2 * Definitions for the 'struct sk_buff' memory handlers.
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/kernel.h>
18 #include <linux/kmemcheck.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/bug.h>
22 #include <linux/cache.h>
23 #include <linux/rbtree.h>
25 #include <linux/atomic.h>
26 #include <asm/types.h>
27 #include <linux/spinlock.h>
28 #include <linux/net.h>
29 #include <linux/textsearch.h>
30 #include <net/checksum.h>
31 #include <linux/rcupdate.h>
32 #include <linux/hrtimer.h>
33 #include <linux/dma-mapping.h>
34 #include <linux/netdev_features.h>
35 #include <linux/sched.h>
36 #include <net/flow_keys.h>
38 /* A. Checksumming of received packets by device.
42 * Device failed to checksum this packet e.g. due to lack of capabilities.
43 * The packet contains full (though not verified) checksum in packet but
44 * not in skb->csum. Thus, skb->csum is undefined in this case.
46 * CHECKSUM_UNNECESSARY:
48 * The hardware you're dealing with doesn't calculate the full checksum
49 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
50 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
51 * if their checksums are okay. skb->csum is still undefined in this case
52 * though. It is a bad option, but, unfortunately, nowadays most vendors do
53 * this. Apparently with the secret goal to sell you new devices, when you
54 * will add new protocol to your host, f.e. IPv6 8)
56 * CHECKSUM_UNNECESSARY is applicable to following protocols:
58 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
59 * zero UDP checksum for either IPv4 or IPv6, the networking stack
60 * may perform further validation in this case.
61 * GRE: only if the checksum is present in the header.
62 * SCTP: indicates the CRC in SCTP header has been validated.
64 * skb->csum_level indicates the number of consecutive checksums found in
65 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
66 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
67 * and a device is able to verify the checksums for UDP (possibly zero),
68 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
69 * two. If the device were only able to verify the UDP checksum and not
70 * GRE, either because it doesn't support GRE checksum of because GRE
71 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
72 * not considered in this case).
76 * This is the most generic way. The device supplied checksum of the _whole_
77 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
78 * hardware doesn't need to parse L3/L4 headers to implement this.
80 * Note: Even if device supports only some protocols, but is able to produce
81 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
85 * This is identical to the case for output below. This may occur on a packet
86 * received directly from another Linux OS, e.g., a virtualized Linux kernel
87 * on the same host. The packet can be treated in the same way as
88 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
89 * checksum must be filled in by the OS or the hardware.
91 * B. Checksumming on output.
95 * The skb was already checksummed by the protocol, or a checksum is not
100 * The device is required to checksum the packet as seen by hard_start_xmit()
101 * from skb->csum_start up to the end, and to record/write the checksum at
102 * offset skb->csum_start + skb->csum_offset.
104 * The device must show its capabilities in dev->features, set up at device
105 * setup time, e.g. netdev_features.h:
107 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
108 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
109 * IPv4. Sigh. Vendors like this way for an unknown reason.
110 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
111 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
112 * NETIF_F_... - Well, you get the picture.
114 * CHECKSUM_UNNECESSARY:
116 * Normally, the device will do per protocol specific checksumming. Protocol
117 * implementations that do not want the NIC to perform the checksum
118 * calculation should use this flag in their outgoing skbs.
120 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
121 * offload. Correspondingly, the FCoE protocol driver
122 * stack should use CHECKSUM_UNNECESSARY.
124 * Any questions? No questions, good. --ANK
127 /* Don't change this without changing skb_csum_unnecessary! */
128 #define CHECKSUM_NONE 0
129 #define CHECKSUM_UNNECESSARY 1
130 #define CHECKSUM_COMPLETE 2
131 #define CHECKSUM_PARTIAL 3
133 /* Maximum value in skb->csum_level */
134 #define SKB_MAX_CSUM_LEVEL 3
136 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
137 #define SKB_WITH_OVERHEAD(X) \
138 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
139 #define SKB_MAX_ORDER(X, ORDER) \
140 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
141 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
142 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
144 /* return minimum truesize of one skb containing X bytes of data */
145 #define SKB_TRUESIZE(X) ((X) + \
146 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
147 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
151 struct pipe_inode_info
;
153 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
154 struct nf_conntrack
{
159 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
160 struct nf_bridge_info
{
163 struct net_device
*physindev
;
164 struct net_device
*physoutdev
;
165 unsigned long data
[32 / sizeof(unsigned long)];
169 struct sk_buff_head
{
170 /* These two members must be first. */
171 struct sk_buff
*next
;
172 struct sk_buff
*prev
;
180 /* To allow 64K frame to be packed as single skb without frag_list we
181 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
182 * buffers which do not start on a page boundary.
184 * Since GRO uses frags we allocate at least 16 regardless of page
187 #if (65536/PAGE_SIZE + 1) < 16
188 #define MAX_SKB_FRAGS 16UL
190 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
193 typedef struct skb_frag_struct skb_frag_t
;
195 struct skb_frag_struct
{
199 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
208 static inline unsigned int skb_frag_size(const skb_frag_t
*frag
)
213 static inline void skb_frag_size_set(skb_frag_t
*frag
, unsigned int size
)
218 static inline void skb_frag_size_add(skb_frag_t
*frag
, int delta
)
223 static inline void skb_frag_size_sub(skb_frag_t
*frag
, int delta
)
228 #define HAVE_HW_TIME_STAMP
231 * struct skb_shared_hwtstamps - hardware time stamps
232 * @hwtstamp: hardware time stamp transformed into duration
233 * since arbitrary point in time
235 * Software time stamps generated by ktime_get_real() are stored in
238 * hwtstamps can only be compared against other hwtstamps from
241 * This structure is attached to packets as part of the
242 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
244 struct skb_shared_hwtstamps
{
248 /* Definitions for tx_flags in struct skb_shared_info */
250 /* generate hardware time stamp */
251 SKBTX_HW_TSTAMP
= 1 << 0,
253 /* generate software time stamp when queueing packet to NIC */
254 SKBTX_SW_TSTAMP
= 1 << 1,
256 /* device driver is going to provide hardware time stamp */
257 SKBTX_IN_PROGRESS
= 1 << 2,
259 /* device driver supports TX zero-copy buffers */
260 SKBTX_DEV_ZEROCOPY
= 1 << 3,
262 /* generate wifi status information (where possible) */
263 SKBTX_WIFI_STATUS
= 1 << 4,
265 /* This indicates at least one fragment might be overwritten
266 * (as in vmsplice(), sendfile() ...)
267 * If we need to compute a TX checksum, we'll need to copy
268 * all frags to avoid possible bad checksum
270 SKBTX_SHARED_FRAG
= 1 << 5,
272 /* generate software time stamp when entering packet scheduling */
273 SKBTX_SCHED_TSTAMP
= 1 << 6,
275 /* generate software timestamp on peer data acknowledgment */
276 SKBTX_ACK_TSTAMP
= 1 << 7,
279 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
280 SKBTX_SCHED_TSTAMP | \
282 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
285 * The callback notifies userspace to release buffers when skb DMA is done in
286 * lower device, the skb last reference should be 0 when calling this.
287 * The zerocopy_success argument is true if zero copy transmit occurred,
288 * false on data copy or out of memory error caused by data copy attempt.
289 * The ctx field is used to track device context.
290 * The desc field is used to track userspace buffer index.
293 void (*callback
)(struct ubuf_info
*, bool zerocopy_success
);
298 /* This data is invariant across clones and lives at
299 * the end of the header data, ie. at skb->end.
301 struct skb_shared_info
{
302 unsigned char nr_frags
;
304 unsigned short gso_size
;
305 /* Warning: this field is not always filled in (UFO)! */
306 unsigned short gso_segs
;
307 unsigned short gso_type
;
308 struct sk_buff
*frag_list
;
309 struct skb_shared_hwtstamps hwtstamps
;
314 * Warning : all fields before dataref are cleared in __alloc_skb()
318 /* Intermediate layers must ensure that destructor_arg
319 * remains valid until skb destructor */
320 void * destructor_arg
;
322 /* must be last field, see pskb_expand_head() */
323 skb_frag_t frags
[MAX_SKB_FRAGS
];
326 /* We divide dataref into two halves. The higher 16 bits hold references
327 * to the payload part of skb->data. The lower 16 bits hold references to
328 * the entire skb->data. A clone of a headerless skb holds the length of
329 * the header in skb->hdr_len.
331 * All users must obey the rule that the skb->data reference count must be
332 * greater than or equal to the payload reference count.
334 * Holding a reference to the payload part means that the user does not
335 * care about modifications to the header part of skb->data.
337 #define SKB_DATAREF_SHIFT 16
338 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
342 SKB_FCLONE_UNAVAILABLE
, /* skb has no fclone (from head_cache) */
343 SKB_FCLONE_ORIG
, /* orig skb (from fclone_cache) */
344 SKB_FCLONE_CLONE
, /* companion fclone skb (from fclone_cache) */
345 SKB_FCLONE_FREE
, /* this companion fclone skb is available */
349 SKB_GSO_TCPV4
= 1 << 0,
350 SKB_GSO_UDP
= 1 << 1,
352 /* This indicates the skb is from an untrusted source. */
353 SKB_GSO_DODGY
= 1 << 2,
355 /* This indicates the tcp segment has CWR set. */
356 SKB_GSO_TCP_ECN
= 1 << 3,
358 SKB_GSO_TCPV6
= 1 << 4,
360 SKB_GSO_FCOE
= 1 << 5,
362 SKB_GSO_GRE
= 1 << 6,
364 SKB_GSO_GRE_CSUM
= 1 << 7,
366 SKB_GSO_IPIP
= 1 << 8,
368 SKB_GSO_SIT
= 1 << 9,
370 SKB_GSO_UDP_TUNNEL
= 1 << 10,
372 SKB_GSO_UDP_TUNNEL_CSUM
= 1 << 11,
374 SKB_GSO_MPLS
= 1 << 12,
378 #if BITS_PER_LONG > 32
379 #define NET_SKBUFF_DATA_USES_OFFSET 1
382 #ifdef NET_SKBUFF_DATA_USES_OFFSET
383 typedef unsigned int sk_buff_data_t
;
385 typedef unsigned char *sk_buff_data_t
;
389 * struct skb_mstamp - multi resolution time stamps
390 * @stamp_us: timestamp in us resolution
391 * @stamp_jiffies: timestamp in jiffies
404 * skb_mstamp_get - get current timestamp
405 * @cl: place to store timestamps
407 static inline void skb_mstamp_get(struct skb_mstamp
*cl
)
409 u64 val
= local_clock();
411 do_div(val
, NSEC_PER_USEC
);
412 cl
->stamp_us
= (u32
)val
;
413 cl
->stamp_jiffies
= (u32
)jiffies
;
417 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
418 * @t1: pointer to newest sample
419 * @t0: pointer to oldest sample
421 static inline u32
skb_mstamp_us_delta(const struct skb_mstamp
*t1
,
422 const struct skb_mstamp
*t0
)
424 s32 delta_us
= t1
->stamp_us
- t0
->stamp_us
;
425 u32 delta_jiffies
= t1
->stamp_jiffies
- t0
->stamp_jiffies
;
427 /* If delta_us is negative, this might be because interval is too big,
428 * or local_clock() drift is too big : fallback using jiffies.
