1666 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1668 skb
->tail
= skb
->data
;
1671 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1673 skb
->tail
= skb
->data
+ offset
;
1676 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1679 * Add data to an sk_buff
1681 unsigned char *pskb_put(struct sk_buff
*skb
, struct sk_buff
*tail
, int len
);
1682 unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1683 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1685 unsigned char *tmp
= skb_tail_pointer(skb
);
1686 SKB_LINEAR_ASSERT(skb
);
1692 unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1693 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1700 unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1701 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1704 BUG_ON(skb
->len
< skb
->data_len
);
1705 return skb
->data
+= len
;
1708 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1710 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1713 unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1715 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1717 if (len
> skb_headlen(skb
) &&
1718 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1721 return skb
->data
+= len
;
1724 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1726 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1729 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1731 if (likely(len
<= skb_headlen(skb
)))
1733 if (unlikely(len
> skb
->len
))
1735 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1739 * skb_headroom - bytes at buffer head
1740 * @skb: buffer to check
1742 * Return the number of bytes of free space at the head of an &sk_buff.
1744 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1746 return skb
->data
- skb
->head
;
1750 * skb_tailroom - bytes at buffer end
1751 * @skb: buffer to check
1753 * Return the number of bytes of free space at the tail of an sk_buff
1755 static inline int skb_tailroom(const struct sk_buff
*skb
)
1757 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1761 * skb_availroom - bytes at buffer end
1762 * @skb: buffer to check
1764 * Return the number of bytes of free space at the tail of an sk_buff
1765 * allocated by sk_stream_alloc()
1767 static inline int skb_availroom(const struct sk_buff
*skb
)
1769 if (skb_is_nonlinear(skb
))
1772 return skb
->end
- skb
->tail
- skb
->reserved_tailroom
;
1776 * skb_reserve - adjust headroom
1777 * @skb: buffer to alter
1778 * @len: bytes to move
1780 * Increase the headroom of an empty &sk_buff by reducing the tail
1781 * room. This is only allowed for an empty buffer.
1783 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
1789 #define ENCAP_TYPE_ETHER 0
1790 #define ENCAP_TYPE_IPPROTO 1
1792 static inline void skb_set_inner_protocol(struct sk_buff
*skb
,
1795 skb
->inner_protocol
= protocol
;
1796 skb
->inner_protocol_type
= ENCAP_TYPE_ETHER
;
1799 static inline void skb_set_inner_ipproto(struct sk_buff
*skb
,
1802 skb
->inner_ipproto
= ipproto
;
1803 skb
->inner_protocol_type
= ENCAP_TYPE_IPPROTO
;
1806 static inline void skb_reset_inner_headers(struct sk_buff
*skb
)
1808 skb
->inner_mac_header
= skb
->mac_header
;
1809 skb
->inner_network_header
= skb
->network_header
;
1810 skb
->inner_transport_header
= skb
->transport_header
;
1813 static inline void skb_reset_mac_len(struct sk_buff
*skb
)
1815 skb
->mac_len
= skb
->network_header
- skb
->mac_header
;
1818 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1821 return skb
->head
+ skb
->inner_transport_header
;
1824 static inline void skb_reset_inner_transport_header(struct sk_buff
*skb
)
1826 skb
->inner_transport_header
= skb
->data
- skb
->head
;
1829 static inline void skb_set_inner_transport_header(struct sk_buff
*skb
,
1832 skb_reset_inner_transport_header(skb
);
1833 skb
->inner_transport_header
+= offset
;
1836 static inline unsigned char *skb_inner_network_header(const struct sk_buff
*skb
)
1838 return skb
->head
+ skb
->inner_network_header
;
1841 static inline void skb_reset_inner_network_header(struct sk_buff
*skb
)
1843 skb
->inner_network_header
= skb
->data
- skb
->head
;
1846 static inline void skb_set_inner_network_header(struct sk_buff
*skb
,
1849 skb_reset_inner_network_header(skb
);
1850 skb
->inner_network_header
+= offset
;
1853 static inline unsigned char *skb_inner_mac_header(const struct sk_buff
*skb
)
1855 return skb
->head
+ skb
->inner_mac_header
;
1858 static inline void skb_reset_inner_mac_header(struct sk_buff
*skb
)
1860 skb
->inner_mac_header
= skb
->data
- skb
->head
;
1863 static inline void skb_set_inner_mac_header(struct sk_buff
*skb
,
1866 skb_reset_inner_mac_header(skb
);
1867 skb
->inner_mac_header
+= offset
;
1869 static inline bool skb_transport_header_was_set(const struct sk_buff
*skb
)
1871 return skb
->transport_header
!= (typeof(skb
->transport_header
))~0U;
1874 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1876 return skb
->head
+ skb
->transport_header
;
1879 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1881 skb
->transport_header
= skb
->data
- skb
->head
;
1884 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1887 skb_reset_transport_header(skb
);
1888 skb
->transport_header
+= offset
;
1891 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1893 return skb
->head
+ skb
->network_header
;
1896 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1898 skb
->network_header
= skb
->data
- skb
->head
;
1901 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1903 skb_reset_network_header(skb
);
1904 skb
->network_header
+= offset
;
1907 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1909 return skb
->head
+ skb
->mac_header
;
1912 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1914 return skb
->mac_header
!= (typeof(skb
->mac_header
))~0U;
1917 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1919 skb
->mac_header
= skb
->data
- skb
->head
;
1922 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1924 skb_reset_mac_header(skb
);
1925 skb
->mac_header
+= offset
;
1928 static inline void skb_pop_mac_header(struct sk_buff
*skb
)
1930 skb
->mac_header
= skb
->network_header
;
1933 static inline void skb_probe_transport_header(struct sk_buff
*skb
,
