2 * Definitions for the 'struct sk_buff' memory handlers.
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/kernel.h>
18 #include <linux/kmemcheck.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/bug.h>
22 #include <linux/cache.h>
23 #include <linux/rbtree.h>
24 #include <linux/socket.h>
26 #include <linux/atomic.h>
27 #include <asm/types.h>
28 #include <linux/spinlock.h>
29 #include <linux/net.h>
30 #include <linux/textsearch.h>
31 #include <net/checksum.h>
32 #include <linux/rcupdate.h>
33 #include <linux/hrtimer.h>
34 #include <linux/dma-mapping.h>
35 #include <linux/netdev_features.h>
36 #include <linux/sched.h>
37 #include <net/flow_keys.h>
39 /* A. Checksumming of received packets by device.
43 * Device failed to checksum this packet e.g. due to lack of capabilities.
44 * The packet contains full (though not verified) checksum in packet but
45 * not in skb->csum. Thus, skb->csum is undefined in this case.
47 * CHECKSUM_UNNECESSARY:
49 * The hardware you're dealing with doesn't calculate the full checksum
50 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
51 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
52 * if their checksums are okay. skb->csum is still undefined in this case
53 * though. It is a bad option, but, unfortunately, nowadays most vendors do
54 * this. Apparently with the secret goal to sell you new devices, when you
55 * will add new protocol to your host, f.e. IPv6 8)
57 * CHECKSUM_UNNECESSARY is applicable to following protocols:
59 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
60 * zero UDP checksum for either IPv4 or IPv6, the networking stack
61 * may perform further validation in this case.
62 * GRE: only if the checksum is present in the header.
63 * SCTP: indicates the CRC in SCTP header has been validated.
65 * skb->csum_level indicates the number of consecutive checksums found in
66 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
67 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
68 * and a device is able to verify the checksums for UDP (possibly zero),
69 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
70 * two. If the device were only able to verify the UDP checksum and not
71 * GRE, either because it doesn't support GRE checksum of because GRE
72 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
73 * not considered in this case).
77 * This is the most generic way. The device supplied checksum of the _whole_
78 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
79 * hardware doesn't need to parse L3/L4 headers to implement this.
81 * Note: Even if device supports only some protocols, but is able to produce
82 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
86 * A checksum is set up to be offloaded to a device as described in the
87 * output description for CHECKSUM_PARTIAL. This may occur on a packet
88 * received directly from another Linux OS, e.g., a virtualized Linux kernel
89 * on the same host, or it may be set in the input path in GRO or remote
90 * checksum offload. For the purposes of checksum verification, the checksum
91 * referred to by skb->csum_start + skb->csum_offset and any preceding
92 * checksums in the packet are considered verified. Any checksums in the
93 * packet that are after the checksum being offloaded are not considered to
96 * B. Checksumming on output.
100 * The skb was already checksummed by the protocol, or a checksum is not
105 * The device is required to checksum the packet as seen by hard_start_xmit()
106 * from skb->csum_start up to the end, and to record/write the checksum at
107 * offset skb->csum_start + skb->csum_offset.
109 * The device must show its capabilities in dev->features, set up at device
110 * setup time, e.g. netdev_features.h:
112 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
113 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
114 * IPv4. Sigh. Vendors like this way for an unknown reason.
115 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
116 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
117 * NETIF_F_... - Well, you get the picture.
119 * CHECKSUM_UNNECESSARY:
121 * Normally, the device will do per protocol specific checksumming. Protocol
122 * implementations that do not want the NIC to perform the checksum
123 * calculation should use this flag in their outgoing skbs.
125 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
126 * offload. Correspondingly, the FCoE protocol driver
127 * stack should use CHECKSUM_UNNECESSARY.
129 * Any questions? No questions, good. --ANK
132 /* Don't change this without changing skb_csum_unnecessary! */
133 #define CHECKSUM_NONE 0
134 #define CHECKSUM_UNNECESSARY 1
135 #define CHECKSUM_COMPLETE 2
136 #define CHECKSUM_PARTIAL 3
138 /* Maximum value in skb->csum_level */
139 #define SKB_MAX_CSUM_LEVEL 3
141 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
142 #define SKB_WITH_OVERHEAD(X) \
143 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
144 #define SKB_MAX_ORDER(X, ORDER) \
145 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
146 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
147 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
149 /* return minimum truesize of one skb containing X bytes of data */
150 #define SKB_TRUESIZE(X) ((X) + \
151 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
152 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
156 struct pipe_inode_info
;
160 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
161 struct nf_conntrack
{
166 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
167 struct nf_bridge_info
{
170 BRNF_PROTO_UNCHANGED
,
176 struct net_device
*physindev
;
177 struct net_device
*physoutdev
;
178 char neigh_header
[8];
183 struct sk_buff_head
{
184 /* These two members must be first. */
185 struct sk_buff
*next
;
186 struct sk_buff
*prev
;
194 /* To allow 64K frame to be packed as single skb without frag_list we
195 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
196 * buffers which do not start on a page boundary.