431 delta_jiffies
>= (INT_MAX
/ (USEC_PER_SEC
/ HZ
)))
433 delta_us
= jiffies_to_usecs(delta_jiffies
);
440 * struct sk_buff - socket buffer
441 * @next: Next buffer in list
442 * @prev: Previous buffer in list
443 * @tstamp: Time we arrived/left
444 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
445 * @sk: Socket we are owned by
446 * @dev: Device we arrived on/are leaving by
447 * @cb: Control buffer. Free for use by every layer. Put private vars here
448 * @_skb_refdst: destination entry (with norefcount bit)
449 * @sp: the security path, used for xfrm
450 * @len: Length of actual data
451 * @data_len: Data length
452 * @mac_len: Length of link layer header
453 * @hdr_len: writable header length of cloned skb
454 * @csum: Checksum (must include start/offset pair)
455 * @csum_start: Offset from skb->head where checksumming should start
456 * @csum_offset: Offset from csum_start where checksum should be stored
457 * @priority: Packet queueing priority
458 * @ignore_df: allow local fragmentation
459 * @cloned: Head may be cloned (check refcnt to be sure)
460 * @ip_summed: Driver fed us an IP checksum
461 * @nohdr: Payload reference only, must not modify header
462 * @nfctinfo: Relationship of this skb to the connection
463 * @pkt_type: Packet class
464 * @fclone: skbuff clone status
465 * @ipvs_property: skbuff is owned by ipvs
466 * @peeked: this packet has been seen already, so stats have been
467 * done for it, don't do them again
468 * @nf_trace: netfilter packet trace flag
469 * @protocol: Packet protocol from driver
470 * @destructor: Destruct function
471 * @nfct: Associated connection, if any
472 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
473 * @skb_iif: ifindex of device we arrived on
474 * @tc_index: Traffic control index
475 * @tc_verd: traffic control verdict
476 * @hash: the packet hash
477 * @queue_mapping: Queue mapping for multiqueue devices
478 * @xmit_more: More SKBs are pending for this queue
479 * @ndisc_nodetype: router type (from link layer)
480 * @ooo_okay: allow the mapping of a socket to a queue to be changed
481 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
483 * @sw_hash: indicates hash was computed in software stack
484 * @wifi_acked_valid: wifi_acked was set
485 * @wifi_acked: whether frame was acked on wifi or not
486 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
487 * @napi_id: id of the NAPI struct this skb came from
488 * @secmark: security marking
489 * @mark: Generic packet mark
490 * @dropcount: total number of sk_receive_queue overflows
491 * @vlan_proto: vlan encapsulation protocol
492 * @vlan_tci: vlan tag control information
493 * @inner_protocol: Protocol (encapsulation)
494 * @inner_transport_header: Inner transport layer header (encapsulation)
495 * @inner_network_header: Network layer header (encapsulation)
496 * @inner_mac_header: Link layer header (encapsulation)
497 * @transport_header: Transport layer header
498 * @network_header: Network layer header
499 * @mac_header: Link layer header
500 * @tail: Tail pointer
502 * @head: Head of buffer
503 * @data: Data head pointer
504 * @truesize: Buffer size
505 * @users: User count - see {datagram,tcp}.c
511 /* These two members must be first. */
512 struct sk_buff
*next
;
513 struct sk_buff
*prev
;
517 struct skb_mstamp skb_mstamp
;
520 struct rb_node rbnode
; /* used in netem & tcp stack */
523 struct net_device
*dev
;
526 * This is the control buffer. It is free to use for every
527 * layer. Please put your private variables there. If you
528 * want to keep them across layers you have to do a skb_clone()
529 * first. This is owned by whoever has the skb queued ATM.
531 char cb
[48] __aligned(8);
533 unsigned long _skb_refdst
;
534 void (*destructor
)(struct sk_buff
*skb
);
538 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
539 struct nf_conntrack
*nfct
;
541 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
542 struct nf_bridge_info
*nf_bridge
;
549 /* Following fields are _not_ copied in __copy_skb_header()
550 * Note that queue_mapping is here mostly to fill a hole.
552 kmemcheck_bitfield_begin(flags1
);
561 kmemcheck_bitfield_end(flags1
);
563 /* fields enclosed in headers_start/headers_end are copied
564 * using a single memcpy() in __copy_skb_header()
567 __u32 headers_start
[0];
570 /* if you move pkt_type around you also must adapt those constants */
571 #ifdef __BIG_ENDIAN_BITFIELD
572 #define PKT_TYPE_MAX (7 << 5)
574 #define PKT_TYPE_MAX 7
576 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
578 __u8 __pkt_type_offset
[0];
589 __u8 wifi_acked_valid
:1;
593 /* Indicates the inner headers are valid in the skbuff. */
594 __u8 encapsulation
:1;
595 __u8 encap_hdr_csum
:1;
597 __u8 csum_complete_sw
:1;
601 #ifdef CONFIG_IPV6_NDISC_NODETYPE
602 __u8 ndisc_nodetype
:2;
604 __u8 ipvs_property
:1;
605 __u8 inner_protocol_type
:1;
606 /* 4 or 6 bit hole */
608 #ifdef CONFIG_NET_SCHED
609 __u16 tc_index
; /* traffic control index */
610 #ifdef CONFIG_NET_CLS_ACT
611 __u16 tc_verd
; /* traffic control verdict */
627 #ifdef CONFIG_NET_RX_BUSY_POLL
628 unsigned int napi_id
;
630 #ifdef CONFIG_NETWORK_SECMARK
636 __u32 reserved_tailroom
;
640 __be16 inner_protocol
;
644 __u16 inner_transport_header
;
645 __u16 inner_network_header
;
646 __u16 inner_mac_header
;
649 __u16 transport_header
;
650 __u16 network_header
;
654 __u32 headers_end
[0];
657 /* These elements must be at the end, see alloc_skb() for details. */
662 unsigned int truesize
;
668 * Handling routines are only of interest to the kernel
670 #include <linux/slab.h>
673 #define SKB_ALLOC_FCLONE 0x01
674 #define SKB_ALLOC_RX 0x02
676 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
677 static inline bool skb_pfmemalloc(const struct sk_buff
*skb
)
679 return unlikely(skb
->pfmemalloc
);
683 * skb might have a dst pointer attached, refcounted or not.
684 * _skb_refdst low order bit is set if refcount was _not_ taken
686 #define SKB_DST_NOREF 1UL
687 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
690 * skb_dst - returns skb dst_entry
693 * Returns skb dst_entry, regardless of reference taken or not.
695 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
697 /* If refdst was not refcounted, check we still are in a
698 * rcu_read_lock section
700 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
701 !rcu_read_lock_held() &&
702 !rcu_read_lock_bh_held());
703 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
707 * skb_dst_set - sets skb dst
711 * Sets skb dst, assuming a reference was taken on dst and should
712 * be released by skb_dst_drop()
714 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
716 skb
->_skb_refdst
= (unsigned long)dst
;
719 void __skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
,
723 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
727 * Sets skb dst, assuming a reference was not taken on dst.
728 * If dst entry is cached, we do not take reference and dst_release
729 * will be avoided by refdst_drop. If dst entry is not cached, we take
730 * reference, so that last dst_release can destroy the dst immediately.
732 static inline void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
)
734 __skb_dst_set_noref(skb
, dst
, false);
738 * skb_dst_set_noref_force - sets skb dst, without taking reference
742 * Sets skb dst, assuming a reference was not taken on dst.
743 * No reference is taken and no dst_release will be called. While for
744 * cached dsts deferred reclaim is a basic feature, for entries that are
745 * not cached it is caller's job to guarantee that last dst_release for
746 * provided dst happens when nobody uses it, eg. after a RCU grace period.
748 static inline void skb_dst_set_noref_force(struct sk_buff
*skb
,
749 struct dst_entry
*dst
)
751 __skb_dst_set_noref(skb
, dst
, true);
755 * skb_dst_is_noref - Test if skb dst isn't refcounted
758 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
760 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
763 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
765 return (struct rtable
*)skb_dst(skb
);
768 void kfree_skb(struct sk_buff
*skb
);
769 void kfree_skb_list(struct sk_buff
*segs
);
770 void skb_tx_error(struct sk_buff
*skb
);
771 void consume_skb(struct sk_buff
*skb
);
772 void __kfree_skb(struct sk_buff
*skb
);
773 extern struct kmem_cache
*skbuff_head_cache
;
775 void kfree_skb_partial(struct sk_buff
*skb
, bool head_stolen
);
776 bool skb_try_coalesce(struct sk_buff
*to
, struct sk_buff
*from
,
777 bool *fragstolen
, int *delta_truesize
);
779 struct sk_buff
*__alloc_skb(unsigned int size
, gfp_t priority
, int flags
,
781 struct sk_buff
*build_skb(void *data
, unsigned int frag_size
);
782 static inline struct sk_buff
*alloc_skb(unsigned int size
,
785 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
788 struct sk_buff
*alloc_skb_with_frags(unsigned long header_len
,
789 unsigned long data_len
,
794 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
795 struct sk_buff_fclones
{
804 * skb_fclone_busy - check if fclone is busy
807 * Returns true is skb is a fast clone, and its clone is not freed.
808 * Some drivers call skb_orphan() in their ndo_start_xmit(),
809 * so we also check that this didnt happen.