1934 const int offset_hint
)
1936 struct flow_keys keys
;
1938 if (skb_transport_header_was_set(skb
))
1940 else if (skb_flow_dissect_flow_keys(skb
, &keys
))
1941 skb_set_transport_header(skb
, keys
.basic
.thoff
);
1943 skb_set_transport_header(skb
, offset_hint
);
1946 static inline void skb_mac_header_rebuild(struct sk_buff
*skb
)
1948 if (skb_mac_header_was_set(skb
)) {
1949 const unsigned char *old_mac
= skb_mac_header(skb
);
1951 skb_set_mac_header(skb
, -skb
->mac_len
);
1952 memmove(skb_mac_header(skb
), old_mac
, skb
->mac_len
);
1956 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
1958 return skb
->csum_start
- skb_headroom(skb
);
1961 static inline int skb_transport_offset(const struct sk_buff
*skb
)
1963 return skb_transport_header(skb
) - skb
->data
;
1966 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
1968 return skb
->transport_header
- skb
->network_header
;
1971 static inline u32
skb_inner_network_header_len(const struct sk_buff
*skb
)
1973 return skb
->inner_transport_header
- skb
->inner_network_header
;
1976 static inline int skb_network_offset(const struct sk_buff
*skb
)
1978 return skb_network_header(skb
) - skb
->data
;
1981 static inline int skb_inner_network_offset(const struct sk_buff
*skb
)
1983 return skb_inner_network_header(skb
) - skb
->data
;
1986 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
1988 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
1992 * CPUs often take a performance hit when accessing unaligned memory
1993 * locations. The actual performance hit varies, it can be small if the
1994 * hardware handles it or large if we have to take an exception and fix it
1997 * Since an ethernet header is 14 bytes network drivers often end up with
1998 * the IP header at an unaligned offset. The IP header can be aligned by
1999 * shifting the start of the packet by 2 bytes. Drivers should do this
2002 * skb_reserve(skb, NET_IP_ALIGN);
2004 * The downside to this alignment of the IP header is that the DMA is now
2005 * unaligned. On some architectures the cost of an unaligned DMA is high
2006 * and this cost outweighs the gains made by aligning the IP header.
2008 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2011 #ifndef NET_IP_ALIGN
2012 #define NET_IP_ALIGN 2
2016 * The networking layer reserves some headroom in skb data (via
2017 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2018 * the header has to grow. In the default case, if the header has to grow
2019 * 32 bytes or less we avoid the reallocation.
2021 * Unfortunately this headroom changes the DMA alignment of the resulting
2022 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2023 * on some architectures. An architecture can override this value,
2024 * perhaps setting it to a cacheline in size (since that will maintain
2025 * cacheline alignment of the DMA). It must be a power of 2.
2027 * Various parts of the networking layer expect at least 32 bytes of
2028 * headroom, you should not reduce this.
2030 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2031 * to reduce average number of cache lines per packet.
2032 * get_rps_cpus() for example only access one 64 bytes aligned block :
2033 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2036 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2039 int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
2041 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
2043 if (unlikely(skb_is_nonlinear(skb
))) {
2048 skb_set_tail_pointer(skb
, len
);
2051 void skb_trim(struct sk_buff
*skb
, unsigned int len
);
2053 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2056 return ___pskb_trim(skb
, len
);
2057 __skb_trim(skb
, len
);
2061 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2063 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
2067 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2068 * @skb: buffer to alter
2071 * This is identical to pskb_trim except that the caller knows that
2072 * the skb is not cloned so we should never get an error due to out-
2075 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
2077 int err
= pskb_trim(skb
, len
);
2082 * skb_orphan - orphan a buffer
2083 * @skb: buffer to orphan
2085 * If a buffer currently has an owner then we call the owner's
2086 * destructor function and make the @skb unowned. The buffer continues
2087 * to exist but is no longer charged to its former owner.
2089 static inline void skb_orphan(struct sk_buff
*skb
)
2091 if (skb
->destructor
) {
2092 skb
->destructor(skb
);
2093 skb
->destructor
= NULL
;
2101 * skb_orphan_frags - orphan the frags contained in a buffer
2102 * @skb: buffer to orphan frags from
2103 * @gfp_mask: allocation mask for replacement pages
2105 * For each frag in the SKB which needs a destructor (i.e. has an
2106 * owner) create a copy of that frag and release the original
2107 * page by calling the destructor.
2109 static inline int skb_orphan_frags(struct sk_buff
*skb
, gfp_t gfp_mask
)
2111 if (likely(!(skb_shinfo(skb
)->tx_flags
& SKBTX_DEV_ZEROCOPY
)))
2113 return skb_copy_ubufs(skb
, gfp_mask
);
2117 * __skb_queue_purge - empty a list
2118 * @list: list to empty
2120 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2121 * the list and one reference dropped. This function does not take the
2122 * list lock and the caller must hold the relevant locks to use it.
2124 void skb_queue_purge(struct sk_buff_head
*list
);
2125 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
2127 struct sk_buff
*skb
;
2128 while ((skb
= __skb_dequeue(list
)) != NULL
)
2132 void *netdev_alloc_frag(unsigned int fragsz
);
2134 struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
, unsigned int length
,
2138 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2139 * @dev: network device to receive on
2140 * @length: length to allocate
2142 * Allocate a new &sk_buff and assign it a usage count of one. The
2143 * buffer has unspecified headroom built in. Users should allocate
2144 * the headroom they think they need without accounting for the
2145 * built in space. The built in space is used for optimisations.
2147 * %NULL is returned if there is no free memory. Although this function
2148 * allocates memory it can be called from an interrupt.