198 * Since GRO uses frags we allocate at least 16 regardless of page
201 #if (65536/PAGE_SIZE + 1) < 16
202 #define MAX_SKB_FRAGS 16UL
204 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
207 typedef struct skb_frag_struct skb_frag_t
;
209 struct skb_frag_struct
{
213 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
222 static inline unsigned int skb_frag_size(const skb_frag_t
*frag
)
227 static inline void skb_frag_size_set(skb_frag_t
*frag
, unsigned int size
)
232 static inline void skb_frag_size_add(skb_frag_t
*frag
, int delta
)
237 static inline void skb_frag_size_sub(skb_frag_t
*frag
, int delta
)
242 #define HAVE_HW_TIME_STAMP
245 * struct skb_shared_hwtstamps - hardware time stamps
246 * @hwtstamp: hardware time stamp transformed into duration
247 * since arbitrary point in time
249 * Software time stamps generated by ktime_get_real() are stored in
252 * hwtstamps can only be compared against other hwtstamps from
255 * This structure is attached to packets as part of the
256 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
258 struct skb_shared_hwtstamps
{
262 /* Definitions for tx_flags in struct skb_shared_info */
264 /* generate hardware time stamp */
265 SKBTX_HW_TSTAMP
= 1 << 0,
267 /* generate software time stamp when queueing packet to NIC */
268 SKBTX_SW_TSTAMP
= 1 << 1,
270 /* device driver is going to provide hardware time stamp */
271 SKBTX_IN_PROGRESS
= 1 << 2,
273 /* device driver supports TX zero-copy buffers */
274 SKBTX_DEV_ZEROCOPY
= 1 << 3,
276 /* generate wifi status information (where possible) */
277 SKBTX_WIFI_STATUS
= 1 << 4,
279 /* This indicates at least one fragment might be overwritten
280 * (as in vmsplice(), sendfile() ...)
281 * If we need to compute a TX checksum, we'll need to copy
282 * all frags to avoid possible bad checksum
284 SKBTX_SHARED_FRAG
= 1 << 5,
286 /* generate software time stamp when entering packet scheduling */
287 SKBTX_SCHED_TSTAMP
= 1 << 6,
289 /* generate software timestamp on peer data acknowledgment */
290 SKBTX_ACK_TSTAMP
= 1 << 7,
293 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
294 SKBTX_SCHED_TSTAMP | \
296 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
299 * The callback notifies userspace to release buffers when skb DMA is done in
300 * lower device, the skb last reference should be 0 when calling this.
301 * The zerocopy_success argument is true if zero copy transmit occurred,
302 * false on data copy or out of memory error caused by data copy attempt.
303 * The ctx field is used to track device context.
304 * The desc field is used to track userspace buffer index.
307 void (*callback
)(struct ubuf_info
*, bool zerocopy_success
);
312 /* This data is invariant across clones and lives at
313 * the end of the header data, ie. at skb->end.
315 struct skb_shared_info
{
316 unsigned char nr_frags
;
318 unsigned short gso_size
;
319 /* Warning: this field is not always filled in (UFO)! */
320 unsigned short gso_segs
;
321 unsigned short gso_type
;
322 struct sk_buff
*frag_list
;
323 struct skb_shared_hwtstamps hwtstamps
;
328 * Warning : all fields before dataref are cleared in __alloc_skb()
332 /* Intermediate layers must ensure that destructor_arg
333 * remains valid until skb destructor */
334 void * destructor_arg
;
336 /* must be last field, see pskb_expand_head() */
337 skb_frag_t frags
[MAX_SKB_FRAGS
];
340 /* We divide dataref into two halves. The higher 16 bits hold references
341 * to the payload part of skb->data. The lower 16 bits hold references to
342 * the entire skb->data. A clone of a headerless skb holds the length of
343 * the header in skb->hdr_len.
345 * All users must obey the rule that the skb->data reference count must be
346 * greater than or equal to the payload reference count.
348 * Holding a reference to the payload part means that the user does not
349 * care about modifications to the header part of skb->data.