811 static inline bool skb_fclone_busy(const struct sock
*sk
,
812 const struct sk_buff
*skb
)
814 const struct sk_buff_fclones
*fclones
;
816 fclones
= container_of(skb
, struct sk_buff_fclones
, skb1
);
818 return skb
->fclone
== SKB_FCLONE_ORIG
&&
819 fclones
->skb2
.fclone
== SKB_FCLONE_CLONE
&&
820 fclones
->skb2
.sk
== sk
;
823 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
826 return __alloc_skb(size
, priority
, SKB_ALLOC_FCLONE
, NUMA_NO_NODE
);
829 struct sk_buff
*__alloc_skb_head(gfp_t priority
, int node
);
830 static inline struct sk_buff
*alloc_skb_head(gfp_t priority
)
832 return __alloc_skb_head(priority
, -1);
835 struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
836 int skb_copy_ubufs(struct sk_buff
*skb
, gfp_t gfp_mask
);
837 struct sk_buff
*skb_clone(struct sk_buff
*skb
, gfp_t priority
);
838 struct sk_buff
*skb_copy(const struct sk_buff
*skb
, gfp_t priority
);
839 struct sk_buff
*__pskb_copy_fclone(struct sk_buff
*skb
, int headroom
,
840 gfp_t gfp_mask
, bool fclone
);
841 static inline struct sk_buff
*__pskb_copy(struct sk_buff
*skb
, int headroom
,
844 return __pskb_copy_fclone(skb
, headroom
, gfp_mask
, false);
847 int pskb_expand_head(struct sk_buff
*skb
, int nhead
, int ntail
, gfp_t gfp_mask
);
848 struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
849 unsigned int headroom
);
850 struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
, int newheadroom
,
851 int newtailroom
, gfp_t priority
);
852 int skb_to_sgvec_nomark(struct sk_buff
*skb
, struct scatterlist
*sg
,
853 int offset
, int len
);
854 int skb_to_sgvec(struct sk_buff
*skb
, struct scatterlist
*sg
, int offset
,
856 int skb_cow_data(struct sk_buff
*skb
, int tailbits
, struct sk_buff
**trailer
);
857 int skb_pad(struct sk_buff
*skb
, int pad
);
858 #define dev_kfree_skb(a) consume_skb(a)
860 int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
861 int getfrag(void *from
, char *to
, int offset
,
862 int len
, int odd
, struct sk_buff
*skb
),
863 void *from
, int length
);
865 struct skb_seq_state
{
869 __u32 stepped_offset
;
870 struct sk_buff
*root_skb
;
871 struct sk_buff
*cur_skb
;
875 void skb_prepare_seq_read(struct sk_buff
*skb
, unsigned int from
,
876 unsigned int to
, struct skb_seq_state
*st
);
877 unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
878 struct skb_seq_state
*st
);
879 void skb_abort_seq_read(struct skb_seq_state
*st
);
881 unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
882 unsigned int to
, struct ts_config
*config
,
883 struct ts_state
*state
);
886 * Packet hash types specify the type of hash in skb_set_hash.
888 * Hash types refer to the protocol layer addresses which are used to
889 * construct a packet's hash. The hashes are used to differentiate or identify
890 * flows of the protocol layer for the hash type. Hash types are either
891 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
893 * Properties of hashes:
895 * 1) Two packets in different flows have different hash values
896 * 2) Two packets in the same flow should have the same hash value
898 * A hash at a higher layer is considered to be more specific. A driver should
899 * set the most specific hash possible.
901 * A driver cannot indicate a more specific hash than the layer at which a hash
902 * was computed. For instance an L3 hash cannot be set as an L4 hash.
904 * A driver may indicate a hash level which is less specific than the
905 * actual layer the hash was computed on. For instance, a hash computed
906 * at L4 may be considered an L3 hash. This should only be done if the
907 * driver can't unambiguously determine that the HW computed the hash at
908 * the higher layer. Note that the "should" in the second property above
911 enum pkt_hash_types
{
912 PKT_HASH_TYPE_NONE
, /* Undefined type */
913 PKT_HASH_TYPE_L2
, /* Input: src_MAC, dest_MAC */
914 PKT_HASH_TYPE_L3
, /* Input: src_IP, dst_IP */
915 PKT_HASH_TYPE_L4
, /* Input: src_IP, dst_IP, src_port, dst_port */
919 skb_set_hash(struct sk_buff
*skb
, __u32 hash
, enum pkt_hash_types type
)
921 skb
->l4_hash
= (type
== PKT_HASH_TYPE_L4
);
926 void __skb_get_hash(struct sk_buff
*skb
);
927 static inline __u32
skb_get_hash(struct sk_buff
*skb
)
929 if (!skb
->l4_hash
&& !skb
->sw_hash
)
935 static inline __u32
skb_get_hash_raw(const struct sk_buff
*skb
)
940 static inline void skb_clear_hash(struct sk_buff
*skb
)
947 static inline void skb_clear_hash_if_not_l4(struct sk_buff
*skb
)
953 static inline void skb_copy_hash(struct sk_buff
*to
, const struct sk_buff
*from
)
955 to
->hash
= from
->hash
;
956 to
->sw_hash
= from
->sw_hash
;
957 to
->l4_hash
= from
->l4_hash
;
960 #ifdef NET_SKBUFF_DATA_USES_OFFSET
961 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
963 return skb
->head
+ skb
->end
;
966 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
971 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
976 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
978 return skb
->end
- skb
->head
;
983 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
985 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
987 return &skb_shinfo(skb
)->hwtstamps
;
991 * skb_queue_empty - check if a queue is empty
994 * Returns true if the queue is empty, false otherwise.
996 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
998 return list
->next
== (const struct sk_buff
*) list
;
1002 * skb_queue_is_last - check if skb is the last entry in the queue
1006 * Returns true if @skb is the last buffer on the list.
1008 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
1009 const struct sk_buff
*skb
)
1011 return skb
->next
== (const struct sk_buff
*) list
;
1015 * skb_queue_is_first - check if skb is the first entry in the queue
1019 * Returns true if @skb is the first buffer on the list.
1021 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
1022 const struct sk_buff
*skb
)
1024 return skb
->prev
== (const struct sk_buff
*) list
;
1028 * skb_queue_next - return the next packet in the queue
1030 * @skb: current buffer
1032 * Return the next packet in @list after @skb. It is only valid to
1033 * call this if skb_queue_is_last() evaluates to false.
1035 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
1036 const struct sk_buff
*skb
)
1038 /* This BUG_ON may seem severe, but if we just return then we
1039 * are going to dereference garbage.
1041 BUG_ON(skb_queue_is_last(list
, skb
));
1046 * skb_queue_prev - return the prev packet in the queue
1048 * @skb: current buffer
1050 * Return the prev packet in @list before @skb. It is only valid to
1051 * call this if skb_queue_is_first() evaluates to false.
1053 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
1054 const struct sk_buff
*skb
)
1056 /* This BUG_ON may seem severe, but if we just return then we
1057 * are going to dereference garbage.
1059 BUG_ON(skb_queue_is_first(list
, skb
));
1064 * skb_get - reference buffer
1065 * @skb: buffer to reference
1067 * Makes another reference to a socket buffer and returns a pointer
1070 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
1072 atomic_inc(&skb
->users
);
1077 * If users == 1, we are the only owner and are can avoid redundant
1082 * skb_cloned - is the buffer a clone
1083 * @skb: buffer to check
1085 * Returns true if the buffer was generated with skb_clone() and is
1086 * one of multiple shared copies of the buffer. Cloned buffers are
1087 * shared data so must not be written to under normal circumstances.
1089 static inline int skb_cloned(const struct sk_buff
*skb
)
1091 return skb
->cloned
&&
1092 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
1095 static inline int skb_unclone(struct sk_buff
*skb
, gfp_t pri
)
1097 might_sleep_if(pri
& __GFP_WAIT
);
1099 if (skb_cloned(skb
))
1100 return pskb_expand_head(skb
, 0, 0, pri
);
1106 * skb_header_cloned - is the header a clone
1107 * @skb: buffer to check
1109 * Returns true if modifying the header part of the buffer requires
1110 * the data to be copied.
1112 static inline int skb_header_cloned(const struct sk_buff
*skb
)
1119 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
1120 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
1121 return dataref
!= 1;
1125 * skb_header_release - release reference to header
1126 * @skb: buffer to operate on
1128 * Drop a reference to the header part of the buffer. This is done
1129 * by acquiring a payload reference. You must not read from the header
1130 * part of skb->data after this.
1131 * Note : Check if you can use __skb_header_release() instead.
1133 static inline void skb_header_release(struct sk_buff
*skb
)
1137 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
1141 * __skb_header_release - release reference to header
1142 * @skb: buffer to operate on
1144 * Variant of skb_header_release() assuming skb is private to caller.
1145 * We can avoid one atomic operation.
1147 static inline void __skb_header_release(struct sk_buff
*skb
)
1150 atomic_set(&skb_shinfo(skb
)->dataref
, 1 + (1 << SKB_DATAREF_SHIFT
));
1155 * skb_shared - is the buffer shared
1156 * @skb: buffer to check
1158 * Returns true if more than one person has a reference to this
1161 static inline int skb_shared(const struct sk_buff
*skb
)
1163 return atomic_read(&skb
->users
) != 1;
1167 * skb_share_check - check if buffer is shared and if so clone it
1168 * @skb: buffer to check
1169 * @pri: priority for memory allocation
1171 * If the buffer is shared the buffer is cloned and the old copy
1172 * drops a reference. A new clone with a single reference is returned.
1173 * If the buffer is not shared the original buffer is returned. When
1174 * being called from interrupt status or with spinlocks held pri must
1177 * NULL is returned on a memory allocation failure.
1179 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
, gfp_t pri
)
1181 might_sleep_if(pri
& __GFP_WAIT
);
1182 if (skb_shared(skb
)) {
1183 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
1195 * Copy shared buffers into a new sk_buff. We effectively do COW on
1196 * packets to handle cases where we have a local reader and forward
1197 * and a couple of other messy ones. The normal one is tcpdumping
1198 * a packet thats being forwarded.
1202 * skb_unshare - make a copy of a shared buffer
1203 * @skb: buffer to check
1204 * @pri: priority for memory allocation
1206 * If the socket buffer is a clone then this function creates a new
1207 * copy of the data, drops a reference count on the old copy and returns
1208 * the new copy with the reference count at 1. If the buffer is not a clone
1209 * the original buffer is returned. When called with a spinlock held or
1210 * from interrupt state @pri must be %GFP_ATOMIC
1212 * %NULL is returned on a memory allocation failure.
1214 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
1217 might_sleep_if(pri
& __GFP_WAIT
);
1218 if (skb_cloned(skb
)) {
1219 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
1221 /* Free our shared copy */
1232 * skb_peek - peek at the head of an &sk_buff_head
1233 * @list_: list to peek at
1235 * Peek an &sk_buff. Unlike most other operations you _MUST_
1236 * be careful with this one. A peek leaves the buffer on the
1237 * list and someone else may run off with it. You must hold
1238 * the appropriate locks or have a private queue to do this.
