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>
24 #include <linux/atomic.h>
25 #include <asm/types.h>
26 #include <linux/spinlock.h>
27 #include <linux/net.h>
28 #include <linux/textsearch.h>
29 #include <net/checksum.h>
30 #include <linux/rcupdate.h>
31 #include <linux/hrtimer.h>
32 #include <linux/dma-mapping.h>
33 #include <linux/netdev_features.h>
34 #include <linux/sched.h>
35 #include <net/flow_keys.h>
37 /* A. Checksumming of received packets by device.
41 * Device failed to checksum this packet e.g. due to lack of capabilities.
42 * The packet contains full (though not verified) checksum in packet but
43 * not in skb->csum. Thus, skb->csum is undefined in this case.
45 * CHECKSUM_UNNECESSARY:
47 * The hardware you're dealing with doesn't calculate the full checksum
48 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
49 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
50 * if their checksums are okay. skb->csum is still undefined in this case
51 * though. It is a bad option, but, unfortunately, nowadays most vendors do
52 * this. Apparently with the secret goal to sell you new devices, when you
53 * will add new protocol to your host, f.e. IPv6 8)
55 * CHECKSUM_UNNECESSARY is applicable to following protocols:
57 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
58 * zero UDP checksum for either IPv4 or IPv6, the networking stack
59 * may perform further validation in this case.
60 * GRE: only if the checksum is present in the header.
61 * SCTP: indicates the CRC in SCTP header has been validated.
63 * skb->csum_level indicates the number of consecutive checksums found in
64 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
65 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
66 * and a device is able to verify the checksums for UDP (possibly zero),
67 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
68 * two. If the device were only able to verify the UDP checksum and not
69 * GRE, either because it doesn't support GRE checksum of because GRE
70 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
71 * not considered in this case).
75 * This is the most generic way. The device supplied checksum of the _whole_
76 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
77 * hardware doesn't need to parse L3/L4 headers to implement this.
79 * Note: Even if device supports only some protocols, but is able to produce
80 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
84 * This is identical to the case for output below. This may occur on a packet
85 * received directly from another Linux OS, e.g., a virtualized Linux kernel
86 * on the same host. The packet can be treated in the same way as
87 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
88 * checksum must be filled in by the OS or the hardware.
90 * B. Checksumming on output.
94 * The skb was already checksummed by the protocol, or a checksum is not
99 * The device is required to checksum the packet as seen by hard_start_xmit()
100 * from skb->csum_start up to the end, and to record/write the checksum at
101 * offset skb->csum_start + skb->csum_offset.
103 * The device must show its capabilities in dev->features, set up at device
104 * setup time, e.g. netdev_features.h:
106 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
107 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
108 * IPv4. Sigh. Vendors like this way for an unknown reason.
109 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
110 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
111 * NETIF_F_... - Well, you get the picture.
113 * CHECKSUM_UNNECESSARY:
115 * Normally, the device will do per protocol specific checksumming. Protocol
116 * implementations that do not want the NIC to perform the checksum
117 * calculation should use this flag in their outgoing skbs.
119 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
120 * offload. Correspondingly, the FCoE protocol driver
121 * stack should use CHECKSUM_UNNECESSARY.
123 * Any questions? No questions, good. --ANK
126 /* Don't change this without changing skb_csum_unnecessary! */
127 #define CHECKSUM_NONE 0
128 #define CHECKSUM_UNNECESSARY 1
129 #define CHECKSUM_COMPLETE 2
130 #define CHECKSUM_PARTIAL 3
132 /* Maximum value in skb->csum_level */
133 #define SKB_MAX_CSUM_LEVEL 3
135 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
136 #define SKB_WITH_OVERHEAD(X) \
137 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
138 #define SKB_MAX_ORDER(X, ORDER) \
139 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
140 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
141 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
143 /* return minimum truesize of one skb containing X bytes of data */
144 #define SKB_TRUESIZE(X) ((X) + \
145 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
146 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
150 struct pipe_inode_info
;
152 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
153 struct nf_conntrack
{
158 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
159 struct nf_bridge_info
{
162 struct net_device
*physindev
;
163 struct net_device
*physoutdev
;
164 unsigned long data
[32 / sizeof(unsigned long)];
168 struct sk_buff_head
{
169 /* These two members must be first. */
170 struct sk_buff
*next
;
171 struct sk_buff
*prev
;
179 /* To allow 64K frame to be packed as single skb without frag_list we
180 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
181 * buffers which do not start on a page boundary.
183 * Since GRO uses frags we allocate at least 16 regardless of page
186 #if (65536/PAGE_SIZE + 1) < 16
187 #define MAX_SKB_FRAGS 16UL
189 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
192 typedef struct skb_frag_struct skb_frag_t
;
194 struct skb_frag_struct
{
198 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
207 static inline unsigned int skb_frag_size(const skb_frag_t
*frag
)
212 static inline void skb_frag_size_set(skb_frag_t
*frag
, unsigned int size
)
217 static inline void skb_frag_size_add(skb_frag_t
*frag
, int delta
)
222 static inline void skb_frag_size_sub(skb_frag_t
*frag
, int delta
)
227 #define HAVE_HW_TIME_STAMP
230 * struct skb_shared_hwtstamps - hardware time stamps
231 * @hwtstamp: hardware time stamp transformed into duration
232 * since arbitrary point in time
234 * Software time stamps generated by ktime_get_real() are stored in
237 * hwtstamps can only be compared against other hwtstamps from
240 * This structure is attached to packets as part of the
241 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
243 struct skb_shared_hwtstamps
{
247 /* Definitions for tx_flags in struct skb_shared_info */
249 /* generate hardware time stamp */
250 SKBTX_HW_TSTAMP
= 1 << 0,
252 /* generate software time stamp when queueing packet to NIC */
253 SKBTX_SW_TSTAMP
= 1 << 1,
255 /* device driver is going to provide hardware time stamp */
256 SKBTX_IN_PROGRESS
= 1 << 2,
258 /* device driver supports TX zero-copy buffers */
259 SKBTX_DEV_ZEROCOPY
= 1 << 3,
261 /* generate wifi status information (where possible) */
262 SKBTX_WIFI_STATUS
= 1 << 4,
264 /* This indicates at least one fragment might be overwritten
265 * (as in vmsplice(), sendfile() ...)
266 * If we need to compute a TX checksum, we'll need to copy
267 * all frags to avoid possible bad checksum
269 SKBTX_SHARED_FRAG
= 1 << 5,
271 /* generate software time stamp when entering packet scheduling */
272 SKBTX_SCHED_TSTAMP
= 1 << 6,
274 /* generate software timestamp on peer data acknowledgment */
275 SKBTX_ACK_TSTAMP
= 1 << 7,
278 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
279 SKBTX_SCHED_TSTAMP | \
281 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
284 * The callback notifies userspace to release buffers when skb DMA is done in
285 * lower device, the skb last reference should be 0 when calling this.
286 * The zerocopy_success argument is true if zero copy transmit occurred,
287 * false on data copy or out of memory error caused by data copy attempt.
288 * The ctx field is used to track device context.
289 * The desc field is used to track userspace buffer index.
292 void (*callback
)(struct ubuf_info
*, bool zerocopy_success
);
297 /* This data is invariant across clones and lives at
298 * the end of the header data, ie. at skb->end.
300 struct skb_shared_info
{
301 unsigned char nr_frags
;
303 unsigned short gso_size
;
304 /* Warning: this field is not always filled in (UFO)! */
305 unsigned short gso_segs
;
306 unsigned short gso_type
;
307 struct sk_buff
*frag_list
;
308 struct skb_shared_hwtstamps hwtstamps
;
313 * Warning : all fields before dataref are cleared in __alloc_skb()
317 /* Intermediate layers must ensure that destructor_arg
318 * remains valid until skb destructor */
319 void * destructor_arg
;
321 /* must be last field, see pskb_expand_head() */
322 skb_frag_t frags
[MAX_SKB_FRAGS
];
325 /* We divide dataref into two halves. The higher 16 bits hold references
326 * to the payload part of skb->data. The lower 16 bits hold references to
327 * the entire skb->data. A clone of a headerless skb holds the length of
328 * the header in skb->hdr_len.
330 * All users must obey the rule that the skb->data reference count must be
331 * greater than or equal to the payload reference count.
333 * Holding a reference to the payload part means that the user does not
334 * care about modifications to the header part of skb->data.
336 #define SKB_DATAREF_SHIFT 16
337 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
341 SKB_FCLONE_UNAVAILABLE
, /* skb has no fclone (from head_cache) */
342 SKB_FCLONE_ORIG
, /* orig skb (from fclone_cache) */
343 SKB_FCLONE_CLONE
, /* companion fclone skb (from fclone_cache) */
344 SKB_FCLONE_FREE
, /* this companion fclone skb is available */
348 SKB_GSO_TCPV4
= 1 << 0,
349 SKB_GSO_UDP
= 1 << 1,
351 /* This indicates the skb is from an untrusted source. */
352 SKB_GSO_DODGY
= 1 << 2,
354 /* This indicates the tcp segment has CWR set. */
355 SKB_GSO_TCP_ECN
= 1 << 3,
357 SKB_GSO_TCPV6
= 1 << 4,
359 SKB_GSO_FCOE
= 1 << 5,
361 SKB_GSO_GRE
= 1 << 6,
363 SKB_GSO_GRE_CSUM
= 1 << 7,
365 SKB_GSO_IPIP
= 1 << 8,
367 SKB_GSO_SIT
= 1 << 9,
369 SKB_GSO_UDP_TUNNEL
= 1 << 10,
371 SKB_GSO_UDP_TUNNEL_CSUM
= 1 << 11,
373 SKB_GSO_MPLS
= 1 << 12,
377 #if BITS_PER_LONG > 32
378 #define NET_SKBUFF_DATA_USES_OFFSET 1
381 #ifdef NET_SKBUFF_DATA_USES_OFFSET
382 typedef unsigned int sk_buff_data_t
;
384 typedef unsigned char *sk_buff_data_t
;
388 * struct skb_mstamp - multi resolution time stamps
389 * @stamp_us: timestamp in us resolution
390 * @stamp_jiffies: timestamp in jiffies
403 * skb_mstamp_get - get current timestamp
404 * @cl: place to store timestamps
406 static inline void skb_mstamp_get(struct skb_mstamp
*cl
)
408 u64 val
= local_clock();
410 do_div(val
, NSEC_PER_USEC
);
411 cl
->stamp_us
= (u32
)val
;
412 cl
->stamp_jiffies
= (u32
)jiffies
;
416 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
417 * @t1: pointer to newest sample
418 * @t0: pointer to oldest sample
420 static inline u32
skb_mstamp_us_delta(const struct skb_mstamp
*t1
,
421 const struct skb_mstamp
*t0
)
423 s32 delta_us
= t1
->stamp_us
- t0
->stamp_us
;
424 u32 delta_jiffies
= t1
->stamp_jiffies
- t0
->stamp_jiffies
;
426 /* If delta_us is negative, this might be because interval is too big,
427 * or local_clock() drift is too big : fallback using jiffies.