2150 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
2151 unsigned int length
)
2153 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
2156 /* legacy helper around __netdev_alloc_skb() */
2157 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
2160 return __netdev_alloc_skb(NULL
, length
, gfp_mask
);
2163 /* legacy helper around netdev_alloc_skb() */
2164 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
2166 return netdev_alloc_skb(NULL
, length
);
2170 static inline struct sk_buff
*__netdev_alloc_skb_ip_align(struct net_device
*dev
,
2171 unsigned int length
, gfp_t gfp
)
2173 struct sk_buff
*skb
= __netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
, gfp
);
2175 if (NET_IP_ALIGN
&& skb
)
2176 skb_reserve(skb
, NET_IP_ALIGN
);
2180 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
2181 unsigned int length
)
2183 return __netdev_alloc_skb_ip_align(dev
, length
, GFP_ATOMIC
);
2186 static inline void skb_free_frag(void *addr
)
2188 __free_page_frag(addr
);
2191 void *napi_alloc_frag(unsigned int fragsz
);
2192 struct sk_buff
*__napi_alloc_skb(struct napi_struct
*napi
,
2193 unsigned int length
, gfp_t gfp_mask
);
2194 static inline struct sk_buff
*napi_alloc_skb(struct napi_struct
*napi
,
2195 unsigned int length
)
2197 return __napi_alloc_skb(napi
, length
, GFP_ATOMIC
);
2201 * __dev_alloc_pages - allocate page for network Rx
2202 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2203 * @order: size of the allocation
2205 * Allocate a new page.
2207 * %NULL is returned if there is no free memory.
2209 static inline struct page
*__dev_alloc_pages(gfp_t gfp_mask
,
2212 /* This piece of code contains several assumptions.
2213 * 1. This is for device Rx, therefor a cold page is preferred.
2214 * 2. The expectation is the user wants a compound page.
2215 * 3. If requesting a order 0 page it will not be compound
2216 * due to the check to see if order has a value in prep_new_page
2217 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2218 * code in gfp_to_alloc_flags that should be enforcing this.
2220 gfp_mask
|= __GFP_COLD
| __GFP_COMP
| __GFP_MEMALLOC
;
2222 return alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, order
);
2225 static inline struct page
*dev_alloc_pages(unsigned int order
)
2227 return __dev_alloc_pages(GFP_ATOMIC
, order
);
2231 * __dev_alloc_page - allocate a page for network Rx
2232 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2234 * Allocate a new page.
2236 * %NULL is returned if there is no free memory.
2238 static inline struct page
*__dev_alloc_page(gfp_t gfp_mask
)
2240 return __dev_alloc_pages(gfp_mask
, 0);
2243 static inline struct page
*dev_alloc_page(void)
2245 return __dev_alloc_page(GFP_ATOMIC
);
2249 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2250 * @page: The page that was allocated from skb_alloc_page
2251 * @skb: The skb that may need pfmemalloc set
2253 static inline void skb_propagate_pfmemalloc(struct page
*page
,
2254 struct sk_buff
*skb
)
2256 if (page
&& page
->pfmemalloc
)
2257 skb
->pfmemalloc
= true;
2261 * skb_frag_page - retrieve the page referred to by a paged fragment
2262 * @frag: the paged fragment
2264 * Returns the &struct page associated with @frag.
2266 static inline struct page
*skb_frag_page(const skb_frag_t
*frag
)
2268 return frag
->page
.p
;
2272 * __skb_frag_ref - take an addition reference on a paged fragment.
2273 * @frag: the paged fragment
2275 * Takes an additional reference on the paged fragment @frag.
2277 static inline void __skb_frag_ref(skb_frag_t
*frag
)
2279 get_page(skb_frag_page(frag
));
2283 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2285 * @f: the fragment offset.
2287 * Takes an additional reference on the @f'th paged fragment of @skb.
2289 static inline void skb_frag_ref(struct sk_buff
*skb
, int f
)
2291 __skb_frag_ref(&skb_shinfo(skb
)->frags
[f
]);
2295 * __skb_frag_unref - release a reference on a paged fragment.
2296 * @frag: the paged fragment
2298 * Releases a reference on the paged fragment @frag.
2300 static inline void __skb_frag_unref(skb_frag_t
*frag
)
2302 put_page(skb_frag_page(frag
));
2306 * skb_frag_unref - release a reference on a paged fragment of an skb.
2308 * @f: the fragment offset
2310 * Releases a reference on the @f'th paged fragment of @skb.
2312 static inline void skb_frag_unref(struct sk_buff
*skb
, int f
)
2314 __skb_frag_unref(&skb_shinfo(skb
)->frags
[f
]);
2318 * skb_frag_address - gets the address of the data contained in a paged fragment
2319 * @frag: the paged fragment buffer
2321 * Returns the address of the data within @frag. The page must already
2324 static inline void *skb_frag_address(const skb_frag_t
*frag
)
2326 return page_address(skb_frag_page(frag
)) + frag
->page_offset
;
2330 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2331 * @frag: the paged fragment buffer
2333 * Returns the address of the data within @frag. Checks that the page
2334 * is mapped and returns %NULL otherwise.
2336 static inline void *skb_frag_address_safe(const skb_frag_t
*frag
)
2338 void *ptr
= page_address(skb_frag_page(frag
));
2342 return ptr
+ frag
->page_offset
;
2346 * __skb_frag_set_page - sets the page contained in a paged fragment
2347 * @frag: the paged fragment
2348 * @page: the page to set
2350 * Sets the fragment @frag to contain @page.
2352 static inline void __skb_frag_set_page(skb_frag_t
*frag
, struct page
*page
)
2354 frag
->page
.p
= page
;
2358 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2360 * @f: the fragment offset
2361 * @page: the page to set
2363 * Sets the @f'th fragment of @skb to contain @page.
2365 static inline void skb_frag_set_page(struct sk_buff
*skb
, int f
,
2368 __skb_frag_set_page(&skb_shinfo(skb
)->frags
[f
], page
);
2371 bool skb_page_frag_refill(unsigned int sz
, struct page_frag
*pfrag
, gfp_t prio
);
2374 * skb_frag_dma_map - maps a paged fragment via the DMA API
2375 * @dev: the device to map the fragment to
2376 * @frag: the paged fragment to map
2377 * @offset: the offset within the fragment (starting at the
2378 * fragment's own offset)
2379 * @size: the number of bytes to map
2380 * @dir: the direction of the mapping (%PCI_DMA_*)
2382 * Maps the page associated with @frag to @device.