1240 * Returns %NULL for an empty list or a pointer to the head element.
1241 * The reference count is not incremented and the reference is therefore
1242 * volatile. Use with caution.
1244 static inline struct sk_buff
*skb_peek(const struct sk_buff_head
*list_
)
1246 struct sk_buff
*skb
= list_
->next
;
1248 if (skb
== (struct sk_buff
*)list_
)
1254 * skb_peek_next - peek skb following the given one from a queue
1255 * @skb: skb to start from
1256 * @list_: list to peek at
1258 * Returns %NULL when the end of the list is met or a pointer to the
1259 * next element. The reference count is not incremented and the
1260 * reference is therefore volatile. Use with caution.
1262 static inline struct sk_buff
*skb_peek_next(struct sk_buff
*skb
,
1263 const struct sk_buff_head
*list_
)
1265 struct sk_buff
*next
= skb
->next
;
1267 if (next
== (struct sk_buff
*)list_
)
1273 * skb_peek_tail - peek at the tail of an &sk_buff_head
1274 * @list_: list to peek at
1276 * Peek an &sk_buff. Unlike most other operations you _MUST_
1277 * be careful with this one. A peek leaves the buffer on the
1278 * list and someone else may run off with it. You must hold
1279 * the appropriate locks or have a private queue to do this.
1281 * Returns %NULL for an empty list or a pointer to the tail element.
1282 * The reference count is not incremented and the reference is therefore
1283 * volatile. Use with caution.
1285 static inline struct sk_buff
*skb_peek_tail(const struct sk_buff_head
*list_
)
1287 struct sk_buff
*skb
= list_
->prev
;
1289 if (skb
== (struct sk_buff
*)list_
)
1296 * skb_queue_len - get queue length
1297 * @list_: list to measure
1299 * Return the length of an &sk_buff queue.
1301 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
1307 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1308 * @list: queue to initialize
1310 * This initializes only the list and queue length aspects of
1311 * an sk_buff_head object. This allows to initialize the list
1312 * aspects of an sk_buff_head without reinitializing things like
1313 * the spinlock. It can also be used for on-stack sk_buff_head
1314 * objects where the spinlock is known to not be used.
1316 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
1318 list
->prev
= list
->next
= (struct sk_buff
*)list
;
1323 * This function creates a split out lock class for each invocation;
1324 * this is needed for now since a whole lot of users of the skb-queue
1325 * infrastructure in drivers have different locking usage (in hardirq)
1326 * than the networking core (in softirq only). In the long run either the
1327 * network layer or drivers should need annotation to consolidate the
1328 * main types of usage into 3 classes.
1330 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
1332 spin_lock_init(&list
->lock
);
1333 __skb_queue_head_init(list
);
1336 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
1337 struct lock_class_key
*class)
1339 skb_queue_head_init(list
);
1340 lockdep_set_class(&list
->lock
, class);
1344 * Insert an sk_buff on a list.
1346 * The "__skb_xxxx()" functions are the non-atomic ones that
1347 * can only be called with interrupts disabled.
1349 void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
,
1350 struct sk_buff_head
*list
);
1351 static inline void __skb_insert(struct sk_buff
*newsk
,
1352 struct sk_buff
*prev
, struct sk_buff
*next
,
1353 struct sk_buff_head
*list
)
1357 next
->prev
= prev
->next
= newsk
;
1361 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
1362 struct sk_buff
*prev
,
1363 struct sk_buff
*next
)
1365 struct sk_buff
*first
= list
->next
;
1366 struct sk_buff
*last
= list
->prev
;
1376 * skb_queue_splice - join two skb lists, this is designed for stacks
1377 * @list: the new list to add
1378 * @head: the place to add it in the first list
1380 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
1381 struct sk_buff_head
*head
)
1383 if (!skb_queue_empty(list
)) {
1384 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1385 head
->qlen
+= list
->qlen
;
1390 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1391 * @list: the new list to add
1392 * @head: the place to add it in the first list
1394 * The list at @list is reinitialised
1396 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
1397 struct sk_buff_head
*head
)
1399 if (!skb_queue_empty(list
)) {
1400 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1401 head
->qlen
+= list
->qlen
;
1402 __skb_queue_head_init(list
);
1407 * skb_queue_splice_tail - join two skb lists, each list being a queue
1408 * @list: the new list to add
1409 * @head: the place to add it in the first list
1411 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
1412 struct sk_buff_head
*head
)
1414 if (!skb_queue_empty(list
)) {
1415 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1416 head
->qlen
+= list
->qlen
;
1421 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1422 * @list: the new list to add
1423 * @head: the place to add it in the first list
1425 * Each of the lists is a queue.
1426 * The list at @list is reinitialised
1428 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
1429 struct sk_buff_head
*head
)
1431 if (!skb_queue_empty(list
)) {
1432 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1433 head
->qlen
+= list
->qlen
;
1434 __skb_queue_head_init(list
);
1439 * __skb_queue_after - queue a buffer at the list head
1440 * @list: list to use
1441 * @prev: place after this buffer
1442 * @newsk: buffer to queue
1444 * Queue a buffer int the middle of a list. This function takes no locks
1445 * and you must therefore hold required locks before calling it.
1447 * A buffer cannot be placed on two lists at the same time.
1449 static inline void __skb_queue_after(struct sk_buff_head
*list
,
1450 struct sk_buff
*prev
,
1451 struct sk_buff
*newsk
)
1453 __skb_insert(newsk
, prev
, prev
->next
, list
);
1456 void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
1457 struct sk_buff_head
*list
);
1459 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1460 struct sk_buff
*next
,
1461 struct sk_buff
*newsk
)
1463 __skb_insert(newsk
, next
->prev
, next
, list
);
1467 * __skb_queue_head - queue a buffer at the list head
1468 * @list: list to use
1469 * @newsk: buffer to queue
1471 * Queue a buffer at the start of a list. This function takes no locks
1472 * and you must therefore hold required locks before calling it.
1474 * A buffer cannot be placed on two lists at the same time.
1476 void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1477 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1478 struct sk_buff
*newsk
)
1480 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1484 * __skb_queue_tail - queue a buffer at the list tail
1485 * @list: list to use
1486 * @newsk: buffer to queue
1488 * Queue a buffer at the end of a list. This function takes no locks
1489 * and you must therefore hold required locks before calling it.
1491 * A buffer cannot be placed on two lists at the same time.
1493 void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1494 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1495 struct sk_buff
*newsk
)
1497 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1501 * remove sk_buff from list. _Must_ be called atomically, and with
1504 void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1505 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1507 struct sk_buff
*next
, *prev
;
1512 skb
->next
= skb
->prev
= NULL
;
1518 * __skb_dequeue - remove from the head of the queue
1519 * @list: list to dequeue from
1521 * Remove the head of the list. This function does not take any locks
1522 * so must be used with appropriate locks held only. The head item is
1523 * returned or %NULL if the list is empty.
1525 struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1526 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1528 struct sk_buff
*skb
= skb_peek(list
);
1530 __skb_unlink(skb
, list
);
1535 * __skb_dequeue_tail - remove from the tail of the queue
1536 * @list: list to dequeue from
1538 * Remove the tail of the list. This function does not take any locks
1539 * so must be used with appropriate locks held only. The tail item is
1540 * returned or %NULL if the list is empty.
1542 struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1543 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1545 struct sk_buff
*skb
= skb_peek_tail(list
);
1547 __skb_unlink(skb
, list
);
1552 static inline bool skb_is_nonlinear(const struct sk_buff
*skb
)
1554 return skb
->data_len
;
1557 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1559 return skb
->len
- skb
->data_len
;
1562 static inline int skb_pagelen(const struct sk_buff
*skb
)
1566 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1567 len
+= skb_frag_size(&skb_shinfo(skb
)->frags
[i
]);
1568 return len
+ skb_headlen(skb
);
1572 * __skb_fill_page_desc - initialise a paged fragment in an skb
1573 * @skb: buffer containing fragment to be initialised
1574 * @i: paged fragment index to initialise
1575 * @page: the page to use for this fragment
1576 * @off: the offset to the data with @page
1577 * @size: the length of the data
1579 * Initialises the @i'th fragment of @skb to point to &size bytes at
1580 * offset @off within @page.
1582 * Does not take any additional reference on the fragment.
1584 static inline void __skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1585 struct page
*page
, int off
, int size
)
1587 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1590 * Propagate page->pfmemalloc to the skb if we can. The problem is
1591 * that not all callers have unique ownership of the page. If
1592 * pfmemalloc is set, we check the mapping as a mapping implies
1593 * page->index is set (index and pfmemalloc share space).
1594 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1595 * do not lose pfmemalloc information as the pages would not be
1596 * allocated using __GFP_MEMALLOC.
1598 frag
->page
.p
= page
;
1599 frag
->page_offset
= off
;
1600 skb_frag_size_set(frag
, size
);
1602 page
= compound_head(page
);
1603 if (page
->pfmemalloc
&& !page
->mapping
)
1604 skb
->pfmemalloc
= true;
1608 * skb_fill_page_desc - initialise a paged fragment in an skb
1609 * @skb: buffer containing fragment to be initialised
1610 * @i: paged fragment index to initialise
1611 * @page: the page to use for this fragment
1612 * @off: the offset to the data with @page
1613 * @size: the length of the data
1615 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1616 * @skb to point to @size bytes at offset @off within @page. In
1617 * addition updates @skb such that @i is the last fragment.
1619 * Does not take any additional reference on the fragment.
1621 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1622 struct page
*page
, int off
, int size
)
1624 __skb_fill_page_desc(skb
, i
, page
, off
, size
);
1625 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1628 void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
, int off
,
1629 int size
, unsigned int truesize
);
1631 void skb_coalesce_rx_frag(struct sk_buff
*skb
, int i
, int size
,
1632 unsigned int truesize
);
1634 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1635 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1636 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1638 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1639 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1641 return skb
->head
+ skb
->tail
;
1644 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1646 skb
->tail
= skb
->data
- skb
->head
;
1649 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1651 skb_reset_tail_pointer(skb
);
1652 skb
->tail
+= offset
;
1655 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1656 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1661 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1663 skb
->tail
= skb
->data
;
1666 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1668 skb
->tail
= skb
->data
+ offset
;
1671 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1674 * Add data to an sk_buff
1676 unsigned char *pskb_put(struct sk_buff
*skb
, struct sk_buff
*tail
, int len
);
1677 unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1678 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1680 unsigned char *tmp
= skb_tail_pointer(skb
);
1681 SKB_LINEAR_ASSERT(skb
);
1687 unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1688 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1695 unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1696 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1699 BUG_ON(skb
->len
< skb
->data_len
);
1700 return skb
->data
+= len
;
1703 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1705 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1708 unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1710 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1712 if (len
> skb_headlen(skb
) &&
1713 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1716 return skb
->data
+= len
;
1719 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1721 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1724 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1726 if (likely(len
<= skb_headlen(skb
)))
1728 if (unlikely(len
> skb
->len
))
1730 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1734 * skb_headroom - bytes at buffer head
1735 * @skb: buffer to check
1737 * Return the number of bytes of free space at the head of an &sk_buff.