351 #define SKB_DATAREF_SHIFT 16
352 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
356 SKB_FCLONE_UNAVAILABLE
, /* skb has no fclone (from head_cache) */
357 SKB_FCLONE_ORIG
, /* orig skb (from fclone_cache) */
358 SKB_FCLONE_CLONE
, /* companion fclone skb (from fclone_cache) */
362 SKB_GSO_TCPV4
= 1 << 0,
363 SKB_GSO_UDP
= 1 << 1,
365 /* This indicates the skb is from an untrusted source. */
366 SKB_GSO_DODGY
= 1 << 2,
368 /* This indicates the tcp segment has CWR set. */
369 SKB_GSO_TCP_ECN
= 1 << 3,
371 SKB_GSO_TCPV6
= 1 << 4,
373 SKB_GSO_FCOE
= 1 << 5,
375 SKB_GSO_GRE
= 1 << 6,
377 SKB_GSO_GRE_CSUM
= 1 << 7,
379 SKB_GSO_IPIP
= 1 << 8,
381 SKB_GSO_SIT
= 1 << 9,
383 SKB_GSO_UDP_TUNNEL
= 1 << 10,
385 SKB_GSO_UDP_TUNNEL_CSUM
= 1 << 11,
387 SKB_GSO_TUNNEL_REMCSUM
= 1 << 12,
390 #if BITS_PER_LONG > 32
391 #define NET_SKBUFF_DATA_USES_OFFSET 1
394 #ifdef NET_SKBUFF_DATA_USES_OFFSET
395 typedef unsigned int sk_buff_data_t
;
397 typedef unsigned char *sk_buff_data_t
;
401 * struct skb_mstamp - multi resolution time stamps
402 * @stamp_us: timestamp in us resolution
403 * @stamp_jiffies: timestamp in jiffies
416 * skb_mstamp_get - get current timestamp
417 * @cl: place to store timestamps
419 static inline void skb_mstamp_get(struct skb_mstamp
*cl
)
421 u64 val
= local_clock();
423 do_div(val
, NSEC_PER_USEC
);
424 cl
->stamp_us
= (u32
)val
;
425 cl
->stamp_jiffies
= (u32
)jiffies
;
429 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
430 * @t1: pointer to newest sample
431 * @t0: pointer to oldest sample
433 static inline u32
skb_mstamp_us_delta(const struct skb_mstamp
*t1
,
434 const struct skb_mstamp
*t0
)
436 s32 delta_us
= t1
->stamp_us
- t0
->stamp_us
;
437 u32 delta_jiffies
= t1
->stamp_jiffies
- t0
->stamp_jiffies
;
439 /* If delta_us is negative, this might be because interval is too big,
440 * or local_clock() drift is too big : fallback using jiffies.
443 delta_jiffies
>= (INT_MAX
/ (USEC_PER_SEC
/ HZ
)))
445 delta_us
= jiffies_to_usecs(delta_jiffies
);
452 * struct sk_buff - socket buffer
453 * @next: Next buffer in list
454 * @prev: Previous buffer in list
455 * @tstamp: Time we arrived/left
456 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
457 * @sk: Socket we are owned by
458 * @dev: Device we arrived on/are leaving by
459 * @cb: Control buffer. Free for use by every layer. Put private vars here
460 * @_skb_refdst: destination entry (with norefcount bit)
461 * @sp: the security path, used for xfrm
462 * @len: Length of actual data
463 * @data_len: Data length
464 * @mac_len: Length of link layer header
465 * @hdr_len: writable header length of cloned skb
466 * @csum: Checksum (must include start/offset pair)
467 * @csum_start: Offset from skb->head where checksumming should start
468 * @csum_offset: Offset from csum_start where checksum should be stored
469 * @priority: Packet queueing priority
470 * @ignore_df: allow local fragmentation
471 * @cloned: Head may be cloned (check refcnt to be sure)
472 * @ip_summed: Driver fed us an IP checksum
473 * @nohdr: Payload reference only, must not modify header
474 * @nfctinfo: Relationship of this skb to the connection
475 * @pkt_type: Packet class
476 * @fclone: skbuff clone status
477 * @ipvs_property: skbuff is owned by ipvs
478 * @peeked: this packet has been seen already, so stats have been
479 * done for it, don't do them again
480 * @nf_trace: netfilter packet trace flag
481 * @protocol: Packet protocol from driver
482 * @destructor: Destruct function
483 * @nfct: Associated connection, if any
484 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
485 * @skb_iif: ifindex of device we arrived on
486 * @tc_index: Traffic control index
487 * @tc_verd: traffic control verdict
488 * @hash: the packet hash
489 * @queue_mapping: Queue mapping for multiqueue devices
490 * @xmit_more: More SKBs are pending for this queue
491 * @ndisc_nodetype: router type (from link layer)
492 * @ooo_okay: allow the mapping of a socket to a queue to be changed
493 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