2384 static inline dma_addr_t
skb_frag_dma_map(struct device
*dev
,
2385 const skb_frag_t
*frag
,
2386 size_t offset
, size_t size
,
2387 enum dma_data_direction dir
)
2389 return dma_map_page(dev
, skb_frag_page(frag
),
2390 frag
->page_offset
+ offset
, size
, dir
);
2393 static inline struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
2396 return __pskb_copy(skb
, skb_headroom(skb
), gfp_mask
);
2400 static inline struct sk_buff
*pskb_copy_for_clone(struct sk_buff
*skb
,
2403 return __pskb_copy_fclone(skb
, skb_headroom(skb
), gfp_mask
, true);
2408 * skb_clone_writable - is the header of a clone writable
2409 * @skb: buffer to check
2410 * @len: length up to which to write
2412 * Returns true if modifying the header part of the cloned buffer
2413 * does not requires the data to be copied.
2415 static inline int skb_clone_writable(const struct sk_buff
*skb
, unsigned int len
)
2417 return !skb_header_cloned(skb
) &&
2418 skb_headroom(skb
) + len
<= skb
->hdr_len
;
2421 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
2426 if (headroom
> skb_headroom(skb
))
2427 delta
= headroom
- skb_headroom(skb
);
2429 if (delta
|| cloned
)
2430 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
2436 * skb_cow - copy header of skb when it is required
2437 * @skb: buffer to cow
2438 * @headroom: needed headroom
2440 * If the skb passed lacks sufficient headroom or its data part
2441 * is shared, data is reallocated. If reallocation fails, an error
2442 * is returned and original skb is not changed.
2444 * The result is skb with writable area skb->head...skb->tail
2445 * and at least @headroom of space at head.
2447 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
2449 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
2453 * skb_cow_head - skb_cow but only making the head writable
2454 * @skb: buffer to cow
2455 * @headroom: needed headroom
2457 * This function is identical to skb_cow except that we replace the
2458 * skb_cloned check by skb_header_cloned. It should be used when
2459 * you only need to push on some header and do not need to modify
2462 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
2464 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
2468 * skb_padto - pad an skbuff up to a minimal size
2469 * @skb: buffer to pad
2470 * @len: minimal length
2472 * Pads up a buffer to ensure the trailing bytes exist and are
2473 * blanked. If the buffer already contains sufficient data it
2474 * is untouched. Otherwise it is extended. Returns zero on
2475 * success. The skb is freed on error.
2477 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
2479 unsigned int size
= skb
->len
;
2480 if (likely(size
>= len
))
2482 return skb_pad(skb
, len
- size
);
2486 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2487 * @skb: buffer to pad
2488 * @len: minimal length
2490 * Pads up a buffer to ensure the trailing bytes exist and are
2491 * blanked. If the buffer already contains sufficient data it
2492 * is untouched. Otherwise it is extended. Returns zero on
2493 * success. The skb is freed on error.
2495 static inline int skb_put_padto(struct sk_buff
*skb
, unsigned int len
)
2497 unsigned int size
= skb
->len
;
2499 if (unlikely(size
< len
)) {
2501 if (skb_pad(skb
, len
))
2503 __skb_put(skb
, len
);
2508 static inline int skb_add_data(struct sk_buff
*skb
,
2509 struct iov_iter
*from
, int copy
)
2511 const int off
= skb
->len
;
2513 if (skb
->ip_summed
== CHECKSUM_NONE
) {
2515 if (csum_and_copy_from_iter(skb_put(skb
, copy
), copy
,
2516 &csum
, from
) == copy
) {
2517 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
2520 } else if (copy_from_iter(skb_put(skb
, copy
), copy
, from
) == copy
)
2523 __skb_trim(skb
, off
);
2527 static inline bool skb_can_coalesce(struct sk_buff
*skb
, int i
,
2528 const struct page
*page
, int off
)
2531 const struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
2533 return page
== skb_frag_page(frag
) &&
2534 off
== frag
->page_offset
+ skb_frag_size(frag
);
2539 static inline int __skb_linearize(struct sk_buff
*skb
)
2541 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
2545 * skb_linearize - convert paged skb to linear one
2546 * @skb: buffer to linarize
2548 * If there is no free memory -ENOMEM is returned, otherwise zero
2549 * is returned and the old skb data released.
2551 static inline int skb_linearize(struct sk_buff
*skb
)
2553 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
2557 * skb_has_shared_frag - can any frag be overwritten
2558 * @skb: buffer to test
2560 * Return true if the skb has at least one frag that might be modified
2561 * by an external entity (as in vmsplice()/sendfile())
2563 static inline bool skb_has_shared_frag(const struct sk_buff
*skb
)
2565 return skb_is_nonlinear(skb
) &&
2566 skb_shinfo(skb
)->tx_flags
& SKBTX_SHARED_FRAG
;
2570 * skb_linearize_cow - make sure skb is linear and writable
2571 * @skb: buffer to process
2573 * If there is no free memory -ENOMEM is returned, otherwise zero
2574 * is returned and the old skb data released.
2576 static inline int skb_linearize_cow(struct sk_buff
*skb
)
2578 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
2579 __skb_linearize(skb
) : 0;
2583 * skb_postpull_rcsum - update checksum for received skb after pull
2584 * @skb: buffer to update
2585 * @start: start of data before pull
2586 * @len: length of data pulled
2588 * After doing a pull on a received packet, you need to call this to
2589 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2590 * CHECKSUM_NONE so that it can be recomputed from scratch.