1739 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1741 return skb
->data
- skb
->head
;
1745 * skb_tailroom - bytes at buffer end
1746 * @skb: buffer to check
1748 * Return the number of bytes of free space at the tail of an sk_buff
1750 static inline int skb_tailroom(const struct sk_buff
*skb
)
1752 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1756 * skb_availroom - bytes at buffer end
1757 * @skb: buffer to check
1759 * Return the number of bytes of free space at the tail of an sk_buff
1760 * allocated by sk_stream_alloc()
1762 static inline int skb_availroom(const struct sk_buff
*skb
)
1764 if (skb_is_nonlinear(skb
))
1767 return skb
->end
- skb
->tail
- skb
->reserved_tailroom
;
1771 * skb_reserve - adjust headroom
1772 * @skb: buffer to alter
1773 * @len: bytes to move
1775 * Increase the headroom of an empty &sk_buff by reducing the tail
1776 * room. This is only allowed for an empty buffer.
1778 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
1784 #define ENCAP_TYPE_ETHER 0
1785 #define ENCAP_TYPE_IPPROTO 1
1787 static inline void skb_set_inner_protocol(struct sk_buff
*skb
,
1790 skb
->inner_protocol
= protocol
;
1791 skb
->inner_protocol_type
= ENCAP_TYPE_ETHER
;
1794 static inline void skb_set_inner_ipproto(struct sk_buff
*skb
,
1797 skb
->inner_ipproto
= ipproto
;
1798 skb
->inner_protocol_type
= ENCAP_TYPE_IPPROTO
;
1801 static inline void skb_reset_inner_headers(struct sk_buff
*skb
)
1803 skb
->inner_mac_header
= skb
->mac_header
;
1804 skb
->inner_network_header
= skb
->network_header
;
1805 skb
->inner_transport_header
= skb
->transport_header
;
1808 static inline void skb_reset_mac_len(struct sk_buff
*skb
)
1810 skb
->mac_len
= skb
->network_header
- skb
->mac_header
;
1813 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1816 return skb
->head
+ skb
->inner_transport_header
;
1819 static inline void skb_reset_inner_transport_header(struct sk_buff
*skb
)
1821 skb
->inner_transport_header
= skb
->data
- skb
->head
;
1824 static inline void skb_set_inner_transport_header(struct sk_buff
*skb
,
1827 skb_reset_inner_transport_header(skb
);
1828 skb
->inner_transport_header
+= offset
;
1831 static inline unsigned char *skb_inner_network_header(const struct sk_buff
*skb
)
1833 return skb
->head
+ skb
->inner_network_header
;
1836 static inline void skb_reset_inner_network_header(struct sk_buff
*skb
)
1838 skb
->inner_network_header
= skb
->data
- skb
->head
;
1841 static inline void skb_set_inner_network_header(struct sk_buff
*skb
,
1844 skb_reset_inner_network_header(skb
);
1845 skb
->inner_network_header
+= offset
;
1848 static inline unsigned char *skb_inner_mac_header(const struct sk_buff
*skb
)
1850 return skb
->head
+ skb
->inner_mac_header
;
1853 static inline void skb_reset_inner_mac_header(struct sk_buff
*skb
)
1855 skb
->inner_mac_header
= skb
->data
- skb
->head
;
1858 static inline void skb_set_inner_mac_header(struct sk_buff
*skb
,
1861 skb_reset_inner_mac_header(skb
);
1862 skb
->inner_mac_header
+= offset
;
1864 static inline bool skb_transport_header_was_set(const struct sk_buff
*skb
)
1866 return skb
->transport_header
!= (typeof(skb
->transport_header
))~0U;
1869 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1871 return skb
->head
+ skb
->transport_header
;
1874 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1876 skb
->transport_header
= skb
->data
- skb
->head
;
1879 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1882 skb_reset_transport_header(skb
);
1883 skb
->transport_header
+= offset
;
1886 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1888 return skb
->head
+ skb
->network_header
;
1891 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1893 skb
->network_header
= skb
->data
- skb
->head
;
1896 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1898 skb_reset_network_header(skb
);
1899 skb
->network_header
+= offset
;
1902 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1904 return skb
->head
+ skb
->mac_header
;
1907 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1909 return skb
->mac_header
!= (typeof(skb
->mac_header
))~0U;
1912 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1914 skb
->mac_header
= skb
->data
- skb
->head
;
1917 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1919 skb_reset_mac_header(skb
);
1920 skb
->mac_header
+= offset
;
1923 static inline void skb_pop_mac_header(struct sk_buff
*skb
)
1925 skb
->mac_header
= skb
->network_header
;
1928 static inline void skb_probe_transport_header(struct sk_buff
*skb
,
1929 const int offset_hint
)
1931 struct flow_keys keys
;
1933 if (skb_transport_header_was_set(skb
))
1935 else if (skb_flow_dissect(skb
, &keys
))
1936 skb_set_transport_header(skb
, keys
.thoff
);
1938 skb_set_transport_header(skb
, offset_hint
);
1941 static inline void skb_mac_header_rebuild(struct sk_buff
*skb
)
1943 if (skb_mac_header_was_set(skb
)) {
1944 const unsigned char *old_mac
= skb_mac_header(skb
);
1946 skb_set_mac_header(skb
, -skb
->mac_len
);
1947 memmove(skb_mac_header(skb
), old_mac
, skb
->mac_len
);
1951 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
1953 return skb
->csum_start
- skb_headroom(skb
);
1956 static inline int skb_transport_offset(const struct sk_buff
*skb
)
1958 return skb_transport_header(skb
) - skb
->data
;
1961 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
1963 return skb
->transport_header
- skb
->network_header
;
1966 static inline u32
skb_inner_network_header_len(const struct sk_buff
*skb
)
1968 return skb
->inner_transport_header
- skb
->inner_network_header
;
1971 static inline int skb_network_offset(const struct sk_buff
*skb
)
1973 return skb_network_header(skb
) - skb
->data
;
1976 static inline int skb_inner_network_offset(const struct sk_buff
*skb
)
1978 return skb_inner_network_header(skb
) - skb
->data
;
1981 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
1983 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
1987 * CPUs often take a performance hit when accessing unaligned memory
1988 * locations. The actual performance hit varies, it can be small if the
1989 * hardware handles it or large if we have to take an exception and fix it
1992 * Since an ethernet header is 14 bytes network drivers often end up with
1993 * the IP header at an unaligned offset. The IP header can be aligned by
1994 * shifting the start of the packet by 2 bytes. Drivers should do this
1997 * skb_reserve(skb, NET_IP_ALIGN);
1999 * The downside to this alignment of the IP header is that the DMA is now
2000 * unaligned. On some architectures the cost of an unaligned DMA is high
2001 * and this cost outweighs the gains made by aligning the IP header.
2003 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2006 #ifndef NET_IP_ALIGN
2007 #define NET_IP_ALIGN 2
2011 * The networking layer reserves some headroom in skb data (via
2012 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2013 * the header has to grow. In the default case, if the header has to grow
2014 * 32 bytes or less we avoid the reallocation.
2016 * Unfortunately this headroom changes the DMA alignment of the resulting
2017 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2018 * on some architectures. An architecture can override this value,
2019 * perhaps setting it to a cacheline in size (since that will maintain
2020 * cacheline alignment of the DMA). It must be a power of 2.
2022 * Various parts of the networking layer expect at least 32 bytes of
2023 * headroom, you should not reduce this.
2025 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2026 * to reduce average number of cache lines per packet.
2027 * get_rps_cpus() for example only access one 64 bytes aligned block :
2028 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2031 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2034 int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
2036 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
2038 if (unlikely(skb_is_nonlinear(skb
))) {
2043 skb_set_tail_pointer(skb
, len
);
2046 void skb_trim(struct sk_buff
*skb
, unsigned int len
);
2048 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2051 return ___pskb_trim(skb
, len
);
2052 __skb_trim(skb
, len
);
2056 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2058 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
2062 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2063 * @skb: buffer to alter
2066 * This is identical to pskb_trim except that the caller knows that
2067 * the skb is not cloned so we should never get an error due to out-
2070 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
2072 int err
= pskb_trim(skb
, len
);
2077 * skb_orphan - orphan a buffer
2078 * @skb: buffer to orphan
2080 * If a buffer currently has an owner then we call the owner's
2081 * destructor function and make the @skb unowned. The buffer continues
2082 * to exist but is no longer charged to its former owner.
2084 static inline void skb_orphan(struct sk_buff
*skb
)
2086 if (skb
->destructor
) {
2087 skb
->destructor(skb
);
2088 skb
->destructor
= NULL
;
2096 * skb_orphan_frags - orphan the frags contained in a buffer
2097 * @skb: buffer to orphan frags from
2098 * @gfp_mask: allocation mask for replacement pages
2100 * For each frag in the SKB which needs a destructor (i.e. has an
2101 * owner) create a copy of that frag and release the original
2102 * page by calling the destructor.
2104 static inline int skb_orphan_frags(struct sk_buff
*skb
, gfp_t gfp_mask
)
2106 if (likely(!(skb_shinfo(skb
)->tx_flags
& SKBTX_DEV_ZEROCOPY
)))
2108 return skb_copy_ubufs(skb
, gfp_mask
);
2112 * __skb_queue_purge - empty a list
2113 * @list: list to empty
2115 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2116 * the list and one reference dropped. This function does not take the
2117 * list lock and the caller must hold the relevant locks to use it.