495 * @sw_hash: indicates hash was computed in software stack
496 * @wifi_acked_valid: wifi_acked was set
497 * @wifi_acked: whether frame was acked on wifi or not
498 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
499 * @napi_id: id of the NAPI struct this skb came from
500 * @secmark: security marking
501 * @mark: Generic packet mark
502 * @vlan_proto: vlan encapsulation protocol
503 * @vlan_tci: vlan tag control information
504 * @inner_protocol: Protocol (encapsulation)
505 * @inner_transport_header: Inner transport layer header (encapsulation)
506 * @inner_network_header: Network layer header (encapsulation)
507 * @inner_mac_header: Link layer header (encapsulation)
508 * @transport_header: Transport layer header
509 * @network_header: Network layer header
510 * @mac_header: Link layer header
511 * @tail: Tail pointer
513 * @head: Head of buffer
514 * @data: Data head pointer
515 * @truesize: Buffer size
516 * @users: User count - see {datagram,tcp}.c
522 /* These two members must be first. */
523 struct sk_buff
*next
;
524 struct sk_buff
*prev
;
528 struct skb_mstamp skb_mstamp
;
531 struct rb_node rbnode
; /* used in netem & tcp stack */
534 struct net_device
*dev
;
537 * This is the control buffer. It is free to use for every
538 * layer. Please put your private variables there. If you
539 * want to keep them across layers you have to do a skb_clone()
540 * first. This is owned by whoever has the skb queued ATM.
542 char cb
[48] __aligned(8);
544 unsigned long _skb_refdst
;
545 void (*destructor
)(struct sk_buff
*skb
);
549 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
550 struct nf_conntrack
*nfct
;
552 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
553 struct nf_bridge_info
*nf_bridge
;
560 /* Following fields are _not_ copied in __copy_skb_header()
561 * Note that queue_mapping is here mostly to fill a hole.
563 kmemcheck_bitfield_begin(flags1
);
572 kmemcheck_bitfield_end(flags1
);
574 /* fields enclosed in headers_start/headers_end are copied
575 * using a single memcpy() in __copy_skb_header()
578 __u32 headers_start
[0];
581 /* if you move pkt_type around you also must adapt those constants */
582 #ifdef __BIG_ENDIAN_BITFIELD
583 #define PKT_TYPE_MAX (7 << 5)
585 #define PKT_TYPE_MAX 7
587 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
589 __u8 __pkt_type_offset
[0];
600 __u8 wifi_acked_valid
:1;
604 /* Indicates the inner headers are valid in the skbuff. */
605 __u8 encapsulation
:1;
606 __u8 encap_hdr_csum
:1;
608 __u8 csum_complete_sw
:1;
612 #ifdef CONFIG_IPV6_NDISC_NODETYPE
613 __u8 ndisc_nodetype
:2;
615 __u8 ipvs_property
:1;
616 __u8 inner_protocol_type
:1;
617 __u8 remcsum_offload
:1;
618 /* 3 or 5 bit hole */
620 #ifdef CONFIG_NET_SCHED
621 __u16 tc_index
; /* traffic control index */
622 #ifdef CONFIG_NET_CLS_ACT
623 __u16 tc_verd
; /* traffic control verdict */
639 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
641 unsigned int napi_id
;
642 unsigned int sender_cpu
;
645 #ifdef CONFIG_NETWORK_SECMARK
650 __u32 reserved_tailroom
;
654 __be16 inner_protocol
;
658 __u16 inner_transport_header
;
659 __u16 inner_network_header
;
660 __u16 inner_mac_header
;
663 __u16 transport_header
;
664 __u16 network_header
;
668 __u32 headers_end
[0];
671 /* These elements must be at the end, see alloc_skb() for details. */
676 unsigned int truesize
;
682 * Handling routines are only of interest to the kernel
684 #include <linux/slab.h>
687 #define SKB_ALLOC_FCLONE 0x01
688 #define SKB_ALLOC_RX 0x02
689 #define SKB_ALLOC_NAPI 0x04
691 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
692 static inline bool skb_pfmemalloc(const struct sk_buff
*skb
)
694 return unlikely(skb
->pfmemalloc
);
698 * skb might have a dst pointer attached, refcounted or not.
699 * _skb_refdst low order bit is set if refcount was _not_ taken
701 #define SKB_DST_NOREF 1UL
702 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
705 * skb_dst - returns skb dst_entry
708 * Returns skb dst_entry, regardless of reference taken or not.