2593 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
2594 const void *start
, unsigned int len
)
2596 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2597 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
2600 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
2603 * pskb_trim_rcsum - trim received skb and update checksum
2604 * @skb: buffer to trim
2607 * This is exactly the same as pskb_trim except that it ensures the
2608 * checksum of received packets are still valid after the operation.
2611 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
2613 if (likely(len
>= skb
->len
))
2615 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2616 skb
->ip_summed
= CHECKSUM_NONE
;
2617 return __pskb_trim(skb
, len
);
2620 #define skb_queue_walk(queue, skb) \
2621 for (skb = (queue)->next; \
2622 skb != (struct sk_buff *)(queue); \
2625 #define skb_queue_walk_safe(queue, skb, tmp) \
2626 for (skb = (queue)->next, tmp = skb->next; \
2627 skb != (struct sk_buff *)(queue); \
2628 skb = tmp, tmp = skb->next)
2630 #define skb_queue_walk_from(queue, skb) \
2631 for (; skb != (struct sk_buff *)(queue); \
2634 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2635 for (tmp = skb->next; \
2636 skb != (struct sk_buff *)(queue); \
2637 skb = tmp, tmp = skb->next)
2639 #define skb_queue_reverse_walk(queue, skb) \
2640 for (skb = (queue)->prev; \
2641 skb != (struct sk_buff *)(queue); \
2644 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2645 for (skb = (queue)->prev, tmp = skb->prev; \
2646 skb != (struct sk_buff *)(queue); \
2647 skb = tmp, tmp = skb->prev)
2649 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2650 for (tmp = skb->prev; \
2651 skb != (struct sk_buff *)(queue); \
2652 skb = tmp, tmp = skb->prev)
2654 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
2656 return skb_shinfo(skb
)->frag_list
!= NULL
;
2659 static inline void skb_frag_list_init(struct sk_buff
*skb
)
2661 skb_shinfo(skb
)->frag_list
= NULL
;
2664 static inline void skb_frag_add_head(struct sk_buff
*skb
, struct sk_buff
*frag
)
2666 frag
->next
= skb_shinfo(skb
)->frag_list
;
2667 skb_shinfo(skb
)->frag_list
= frag
;
2670 #define skb_walk_frags(skb, iter) \
2671 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2673 struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
2674 int *peeked
, int *off
, int *err
);
2675 struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
, int noblock
,
2677 unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
2678 struct poll_table_struct
*wait
);
2679 int skb_copy_datagram_iter(const struct sk_buff
*from
, int offset
,
2680 struct iov_iter
*to
, int size
);
2681 static inline int skb_copy_datagram_msg(const struct sk_buff
*from
, int offset
,
2682 struct msghdr
*msg
, int size
)
2684 return skb_copy_datagram_iter(from
, offset
, &msg
->msg_iter
, size
);
2686 int skb_copy_and_csum_datagram_msg(struct sk_buff
*skb
, int hlen
,
2687 struct msghdr
*msg
);
2688 int skb_copy_datagram_from_iter(struct sk_buff
*skb
, int offset
,
2689 struct iov_iter
*from
, int len
);
2690 int zerocopy_sg_from_iter(struct sk_buff
*skb
, struct iov_iter
*frm
);
2691 void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
2692 void skb_free_datagram_locked(struct sock
*sk
, struct sk_buff
*skb
);
2693 int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
, unsigned int flags
);
2694 int skb_copy_bits(const struct sk_buff
*skb
, int offset
, void *to
, int len
);
2695 int skb_store_bits(struct sk_buff
*skb
, int offset
, const void *from
, int len
);
2696 __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
, int offset
, u8
*to
,
2697 int len
, __wsum csum
);
2698 int skb_splice_bits(struct sk_buff
*skb
, unsigned int offset
,
2699 struct pipe_inode_info
*pipe
, unsigned int len
,
2700 unsigned int flags
);
2701 void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
2702 unsigned int skb_zerocopy_headlen(const struct sk_buff
*from
);
2703 int skb_zerocopy(struct sk_buff
*to
, struct sk_buff
*from
,
2705 void skb_split(struct sk_buff
*skb
, struct sk_buff
*skb1
, const u32 len
);
2706 int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
, int shiftlen
);
2707 void skb_scrub_packet(struct sk_buff
*skb
, bool xnet
);
2708 unsigned int skb_gso_transport_seglen(const struct sk_buff
*skb
);
2709 struct sk_buff
*skb_segment(struct sk_buff
*skb
, netdev_features_t features
);
2710 struct sk_buff
*skb_vlan_untag(struct sk_buff
*skb
);
2711 int skb_ensure_writable(struct sk_buff
*skb
, int write_len
);
2712 int skb_vlan_pop(struct sk_buff
*skb
);
2713 int skb_vlan_push(struct sk_buff
*skb
, __be16 vlan_proto
, u16 vlan_tci
);
2715 static inline int memcpy_from_msg(void *data
, struct msghdr
*msg
, int len
)
2717 return copy_from_iter(data
, len
, &msg
->msg_iter
) == len
? 0 : -EFAULT
;
2720 static inline int memcpy_to_msg(struct msghdr
*msg
, void *data
, int len
)
2722 return copy_to_iter(data
, len
, &msg
->msg_iter
) == len
? 0 : -EFAULT
;
2725 struct skb_checksum_ops
{
2726 __wsum (*update
)(const void *mem
, int len
, __wsum wsum
);
2727 __wsum (*combine
)(__wsum csum
, __wsum csum2
, int offset
, int len
);
2730 __wsum
__skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
2731 __wsum csum
, const struct skb_checksum_ops
*ops
);
2732 __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
2735 static inline void *__skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2736 int len
, void *data
, int hlen
, void *buffer
)
2738 if (hlen
- offset
>= len
)
2739 return data
+ offset
;
2742 skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
2748 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2749 int len
, void *buffer
)
2751 return __skb_header_pointer(skb
, offset
, len
, skb
->data
,
2752 skb_headlen(skb
), buffer
);
2756 * skb_needs_linearize - check if we need to linearize a given skb
2757 * depending on the given device features.