430 delta_jiffies
>= (INT_MAX
/ (USEC_PER_SEC
/ HZ
)))
432 delta_us
= jiffies_to_usecs(delta_jiffies
);
439 * struct sk_buff - socket buffer
440 * @next: Next buffer in list
441 * @prev: Previous buffer in list
442 * @tstamp: Time we arrived/left
443 * @sk: Socket we are owned by
444 * @dev: Device we arrived on/are leaving by
445 * @cb: Control buffer. Free for use by every layer. Put private vars here
446 * @_skb_refdst: destination entry (with norefcount bit)
447 * @sp: the security path, used for xfrm
448 * @len: Length of actual data
449 * @data_len: Data length
450 * @mac_len: Length of link layer header
451 * @hdr_len: writable header length of cloned skb
452 * @csum: Checksum (must include start/offset pair)
453 * @csum_start: Offset from skb->head where checksumming should start
454 * @csum_offset: Offset from csum_start where checksum should be stored
455 * @priority: Packet queueing priority
456 * @ignore_df: allow local fragmentation
457 * @cloned: Head may be cloned (check refcnt to be sure)
458 * @ip_summed: Driver fed us an IP checksum
459 * @nohdr: Payload reference only, must not modify header
460 * @nfctinfo: Relationship of this skb to the connection
461 * @pkt_type: Packet class
462 * @fclone: skbuff clone status
463 * @ipvs_property: skbuff is owned by ipvs
464 * @peeked: this packet has been seen already, so stats have been
465 * done for it, don't do them again
466 * @nf_trace: netfilter packet trace flag
467 * @protocol: Packet protocol from driver
468 * @destructor: Destruct function
469 * @nfct: Associated connection, if any
470 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
471 * @skb_iif: ifindex of device we arrived on
472 * @tc_index: Traffic control index
473 * @tc_verd: traffic control verdict
474 * @hash: the packet hash
475 * @queue_mapping: Queue mapping for multiqueue devices
476 * @xmit_more: More SKBs are pending for this queue
477 * @ndisc_nodetype: router type (from link layer)
478 * @ooo_okay: allow the mapping of a socket to a queue to be changed
479 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
481 * @sw_hash: indicates hash was computed in software stack
482 * @wifi_acked_valid: wifi_acked was set
483 * @wifi_acked: whether frame was acked on wifi or not
484 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
485 * @napi_id: id of the NAPI struct this skb came from
486 * @secmark: security marking
487 * @mark: Generic packet mark
488 * @dropcount: total number of sk_receive_queue overflows
489 * @vlan_proto: vlan encapsulation protocol
490 * @vlan_tci: vlan tag control information
491 * @inner_protocol: Protocol (encapsulation)
492 * @inner_transport_header: Inner transport layer header (encapsulation)
493 * @inner_network_header: Network layer header (encapsulation)
494 * @inner_mac_header: Link layer header (encapsulation)
495 * @transport_header: Transport layer header
496 * @network_header: Network layer header
497 * @mac_header: Link layer header
498 * @tail: Tail pointer
500 * @head: Head of buffer
501 * @data: Data head pointer
502 * @truesize: Buffer size
503 * @users: User count - see {datagram,tcp}.c
507 /* These two members must be first. */
508 struct sk_buff
*next
;
509 struct sk_buff
*prev
;
513 struct skb_mstamp skb_mstamp
;
517 struct net_device
*dev
;
520 * This is the control buffer. It is free to use for every
521 * layer. Please put your private variables there. If you
522 * want to keep them across layers you have to do a skb_clone()
523 * first. This is owned by whoever has the skb queued ATM.
525 char cb
[48] __aligned(8);
527 unsigned long _skb_refdst
;
528 void (*destructor
)(struct sk_buff
*skb
);
532 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
533 struct nf_conntrack
*nfct
;
535 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
536 struct nf_bridge_info
*nf_bridge
;
543 /* Following fields are _not_ copied in __copy_skb_header()
544 * Note that queue_mapping is here mostly to fill a hole.
546 kmemcheck_bitfield_begin(flags1
);
555 kmemcheck_bitfield_end(flags1
);
557 /* fields enclosed in headers_start/headers_end are copied
558 * using a single memcpy() in __copy_skb_header()
560 __u32 headers_start
[0];
562 /* if you move pkt_type around you also must adapt those constants */
563 #ifdef __BIG_ENDIAN_BITFIELD
564 #define PKT_TYPE_MAX (7 << 5)
566 #define PKT_TYPE_MAX 7
568 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
570 __u8 __pkt_type_offset
[0];
581 __u8 wifi_acked_valid
:1;
585 /* Indicates the inner headers are valid in the skbuff. */
586 __u8 encapsulation
:1;
587 __u8 encap_hdr_csum
:1;
589 __u8 csum_complete_sw
:1;
593 #ifdef CONFIG_IPV6_NDISC_NODETYPE
594 __u8 ndisc_nodetype
:2;
596 __u8 ipvs_property
:1;
597 __u8 inner_protocol_type
:1;
598 /* 4 or 6 bit hole */
600 #ifdef CONFIG_NET_SCHED
601 __u16 tc_index
; /* traffic control index */
602 #ifdef CONFIG_NET_CLS_ACT
603 __u16 tc_verd
; /* traffic control verdict */
619 #ifdef CONFIG_NET_RX_BUSY_POLL
620 unsigned int napi_id
;
622 #ifdef CONFIG_NETWORK_SECMARK
628 __u32 reserved_tailroom
;
632 __be16 inner_protocol
;
636 __u16 inner_transport_header
;
637 __u16 inner_network_header
;
638 __u16 inner_mac_header
;
641 __u16 transport_header
;
642 __u16 network_header
;
645 __u32 headers_end
[0];
647 /* These elements must be at the end, see alloc_skb() for details. */
652 unsigned int truesize
;
658 * Handling routines are only of interest to the kernel
660 #include <linux/slab.h>
663 #define SKB_ALLOC_FCLONE 0x01
664 #define SKB_ALLOC_RX 0x02
666 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
667 static inline bool skb_pfmemalloc(const struct sk_buff
*skb
)
669 return unlikely(skb
->pfmemalloc
);
673 * skb might have a dst pointer attached, refcounted or not.
674 * _skb_refdst low order bit is set if refcount was _not_ taken
676 #define SKB_DST_NOREF 1UL
677 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
680 * skb_dst - returns skb dst_entry
683 * Returns skb dst_entry, regardless of reference taken or not.
685 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
687 /* If refdst was not refcounted, check we still are in a
688 * rcu_read_lock section
690 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
691 !rcu_read_lock_held() &&
692 !rcu_read_lock_bh_held());
693 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
697 * skb_dst_set - sets skb dst
701 * Sets skb dst, assuming a reference was taken on dst and should
702 * be released by skb_dst_drop()
704 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
706 skb
->_skb_refdst
= (unsigned long)dst
;
709 void __skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
,
713 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
717 * Sets skb dst, assuming a reference was not taken on dst.
718 * If dst entry is cached, we do not take reference and dst_release
719 * will be avoided by refdst_drop. If dst entry is not cached, we take
720 * reference, so that last dst_release can destroy the dst immediately.
722 static inline void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
)
724 __skb_dst_set_noref(skb
, dst
, false);
728 * skb_dst_set_noref_force - sets skb dst, without taking reference
732 * Sets skb dst, assuming a reference was not taken on dst.
733 * No reference is taken and no dst_release will be called. While for
734 * cached dsts deferred reclaim is a basic feature, for entries that are
735 * not cached it is caller's job to guarantee that last dst_release for
736 * provided dst happens when nobody uses it, eg. after a RCU grace period.
738 static inline void skb_dst_set_noref_force(struct sk_buff
*skb
,
739 struct dst_entry
*dst
)
741 __skb_dst_set_noref(skb
, dst
, true);
745 * skb_dst_is_noref - Test if skb dst isn't refcounted
748 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
750 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
753 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
755 return (struct rtable
*)skb_dst(skb
);
758 void kfree_skb(struct sk_buff
*skb
);
759 void kfree_skb_list(struct sk_buff
*segs
);
760 void skb_tx_error(struct sk_buff
*skb
);
761 void consume_skb(struct sk_buff
*skb
);
762 void __kfree_skb(struct sk_buff
*skb
);
763 extern struct kmem_cache
*skbuff_head_cache
;
765 void kfree_skb_partial(struct sk_buff
*skb
, bool head_stolen
);
766 bool skb_try_coalesce(struct sk_buff
*to
, struct sk_buff
*from
,
767 bool *fragstolen
, int *delta_truesize
);
769 struct sk_buff
*__alloc_skb(unsigned int size
, gfp_t priority
, int flags
,
771 struct sk_buff
*build_skb(void *data
, unsigned int frag_size
);
772 static inline struct sk_buff
*alloc_skb(unsigned int size
,
775 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
778 struct sk_buff
*alloc_skb_with_frags(unsigned long header_len
,
779 unsigned long data_len
,
784 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
785 struct sk_buff_fclones
{
794 * skb_fclone_busy - check if fclone is busy
797 * Returns true is skb is a fast clone, and its clone is not freed.