710 static inline struct dst_entry
*skb_dst(const struct sk_buff
*skb
)
712 /* If refdst was not refcounted, check we still are in a
713 * rcu_read_lock section
715 WARN_ON((skb
->_skb_refdst
& SKB_DST_NOREF
) &&
716 !rcu_read_lock_held() &&
717 !rcu_read_lock_bh_held());
718 return (struct dst_entry
*)(skb
->_skb_refdst
& SKB_DST_PTRMASK
);
722 * skb_dst_set - sets skb dst
726 * Sets skb dst, assuming a reference was taken on dst and should
727 * be released by skb_dst_drop()
729 static inline void skb_dst_set(struct sk_buff
*skb
, struct dst_entry
*dst
)
731 skb
->_skb_refdst
= (unsigned long)dst
;
735 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
739 * Sets skb dst, assuming a reference was not taken on dst.
740 * If dst entry is cached, we do not take reference and dst_release
741 * will be avoided by refdst_drop. If dst entry is not cached, we take
742 * reference, so that last dst_release can destroy the dst immediately.
744 static inline void skb_dst_set_noref(struct sk_buff
*skb
, struct dst_entry
*dst
)
746 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
747 skb
->_skb_refdst
= (unsigned long)dst
| SKB_DST_NOREF
;
751 * skb_dst_is_noref - Test if skb dst isn't refcounted
754 static inline bool skb_dst_is_noref(const struct sk_buff
*skb
)
756 return (skb
->_skb_refdst
& SKB_DST_NOREF
) && skb_dst(skb
);
759 static inline struct rtable
*skb_rtable(const struct sk_buff
*skb
)
761 return (struct rtable
*)skb_dst(skb
);
764 void kfree_skb(struct sk_buff
*skb
);
765 void kfree_skb_list(struct sk_buff
*segs
);
766 void skb_tx_error(struct sk_buff
*skb
);
767 void consume_skb(struct sk_buff
*skb
);
768 void __kfree_skb(struct sk_buff
*skb
);
769 extern struct kmem_cache
*skbuff_head_cache
;
771 void kfree_skb_partial(struct sk_buff
*skb
, bool head_stolen
);
772 bool skb_try_coalesce(struct sk_buff
*to
, struct sk_buff
*from
,
773 bool *fragstolen
, int *delta_truesize
);
775 struct sk_buff
*__alloc_skb(unsigned int size
, gfp_t priority
, int flags
,
777 struct sk_buff
*__build_skb(void *data
, unsigned int frag_size
);
778 struct sk_buff
*build_skb(void *data
, unsigned int frag_size
);
779 static inline struct sk_buff
*alloc_skb(unsigned int size
,
782 return __alloc_skb(size
, priority
, 0, NUMA_NO_NODE
);
785 struct sk_buff
*alloc_skb_with_frags(unsigned long header_len
,
786 unsigned long data_len
,
791 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
792 struct sk_buff_fclones
{
801 * skb_fclone_busy - check if fclone is busy
804 * Returns true is skb is a fast clone, and its clone is not freed.
805 * Some drivers call skb_orphan() in their ndo_start_xmit(),
806 * so we also check that this didnt happen.
808 static inline bool skb_fclone_busy(const struct sock
*sk
,
809 const struct sk_buff
*skb
)
811 const struct sk_buff_fclones
*fclones
;
813 fclones
= container_of(skb
, struct sk_buff_fclones
, skb1
);
815 return skb
->fclone
== SKB_FCLONE_ORIG
&&
816 atomic_read(&fclones
->fclone_ref
) > 1 &&
817 fclones
->skb2
.sk
== sk
;
820 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
823 return __alloc_skb(size
, priority
, SKB_ALLOC_FCLONE
, NUMA_NO_NODE
);
826 struct sk_buff
*__alloc_skb_head(gfp_t priority
, int node
);
827 static inline struct sk_buff
*alloc_skb_head(gfp_t priority
)
829 return __alloc_skb_head(priority
, -1);
832 struct sk_buff
*skb_morph(struct sk_buff
*dst
, struct sk_buff
*src
);
833 int skb_copy_ubufs(struct sk_buff
*skb
, gfp_t gfp_mask
);
834 struct sk_buff
*skb_clone(struct sk_buff
*skb
, gfp_t priority
);
835 struct sk_buff
*skb_copy(const struct sk_buff
*skb
, gfp_t priority
);
836 struct sk_buff
*__pskb_copy_fclone(struct sk_buff
*skb
, int headroom
,
837 gfp_t gfp_mask
, bool fclone
);
838 static inline struct sk_buff
*__pskb_copy(struct sk_buff
*skb
, int headroom
,
841 return __pskb_copy_fclone(skb
, headroom
, gfp_mask
, false);
844 int pskb_expand_head(struct sk_buff
*skb
, int nhead
, int ntail
, gfp_t gfp_mask
);
845 struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
846 unsigned int headroom
);
847 struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
, int newheadroom
,
848 int newtailroom