2758 * @skb: socket buffer to check
2759 * @features: net device features
2761 * Returns true if either:
2762 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2763 * 2. skb is fragmented and the device does not support SG.
2765 static inline bool skb_needs_linearize(struct sk_buff
*skb
,
2766 netdev_features_t features
)
2768 return skb_is_nonlinear(skb
) &&
2769 ((skb_has_frag_list(skb
) && !(features
& NETIF_F_FRAGLIST
)) ||
2770 (skb_shinfo(skb
)->nr_frags
&& !(features
& NETIF_F_SG
)));
2773 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
2775 const unsigned int len
)
2777 memcpy(to
, skb
->data
, len
);
2780 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
2781 const int offset
, void *to
,
2782 const unsigned int len
)
2784 memcpy(to
, skb
->data
+ offset
, len
);
2787 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
2789 const unsigned int len
)
2791 memcpy(skb
->data
, from
, len
);
2794 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
2797 const unsigned int len
)
2799 memcpy(skb
->data
+ offset
, from
, len
);
2802 void skb_init(void);
2804 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
2810 * skb_get_timestamp - get timestamp from a skb
2811 * @skb: skb to get stamp from
2812 * @stamp: pointer to struct timeval to store stamp in
2814 * Timestamps are stored in the skb as offsets to a base timestamp.
2815 * This function converts the offset back to a struct timeval and stores
2818 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
2819 struct timeval
*stamp
)
2821 *stamp
= ktime_to_timeval(skb
->tstamp
);
2824 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
2825 struct timespec
*stamp
)
2827 *stamp
= ktime_to_timespec(skb
->tstamp
);
2830 static inline void __net_timestamp(struct sk_buff
*skb
)
2832 skb
->tstamp
= ktime_get_real();
2835 static inline ktime_t
net_timedelta(ktime_t t
)
2837 return ktime_sub(ktime_get_real(), t
);
2840 static inline ktime_t
net_invalid_timestamp(void)
2842 return ktime_set(0, 0);
2845 struct sk_buff
*skb_clone_sk(struct sk_buff
*skb
);
2847 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2849 void skb_clone_tx_timestamp(struct sk_buff
*skb
);
2850 bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
2852 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2854 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
2858 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
2863 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2866 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2868 * PHY drivers may accept clones of transmitted packets for
2869 * timestamping via their phy_driver.txtstamp method. These drivers
2870 * must call this function to return the skb back to the stack, with
2871 * or without a timestamp.
2873 * @skb: clone of the the original outgoing packet
2874 * @hwtstamps: hardware time stamps, may be NULL if not available
2877 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
2878 struct skb_shared_hwtstamps
*hwtstamps
);
2880 void __skb_tstamp_tx(struct sk_buff
*orig_skb
,
2881 struct skb_shared_hwtstamps
*hwtstamps
,
2882 struct sock
*sk
, int tstype
);
2885 * skb_tstamp_tx - queue clone of skb with send time stamps
2886 * @orig_skb: the original outgoing packet
2887 * @hwtstamps: hardware time stamps, may be NULL if not available
2889 * If the skb has a socket associated, then this function clones the
2890 * skb (thus sharing the actual data and optional structures), stores
2891 * the optional hardware time stamping information (if non NULL) or
2892 * generates a software time stamp (otherwise), then queues the clone
2893 * to the error queue of the socket. Errors are silently ignored.
2895 void skb_tstamp_tx(struct sk_buff
*orig_skb
,
2896 struct skb_shared_hwtstamps
*hwtstamps
);
2898 static inline void sw_tx_timestamp(struct sk_buff
*skb
)
2900 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
&&
2901 !(skb_shinfo(skb
)->tx_flags
& SKBTX_IN_PROGRESS
))
2902 skb_tstamp_tx(skb
, NULL
);
2906 * skb_tx_timestamp() - Driver hook for transmit timestamping
2908 * Ethernet MAC Drivers should call this function in their hard_xmit()
2909 * function immediately before giving the sk_buff to the MAC hardware.
2911 * Specifically, one should make absolutely sure that this function is
2912 * called before TX completion of this packet can trigger. Otherwise
2913 * the packet could potentially already be freed.
2915 * @skb: A socket buffer.
2917 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
2919 skb_clone_tx_timestamp(skb
);
2920 sw_tx_timestamp(skb
);
2924 * skb_complete_wifi_ack - deliver skb with wifi status
2926 * @skb: the original outgoing packet
2927 * @acked: ack status
2930 void skb_complete_wifi_ack(struct sk_buff
*skb
, bool acked
);
2932 __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
2933 __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
2935 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
2937 return ((skb
->ip_summed
== CHECKSUM_UNNECESSARY
) ||
2939 (skb
->ip_summed
== CHECKSUM_PARTIAL
&&
2940 skb_checksum_start_offset(skb
) >= 0));
2944 * skb_checksum_complete - Calculate checksum of an entire packet
2945 * @skb: packet to process
2947 * This function calculates the checksum over the entire packet plus
2948 * the value of skb->csum. The latter can be used to supply the
2949 * checksum of a pseudo header as used by TCP/UDP. It returns the
2952 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2953 * this function can be used to verify that checksum on received
2954 * packets. In that case the function should return zero if the
2955 * checksum is correct. In particular, this function will return zero
2956 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2957 * hardware has already verified the correctness of the checksum.
2959 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
2961 return skb_csum_unnecessary(skb
) ?