799 static inline bool skb_fclone_busy(const struct sk_buff
*skb
)
801 const struct sk_buff_fclones
*fclones
;
803 fclones
= container_of(skb
, struct sk_buff_fclones
, skb1
);
805 return skb
->fclone
== SKB_FCLONE_ORIG
&&
806 fclones
->skb2
.fclone
== SKB_FCLONE_CLONE
;
809 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
812 return __alloc_skb(size
, priority
, SKB_ALLOC_FCLONE
, NUMA_NO_NODE
);
815 struct sk_buff
*__alloc_skb_head(gfp_t priority
, int node
);
816 static inline struct sk_buff
*alloc_skb_head(gfp_t priority
)
818 return __alloc_skb_head(priority
, -1);
821 struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
822 int skb_copy_ubufs(struct sk_buff
*skb
, gfp_t gfp_mask
);
823 struct sk_buff
*skb_clone(struct sk_buff
*skb
, gfp_t priority
);
824 struct sk_buff
*skb_copy(const struct sk_buff
*skb
, gfp_t priority
);
825 struct sk_buff
*__pskb_copy_fclone(struct sk_buff
*skb
, int headroom
,
826 gfp_t gfp_mask
, bool fclone
);
827 static inline struct sk_buff
*__pskb_copy(struct sk_buff
*skb
, int headroom
,
830 return __pskb_copy_fclone(skb
, headroom
, gfp_mask
, false);
833 int pskb_expand_head(struct sk_buff
*skb
, int nhead
, int ntail
, gfp_t gfp_mask
);
834 struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
835 unsigned int headroom
);
836 struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
, int newheadroom
,
837 int newtailroom
, gfp_t priority
);
838 int skb_to_sgvec_nomark(struct sk_buff
*skb
, struct scatterlist
*sg
,
839 int offset
, int len
);
840 int skb_to_sgvec(struct sk_buff
*skb
, struct scatterlist
*sg
, int offset
,
842 int skb_cow_data(struct sk_buff
*skb
, int tailbits
, struct sk_buff
**trailer
);
843 int skb_pad(struct sk_buff
*skb
, int pad
);
844 #define dev_kfree_skb(a) consume_skb(a)
846 int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
847 int getfrag(void *from
, char *to
, int offset
,
848 int len
, int odd
, struct sk_buff
*skb
),
849 void *from
, int length
);
851 struct skb_seq_state
{
855 __u32 stepped_offset
;
856 struct sk_buff
*root_skb
;
857 struct sk_buff
*cur_skb
;
861 void skb_prepare_seq_read(struct sk_buff
*skb
, unsigned int from
,
862 unsigned int to
, struct skb_seq_state
*st
);
863 unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
864 struct skb_seq_state
*st
);
865 void skb_abort_seq_read(struct skb_seq_state
*st
);
867 unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
868 unsigned int to
, struct ts_config
*config
,
869 struct ts_state
*state
);
872 * Packet hash types specify the type of hash in skb_set_hash.
874 * Hash types refer to the protocol layer addresses which are used to
875 * construct a packet's hash. The hashes are used to differentiate or identify
876 * flows of the protocol layer for the hash type. Hash types are either
877 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
879 * Properties of hashes:
881 * 1) Two packets in different flows have different hash values
882 * 2) Two packets in the same flow should have the same hash value
884 * A hash at a higher layer is considered to be more specific. A driver should
885 * set the most specific hash possible.
887 * A driver cannot indicate a more specific hash than the layer at which a hash
888 * was computed. For instance an L3 hash cannot be set as an L4 hash.
890 * A driver may indicate a hash level which is less specific than the
891 * actual layer the hash was computed on. For instance, a hash computed
892 * at L4 may be considered an L3 hash. This should only be done if the
893 * driver can't unambiguously determine that the HW computed the hash at
894 * the higher layer. Note that the "should" in the second property above
897 enum pkt_hash_types
{
898 PKT_HASH_TYPE_NONE
, /* Undefined type */
899 PKT_HASH_TYPE_L2
, /* Input: src_MAC, dest_MAC */
900 PKT_HASH_TYPE_L3
, /* Input: src_IP, dst_IP */
901 PKT_HASH_TYPE_L4
, /* Input: src_IP, dst_IP, src_port, dst_port */
905 skb_set_hash(struct sk_buff
*skb
, __u32 hash
, enum pkt_hash_types type
)
907 skb
->l4_hash
= (type
== PKT_HASH_TYPE_L4
);
912 void __skb_get_hash(struct sk_buff
*skb
);
913 static inline __u32
skb_get_hash(struct sk_buff
*skb
)
915 if (!skb
->l4_hash
&& !skb
->sw_hash
)
921 static inline __u32
skb_get_hash_raw(const struct sk_buff
*skb
)
926 static inline void skb_clear_hash(struct sk_buff
*skb
)
933 static inline void skb_clear_hash_if_not_l4(struct sk_buff
*skb
)
939 static inline void skb_copy_hash(struct sk_buff
*to
, const struct sk_buff
*from
)
941 to
->hash
= from
->hash
;
942 to
->sw_hash
= from
->sw_hash
;
943 to
->l4_hash
= from
->l4_hash
;
946 #ifdef NET_SKBUFF_DATA_USES_OFFSET
947 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
949 return skb
->head
+ skb
->end
;
952 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
957 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
962 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
964 return skb
->end
- skb
->head
;
969 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
971 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
973 return &skb_shinfo(skb
)->hwtstamps
;
977 * skb_queue_empty - check if a queue is empty
980 * Returns true if the queue is empty, false otherwise.
982 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
984 return list
->next
== (const struct sk_buff
*) list
;
988 * skb_queue_is_last - check if skb is the last entry in the queue
992 * Returns true if @skb is the last buffer on the list.
994 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
995 const struct sk_buff
*skb
)
997 return skb
->next
== (const struct sk_buff
*) list
;
1001 * skb_queue_is_first - check if skb is the first entry in the queue
1005 * Returns true if @skb is the first buffer on the list.
1007 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
1008 const struct sk_buff
*skb
)
1010 return skb
->prev
== (const struct sk_buff
*) list
;
1014 * skb_queue_next - return the next packet in the queue
1016 * @skb: current buffer
1018 * Return the next packet in @list after @skb. It is only valid to
1019 * call this if skb_queue_is_last() evaluates to false.
1021 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
1022 const struct sk_buff
*skb
)
1024 /* This BUG_ON may seem severe, but if we just return then we
1025 * are going to dereference garbage.
1027 BUG_ON(skb_queue_is_last(list
, skb
));
1032 * skb_queue_prev - return the prev packet in the queue
1034 * @skb: current buffer
1036 * Return the prev packet in @list before @skb. It is only valid to
1037 * call this if skb_queue_is_first() evaluates to false.
1039 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
1040 const struct sk_buff
*skb
)
1042 /* This BUG_ON may seem severe, but if we just return then we
1043 * are going to dereference garbage.
1045 BUG_ON(skb_queue_is_first(list
, skb
));
1050 * skb_get - reference buffer
1051 * @skb: buffer to reference
1053 * Makes another reference to a socket buffer and returns a pointer
1056 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
1058 atomic_inc(&skb
->users
);
1063 * If users == 1, we are the only owner and are can avoid redundant
1068 * skb_cloned - is the buffer a clone
1069 * @skb: buffer to check
1071 * Returns true if the buffer was generated with skb_clone() and is
1072 * one of multiple shared copies of the buffer. Cloned buffers are
1073 * shared data so must not be written to under normal circumstances.
1075 static inline int skb_cloned(const struct sk_buff
*skb
)
1077 return skb
->cloned
&&
1078 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
1081 static inline int skb_unclone(struct sk_buff
*skb
, gfp_t pri
)
1083 might_sleep_if(pri
& __GFP_WAIT
);
1085 if (skb_cloned(skb
))
1086 return pskb_expand_head(skb
, 0, 0, pri
);
1092 * skb_header_cloned - is the header a clone
1093 * @skb: buffer to check
1095 * Returns true if modifying the header part of the buffer requires
1096 * the data to be copied.
1098 static inline int skb_header_cloned(const struct sk_buff
*skb
)
1105 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
1106 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
1107 return dataref
!= 1;
1111 * skb_header_release - release reference to header
1112 * @skb: buffer to operate on
1114 * Drop a reference to the header part of the buffer. This is done
1115 * by acquiring a payload reference. You must not read from the header
1116 * part of skb->data after this.
1117 * Note : Check if you can use __skb_header_release() instead.
1119 static inline void skb_header_release(struct sk_buff
*skb
)
1123 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
1127 * __skb_header_release - release reference to header
1128 * @skb: buffer to operate on
1130 * Variant of skb_header_release() assuming skb is private to caller.
1131 * We can avoid one atomic operation.
1133 static inline void __skb_header_release(struct sk_buff
*skb
)
1136 atomic_set(&skb_shinfo(skb
)->dataref
, 1 + (1 << SKB_DATAREF_SHIFT
));
1141 * skb_shared - is the buffer shared
1142 * @skb: buffer to check
1144 * Returns true if more than one person has a reference to this
1147 static inline int skb_shared(const struct sk_buff
*skb
)
1149 return atomic_read(&skb
->users
) != 1;
1153 * skb_share_check - check if buffer is shared and if so clone it
1154 * @skb: buffer to check
1155 * @pri: priority for memory allocation
1157 * If the buffer is shared the buffer is cloned and the old copy
1158 * drops a reference. A new clone with a single reference is returned.
1159 * If the buffer is not shared the original buffer is returned. When
1160 * being called from interrupt status or with spinlocks held pri must
1163 * NULL is returned on a memory allocation failure.
1165 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
, gfp_t pri
)
1167 might_sleep_if(pri
& __GFP_WAIT
);
1168 if (skb_shared(skb
)) {
1169 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
1181 * Copy shared buffers into a new sk_buff. We effectively do COW on
1182 * packets to handle cases where we have a local reader and forward
1183 * and a couple of other messy ones. The normal one is tcpdumping
1184 * a packet thats being forwarded.
1188 * skb_unshare - make a copy of a shared buffer
1189 * @skb: buffer to check
1190 * @pri: priority for memory allocation
1192 * If the socket buffer is a clone then this function creates a new
1193 * copy of the data, drops a reference count on the old copy and returns
1194 * the new copy with the reference count at 1. If the buffer is not a clone
1195 * the original buffer is returned. When called with a spinlock held or
1196 * from interrupt state @pri must be %GFP_ATOMIC
1198 * %NULL is returned on a memory allocation failure.
1200 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
1203 might_sleep_if(pri
& __GFP_WAIT
);
1204 if (skb_cloned(skb
)) {
1205 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
1207 /* Free our shared copy */
1218 * skb_peek - peek at the head of an &sk_buff_head
1219 * @list_: list to peek at
1221 * Peek an &sk_buff. Unlike most other operations you _MUST_
1222 * be careful with this one. A peek leaves the buffer on the
1223 * list and someone else may run off with it. You must hold
1224 * the appropriate locks or have a private queue to do this.
1226 * Returns %NULL for an empty list or a pointer to the head element.
1227 * The reference count is not incremented and the reference is therefore
1228 * volatile. Use with caution.