, gfp_t priority
);
849 int skb_to_sgvec_nomark(struct sk_buff
*skb
, struct scatterlist
*sg
,
850 int offset
, int len
);
851 int skb_to_sgvec(struct sk_buff
*skb
, struct scatterlist
*sg
, int offset
,
853 int skb_cow_data(struct sk_buff
*skb
, int tailbits
, struct sk_buff
**trailer
);
854 int skb_pad(struct sk_buff
*skb
, int pad
);
855 #define dev_kfree_skb(a) consume_skb(a)
857 int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
858 int getfrag(void *from
, char *to
, int offset
,
859 int len
, int odd
, struct sk_buff
*skb
),
860 void *from
, int length
);
862 struct skb_seq_state
{
866 __u32 stepped_offset
;
867 struct sk_buff
*root_skb
;
868 struct sk_buff
*cur_skb
;
872 void skb_prepare_seq_read(struct sk_buff
*skb
, unsigned int from
,
873 unsigned int to
, struct skb_seq_state
*st
);
874 unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
875 struct skb_seq_state
*st
);
876 void skb_abort_seq_read(struct skb_seq_state
*st
);
878 unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
879 unsigned int to
, struct ts_config
*config
);
882 * Packet hash types specify the type of hash in skb_set_hash.
884 * Hash types refer to the protocol layer addresses which are used to
885 * construct a packet's hash. The hashes are used to differentiate or identify
886 * flows of the protocol layer for the hash type. Hash types are either
887 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
889 * Properties of hashes:
891 * 1) Two packets in different flows have different hash values
892 * 2) Two packets in the same flow should have the same hash value
894 * A hash at a higher layer is considered to be more specific. A driver should
895 * set the most specific hash possible.
897 * A driver cannot indicate a more specific hash than the layer at which a hash
898 * was computed. For instance an L3 hash cannot be set as an L4 hash.
900 * A driver may indicate a hash level which is less specific than the
901 * actual layer the hash was computed on. For instance, a hash computed
902 * at L4 may be considered an L3 hash. This should only be done if the
903 * driver can't unambiguously determine that the HW computed the hash at
904 * the higher layer. Note that the "should" in the second property above
907 enum pkt_hash_types
{
908 PKT_HASH_TYPE_NONE
, /* Undefined type */
909 PKT_HASH_TYPE_L2
, /* Input: src_MAC, dest_MAC */
910 PKT_HASH_TYPE_L3
, /* Input: src_IP, dst_IP */
911 PKT_HASH_TYPE_L4
, /* Input: src_IP, dst_IP, src_port, dst_port */
915 skb_set_hash(struct sk_buff
*skb
, __u32 hash
, enum pkt_hash_types type
)
917 skb
->l4_hash
= (type
== PKT_HASH_TYPE_L4
);
922 void __skb_get_hash(struct sk_buff
*skb
);
923 static inline __u32
skb_get_hash(struct sk_buff
*skb
)
925 if (!skb
->l4_hash
&& !skb
->sw_hash
)
931 static inline __u32
skb_get_hash_raw(const struct sk_buff
*skb
)
936 static inline void skb_clear_hash(struct sk_buff
*skb
)
943 static inline void skb_clear_hash_if_not_l4(struct sk_buff
*skb
)
949 static inline void skb_copy_hash(struct sk_buff
*to
, const struct sk_buff
*from
)
951 to
->hash
= from
->hash
;
952 to
->sw_hash
= from
->sw_hash
;
953 to
->l4_hash
= from
->l4_hash
;
956 static inline void skb_sender_cpu_clear(struct sk_buff
*skb
)
963 #ifdef NET_SKBUFF_DATA_USES_OFFSET
964 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
966 return skb
->head
+ skb
->end
;
969 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
974 static inline unsigned char *skb_end_pointer(const struct sk_buff
*skb
)
979 static inline unsigned int skb_end_offset(const struct sk_buff
*skb
)
981 return skb
->end
- skb
->head
;
986 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
988 static inline struct skb_shared_hwtstamps
*skb_hwtstamps(struct sk_buff
*skb
)
990 return &skb_shinfo(skb
)->hwtstamps
;
994 * skb_queue_empty - check if a queue is empty
997 * Returns true if the queue is empty, false otherwise.
999 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
1001 return list
->next
== (const struct sk_buff
*) list
;
1005 * skb_queue_is_last - check if skb is the last entry in the queue
1009 * Returns true if @skb is the last buffer on the list.