2962 0 : __skb_checksum_complete(skb
);
2965 static inline void __skb_decr_checksum_unnecessary(struct sk_buff
*skb
)
2967 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
2968 if (skb
->csum_level
== 0)
2969 skb
->ip_summed
= CHECKSUM_NONE
;
2975 static inline void __skb_incr_checksum_unnecessary(struct sk_buff
*skb
)
2977 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
2978 if (skb
->csum_level
< SKB_MAX_CSUM_LEVEL
)
2980 } else if (skb
->ip_summed
== CHECKSUM_NONE
) {
2981 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
2982 skb
->csum_level
= 0;
2986 static inline void __skb_mark_checksum_bad(struct sk_buff
*skb
)
2988 /* Mark current checksum as bad (typically called from GRO
2989 * path). In the case that ip_summed is CHECKSUM_NONE
2990 * this must be the first checksum encountered in the packet.
2991 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2992 * checksum after the last one validated. For UDP, a zero
2993 * checksum can not be marked as bad.
2996 if (skb
->ip_summed
== CHECKSUM_NONE
||
2997 skb
->ip_summed
== CHECKSUM_UNNECESSARY
)
3001 /* Check if we need to perform checksum complete validation.
3003 * Returns true if checksum complete is needed, false otherwise
3004 * (either checksum is unnecessary or zero checksum is allowed).
3006 static inline bool __skb_checksum_validate_needed(struct sk_buff
*skb
,
3010 if (skb_csum_unnecessary(skb
) || (zero_okay
&& !check
)) {
3011 skb
->csum_valid
= 1;
3012 __skb_decr_checksum_unnecessary(skb
);
3019 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3022 #define CHECKSUM_BREAK 76
3024 /* Unset checksum-complete
3026 * Unset checksum complete can be done when packet is being modified
3027 * (uncompressed for instance) and checksum-complete value is
3030 static inline void skb_checksum_complete_unset(struct sk_buff
*skb
)
3032 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3033 skb
->ip_summed
= CHECKSUM_NONE
;
3036 /* Validate (init) checksum based on checksum complete.
3039 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3040 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3041 * checksum is stored in skb->csum for use in __skb_checksum_complete
3042 * non-zero: value of invalid checksum
3045 static inline __sum16
__skb_checksum_validate_complete(struct sk_buff
*skb
,
3049 if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
3050 if (!csum_fold(csum_add(psum
, skb
->csum
))) {
3051 skb
->csum_valid
= 1;
3054 } else if (skb
->csum_bad
) {
3055 /* ip_summed == CHECKSUM_NONE in this case */
3056 return (__force __sum16
)1;
3061 if (complete
|| skb
->len
<= CHECKSUM_BREAK
) {
3064 csum
= __skb_checksum_complete(skb
);
3065 skb
->csum_valid
= !csum
;
3072 static inline __wsum
null_compute_pseudo(struct sk_buff
*skb
, int proto
)
3077 /* Perform checksum validate (init). Note that this is a macro since we only
3078 * want to calculate the pseudo header which is an input function if necessary.
3079 * First we try to validate without any computation (checksum unnecessary) and
3080 * then calculate based on checksum complete calling the function to compute
3084 * 0: checksum is validated or try to in skb_checksum_complete
3085 * non-zero: value of invalid checksum
3087 #define __skb_checksum_validate(skb, proto, complete, \
3088 zero_okay, check, compute_pseudo) \
3090 __sum16 __ret = 0; \
3091 skb->csum_valid = 0; \
3092 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3093 __ret = __skb_checksum_validate_complete(skb, \
3094 complete, compute_pseudo(skb, proto)); \
3098 #define skb_checksum_init(skb, proto, compute_pseudo) \
3099 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3101 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3102 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3104 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3105 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3107 #define skb_checksum_validate_zero_check(skb, proto, check, \
3109 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3111 #define skb_checksum_simple_validate(skb) \
3112 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3114 static inline bool __skb_checksum_convert_check(struct sk_buff
*skb
)
3116 return (skb
->ip_summed
== CHECKSUM_NONE
&&
3117 skb
->csum_valid
&& !skb
->csum_bad
);
3120 static inline void __skb_checksum_convert(struct sk_buff
*skb
,
3121 __sum16 check
, __wsum pseudo
)
3123 skb
->csum
= ~pseudo
;
3124 skb
->ip_summed
= CHECKSUM_COMPLETE
;
3127 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3129 if (__skb_checksum_convert_check(skb)) \
3130 __skb_checksum_convert(skb, check, \
3131 compute_pseudo(skb, proto)); \
3134 static inline void skb_remcsum_adjust_partial(struct sk_buff
*skb
, void *ptr
,
3135 u16 start
, u16 offset
)
3137 skb
->ip_summed
= CHECKSUM_PARTIAL
;
3138 skb
->csum_start
= ((unsigned char *)ptr
+ start
) - skb
->head
;
3139 skb
->csum_offset
= offset
- start
;
3142 /* Update skbuf and packet to reflect the remote checksum offload operation.