1230 static inline struct sk_buff
*skb_peek(const struct sk_buff_head
*list_
)
1232 struct sk_buff
*skb
= list_
->next
;
1234 if (skb
== (struct sk_buff
*)list_
)
1240 * skb_peek_next - peek skb following the given one from a queue
1241 * @skb: skb to start from
1242 * @list_: list to peek at
1244 * Returns %NULL when the end of the list is met or a pointer to the
1245 * next element. The reference count is not incremented and the
1246 * reference is therefore volatile. Use with caution.
1248 static inline struct sk_buff
*skb_peek_next(struct sk_buff
*skb
,
1249 const struct sk_buff_head
*list_
)
1251 struct sk_buff
*next
= skb
->next
;
1253 if (next
== (struct sk_buff
*)list_
)
1259 * skb_peek_tail - peek at the tail of an &sk_buff_head
1260 * @list_: list to peek at
1262 * Peek an &sk_buff. Unlike most other operations you _MUST_
1263 * be careful with this one. A peek leaves the buffer on the
1264 * list and someone else may run off with it. You must hold
1265 * the appropriate locks or have a private queue to do this.
1267 * Returns %NULL for an empty list or a pointer to the tail element.
1268 * The reference count is not incremented and the reference is therefore
1269 * volatile. Use with caution.
1271 static inline struct sk_buff
*skb_peek_tail(const struct sk_buff_head
*list_
)
1273 struct sk_buff
*skb
= list_
->prev
;
1275 if (skb
== (struct sk_buff
*)list_
)
1282 * skb_queue_len - get queue length
1283 * @list_: list to measure
1285 * Return the length of an &sk_buff queue.
1287 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
1293 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1294 * @list: queue to initialize
1296 * This initializes only the list and queue length aspects of
1297 * an sk_buff_head object. This allows to initialize the list
1298 * aspects of an sk_buff_head without reinitializing things like
1299 * the spinlock. It can also be used for on-stack sk_buff_head
1300 * objects where the spinlock is known to not be used.
1302 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
1304 list
->prev
= list
->next
= (struct sk_buff
*)list
;
1309 * This function creates a split out lock class for each invocation;
1310 * this is needed for now since a whole lot of users of the skb-queue
1311 * infrastructure in drivers have different locking usage (in hardirq)
1312 * than the networking core (in softirq only). In the long run either the
1313 * network layer or drivers should need annotation to consolidate the
1314 * main types of usage into 3 classes.
1316 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
1318 spin_lock_init(&list
->lock
);
1319 __skb_queue_head_init(list
);
1322 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
1323 struct lock_class_key
*class)
1325 skb_queue_head_init(list
);
1326 lockdep_set_class(&list
->lock
, class);
1330 * Insert an sk_buff on a list.
1332 * The "__skb_xxxx()" functions are the non-atomic ones that
1333 * can only be called with interrupts disabled.
1335 void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
,
1336 struct sk_buff_head
*list
);
1337 static inline void __skb_insert(struct sk_buff
*newsk
,
1338 struct sk_buff
*prev
, struct sk_buff
*next
,
1339 struct sk_buff_head
*list
)
1343 next
->prev
= prev
->next
= newsk
;
1347 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
1348 struct sk_buff
*prev
,
1349 struct sk_buff
*next
)
1351 struct sk_buff
*first
= list
->next
;
1352 struct sk_buff
*last
= list
->prev
;
1362 * skb_queue_splice - join two skb lists, this is designed for stacks
1363 * @list: the new list to add
1364 * @head: the place to add it in the first list
1366 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
1367 struct sk_buff_head
*head
)
1369 if (!skb_queue_empty(list
)) {
1370 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1371 head
->qlen
+= list
->qlen
;
1376 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1377 * @list: the new list to add
1378 * @head: the place to add it in the first list
1380 * The list at @list is reinitialised
1382 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
1383 struct sk_buff_head
*head
)
1385 if (!skb_queue_empty(list
)) {
1386 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1387 head
->qlen
+= list
->qlen
;
1388 __skb_queue_head_init(list
);
1393 * skb_queue_splice_tail - join two skb lists, each list being a queue
1394 * @list: the new list to add
1395 * @head: the place to add it in the first list
1397 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
1398 struct sk_buff_head
*head
)
1400 if (!skb_queue_empty(list
)) {
1401 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1402 head
->qlen
+= list
->qlen
;
1407 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1408 * @list: the new list to add
1409 * @head: the place to add it in the first list
1411 * Each of the lists is a queue.
1412 * The list at @list is reinitialised
1414 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
1415 struct sk_buff_head
*head
)
1417 if (!skb_queue_empty(list
)) {
1418 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1419 head
->qlen
+= list
->qlen
;
1420 __skb_queue_head_init(list
);
1425 * __skb_queue_after - queue a buffer at the list head
1426 * @list: list to use
1427 * @prev: place after this buffer
1428 * @newsk: buffer to queue
1430 * Queue a buffer int the middle of a list. This function takes no locks
1431 * and you must therefore hold required locks before calling it.
1433 * A buffer cannot be placed on two lists at the same time.
1435 static inline void __skb_queue_after(struct sk_buff_head
*list
,
1436 struct sk_buff
*prev
,
1437 struct sk_buff
*newsk
)
1439 __skb_insert(newsk
, prev
, prev
->next
, list
);
1442 void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
1443 struct sk_buff_head
*list
);
1445 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1446 struct sk_buff
*next
,
1447 struct sk_buff
*newsk
)
1449 __skb_insert(newsk
, next
->prev
, next
, list
);
1453 * __skb_queue_head - queue a buffer at the list head
1454 * @list: list to use
1455 * @newsk: buffer to queue
1457 * Queue a buffer at the start of a list. This function takes no locks
1458 * and you must therefore hold required locks before calling it.
1460 * A buffer cannot be placed on two lists at the same time.
1462 void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1463 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1464 struct sk_buff
*newsk
)
1466 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1470 * __skb_queue_tail - queue a buffer at the list tail
1471 * @list: list to use
1472 * @newsk: buffer to queue
1474 * Queue a buffer at the end of a list. This function takes no locks
1475 * and you must therefore hold required locks before calling it.
1477 * A buffer cannot be placed on two lists at the same time.
1479 void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1480 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1481 struct sk_buff
*newsk
)
1483 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1487 * remove sk_buff from list. _Must_ be called atomically, and with
1490 void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1491 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1493 struct sk_buff
*next
, *prev
;
1498 skb
->next
= skb
->prev
= NULL
;
1504 * __skb_dequeue - remove from the head of the queue
1505 * @list: list to dequeue from
1507 * Remove the head of the list. This function does not take any locks
1508 * so must be used with appropriate locks held only. The head item is
1509 * returned or %NULL if the list is empty.
1511 struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1512 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1514 struct sk_buff
*skb
= skb_peek(list
);
1516 __skb_unlink(skb
, list
);
1521 * __skb_dequeue_tail - remove from the tail of the queue
1522 * @list: list to dequeue from
1524 * Remove the tail of the list. This function does not take any locks
1525 * so must be used with appropriate locks held only. The tail item is
1526 * returned or %NULL if the list is empty.
1528 struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1529 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1531 struct sk_buff
*skb
= skb_peek_tail(list
);
1533 __skb_unlink(skb
, list
);
1538 static inline bool skb_is_nonlinear(const struct sk_buff
*skb
)
1540 return skb
->data_len
;
1543 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1545 return skb
->len
- skb
->data_len
;
1548 static inline int skb_pagelen(const struct sk_buff
*skb
)
1552 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1553 len
+= skb_frag_size(&skb_shinfo(skb
)->frags
[i
]);
1554 return len
+ skb_headlen(skb
);
1558 * __skb_fill_page_desc - initialise a paged fragment in an skb
1559 * @skb: buffer containing fragment to be initialised
1560 * @i: paged fragment index to initialise
1561 * @page: the page to use for this fragment
1562 * @off: the offset to the data with @page
1563 * @size: the length of the data
1565 * Initialises the @i'th fragment of @skb to point to &size bytes at
1566 * offset @off within @page.
1568 * Does not take any additional reference on the fragment.
1570 static inline void __skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1571 struct page
*page
, int off
, int size
)
1573 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1576 * Propagate page->pfmemalloc to the skb if we can. The problem is
1577 * that not all callers have unique ownership of the page. If
1578 * pfmemalloc is set, we check the mapping as a mapping implies
1579 * page->index is set (index and pfmemalloc share space).
1580 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1581 * do not lose pfmemalloc information as the pages would not be
1582 * allocated using __GFP_MEMALLOC.
1584 frag
->page
.p
= page
;
1585 frag
->page_offset
= off
;
1586 skb_frag_size_set(frag
, size
);
1588 page
= compound_head(page
);
1589 if (page
->pfmemalloc
&& !page
->mapping
)
1590 skb
->pfmemalloc
= true;
1594 * skb_fill_page_desc - initialise a paged fragment in an skb
1595 * @skb: buffer containing fragment to be initialised
1596 * @i: paged fragment index to initialise
1597 * @page: the page to use for this fragment
1598 * @off: the offset to the data with @page
1599 * @size: the length of the data
1601 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1602 * @skb to point to @size bytes at offset @off within @page. In
1603 * addition updates @skb such that @i is the last fragment.
1605 * Does not take any additional reference on the fragment.
1607 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1608 struct page
*page
, int off
, int size
)
1610 __skb_fill_page_desc(skb
, i
, page
, off
, size
);
1611 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1614 void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
, int off
,
1615 int size
, unsigned int truesize
);
1617 void skb_coalesce_rx_frag(struct sk_buff
*skb
, int i
, int size
,
1618 unsigned int truesize
);
1620 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1621 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1622 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1624 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1625 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1627 return skb
->head
+ skb
->tail
;
1630 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1632 skb
->tail
= skb
->data
- skb
->head
;
1635 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1637 skb_reset_tail_pointer(skb
);
1638 skb
->tail
+= offset
;
1641 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1642 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1647 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1649 skb
->tail
= skb
->data
;
1652 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1654 skb
->tail
= skb
->data
+ offset
;
1657 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1660 * Add data to an sk_buff
1662 unsigned char *pskb_put(struct sk_buff
*skb
, struct sk_buff
*tail
, int len
);
1663 unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1664 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1666 unsigned char *tmp
= skb_tail_pointer(skb
);
1667 SKB_LINEAR_ASSERT(skb
);
1673 unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1674 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1681 unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1682 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1685 BUG_ON(skb
->len
< skb
->data_len
);
1686 return skb
->data
+= len
;
1689 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1691 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1694 unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1696 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1698 if (len
> skb_headlen(skb
) &&
1699 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1702 return skb
->data
+= len
;
1705 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1707 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1710 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1712 if (likely(len
<= skb_headlen(skb
)))
1714 if (unlikely(len
> skb
->len
))
1716 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1720 * skb_headroom - bytes at buffer head
1721 * @skb: buffer to check
1723 * Return the number of bytes of free space at the head of an &sk_buff.