1011 static inline bool skb_queue_is_last(const struct sk_buff_head
*list
,
1012 const struct sk_buff
*skb
)
1014 return skb
->next
== (const struct sk_buff
*) list
;
1018 * skb_queue_is_first - check if skb is the first entry in the queue
1022 * Returns true if @skb is the first buffer on the list.
1024 static inline bool skb_queue_is_first(const struct sk_buff_head
*list
,
1025 const struct sk_buff
*skb
)
1027 return skb
->prev
== (const struct sk_buff
*) list
;
1031 * skb_queue_next - return the next packet in the queue
1033 * @skb: current buffer
1035 * Return the next packet in @list after @skb. It is only valid to
1036 * call this if skb_queue_is_last() evaluates to false.
1038 static inline struct sk_buff
*skb_queue_next(const struct sk_buff_head
*list
,
1039 const struct sk_buff
*skb
)
1041 /* This BUG_ON may seem severe, but if we just return then we
1042 * are going to dereference garbage.
1044 BUG_ON(skb_queue_is_last(list
, skb
));
1049 * skb_queue_prev - return the prev packet in the queue
1051 * @skb: current buffer
1053 * Return the prev packet in @list before @skb. It is only valid to
1054 * call this if skb_queue_is_first() evaluates to false.
1056 static inline struct sk_buff
*skb_queue_prev(const struct sk_buff_head
*list
,
1057 const struct sk_buff
*skb
)
1059 /* This BUG_ON may seem severe, but if we just return then we
1060 * are going to dereference garbage.
1062 BUG_ON(skb_queue_is_first(list
, skb
));
1067 * skb_get - reference buffer
1068 * @skb: buffer to reference
1070 * Makes another reference to a socket buffer and returns a pointer
1073 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
1075 atomic_inc(&skb
->users
);
1080 * If users == 1, we are the only owner and are can avoid redundant
1085 * skb_cloned - is the buffer a clone
1086 * @skb: buffer to check
1088 * Returns true if the buffer was generated with skb_clone() and is
1089 * one of multiple shared copies of the buffer. Cloned buffers are
1090 * shared data so must not be written to under normal circumstances.
1092 static inline int skb_cloned(const struct sk_buff
*skb
)
1094 return skb
->cloned
&&
1095 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
1098 static inline int skb_unclone(struct sk_buff
*skb
, gfp_t pri
)
1100 might_sleep_if(pri
& __GFP_WAIT
);
1102 if (skb_cloned(skb
))
1103 return pskb_expand_head(skb
, 0, 0, pri
);
1109 * skb_header_cloned - is the header a clone
1110 * @skb: buffer to check
1112 * Returns true if modifying the header part of the buffer requires
1113 * the data to be copied.
1115 static inline int skb_header_cloned(const struct sk_buff
*skb
)
1122 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
1123 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
1124 return dataref
!= 1;
1128 * skb_header_release - release reference to header
1129 * @skb: buffer to operate on
1131 * Drop a reference to the header part of the buffer. This is done
1132 * by acquiring a payload reference. You must not read from the header
1133 * part of skb->data after this.
1134 * Note : Check if you can use __skb_header_release() instead.
1136 static inline void skb_header_release(struct sk_buff
*skb
)
1140 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
1144 * __skb_header_release - release reference to header
1145 * @skb: buffer to operate on
1147 * Variant of skb_header_release() assuming skb is private to caller.
1148 * We can avoid one atomic operation.
1150 static inline void __skb_header_release(struct sk_buff
*skb
)
1153 atomic_set(&skb_shinfo(skb
)->dataref
, 1 + (1 << SKB_DATAREF_SHIFT
));
1158 * skb_shared - is the buffer shared
1159 * @skb: buffer to check
1161 * Returns true if more than one person has a reference to this
1164 static inline int skb_shared(const struct sk_buff
*skb
)
1166 return atomic_read(&skb
->users
) != 1;
1170 * skb_share_check - check if buffer is shared and if so clone it
1171 * @skb: buffer to check
1172 * @pri: priority for memory allocation
1174 * If the buffer is shared the buffer is cloned and the old copy
1175 * drops a reference. A new clone with a single reference is returned.
1176 * If the buffer is not shared the original buffer is returned. When
1177 * being called from interrupt status or with spinlocks held pri must
1180 * NULL is returned on a memory allocation failure.
1182 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
, gfp_t pri
)
1184 might_sleep_if(pri
& __GFP_WAIT
);
1185 if (skb_shared(skb
)) {
1186 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
1198 * Copy shared buffers into a new sk_buff. We effectively do COW on
1199 * packets to handle cases where we have a local reader and forward
1200 * and a couple of other messy ones. The normal one is tcpdumping
1201 * a packet thats being forwarded.