2119 void skb_queue_purge(struct sk_buff_head
*list
);
2120 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
2122 struct sk_buff
*skb
;
2123 while ((skb
= __skb_dequeue(list
)) != NULL
)
2127 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2128 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2129 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2131 void *netdev_alloc_frag(unsigned int fragsz
);
2133 struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
, unsigned int length
,
2137 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2138 * @dev: network device to receive on
2139 * @length: length to allocate
2141 * Allocate a new &sk_buff and assign it a usage count of one. The
2142 * buffer has unspecified headroom built in. Users should allocate
2143 * the headroom they think they need without accounting for the
2144 * built in space. The built in space is used for optimisations.
2146 * %NULL is returned if there is no free memory. Although this function
2147 * allocates memory it can be called from an interrupt.
2149 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
2150 unsigned int length
)
2152 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
2155 /* legacy helper around __netdev_alloc_skb() */
2156 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
2159 return __netdev_alloc_skb(NULL
, length
, gfp_mask
);
2162 /* legacy helper around netdev_alloc_skb() */
2163 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
2165 return netdev_alloc_skb(NULL
, length
);
2169 static inline struct sk_buff
*__netdev_alloc_skb_ip_align(struct net_device
*dev
,
2170 unsigned int length
, gfp_t gfp
)
2172 struct sk_buff
*skb
= __netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
, gfp
);
2174 if (NET_IP_ALIGN
&& skb
)
2175 skb_reserve(skb
, NET_IP_ALIGN
);
2179 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
2180 unsigned int length
)
2182 return __netdev_alloc_skb_ip_align(dev
, length
, GFP_ATOMIC
);
2186 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
2187 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2188 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2189 * @order: size of the allocation
2191 * Allocate a new page.
2193 * %NULL is returned if there is no free memory.
2195 static inline struct page
*__skb_alloc_pages(gfp_t gfp_mask
,
2196 struct sk_buff
*skb
,
2201 gfp_mask
|= __GFP_COLD
;
2203 if (!(gfp_mask
& __GFP_NOMEMALLOC
))
2204 gfp_mask
|= __GFP_MEMALLOC
;
2206 page
= alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, order
);
2207 if (skb
&& page
&& page
->pfmemalloc
)
2208 skb
->pfmemalloc
= true;
2214 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2215 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2216 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2218 * Allocate a new page.
2220 * %NULL is returned if there is no free memory.
2222 static inline struct page
*__skb_alloc_page(gfp_t gfp_mask
,
2223 struct sk_buff
*skb
)
2225 return __skb_alloc_pages(gfp_mask
, skb
, 0);
2229 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2230 * @page: The page that was allocated from skb_alloc_page
2231 * @skb: The skb that may need pfmemalloc set
2233 static inline void skb_propagate_pfmemalloc(struct page
*page
,
2234 struct sk_buff
*skb
)
2236 if (page
&& page
->pfmemalloc
)
2237 skb
->pfmemalloc
= true;
2241 * skb_frag_page - retrieve the page referred to by a paged fragment
2242 * @frag: the paged fragment
2244 * Returns the &struct page associated with @frag.
2246 static inline struct page
*skb_frag_page(const skb_frag_t
*frag
)
2248 return frag
->page
.p
;
2252 * __skb_frag_ref - take an addition reference on a paged fragment.
2253 * @frag: the paged fragment
2255 * Takes an additional reference on the paged fragment @frag.
2257 static inline void __skb_frag_ref(skb_frag_t
*frag
)
2259 get_page(skb_frag_page(frag
));
2263 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2265 * @f: the fragment offset.
2267 * Takes an additional reference on the @f'th paged fragment of @skb.
2269 static inline void skb_frag_ref(struct sk_buff
*skb
, int f
)
2271 __skb_frag_ref(&skb_shinfo(skb
)->frags
[f
]);
2275 * __skb_frag_unref - release a reference on a paged fragment.
2276 * @frag: the paged fragment
2278 * Releases a reference on the paged fragment @frag.
2280 static inline void __skb_frag_unref(skb_frag_t
*frag
)
2282 put_page(skb_frag_page(frag
));
2286 * skb_frag_unref - release a reference on a paged fragment of an skb.
2288 * @f: the fragment offset
2290 * Releases a reference on the @f'th paged fragment of @skb.
2292 static inline void skb_frag_unref(struct sk_buff
*skb
, int f
)
2294 __skb_frag_unref(&skb_shinfo(skb
)->frags
[f
]);
2298 * skb_frag_address - gets the address of the data contained in a paged fragment
2299 * @frag: the paged fragment buffer
2301 * Returns the address of the data within @frag. The page must already
2304 static inline void *skb_frag_address(const skb_frag_t
*frag
)
2306 return page_address(skb_frag_page(frag
)) + frag
->page_offset
;
2310 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2311 * @frag: the paged fragment buffer
2313 * Returns the address of the data within @frag. Checks that the page
2314 * is mapped and returns %NULL otherwise.
2316 static inline void *skb_frag_address_safe(const skb_frag_t
*frag
)
2318 void *ptr
= page_address(skb_frag_page(frag
));
2322 return ptr
+ frag
->page_offset
;
2326 * __skb_frag_set_page - sets the page contained in a paged fragment
2327 * @frag: the paged fragment
2328 * @page: the page to set
2330 * Sets the fragment @frag to contain @page.
2332 static inline void __skb_frag_set_page(skb_frag_t
*frag
, struct page
*page
)
2334 frag
->page
.p
= page
;
2338 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2340 * @f: the fragment offset
2341 * @page: the page to set
2343 * Sets the @f'th fragment of @skb to contain @page.
2345 static inline void skb_frag_set_page(struct sk_buff
*skb
, int f
,
2348 __skb_frag_set_page(&skb_shinfo(skb
)->frags
[f
], page
);
2351 bool skb_page_frag_refill(unsigned int sz
, struct page_frag
*pfrag
, gfp_t prio
);
2354 * skb_frag_dma_map - maps a paged fragment via the DMA API
2355 * @dev: the device to map the fragment to
2356 * @frag: the paged fragment to map
2357 * @offset: the offset within the fragment (starting at the
2358 * fragment's own offset)
2359 * @size: the number of bytes to map
2360 * @dir: the direction of the mapping (%PCI_DMA_*)
2362 * Maps the page associated with @frag to @device.
2364 static inline dma_addr_t
skb_frag_dma_map(struct device
*dev
,
2365 const skb_frag_t
*frag
,
2366 size_t offset
, size_t size
,
2367 enum dma_data_direction dir
)
2369 return dma_map_page(dev
, skb_frag_page(frag
),
2370 frag
->page_offset
+ offset
, size
, dir
);
2373 static inline struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
2376 return __pskb_copy(skb
, skb_headroom(skb
), gfp_mask
);
2380 static inline struct sk_buff
*pskb_copy_for_clone(struct sk_buff
*skb
,
2383 return __pskb_copy_fclone(skb
, skb_headroom(skb
), gfp_mask
, true);
2388 * skb_clone_writable - is the header of a clone writable
2389 * @skb: buffer to check
2390 * @len: length up to which to write
2392 * Returns true if modifying the header part of the cloned buffer
2393 * does not requires the data to be copied.
2395 static inline int skb_clone_writable(const struct sk_buff
*skb
, unsigned int len
)
2397 return !skb_header_cloned(skb
) &&
2398 skb_headroom(skb
) + len
<= skb
->hdr_len
;
2401 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
2406 if (headroom
> skb_headroom(skb
))
2407 delta
= headroom
- skb_headroom(skb
);
2409 if (delta
|| cloned
)
2410 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
2416 * skb_cow - copy header of skb when it is required
2417 * @skb: buffer to cow
2418 * @headroom: needed headroom
2420 * If the skb passed lacks sufficient headroom or its data part
2421 * is shared, data is reallocated. If reallocation fails, an error
2422 * is returned and original skb is not changed.
2424 * The result is skb with writable area skb->head...skb->tail
2425 * and at least @headroom of space at head.
2427 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
2429 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
2433 * skb_cow_head - skb_cow but only making the head writable
2434 * @skb: buffer to cow
2435 * @headroom: needed headroom
2437 * This function is identical to skb_cow except that we replace the
2438 * skb_cloned check by skb_header_cloned. It should be used when
2439 * you only need to push on some header and do not need to modify
2442 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
2444 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
2448 * skb_padto - pad an skbuff up to a minimal size
2449 * @skb: buffer to pad
2450 * @len: minimal length
2452 * Pads up a buffer to ensure the trailing bytes exist and are
2453 * blanked. If the buffer already contains sufficient data it
2454 * is untouched. Otherwise it is extended. Returns zero on
2455 * success. The skb is freed on error.
2458 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
2460 unsigned int size
= skb
->len
;
2461 if (likely(size
>= len
))
2463 return skb_pad(skb
, len
- size
);
2466 static inline int skb_add_data(struct sk_buff
*skb
,
2467 char __user
*from
, int copy
)
2469 const int off
= skb
->len
;
2471 if (skb
->ip_summed
== CHECKSUM_NONE
) {
2473 __wsum csum
= csum_and_copy_from_user(from
, skb_put(skb
, copy
),
2476 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
2479 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
2482 __skb_trim(skb
, off
);
2486 static inline bool skb_can_coalesce(struct sk_buff
*skb
, int i
,
2487 const struct page
*page
, int off
)
2490 const struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
2492 return page
== skb_frag_page(frag
) &&
2493 off
== frag
->page_offset
+ skb_frag_size(frag
);
2498 static inline int __skb_linearize(struct sk_buff
*skb
)
2500 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
2504 * skb_linearize - convert paged skb to linear one
2505 * @skb: buffer to linarize
2507 * If there is no free memory -ENOMEM is returned, otherwise zero
2508 * is returned and the old skb data released.
2510 static inline int skb_linearize(struct sk_buff
*skb
)
2512 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
2516 * skb_has_shared_frag - can any frag be overwritten
2517 * @skb: buffer to test
2519 * Return true if the skb has at least one frag that might be modified
2520 * by an external entity (as in vmsplice()/sendfile())
2522 static inline bool skb_has_shared_frag(const struct sk_buff
*skb
)
2524 return skb_is_nonlinear(skb
) &&
2525 skb_shinfo(skb
)->tx_flags
& SKBTX_SHARED_FRAG
;
2529 * skb_linearize_cow - make sure skb is linear and writable
2530 * @skb: buffer to process
2532 * If there is no free memory -ENOMEM is returned, otherwise zero
2533 * is returned and the old skb data released.
2535 static inline int skb_linearize_cow(struct sk_buff
*skb
)
2537 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
2538 __skb_linearize(skb
) : 0;
2542 * skb_postpull_rcsum - update checksum for received skb after pull
2543 * @skb: buffer to update
2544 * @start: start of data before pull
2545 * @len: length of data pulled
2547 * After doing a pull on a received packet, you need to call this to
2548 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2549 * CHECKSUM_NONE so that it can be recomputed from scratch.