1725 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1727 return skb
->data
- skb
->head
;
1731 * skb_tailroom - bytes at buffer end
1732 * @skb: buffer to check
1734 * Return the number of bytes of free space at the tail of an sk_buff
1736 static inline int skb_tailroom(const struct sk_buff
*skb
)
1738 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1742 * skb_availroom - bytes at buffer end
1743 * @skb: buffer to check
1745 * Return the number of bytes of free space at the tail of an sk_buff
1746 * allocated by sk_stream_alloc()
1748 static inline int skb_availroom(const struct sk_buff
*skb
)
1750 if (skb_is_nonlinear(skb
))
1753 return skb
->end
- skb
->tail
- skb
->reserved_tailroom
;
1757 * skb_reserve - adjust headroom
1758 * @skb: buffer to alter
1759 * @len: bytes to move
1761 * Increase the headroom of an empty &sk_buff by reducing the tail
1762 * room. This is only allowed for an empty buffer.
1764 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
1770 #define ENCAP_TYPE_ETHER 0
1771 #define ENCAP_TYPE_IPPROTO 1
1773 static inline void skb_set_inner_protocol(struct sk_buff
*skb
,
1776 skb
->inner_protocol
= protocol
;
1777 skb
->inner_protocol_type
= ENCAP_TYPE_ETHER
;
1780 static inline void skb_set_inner_ipproto(struct sk_buff
*skb
,
1783 skb
->inner_ipproto
= ipproto
;
1784 skb
->inner_protocol_type
= ENCAP_TYPE_IPPROTO
;
1787 static inline void skb_reset_inner_headers(struct sk_buff
*skb
)
1789 skb
->inner_mac_header
= skb
->mac_header
;
1790 skb
->inner_network_header
= skb
->network_header
;
1791 skb
->inner_transport_header
= skb
->transport_header
;
1794 static inline void skb_reset_mac_len(struct sk_buff
*skb
)
1796 skb
->mac_len
= skb
->network_header
- skb
->mac_header
;
1799 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1802 return skb
->head
+ skb
->inner_transport_header
;
1805 static inline void skb_reset_inner_transport_header(struct sk_buff
*skb
)
1807 skb
->inner_transport_header
= skb
->data
- skb
->head
;
1810 static inline void skb_set_inner_transport_header(struct sk_buff
*skb
,
1813 skb_reset_inner_transport_header(skb
);
1814 skb
->inner_transport_header
+= offset
;
1817 static inline unsigned char *skb_inner_network_header(const struct sk_buff
*skb
)
1819 return skb
->head
+ skb
->inner_network_header
;
1822 static inline void skb_reset_inner_network_header(struct sk_buff
*skb
)
1824 skb
->inner_network_header
= skb
->data
- skb
->head
;
1827 static inline void skb_set_inner_network_header(struct sk_buff
*skb
,
1830 skb_reset_inner_network_header(skb
);
1831 skb
->inner_network_header
+= offset
;
1834 static inline unsigned char *skb_inner_mac_header(const struct sk_buff
*skb
)
1836 return skb
->head
+ skb
->inner_mac_header
;
1839 static inline void skb_reset_inner_mac_header(struct sk_buff
*skb
)
1841 skb
->inner_mac_header
= skb
->data
- skb
->head
;
1844 static inline void skb_set_inner_mac_header(struct sk_buff
*skb
,
1847 skb_reset_inner_mac_header(skb
);
1848 skb
->inner_mac_header
+= offset
;
1850 static inline bool skb_transport_header_was_set(const struct sk_buff
*skb
)
1852 return skb
->transport_header
!= (typeof(skb
->transport_header
))~0U;
1855 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1857 return skb
->head
+ skb
->transport_header
;
1860 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1862 skb
->transport_header
= skb
->data
- skb
->head
;
1865 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1868 skb_reset_transport_header(skb
);
1869 skb
->transport_header
+= offset
;
1872 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1874 return skb
->head
+ skb
->network_header
;
1877 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1879 skb
->network_header
= skb
->data
- skb
->head
;
1882 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1884 skb_reset_network_header(skb
);
1885 skb
->network_header
+= offset
;
1888 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1890 return skb
->head
+ skb
->mac_header
;
1893 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1895 return skb
->mac_header
!= (typeof(skb
->mac_header
))~0U;
1898 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1900 skb
->mac_header
= skb
->data
- skb
->head
;
1903 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1905 skb_reset_mac_header(skb
);
1906 skb
->mac_header
+= offset
;
1909 static inline void skb_pop_mac_header(struct sk_buff
*skb
)
1911 skb
->mac_header
= skb
->network_header
;
1914 static inline void skb_probe_transport_header(struct sk_buff
*skb
,
1915 const int offset_hint
)
1917 struct flow_keys keys
;
1919 if (skb_transport_header_was_set(skb
))
1921 else if (skb_flow_dissect(skb
, &keys
))
1922 skb_set_transport_header(skb
, keys
.thoff
);
1924 skb_set_transport_header(skb
, offset_hint
);
1927 static inline void skb_mac_header_rebuild(struct sk_buff
*skb
)
1929 if (skb_mac_header_was_set(skb
)) {
1930 const unsigned char *old_mac
= skb_mac_header(skb
);
1932 skb_set_mac_header(skb
, -skb
->mac_len
);
1933 memmove(skb_mac_header(skb
), old_mac
, skb
->mac_len
);
1937 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
1939 return skb
->csum_start
- skb_headroom(skb
);
1942 static inline int skb_transport_offset(const struct sk_buff
*skb
)
1944 return skb_transport_header(skb
) - skb
->data
;
1947 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
1949 return skb
->transport_header
- skb
->network_header
;
1952 static inline u32
skb_inner_network_header_len(const struct sk_buff
*skb
)
1954 return skb
->inner_transport_header
- skb
->inner_network_header
;
1957 static inline int skb_network_offset(const struct sk_buff
*skb
)
1959 return skb_network_header(skb
) - skb
->data
;
1962 static inline int skb_inner_network_offset(const struct sk_buff
*skb
)
1964 return skb_inner_network_header(skb
) - skb
->data
;
1967 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
1969 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
1973 * CPUs often take a performance hit when accessing unaligned memory
1974 * locations. The actual performance hit varies, it can be small if the
1975 * hardware handles it or large if we have to take an exception and fix it
1978 * Since an ethernet header is 14 bytes network drivers often end up with
1979 * the IP header at an unaligned offset. The IP header can be aligned by
1980 * shifting the start of the packet by 2 bytes. Drivers should do this
1983 * skb_reserve(skb, NET_IP_ALIGN);
1985 * The downside to this alignment of the IP header is that the DMA is now
1986 * unaligned. On some architectures the cost of an unaligned DMA is high
1987 * and this cost outweighs the gains made by aligning the IP header.
1989 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1992 #ifndef NET_IP_ALIGN
1993 #define NET_IP_ALIGN 2
1997 * The networking layer reserves some headroom in skb data (via
1998 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1999 * the header has to grow. In the default case, if the header has to grow
2000 * 32 bytes or less we avoid the reallocation.
2002 * Unfortunately this headroom changes the DMA alignment of the resulting
2003 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2004 * on some architectures. An architecture can override this value,
2005 * perhaps setting it to a cacheline in size (since that will maintain
2006 * cacheline alignment of the DMA). It must be a power of 2.
2008 * Various parts of the networking layer expect at least 32 bytes of
2009 * headroom, you should not reduce this.
2011 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2012 * to reduce average number of cache lines per packet.
2013 * get_rps_cpus() for example only access one 64 bytes aligned block :
2014 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2017 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2020 int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
2022 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
2024 if (unlikely(skb_is_nonlinear(skb
))) {
2029 skb_set_tail_pointer(skb
, len
);
2032 void skb_trim(struct sk_buff
*skb
, unsigned int len
);
2034 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2037 return ___pskb_trim(skb
, len
);
2038 __skb_trim(skb
, len
);
2042 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2044 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
2048 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2049 * @skb: buffer to alter
2052 * This is identical to pskb_trim except that the caller knows that
2053 * the skb is not cloned so we should never get an error due to out-
2056 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
2058 int err
= pskb_trim(skb
, len
);
2063 * skb_orphan - orphan a buffer
2064 * @skb: buffer to orphan
2066 * If a buffer currently has an owner then we call the owner's
2067 * destructor function and make the @skb unowned. The buffer continues
2068 * to exist but is no longer charged to its former owner.
2070 static inline void skb_orphan(struct sk_buff
*skb
)
2072 if (skb
->destructor
) {
2073 skb
->destructor(skb
);
2074 skb
->destructor
= NULL
;
2082 * skb_orphan_frags - orphan the frags contained in a buffer
2083 * @skb: buffer to orphan frags from
2084 * @gfp_mask: allocation mask for replacement pages
2086 * For each frag in the SKB which needs a destructor (i.e. has an
2087 * owner) create a copy of that frag and release the original
2088 * page by calling the destructor.
2090 static inline int skb_orphan_frags(struct sk_buff
*skb
, gfp_t gfp_mask
)
2092 if (likely(!(skb_shinfo(skb
)->tx_flags
& SKBTX_DEV_ZEROCOPY
)))
2094 return skb_copy_ubufs(skb
, gfp_mask
);
2098 * __skb_queue_purge - empty a list
2099 * @list: list to empty
2101 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2102 * the list and one reference dropped. This function does not take the
2103 * list lock and the caller must hold the relevant locks to use it.
2105 void skb_queue_purge(struct sk_buff_head
*list
);
2106 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
2108 struct sk_buff
*skb
;
2109 while ((skb
= __skb_dequeue(list
)) != NULL
)
2113 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2114 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2115 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2117 void *netdev_alloc_frag(unsigned int fragsz
);
2119 struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
, unsigned int length
,
2123 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2124 * @dev: network device to receive on
2125 * @length: length to allocate
2127 * Allocate a new &sk_buff and assign it a usage count of one. The
2128 * buffer has unspecified headroom built in. Users should allocate
2129 * the headroom they think they need without accounting for the
2130 * built in space. The built in space is used for optimisations.