3143 * When called, ptr indicates the starting point for skb->csum when
3144 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3145 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3147 static inline void skb_remcsum_process(struct sk_buff
*skb
, void *ptr
,
3148 int start
, int offset
, bool nopartial
)
3153 skb_remcsum_adjust_partial(skb
, ptr
, start
, offset
);
3157 if (unlikely(skb
->ip_summed
!= CHECKSUM_COMPLETE
)) {
3158 __skb_checksum_complete(skb
);
3159 skb_postpull_rcsum(skb
, skb
->data
, ptr
- (void *)skb
->data
);
3162 delta
= remcsum_adjust(ptr
, skb
->csum
, start
, offset
);
3164 /* Adjust skb->csum since we changed the packet */
3165 skb
->csum
= csum_add(skb
->csum
, delta
);
3168 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3169 void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
3170 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
3172 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
3173 nf_conntrack_destroy(nfct
);
3175 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
3178 atomic_inc(&nfct
->use
);
3181 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3182 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
3184 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
3187 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
3190 atomic_inc(&nf_bridge
->use
);
3192 #endif /* CONFIG_BRIDGE_NETFILTER */
3193 static inline void nf_reset(struct sk_buff
*skb
)
3195 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3196 nf_conntrack_put(skb
->nfct
);
3199 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3200 nf_bridge_put(skb
->nf_bridge
);
3201 skb
->nf_bridge
= NULL
;
3205 static inline void nf_reset_trace(struct sk_buff
*skb
)
3207 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3212 /* Note: This doesn't put any conntrack and bridge info in dst. */
3213 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
,
3216 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3217 dst
->nfct
= src
->nfct
;
3218 nf_conntrack_get(src
->nfct
);
3220 dst
->nfctinfo
= src
->nfctinfo
;
3222 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3223 dst
->nf_bridge
= src
->nf_bridge
;
3224 nf_bridge_get(src
->nf_bridge
);
3226 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3228 dst
->nf_trace
= src
->nf_trace
;
3232 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
3234 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3235 nf_conntrack_put(dst
->nfct
);
3237 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3238 nf_bridge_put(dst
->nf_bridge
);
3240 __nf_copy(dst
, src
, true);
3243 #ifdef CONFIG_NETWORK_SECMARK
3244 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3246 to
->secmark
= from
->secmark
;
3249 static inline void skb_init_secmark(struct sk_buff
*skb
)
3254 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3257 static inline void skb_init_secmark(struct sk_buff
*skb
)
3261 static inline bool skb_irq_freeable(const struct sk_buff
*skb
)
3263 return !skb
->destructor
&&
3264 #if IS_ENABLED(CONFIG_XFRM)
3267 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3270 !skb
->_skb_refdst
&&
3271 !skb_has_frag_list(skb
);
3274 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
3276 skb
->queue_mapping
= queue_mapping
;
3279 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
3281 return skb
->queue_mapping
;
3284 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
3286 to
->queue_mapping
= from
->queue_mapping
;
3289 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
3291 skb
->queue_mapping
= rx_queue
+ 1;
3294 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
3296 return skb
->queue_mapping
- 1;
3299 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
3301 return skb
->queue_mapping
!= 0;
3304 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
3313 /* Keeps track of mac header offset relative to skb->head.
3314 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3315 * For non-tunnel skb it points to skb_mac_header() and for
3316 * tunnel skb it points to outer mac header.
3317 * Keeps track of level of encapsulation of network headers.
3324 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3326 static inline int skb_tnl_header_len(const struct sk_buff
*inner_skb
)
3328 return (skb_mac_header(inner_skb
) - inner_skb
->head
) -
3329 SKB_GSO_CB(inner_skb
)->mac_offset
;
3332 static inline int gso_pskb_expand_head(struct sk_buff
*skb
, int extra
)
3334 int new_headroom
, headroom
;
3337 headroom
= skb_headroom(skb
);
3338 ret
= pskb_expand_head(skb
, extra
, 0, GFP_ATOMIC
);
3342 new_headroom
= skb_headroom(skb
);
3343 SKB_GSO_CB(skb
)->mac_offset
+= (new_headroom
- headroom
);
3347 /* Compute the checksum for a gso segment. First compute the checksum value
3348 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3349 * then add in skb->csum (checksum from csum_start to end of packet).
3350 * skb->csum and csum_start are then updated to reflect the checksum of the
3351 * resultant packet starting from the transport header-- the resultant checksum
3352 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3355 static inline __sum16
gso_make_checksum(struct sk_buff
*skb
, __wsum res
)
3357 int plen
= SKB_GSO_CB(skb
)->csum_start
- skb_headroom(skb
) -
3358 skb_transport_offset(skb
);
3361 partial
= csum_partial(skb_transport_header(skb
), plen
, skb
->csum
);
3363 SKB_GSO_CB(skb
)->csum_start
-= plen
;
3365 return csum_fold(partial
);
3368 static inline bool skb_is_gso(const struct sk_buff
*skb
)
3370 return skb_shinfo(skb
)->gso_size
;
3373 /* Note: Should be called only if skb_is_gso(skb) is true */
3374 static inline bool skb_is_gso_v6(const struct sk_buff
*skb
)
3376 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
3379 void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
3381 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
3383 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3384 * wanted then gso_type will be set. */
3385 const struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
3387 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
3388 unlikely(shinfo
->gso_type
== 0)) {
3389 __skb_warn_lro_forwarding(skb
);
3395 static inline void skb_forward_csum(struct sk_buff
*skb
)
3397 /* Unfortunately we don't support this one. Any brave souls? */
3398 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3399 skb
->ip_summed
= CHECKSUM_NONE
;
3403 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3404 * @skb: skb to check
3406 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3407 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3408 * use this helper, to document places where we make this assertion.
3410 static inline void skb_checksum_none_assert(const struct sk_buff
*skb
)
3413 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
3417 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
3419 int skb_checksum_setup(struct sk_buff
*skb
, bool recalculate
);
3420 struct sk_buff
*skb_checksum_trimmed(struct sk_buff
*skb
,
3421 unsigned int transport_len
,
3422 __sum16(*skb_chkf
)(struct sk_buff
*skb
));
3425 * skb_head_is_locked - Determine if the skb->head is locked down
3426 * @skb: skb to check
3428 * The head on skbs build around a head frag can be removed if they are
3429 * not cloned. This function returns true if the skb head is locked down
3430 * due to either being allocated via kmalloc, or by being a clone with
3431 * multiple references to the head.
3433 static inline bool skb_head_is_locked(const struct sk_buff
*skb
)
3435 return !skb
->head_frag
|| skb_cloned(skb
);
3439 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3443 * skb_gso_network_seglen is used to determine the real size of the
3444 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3446 * The MAC/L2 header is not accounted for.
3448 static inline unsigned int skb_gso_network_seglen(const struct sk_buff
*skb
)
3450 unsigned int hdr_len
= skb_transport_header(skb
) -
3451 skb_network_header(skb
);
3452 return hdr_len
+ skb_gso_transport_seglen(skb
);
3454 #endif /* __KERNEL__ */
3455 #endif /* _LINUX_SKBUFF_H */