2552 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
2553 const void *start
, unsigned int len
)
2555 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2556 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
2559 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
2562 * pskb_trim_rcsum - trim received skb and update checksum
2563 * @skb: buffer to trim
2566 * This is exactly the same as pskb_trim except that it ensures the
2567 * checksum of received packets are still valid after the operation.
2570 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
2572 if (likely(len
>= skb
->len
))
2574 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2575 skb
->ip_summed
= CHECKSUM_NONE
;
2576 return __pskb_trim(skb
, len
);
2579 #define skb_queue_walk(queue, skb) \
2580 for (skb = (queue)->next; \
2581 skb != (struct sk_buff *)(queue); \
2584 #define skb_queue_walk_safe(queue, skb, tmp) \
2585 for (skb = (queue)->next, tmp = skb->next; \
2586 skb != (struct sk_buff *)(queue); \
2587 skb = tmp, tmp = skb->next)
2589 #define skb_queue_walk_from(queue, skb) \
2590 for (; skb != (struct sk_buff *)(queue); \
2593 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2594 for (tmp = skb->next; \
2595 skb != (struct sk_buff *)(queue); \
2596 skb = tmp, tmp = skb->next)
2598 #define skb_queue_reverse_walk(queue, skb) \
2599 for (skb = (queue)->prev; \
2600 skb != (struct sk_buff *)(queue); \
2603 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2604 for (skb = (queue)->prev, tmp = skb->prev; \
2605 skb != (struct sk_buff *)(queue); \
2606 skb = tmp, tmp = skb->prev)
2608 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2609 for (tmp = skb->prev; \
2610 skb != (struct sk_buff *)(queue); \
2611 skb = tmp, tmp = skb->prev)
2613 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
2615 return skb_shinfo(skb
)->frag_list
!= NULL
;
2618 static inline void skb_frag_list_init(struct sk_buff
*skb
)
2620 skb_shinfo(skb
)->frag_list
= NULL
;
2623 static inline void skb_frag_add_head(struct sk_buff
*skb
, struct sk_buff
*frag
)
2625 frag
->next
= skb_shinfo(skb
)->frag_list
;
2626 skb_shinfo(skb
)->frag_list
= frag
;
2629 #define skb_walk_frags(skb, iter) \
2630 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2632 struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
2633 int *peeked
, int *off
, int *err
);
2634 struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
, int noblock
,
2636 unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
2637 struct poll_table_struct
*wait
);
2638 int skb_copy_datagram_iovec(const struct sk_buff
*from
, int offset
,
2639 struct iovec
*to
, int size
);
2640 int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
, int hlen
,
2642 int skb_copy_datagram_from_iovec(struct sk_buff
*skb
, int offset
,
2643 const struct iovec
*from
, int from_offset
,
2645 int zerocopy_sg_from_iovec(struct sk_buff
*skb
, const struct iovec
*frm
,
2646 int offset
, size_t count
);
2647 int skb_copy_datagram_const_iovec(const struct sk_buff
*from
, int offset
,
2648 const struct iovec
*to
, int to_offset
,
2650 void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
2651 void skb_free_datagram_locked(struct sock
*sk
, struct sk_buff
*skb
);
2652 int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
, unsigned int flags
);
2653 int skb_copy_bits(const struct sk_buff
*skb
, int offset
, void *to
, int len
);
2654 int skb_store_bits(struct sk_buff
*skb
, int offset
, const void *from
, int len
);
2655 __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
, int offset
, u8
*to
,
2656 int len
, __wsum csum
);
2657 int skb_splice_bits(struct sk_buff
*skb
, unsigned int offset
,
2658 struct pipe_inode_info
*pipe
, unsigned int len
,
2659 unsigned int flags
);
2660 void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
2661 unsigned int skb_zerocopy_headlen(const struct sk_buff
*from
);
2662 int skb_zerocopy(struct sk_buff
*to
, struct sk_buff
*from
,
2664 void skb_split(struct sk_buff
*skb
, struct sk_buff
*skb1
, const u32 len
);
2665 int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
, int shiftlen
);
2666 void skb_scrub_packet(struct sk_buff
*skb
, bool xnet
);
2667 unsigned int skb_gso_transport_seglen(const struct sk_buff
*skb
);
2668 struct sk_buff
*skb_segment(struct sk_buff
*skb
, netdev_features_t features
);
2669 struct sk_buff
*skb_vlan_untag(struct sk_buff
*skb
);
2671 struct skb_checksum_ops
{
2672 __wsum (*update
)(const void *mem
, int len
, __wsum wsum
);
2673 __wsum (*combine
)(__wsum csum
, __wsum csum2
, int offset
, int len
);
2676 __wsum
__skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
2677 __wsum csum
, const struct skb_checksum_ops
*ops
);
2678 __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
2681 static inline void *__skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2682 int len
, void *data
, int hlen
, void *buffer
)
2684 if (hlen
- offset
>= len
)
2685 return data
+ offset
;
2688 skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
2694 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2695 int len
, void *buffer
)
2697 return __skb_header_pointer(skb
, offset
, len
, skb
->data
,
2698 skb_headlen(skb
), buffer
);
2702 * skb_needs_linearize - check if we need to linearize a given skb
2703 * depending on the given device features.
2704 * @skb: socket buffer to check
2705 * @features: net device features
2707 * Returns true if either:
2708 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2709 * 2. skb is fragmented and the device does not support SG.
2711 static inline bool skb_needs_linearize(struct sk_buff
*skb
,
2712 netdev_features_t features
)
2714 return skb_is_nonlinear(skb
) &&
2715 ((skb_has_frag_list(skb
) && !(features
& NETIF_F_FRAGLIST
)) ||
2716 (skb_shinfo(skb
)->nr_frags
&& !(features
& NETIF_F_SG
)));
2719 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
2721 const unsigned int len
)
2723 memcpy(to
, skb
->data
, len
);
2726 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
2727 const int offset
, void *to
,
2728 const unsigned int len
)
2730 memcpy(to
, skb
->data
+ offset
, len
);
2733 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
2735 const unsigned int len
)
2737 memcpy(skb
->data
, from
, len
);
2740 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
2743 const unsigned int len
)
2745 memcpy(skb
->data
+ offset
, from
, len
);
2748 void skb_init(void);
2750 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
2756 * skb_get_timestamp - get timestamp from a skb
2757 * @skb: skb to get stamp from
2758 * @stamp: pointer to struct timeval to store stamp in
2760 * Timestamps are stored in the skb as offsets to a base timestamp.
2761 * This function converts the offset back to a struct timeval and stores
2764 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
2765 struct timeval
*stamp
)
2767 *stamp
= ktime_to_timeval(skb
->tstamp
);
2770 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
2771 struct timespec
*stamp
)
2773 *stamp
= ktime_to_timespec(skb
->tstamp
);
2776 static inline void __net_timestamp(struct sk_buff
*skb
)
2778 skb
->tstamp
= ktime_get_real();
2781 static inline ktime_t
net_timedelta(ktime_t t
)
2783 return ktime_sub(ktime_get_real(), t
);
2786 static inline ktime_t
net_invalid_timestamp(void)
2788 return ktime_set(0, 0);
2791 struct sk_buff
*skb_clone_sk(struct sk_buff
*skb
);
2793 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2795 void skb_clone_tx_timestamp(struct sk_buff
*skb
);
2796 bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
2798 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2800 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
2804 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
2809 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2812 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2814 * PHY drivers may accept clones of transmitted packets for
2815 * timestamping via their phy_driver.txtstamp method. These drivers
2816 * must call this function to return the skb back to the stack, with
2817 * or without a timestamp.
2819 * @skb: clone of the the original outgoing packet
2820 * @hwtstamps: hardware time stamps, may be NULL if not available
2823 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
2824 struct skb_shared_hwtstamps
*hwtstamps
);
2826 void __skb_tstamp_tx(struct sk_buff
*orig_skb
,
2827 struct skb_shared_hwtstamps
*hwtstamps
,
2828 struct sock
*sk
, int tstype
);
2831 * skb_tstamp_tx - queue clone of skb with send time stamps
2832 * @orig_skb: the original outgoing packet
2833 * @hwtstamps: hardware time stamps, may be NULL if not available
2835 * If the skb has a socket associated, then this function clones the
2836 * skb (thus sharing the actual data and optional structures), stores
2837 * the optional hardware time stamping information (if non NULL) or
2838 * generates a software time stamp (otherwise), then queues the clone
2839 * to the error queue of the socket. Errors are silently ignored.
2841 void skb_tstamp_tx(struct sk_buff
*orig_skb
,
2842 struct skb_shared_hwtstamps
*hwtstamps
);
2844 static inline void sw_tx_timestamp(struct sk_buff
*skb
)
2846 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
&&
2847 !(skb_shinfo(skb
)->tx_flags
& SKBTX_IN_PROGRESS
))
2848 skb_tstamp_tx(skb
, NULL
);
2852 * skb_tx_timestamp() - Driver hook for transmit timestamping
2854 * Ethernet MAC Drivers should call this function in their hard_xmit()
2855 * function immediately before giving the sk_buff to the MAC hardware.
2857 * Specifically, one should make absolutely sure that this function is
2858 * called before TX completion of this packet can trigger. Otherwise
2859 * the packet could potentially already be freed.
2861 * @skb: A socket buffer.
2863 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
2865 skb_clone_tx_timestamp(skb
);
2866 sw_tx_timestamp(skb
);
2870 * skb_complete_wifi_ack - deliver skb with wifi status
2872 * @skb: the original outgoing packet
2873 * @acked: ack status
2876 void skb_complete_wifi_ack(struct sk_buff
*skb
, bool acked
);
2878 __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
2879 __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
2881 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
2883 return ((skb
->ip_summed
& CHECKSUM_UNNECESSARY
) || skb
->csum_valid
);
2887 * skb_checksum_complete - Calculate checksum of an entire packet
2888 * @skb: packet to process
2890 * This function calculates the checksum over the entire packet plus
2891 * the value of skb->csum. The latter can be used to supply the
2892 * checksum of a pseudo header as used by TCP/UDP. It returns the
2895 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2896 * this function can be used to verify that checksum on received
2897 * packets. In that case the function should return zero if the
2898 * checksum is correct. In particular, this function will return zero
2899 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2900 * hardware has already verified the correctness of the checksum.
2902 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
2904 return skb_csum_unnecessary(skb
) ?