2132 * %NULL is returned if there is no free memory. Although this function
2133 * allocates memory it can be called from an interrupt.
2135 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
2136 unsigned int length
)
2138 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
2141 /* legacy helper around __netdev_alloc_skb() */
2142 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
2145 return __netdev_alloc_skb(NULL
, length
, gfp_mask
);
2148 /* legacy helper around netdev_alloc_skb() */
2149 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
2151 return netdev_alloc_skb(NULL
, length
);
2155 static inline struct sk_buff
*__netdev_alloc_skb_ip_align(struct net_device
*dev
,
2156 unsigned int length
, gfp_t gfp
)
2158 struct sk_buff
*skb
= __netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
, gfp
);
2160 if (NET_IP_ALIGN
&& skb
)
2161 skb_reserve(skb
, NET_IP_ALIGN
);
2165 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
2166 unsigned int length
)
2168 return __netdev_alloc_skb_ip_align(dev
, length
, GFP_ATOMIC
);
2172 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
2173 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2174 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2175 * @order: size of the allocation
2177 * Allocate a new page.
2179 * %NULL is returned if there is no free memory.
2181 static inline struct page
*__skb_alloc_pages(gfp_t gfp_mask
,
2182 struct sk_buff
*skb
,
2187 gfp_mask
|= __GFP_COLD
;
2189 if (!(gfp_mask
& __GFP_NOMEMALLOC
))
2190 gfp_mask
|= __GFP_MEMALLOC
;
2192 page
= alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, order
);
2193 if (skb
&& page
&& page
->pfmemalloc
)
2194 skb
->pfmemalloc
= true;
2200 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2201 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2202 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2204 * Allocate a new page.
2206 * %NULL is returned if there is no free memory.
2208 static inline struct page
*__skb_alloc_page(gfp_t gfp_mask
,
2209 struct sk_buff
*skb
)
2211 return __skb_alloc_pages(gfp_mask
, skb
, 0);
2215 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2216 * @page: The page that was allocated from skb_alloc_page
2217 * @skb: The skb that may need pfmemalloc set
2219 static inline void skb_propagate_pfmemalloc(struct page
*page
,
2220 struct sk_buff
*skb
)
2222 if (page
&& page
->pfmemalloc
)
2223 skb
->pfmemalloc
= true;
2227 * skb_frag_page - retrieve the page referred to by a paged fragment
2228 * @frag: the paged fragment
2230 * Returns the &struct page associated with @frag.
2232 static inline struct page
*skb_frag_page(const skb_frag_t
*frag
)
2234 return frag
->page
.p
;
2238 * __skb_frag_ref - take an addition reference on a paged fragment.
2239 * @frag: the paged fragment
2241 * Takes an additional reference on the paged fragment @frag.
2243 static inline void __skb_frag_ref(skb_frag_t
*frag
)
2245 get_page(skb_frag_page(frag
));
2249 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2251 * @f: the fragment offset.
2253 * Takes an additional reference on the @f'th paged fragment of @skb.
2255 static inline void skb_frag_ref(struct sk_buff
*skb
, int f
)
2257 __skb_frag_ref(&skb_shinfo(skb
)->frags
[f
]);
2261 * __skb_frag_unref - release a reference on a paged fragment.
2262 * @frag: the paged fragment
2264 * Releases a reference on the paged fragment @frag.
2266 static inline void __skb_frag_unref(skb_frag_t
*frag
)
2268 put_page(skb_frag_page(frag
));
2272 * skb_frag_unref - release a reference on a paged fragment of an skb.
2274 * @f: the fragment offset
2276 * Releases a reference on the @f'th paged fragment of @skb.
2278 static inline void skb_frag_unref(struct sk_buff
*skb
, int f
)
2280 __skb_frag_unref(&skb_shinfo(skb
)->frags
[f
]);
2284 * skb_frag_address - gets the address of the data contained in a paged fragment
2285 * @frag: the paged fragment buffer
2287 * Returns the address of the data within @frag. The page must already
2290 static inline void *skb_frag_address(const skb_frag_t
*frag
)
2292 return page_address(skb_frag_page(frag
)) + frag
->page_offset
;
2296 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2297 * @frag: the paged fragment buffer
2299 * Returns the address of the data within @frag. Checks that the page
2300 * is mapped and returns %NULL otherwise.
2302 static inline void *skb_frag_address_safe(const skb_frag_t
*frag
)
2304 void *ptr
= page_address(skb_frag_page(frag
));
2308 return ptr
+ frag
->page_offset
;
2312 * __skb_frag_set_page - sets the page contained in a paged fragment
2313 * @frag: the paged fragment
2314 * @page: the page to set
2316 * Sets the fragment @frag to contain @page.
2318 static inline void __skb_frag_set_page(skb_frag_t
*frag
, struct page
*page
)
2320 frag
->page
.p
= page
;
2324 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2326 * @f: the fragment offset
2327 * @page: the page to set
2329 * Sets the @f'th fragment of @skb to contain @page.
2331 static inline void skb_frag_set_page(struct sk_buff
*skb
, int f
,
2334 __skb_frag_set_page(&skb_shinfo(skb
)->frags
[f
], page
);
2337 bool skb_page_frag_refill(unsigned int sz
, struct page_frag
*pfrag
, gfp_t prio
);
2340 * skb_frag_dma_map - maps a paged fragment via the DMA API
2341 * @dev: the device to map the fragment to
2342 * @frag: the paged fragment to map
2343 * @offset: the offset within the fragment (starting at the
2344 * fragment's own offset)
2345 * @size: the number of bytes to map
2346 * @dir: the direction of the mapping (%PCI_DMA_*)
2348 * Maps the page associated with @frag to @device.
2350 static inline dma_addr_t
skb_frag_dma_map(struct device
*dev
,
2351 const skb_frag_t
*frag
,
2352 size_t offset
, size_t size
,
2353 enum dma_data_direction dir
)
2355 return dma_map_page(dev
, skb_frag_page(frag
),
2356 frag
->page_offset
+ offset
, size
, dir
);
2359 static inline struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
2362 return __pskb_copy(skb
, skb_headroom(skb
), gfp_mask
);
2366 static inline struct sk_buff
*pskb_copy_for_clone(struct sk_buff
*skb
,
2369 return __pskb_copy_fclone(skb
, skb_headroom(skb
), gfp_mask
, true);
2374 * skb_clone_writable - is the header of a clone writable
2375 * @skb: buffer to check
2376 * @len: length up to which to write
2378 * Returns true if modifying the header part of the cloned buffer
2379 * does not requires the data to be copied.
2381 static inline int skb_clone_writable(const struct sk_buff
*skb
, unsigned int len
)
2383 return !skb_header_cloned(skb
) &&
2384 skb_headroom(skb
) + len
<= skb
->hdr_len
;
2387 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
2392 if (headroom
> skb_headroom(skb
))
2393 delta
= headroom
- skb_headroom(skb
);
2395 if (delta
|| cloned
)
2396 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
2402 * skb_cow - copy header of skb when it is required
2403 * @skb: buffer to cow
2404 * @headroom: needed headroom
2406 * If the skb passed lacks sufficient headroom or its data part
2407 * is shared, data is reallocated. If reallocation fails, an error
2408 * is returned and original skb is not changed.
2410 * The result is skb with writable area skb->head...skb->tail
2411 * and at least @headroom of space at head.
2413 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
2415 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
2419 * skb_cow_head - skb_cow but only making the head writable
2420 * @skb: buffer to cow
2421 * @headroom: needed headroom
2423 * This function is identical to skb_cow except that we replace the
2424 * skb_cloned check by skb_header_cloned. It should be used when
2425 * you only need to push on some header and do not need to modify
2428 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
2430 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
2434 * skb_padto - pad an skbuff up to a minimal size
2435 * @skb: buffer to pad
2436 * @len: minimal length
2438 * Pads up a buffer to ensure the trailing bytes exist and are
2439 * blanked. If the buffer already contains sufficient data it
2440 * is untouched. Otherwise it is extended. Returns zero on
2441 * success. The skb is freed on error.
2444 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
2446 unsigned int size
= skb
->len
;
2447 if (likely(size
>= len
))
2449 return skb_pad(skb
, len
- size
);
2452 static inline int skb_add_data(struct sk_buff
*skb
,
2453 char __user
*from
, int copy
)
2455 const int off
= skb
->len
;
2457 if (skb
->ip_summed
== CHECKSUM_NONE
) {
2459 __wsum csum
= csum_and_copy_from_user(from
, skb_put(skb
, copy
),
2462 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
2465 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
2468 __skb_trim(skb
, off
);
2472 static inline bool skb_can_coalesce(struct sk_buff
*skb
, int i
,
2473 const struct page
*page
, int off
)
2476 const struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
2478 return page
== skb_frag_page(frag
) &&
2479 off
== frag
->page_offset
+ skb_frag_size(frag
);
2484 static inline int __skb_linearize(struct sk_buff
*skb
)
2486 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
2490 * skb_linearize - convert paged skb to linear one
2491 * @skb: buffer to linarize
2493 * If there is no free memory -ENOMEM is returned, otherwise zero
2494 * is returned and the old skb data released.
2496 static inline int skb_linearize(struct sk_buff
*skb
)
2498 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
2502 * skb_has_shared_frag - can any frag be overwritten
2503 * @skb: buffer to test
2505 * Return true if the skb has at least one frag that might be modified
2506 * by an external entity (as in vmsplice()/sendfile())
2508 static inline bool skb_has_shared_frag(const struct sk_buff
*skb
)
2510 return skb_is_nonlinear(skb
) &&
2511 skb_shinfo(skb
)->tx_flags
& SKBTX_SHARED_FRAG
;
2515 * skb_linearize_cow - make sure skb is linear and writable
2516 * @skb: buffer to process
2518 * If there is no free memory -ENOMEM is returned, otherwise zero
2519 * is returned and the old skb data released.
2521 static inline int skb_linearize_cow(struct sk_buff
*skb
)
2523 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
2524 __skb_linearize(skb
) : 0;
2528 * skb_postpull_rcsum - update checksum for received skb after pull
2529 * @skb: buffer to update
2530 * @start: start of data before pull
2531 * @len: length of data pulled
2533 * After doing a pull on a received packet, you need to call this to
2534 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2535 * CHECKSUM_NONE so that it can be recomputed from scratch.
2538 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
2539 const void *start
, unsigned int len
)
2541 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2542 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
2545 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
2548 * pskb_trim_rcsum - trim received skb and update checksum
2549 * @skb: buffer to trim
2552 * This is exactly the same as pskb_trim except that it ensures the
2553 * checksum of received packets are still valid after the operation.
2556 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
2558 if (likely(len
>= skb
->len
))
2560 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2561 skb
->ip_summed
= CHECKSUM_NONE
;
2562 return __pskb_trim(skb
, len
);
2565 #define skb_queue_walk(queue, skb) \
2566 for (skb = (queue)->next; \
2567 skb != (struct sk_buff *)(queue); \
2570 #define skb_queue_walk_safe(queue, skb, tmp) \
2571 for (skb = (queue)->next, tmp = skb->next; \
2572 skb != (struct sk_buff *)(queue); \
2573 skb = tmp, tmp = skb->next)
2575 #define skb_queue_walk_from(queue, skb) \
2576 for (; skb != (struct sk_buff *)(queue); \
2579 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2580 for (tmp = skb->next; \
2581 skb != (struct sk_buff *)(queue); \
2582 skb = tmp, tmp = skb->next)
2584 #define skb_queue_reverse_walk(queue, skb) \
2585 for (skb = (queue)->prev; \
2586 skb != (struct sk_buff *)(queue); \
2589 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2590 for (skb = (queue)->prev, tmp = skb->prev; \
2591 skb != (struct sk_buff *)(queue); \
2592 skb = tmp, tmp = skb->prev)
2594 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2595 for (tmp = skb->prev; \
2596 skb != (struct sk_buff *)(queue); \
2597 skb = tmp, tmp = skb->prev)
2599 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
2601 return skb_shinfo(skb
)->frag_list
!= NULL
;
2604 static inline void skb_frag_list_init(struct sk_buff
*skb
)
2606 skb_shinfo(skb
)->frag_list
= NULL
;
2609 static inline void skb_frag_add_head(struct sk_buff
*skb
, struct sk_buff
*frag
)
2611 frag
->next
= skb_shinfo(skb
)->frag_list
;
2612 skb_shinfo(skb
)->frag_list
= frag
;
2615 #define skb_walk_frags(skb, iter) \
2616 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2618 struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
2619 int *peeked
, int *off
, int *err
);
2620 struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
, int noblock
,
2622 unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
2623 struct poll_table_struct
*wait
);
2624 int skb_copy_datagram_iovec(const struct sk_buff
*from
, int offset
,
2625 struct iovec
*to
, int size
);
2626 int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
, int hlen
,
2628 int skb_copy_datagram_from_iovec(struct sk_buff
*skb
, int offset
,
2629 const struct iovec
*from
, int from_offset
,
2631 int zerocopy_sg_from_iovec(struct sk_buff
*skb
, const struct iovec
*frm
,
2632 int offset
, size_t count
);
2633 int skb_copy_datagram_const_iovec(const struct sk_buff
*from
, int offset
,
2634 const struct iovec
*to
, int to_offset
,
2636 void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
2637 void skb_free_datagram_locked(struct sock
*sk
, struct sk_buff
*skb
);
2638 int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
, unsigned int flags
);
2639 int skb_copy_bits(const struct sk_buff
*skb
, int offset
, void *to
, int len
);
2640 int skb_store_bits(struct sk_buff
*skb
, int offset
, const void *from
, int len
);
2641 __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
, int offset
, u8
*to
,
2642 int len
, __wsum csum
);
2643 int skb_splice_bits(struct sk_buff
*skb
, unsigned int offset
,
2644 struct pipe_inode_info
*pipe
, unsigned int len
,
2645 unsigned int flags
);
2646 void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
2647 unsigned int skb_zerocopy_headlen(const struct sk_buff
*from
);
2648 int skb_zerocopy(struct sk_buff
*to
, struct sk_buff
*from
,
2650 void skb_split(struct sk_buff
*skb
, struct sk_buff
*skb1
, const u32 len
);
2651 int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
, int shiftlen
);
2652 void skb_scrub_packet(struct sk_buff
*skb
, bool xnet
);
2653 unsigned int skb_gso_transport_seglen(const struct sk_buff
*skb
);
2654 struct sk_buff
*skb_segment(struct sk_buff
*skb
, netdev_features_t features
);
2655 struct sk_buff
*skb_vlan_untag(struct sk_buff
*skb
);
2657 struct skb_checksum_ops
{
2658 __wsum (*update
)(const void *mem
, int len
, __wsum wsum
);
2659 __wsum (*combine
)(__wsum csum
, __wsum csum2
, int offset
, int len
);
2662 __wsum
__skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
2663 __wsum csum
, const struct skb_checksum_ops
*ops
);
2664 __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
2667 static inline void *__skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2668 int len
, void *data
, int hlen
, void *buffer
)
2670 if (hlen
- offset
>= len
)
2671 return data
+ offset
;
2674 skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
2680 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2681 int len
, void *buffer
)
2683 return __skb_header_pointer(skb
, offset
, len
, skb
->data
,
2684 skb_headlen(skb
), buffer
);
2688 * skb_needs_linearize - check if we need to linearize a given skb
2689 * depending on the given device features.
2690 * @skb: socket buffer to check
2691 * @features: net device features
2693 * Returns true if either:
2694 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2695 * 2. skb is fragmented and the device does not support SG.
2697 static inline bool skb_needs_linearize(struct sk_buff
*skb
,
2698 netdev_features_t features
)
2700 return skb_is_nonlinear(skb
) &&
2701 ((skb_has_frag_list(skb
) && !(features
& NETIF_F_FRAGLIST
)) ||
2702 (skb_shinfo(skb
)->nr_frags
&& !(features
& NETIF_F_SG
)));
2705 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
2707 const unsigned int len
)
2709 memcpy(to
, skb
->data
, len
);
2712 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
2713 const int offset
, void *to
,
2714 const unsigned int len
)
2716 memcpy(to
, skb
->data
+ offset
, len
);
2719 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
2721 const unsigned int len
)
2723 memcpy(skb
->data
, from
, len
);
2726 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
2729 const unsigned int len
)
2731 memcpy(skb
->data
+ offset
, from
, len
);
2734 void skb_init(void);
2736 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
2742 * skb_get_timestamp - get timestamp from a skb
2743 * @skb: skb to get stamp from
2744 * @stamp: pointer to struct timeval to store stamp in
2746 * Timestamps are stored in the skb as offsets to a base timestamp.
2747 * This function converts the offset back to a struct timeval and stores
2750 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
2751 struct timeval
*stamp
)
2753 *stamp
= ktime_to_timeval(skb
->tstamp
);
2756 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
2757 struct timespec
*stamp
)
2759 *stamp
= ktime_to_timespec(skb
->tstamp
);
2762 static inline void __net_timestamp(struct sk_buff
*skb
)
2764 skb
->tstamp
= ktime_get_real();
2767 static inline ktime_t
net_timedelta(ktime_t t
)
2769 return ktime_sub(ktime_get_real(), t
);
2772 static inline ktime_t
net_invalid_timestamp(void)
2774 return ktime_set(0, 0);
2777 struct sk_buff
*skb_clone_sk(struct sk_buff
*skb
);
2779 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2781 void skb_clone_tx_timestamp(struct sk_buff
*skb
);
2782 bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
2784 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2786 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
2790 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
2795 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2798 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2800 * PHY drivers may accept clones of transmitted packets for
2801 * timestamping via their phy_driver.txtstamp method. These drivers
2802 * must call this function to return the skb back to the stack, with
2803 * or without a timestamp.
2805 * @skb: clone of the the original outgoing packet
2806 * @hwtstamps: hardware time stamps, may be NULL if not available
2809 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
2810 struct skb_shared_hwtstamps
*hwtstamps
);
2812 void __skb_tstamp_tx(struct sk_buff
*orig_skb
,
2813 struct skb_shared_hwtstamps
*hwtstamps
,
2814 struct sock
*sk
, int tstype
);
2817 * skb_tstamp_tx - queue clone of skb with send time stamps
2818 * @orig_skb: the original outgoing packet
2819 * @hwtstamps: hardware time stamps, may be NULL if not available
2821 * If the skb has a socket associated, then this function clones the
2822 * skb (thus sharing the actual data and optional structures), stores
2823 * the optional hardware time stamping information (if non NULL) or
2824 * generates a software time stamp (otherwise), then queues the clone
2825 * to the error queue of the socket. Errors are silently ignored.
2827 void skb_tstamp_tx(struct sk_buff
*orig_skb
,
2828 struct skb_shared_hwtstamps
*hwtstamps
);
2830 static inline void sw_tx_timestamp(struct sk_buff
*skb
)
2832 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
&&
2833 !(skb_shinfo(skb
)->tx_flags
& SKBTX_IN_PROGRESS
))
2834 skb_tstamp_tx(skb
, NULL
);
2838 * skb_tx_timestamp() - Driver hook for transmit timestamping
2840 * Ethernet MAC Drivers should call this function in their hard_xmit()
2841 * function immediately before giving the sk_buff to the MAC hardware.
2843 * Specifically, one should make absolutely sure that this function is
2844 * called before TX completion of this packet can trigger. Otherwise
2845 * the packet could potentially already be freed.
2847 * @skb: A socket buffer.
2849 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
2851 skb_clone_tx_timestamp(skb
);
2852 sw_tx_timestamp(skb
);
2856 * skb_complete_wifi_ack - deliver skb with wifi status
2858 * @skb: the original outgoing packet
2859 * @acked: ack status
2862 void skb_complete_wifi_ack(struct sk_buff
*skb
, bool acked
);
2864 __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
2865 __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
2867 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
2869 return ((skb
->ip_summed
& CHECKSUM_UNNECESSARY
) || skb
->csum_valid
);
2873 * skb_checksum_complete - Calculate checksum of an entire packet
2874 * @skb: packet to process
2876 * This function calculates the checksum over the entire packet plus
2877 * the value of skb->csum. The latter can be used to supply the
2878 * checksum of a pseudo header as used by TCP/UDP. It returns the
2881 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2882 * this function can be used to verify that checksum on received
2883 * packets. In that case the function should return zero if the
2884 * checksum is correct. In particular, this function will return zero
2885 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2886 * hardware has already verified the correctness of the checksum.
2888 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
2890 return skb_csum_unnecessary(skb
) ?