2905 0 : __skb_checksum_complete(skb
);
2908 static inline void __skb_decr_checksum_unnecessary(struct sk_buff
*skb
)
2910 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
2911 if (skb
->csum_level
== 0)
2912 skb
->ip_summed
= CHECKSUM_NONE
;
2918 static inline void __skb_incr_checksum_unnecessary(struct sk_buff
*skb
)
2920 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
2921 if (skb
->csum_level
< SKB_MAX_CSUM_LEVEL
)
2923 } else if (skb
->ip_summed
== CHECKSUM_NONE
) {
2924 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
2925 skb
->csum_level
= 0;
2929 static inline void __skb_mark_checksum_bad(struct sk_buff
*skb
)
2931 /* Mark current checksum as bad (typically called from GRO
2932 * path). In the case that ip_summed is CHECKSUM_NONE
2933 * this must be the first checksum encountered in the packet.
2934 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2935 * checksum after the last one validated. For UDP, a zero
2936 * checksum can not be marked as bad.
2939 if (skb
->ip_summed
== CHECKSUM_NONE
||
2940 skb
->ip_summed
== CHECKSUM_UNNECESSARY
)
2944 /* Check if we need to perform checksum complete validation.
2946 * Returns true if checksum complete is needed, false otherwise
2947 * (either checksum is unnecessary or zero checksum is allowed).
2949 static inline bool __skb_checksum_validate_needed(struct sk_buff
*skb
,
2953 if (skb_csum_unnecessary(skb
) || (zero_okay
&& !check
)) {
2954 skb
->csum_valid
= 1;
2955 __skb_decr_checksum_unnecessary(skb
);
2962 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2965 #define CHECKSUM_BREAK 76
2967 /* Validate (init) checksum based on checksum complete.
2970 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2971 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2972 * checksum is stored in skb->csum for use in __skb_checksum_complete
2973 * non-zero: value of invalid checksum
2976 static inline __sum16
__skb_checksum_validate_complete(struct sk_buff
*skb
,
2980 if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
2981 if (!csum_fold(csum_add(psum
, skb
->csum
))) {
2982 skb
->csum_valid
= 1;
2985 } else if (skb
->csum_bad
) {
2986 /* ip_summed == CHECKSUM_NONE in this case */
2992 if (complete
|| skb
->len
<= CHECKSUM_BREAK
) {
2995 csum
= __skb_checksum_complete(skb
);
2996 skb
->csum_valid
= !csum
;
3003 static inline __wsum
null_compute_pseudo(struct sk_buff
*skb
, int proto
)
3008 /* Perform checksum validate (init). Note that this is a macro since we only
3009 * want to calculate the pseudo header which is an input function if necessary.
3010 * First we try to validate without any computation (checksum unnecessary) and
3011 * then calculate based on checksum complete calling the function to compute
3015 * 0: checksum is validated or try to in skb_checksum_complete
3016 * non-zero: value of invalid checksum
3018 #define __skb_checksum_validate(skb, proto, complete, \
3019 zero_okay, check, compute_pseudo) \
3021 __sum16 __ret = 0; \
3022 skb->csum_valid = 0; \
3023 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3024 __ret = __skb_checksum_validate_complete(skb, \
3025 complete, compute_pseudo(skb, proto)); \
3029 #define skb_checksum_init(skb, proto, compute_pseudo) \
3030 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3032 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3033 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3035 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3036 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3038 #define skb_checksum_validate_zero_check(skb, proto, check, \
3040 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
3042 #define skb_checksum_simple_validate(skb) \
3043 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3045 static inline bool __skb_checksum_convert_check(struct sk_buff
*skb
)
3047 return (skb
->ip_summed
== CHECKSUM_NONE
&&
3048 skb
->csum_valid
&& !skb
->csum_bad
);
3051 static inline void __skb_checksum_convert(struct sk_buff
*skb
,
3052 __sum16 check
, __wsum pseudo
)
3054 skb
->csum
= ~pseudo
;
3055 skb
->ip_summed
= CHECKSUM_COMPLETE
;
3058 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3060 if (__skb_checksum_convert_check(skb)) \
3061 __skb_checksum_convert(skb, check, \
3062 compute_pseudo(skb, proto)); \
3065 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3066 void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
3067 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
3069 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
3070 nf_conntrack_destroy(nfct
);
3072 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
3075 atomic_inc(&nfct
->use
);
3078 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3079 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
3081 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
3084 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
3087 atomic_inc(&nf_bridge
->use
);
3089 #endif /* CONFIG_BRIDGE_NETFILTER */
3090 static inline void nf_reset(struct sk_buff
*skb
)
3092 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3093 nf_conntrack_put(skb
->nfct
);
3096 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3097 nf_bridge_put(skb
->nf_bridge
);
3098 skb
->nf_bridge
= NULL
;
3102 static inline void nf_reset_trace(struct sk_buff
*skb
)
3104 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3109 /* Note: This doesn't put any conntrack and bridge info in dst. */
3110 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
,
3113 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3114 dst
->nfct
= src
->nfct
;
3115 nf_conntrack_get(src
->nfct
);
3117 dst
->nfctinfo
= src
->nfctinfo
;
3119 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3120 dst
->nf_bridge
= src
->nf_bridge
;
3121 nf_bridge_get(src
->nf_bridge
);
3123 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3125 dst
->nf_trace
= src
->nf_trace
;
3129 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
3131 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3132 nf_conntrack_put(dst
->nfct
);
3134 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3135 nf_bridge_put(dst
->nf_bridge
);
3137 __nf_copy(dst
, src
, true);
3140 #ifdef CONFIG_NETWORK_SECMARK
3141 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3143 to
->secmark
= from
->secmark
;
3146 static inline void skb_init_secmark(struct sk_buff
*skb
)
3151 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3154 static inline void skb_init_secmark(struct sk_buff
*skb
)
3158 static inline bool skb_irq_freeable(const struct sk_buff
*skb
)
3160 return !skb
->destructor
&&
3161 #if IS_ENABLED(CONFIG_XFRM)
3164 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3167 !skb
->_skb_refdst
&&
3168 !skb_has_frag_list(skb
);
3171 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
3173 skb
->queue_mapping
= queue_mapping
;
3176 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
3178 return skb
->queue_mapping
;
3181 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
3183 to
->queue_mapping
= from
->queue_mapping
;
3186 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
3188 skb
->queue_mapping
= rx_queue
+ 1;
3191 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
3193 return skb
->queue_mapping
- 1;
3196 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
3198 return skb
->queue_mapping
!= 0;
3201 u16
__skb_tx_hash(const struct net_device
*dev
, struct sk_buff
*skb
,
3202 unsigned int num_tx_queues
);
3204 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
3213 /* Keeps track of mac header offset relative to skb->head.
3214 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3215 * For non-tunnel skb it points to skb_mac_header() and for
3216 * tunnel skb it points to outer mac header.
3217 * Keeps track of level of encapsulation of network headers.
3224 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3226 static inline int skb_tnl_header_len(const struct sk_buff
*inner_skb
)
3228 return (skb_mac_header(inner_skb
) - inner_skb
->head
) -
3229 SKB_GSO_CB(inner_skb
)->mac_offset
;
3232 static inline int gso_pskb_expand_head(struct sk_buff
*skb
, int extra
)
3234 int new_headroom
, headroom
;
3237 headroom
= skb_headroom(skb
);
3238 ret
= pskb_expand_head(skb
, extra
, 0, GFP_ATOMIC
);
3242 new_headroom
= skb_headroom(skb
);
3243 SKB_GSO_CB(skb
)->mac_offset
+= (new_headroom
- headroom
);
3247 /* Compute the checksum for a gso segment. First compute the checksum value
3248 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3249 * then add in skb->csum (checksum from csum_start to end of packet).
3250 * skb->csum and csum_start are then updated to reflect the checksum of the
3251 * resultant packet starting from the transport header-- the resultant checksum
3252 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3255 static inline __sum16
gso_make_checksum(struct sk_buff
*skb
, __wsum res
)
3257 int plen
= SKB_GSO_CB(skb
)->csum_start
- skb_headroom(skb
) -
3258 skb_transport_offset(skb
);
3261 csum
= csum_fold(csum_partial(skb_transport_header(skb
),
3264 SKB_GSO_CB(skb
)->csum_start
-= plen
;
3269 static inline bool skb_is_gso(const struct sk_buff
*skb
)
3271 return skb_shinfo(skb
)->gso_size
;
3274 /* Note: Should be called only if skb_is_gso(skb) is true */
3275 static inline bool skb_is_gso_v6(const struct sk_buff
*skb
)
3277 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
3280 void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
3282 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
3284 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3285 * wanted then gso_type will be set. */
3286 const struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
3288 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
3289 unlikely(shinfo
->gso_type
== 0)) {
3290 __skb_warn_lro_forwarding(skb
);
3296 static inline void skb_forward_csum(struct sk_buff
*skb
)
3298 /* Unfortunately we don't support this one. Any brave souls? */
3299 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3300 skb
->ip_summed
= CHECKSUM_NONE
;
3304 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3305 * @skb: skb to check
3307 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3308 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3309 * use this helper, to document places where we make this assertion.
3311 static inline void skb_checksum_none_assert(const struct sk_buff
*skb
)
3314 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
3318 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
3320 int skb_checksum_setup(struct sk_buff
*skb
, bool recalculate
);
3322 u32
skb_get_poff(const struct sk_buff
*skb
);
3323 u32
__skb_get_poff(const struct sk_buff
*skb
, void *data
,
3324 const struct flow_keys
*keys
, int hlen
);
3327 * skb_head_is_locked - Determine if the skb->head is locked down
3328 * @skb: skb to check
3330 * The head on skbs build around a head frag can be removed if they are
3331 * not cloned. This function returns true if the skb head is locked down
3332 * due to either being allocated via kmalloc, or by being a clone with
3333 * multiple references to the head.
3335 static inline bool skb_head_is_locked(const struct sk_buff
*skb
)
3337 return !skb
->head_frag
|| skb_cloned(skb
);
3341 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3345 * skb_gso_network_seglen is used to determine the real size of the
3346 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3348 * The MAC/L2 header is not accounted for.
3350 static inline unsigned int skb_gso_network_seglen(const struct sk_buff
*skb
)
3352 unsigned int hdr_len
= skb_transport_header(skb
) -
3353 skb_network_header(skb
);
3354 return hdr_len
+ skb_gso_transport_seglen(skb
);
3356 #endif /* __KERNEL__ */
3357 #endif /* _LINUX_SKBUFF_H */