2891 0 : __skb_checksum_complete(skb
);
2894 static inline void __skb_decr_checksum_unnecessary(struct sk_buff
*skb
)
2896 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
2897 if (skb
->csum_level
== 0)
2898 skb
->ip_summed
= CHECKSUM_NONE
;
2904 static inline void __skb_incr_checksum_unnecessary(struct sk_buff
*skb
)
2906 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
2907 if (skb
->csum_level
< SKB_MAX_CSUM_LEVEL
)
2909 } else if (skb
->ip_summed
== CHECKSUM_NONE
) {
2910 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
2911 skb
->csum_level
= 0;
2915 static inline void __skb_mark_checksum_bad(struct sk_buff
*skb
)
2917 /* Mark current checksum as bad (typically called from GRO
2918 * path). In the case that ip_summed is CHECKSUM_NONE
2919 * this must be the first checksum encountered in the packet.
2920 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2921 * checksum after the last one validated. For UDP, a zero
2922 * checksum can not be marked as bad.
2925 if (skb
->ip_summed
== CHECKSUM_NONE
||
2926 skb
->ip_summed
== CHECKSUM_UNNECESSARY
)
2930 /* Check if we need to perform checksum complete validation.
2932 * Returns true if checksum complete is needed, false otherwise
2933 * (either checksum is unnecessary or zero checksum is allowed).
2935 static inline bool __skb_checksum_validate_needed(struct sk_buff
*skb
,
2939 if (skb_csum_unnecessary(skb
) || (zero_okay
&& !check
)) {
2940 skb
->csum_valid
= 1;
2941 __skb_decr_checksum_unnecessary(skb
);
2948 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2951 #define CHECKSUM_BREAK 76
2953 /* Validate (init) checksum based on checksum complete.
2956 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2957 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2958 * checksum is stored in skb->csum for use in __skb_checksum_complete
2959 * non-zero: value of invalid checksum
2962 static inline __sum16
__skb_checksum_validate_complete(struct sk_buff
*skb
,
2966 if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
2967 if (!csum_fold(csum_add(psum
, skb
->csum
))) {
2968 skb
->csum_valid
= 1;
2971 } else if (skb
->csum_bad
) {
2972 /* ip_summed == CHECKSUM_NONE in this case */
2978 if (complete
|| skb
->len
<= CHECKSUM_BREAK
) {
2981 csum
= __skb_checksum_complete(skb
);
2982 skb
->csum_valid
= !csum
;
2989 static inline __wsum
null_compute_pseudo(struct sk_buff
*skb
, int proto
)
2994 /* Perform checksum validate (init). Note that this is a macro since we only
2995 * want to calculate the pseudo header which is an input function if necessary.
2996 * First we try to validate without any computation (checksum unnecessary) and
2997 * then calculate based on checksum complete calling the function to compute
3001 * 0: checksum is validated or try to in skb_checksum_complete
3002 * non-zero: value of invalid checksum
3004 #define __skb_checksum_validate(skb, proto, complete, \
3005 zero_okay, check, compute_pseudo) \
3007 __sum16 __ret = 0; \
3008 skb->csum_valid = 0; \
3009 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3010 __ret = __skb_checksum_validate_complete(skb, \
3011 complete, compute_pseudo(skb, proto)); \
3015 #define skb_checksum_init(skb, proto, compute_pseudo) \
3016 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3018 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3019 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3021 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3022 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3024 #define skb_checksum_validate_zero_check(skb, proto, check, \
3026 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
3028 #define skb_checksum_simple_validate(skb) \
3029 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3031 static inline bool __skb_checksum_convert_check(struct sk_buff
*skb
)
3033 return (skb
->ip_summed
== CHECKSUM_NONE
&&
3034 skb
->csum_valid
&& !skb
->csum_bad
);
3037 static inline void __skb_checksum_convert(struct sk_buff
*skb
,
3038 __sum16 check
, __wsum pseudo
)
3040 skb
->csum
= ~pseudo
;
3041 skb
->ip_summed
= CHECKSUM_COMPLETE
;
3044 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3046 if (__skb_checksum_convert_check(skb)) \
3047 __skb_checksum_convert(skb, check, \
3048 compute_pseudo(skb, proto)); \
3051 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3052 void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
3053 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
3055 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
3056 nf_conntrack_destroy(nfct
);
3058 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
3061 atomic_inc(&nfct
->use
);
3064 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3065 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
3067 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
3070 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
3073 atomic_inc(&nf_bridge
->use
);
3075 #endif /* CONFIG_BRIDGE_NETFILTER */
3076 static inline void nf_reset(struct sk_buff
*skb
)
3078 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3079 nf_conntrack_put(skb
->nfct
);
3082 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3083 nf_bridge_put(skb
->nf_bridge
);
3084 skb
->nf_bridge
= NULL
;
3088 static inline void nf_reset_trace(struct sk_buff
*skb
)
3090 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3095 /* Note: This doesn't put any conntrack and bridge info in dst. */
3096 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
,
3099 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3100 dst
->nfct
= src
->nfct
;
3101 nf_conntrack_get(src
->nfct
);
3103 dst
->nfctinfo
= src
->nfctinfo
;
3105 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3106 dst
->nf_bridge
= src
->nf_bridge
;
3107 nf_bridge_get(src
->nf_bridge
);
3109 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3111 dst
->nf_trace
= src
->nf_trace
;
3115 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
3117 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3118 nf_conntrack_put(dst
->nfct
);
3120 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3121 nf_bridge_put(dst
->nf_bridge
);
3123 __nf_copy(dst
, src
, true);
3126 #ifdef CONFIG_NETWORK_SECMARK
3127 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3129 to
->secmark
= from
->secmark
;
3132 static inline void skb_init_secmark(struct sk_buff
*skb
)
3137 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3140 static inline void skb_init_secmark(struct sk_buff
*skb
)
3144 static inline bool skb_irq_freeable(const struct sk_buff
*skb
)
3146 return !skb
->destructor
&&
3147 #if IS_ENABLED(CONFIG_XFRM)
3150 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3153 !skb
->_skb_refdst
&&
3154 !skb_has_frag_list(skb
);
3157 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
3159 skb
->queue_mapping
= queue_mapping
;
3162 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
3164 return skb
->queue_mapping
;
3167 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
3169 to
->queue_mapping
= from
->queue_mapping
;
3172 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
3174 skb
->queue_mapping
= rx_queue
+ 1;
3177 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
3179 return skb
->queue_mapping
- 1;
3182 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
3184 return skb
->queue_mapping
!= 0;
3187 u16
__skb_tx_hash(const struct net_device
*dev
, struct sk_buff
*skb
,
3188 unsigned int num_tx_queues
);
3190 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
3199 /* Keeps track of mac header offset relative to skb->head.
3200 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3201 * For non-tunnel skb it points to skb_mac_header() and for
3202 * tunnel skb it points to outer mac header.
3203 * Keeps track of level of encapsulation of network headers.
3210 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3212 static inline int skb_tnl_header_len(const struct sk_buff
*inner_skb
)
3214 return (skb_mac_header(inner_skb
) - inner_skb
->head
) -
3215 SKB_GSO_CB(inner_skb
)->mac_offset
;
3218 static inline int gso_pskb_expand_head(struct sk_buff
*skb
, int extra
)
3220 int new_headroom
, headroom
;
3223 headroom
= skb_headroom(skb
);
3224 ret
= pskb_expand_head(skb
, extra
, 0, GFP_ATOMIC
);
3228 new_headroom
= skb_headroom(skb
);
3229 SKB_GSO_CB(skb
)->mac_offset
+= (new_headroom
- headroom
);
3233 /* Compute the checksum for a gso segment. First compute the checksum value
3234 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3235 * then add in skb->csum (checksum from csum_start to end of packet).
3236 * skb->csum and csum_start are then updated to reflect the checksum of the
3237 * resultant packet starting from the transport header-- the resultant checksum
3238 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3241 static inline __sum16
gso_make_checksum(struct sk_buff
*skb
, __wsum res
)
3243 int plen
= SKB_GSO_CB(skb
)->csum_start
- skb_headroom(skb
) -
3244 skb_transport_offset(skb
);
3247 csum
= csum_fold(csum_partial(skb_transport_header(skb
),
3250 SKB_GSO_CB(skb
)->csum_start
-= plen
;
3255 static inline bool skb_is_gso(const struct sk_buff
*skb
)
3257 return skb_shinfo(skb
)->gso_size
;
3260 /* Note: Should be called only if skb_is_gso(skb) is true */
3261 static inline bool skb_is_gso_v6(const struct sk_buff
*skb
)
3263 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
3266 void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
3268 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
3270 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3271 * wanted then gso_type will be set. */
3272 const struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
3274 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
3275 unlikely(shinfo
->gso_type
== 0)) {
3276 __skb_warn_lro_forwarding(skb
);
3282 static inline void skb_forward_csum(struct sk_buff
*skb
)
3284 /* Unfortunately we don't support this one. Any brave souls? */
3285 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3286 skb
->ip_summed
= CHECKSUM_NONE
;
3290 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3291 * @skb: skb to check
3293 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3294 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3295 * use this helper, to document places where we make this assertion.
3297 static inline void skb_checksum_none_assert(const struct sk_buff
*skb
)
3300 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
3304 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
3306 int skb_checksum_setup(struct sk_buff
*skb
, bool recalculate
);
3308 u32
skb_get_poff(const struct sk_buff
*skb
);
3309 u32
__skb_get_poff(const struct sk_buff
*skb
, void *data
,
3310 const struct flow_keys
*keys
, int hlen
);
3313 * skb_head_is_locked - Determine if the skb->head is locked down
3314 * @skb: skb to check
3316 * The head on skbs build around a head frag can be removed if they are
3317 * not cloned. This function returns true if the skb head is locked down
3318 * due to either being allocated via kmalloc, or by being a clone with
3319 * multiple references to the head.
3321 static inline bool skb_head_is_locked(const struct sk_buff
*skb
)
3323 return !skb
->head_frag
|| skb_cloned(skb
);
3327 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3331 * skb_gso_network_seglen is used to determine the real size of the
3332 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3334 * The MAC/L2 header is not accounted for.
3336 static inline unsigned int skb_gso_network_seglen(const struct sk_buff
*skb
)
3338 unsigned int hdr_len
= skb_transport_header(skb
) -
3339 skb_network_header(skb
);
3340 return hdr_len
+ skb_gso_transport_seglen(skb
);
3342 #endif /* __KERNEL__ */
3343 #endif /* _LINUX_SKBUFF_H */