1205 * skb_unshare - make a copy of a shared buffer
1206 * @skb: buffer to check
1207 * @pri: priority for memory allocation
1209 * If the socket buffer is a clone then this function creates a new
1210 * copy of the data, drops a reference count on the old copy and returns
1211 * the new copy with the reference count at 1. If the buffer is not a clone
1212 * the original buffer is returned. When called with a spinlock held or
1213 * from interrupt state @pri must be %GFP_ATOMIC
1215 * %NULL is returned on a memory allocation failure.
1217 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
1220 might_sleep_if(pri
& __GFP_WAIT
);
1221 if (skb_cloned(skb
)) {
1222 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
1224 /* Free our shared copy */
1235 * skb_peek - peek at the head of an &sk_buff_head
1236 * @list_: list to peek at
1238 * Peek an &sk_buff. Unlike most other operations you _MUST_
1239 * be careful with this one. A peek leaves the buffer on the
1240 * list and someone else may run off with it. You must hold
1241 * the appropriate locks or have a private queue to do this.
1243 * Returns %NULL for an empty list or a pointer to the head element.
1244 * The reference count is not incremented and the reference is therefore
1245 * volatile. Use with caution.
1247 static inline struct sk_buff
*skb_peek(const struct sk_buff_head
*list_
)
1249 struct sk_buff
*skb
= list_
->next
;
1251 if (skb
== (struct sk_buff
*)list_
)
1257 * skb_peek_next - peek skb following the given one from a queue
1258 * @skb: skb to start from
1259 * @list_: list to peek at
1261 * Returns %NULL when the end of the list is met or a pointer to the
1262 * next element. The reference count is not incremented and the
1263 * reference is therefore volatile. Use with caution.
1265 static inline struct sk_buff
*skb_peek_next(struct sk_buff
*skb
,
1266 const struct sk_buff_head
*list_
)
1268 struct sk_buff
*next
= skb
->next
;
1270 if (next
== (struct sk_buff
*)list_
)
1276 * skb_peek_tail - peek at the tail of an &sk_buff_head
1277 * @list_: list to peek at
1279 * Peek an &sk_buff. Unlike most other operations you _MUST_
1280 * be careful with this one. A peek leaves the buffer on the
1281 * list and someone else may run off with it. You must hold
1282 * the appropriate locks or have a private queue to do this.
1284 * Returns %NULL for an empty list or a pointer to the tail element.
1285 * The reference count is not incremented and the reference is therefore
1286 * volatile. Use with caution.
1288 static inline struct sk_buff
*skb_peek_tail(const struct sk_buff_head
*list_
)
1290 struct sk_buff
*skb
= list_
->prev
;
1292 if (skb
== (struct sk_buff
*)list_
)
1299 * skb_queue_len - get queue length
1300 * @list_: list to measure
1302 * Return the length of an &sk_buff queue.
1304 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
1310 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1311 * @list: queue to initialize
1313 * This initializes only the list and queue length aspects of
1314 * an sk_buff_head object. This allows to initialize the list
1315 * aspects of an sk_buff_head without reinitializing things like
1316 * the spinlock. It can also be used for on-stack sk_buff_head
1317 * objects where the spinlock is known to not be used.
1319 static inline void __skb_queue_head_init(struct sk_buff_head
*list
)
1321 list
->prev
= list
->next
= (struct sk_buff
*)list
;
1326 * This function creates a split out lock class for each invocation;
1327 * this is needed for now since a whole lot of users of the skb-queue
1328 * infrastructure in drivers have different locking usage (in hardirq)
1329 * than the networking core (in softirq only). In the long run either the
1330 * network layer or drivers should need annotation to consolidate the
1331 * main types of usage into 3 classes.
1333 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
1335 spin_lock_init(&list
->lock
);
1336 __skb_queue_head_init(list
);
1339 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
1340 struct lock_class_key
*class)
1342 skb_queue_head_init(list
);
1343 lockdep_set_class(&list
->lock
, class);
1347 * Insert an sk_buff on a list.
1349 * The "__skb_xxxx()" functions are the non-atomic ones that
1350 * can only be called with interrupts disabled.
1352 void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
,
1353 struct sk_buff_head
*list
);
1354 static inline void __skb_insert(struct sk_buff
*newsk
,
1355 struct sk_buff
*prev
, struct sk_buff
*next
,
1356 struct sk_buff_head
*list
)
1360 next
->prev
= prev
->next
= newsk
;
1364 static inline void __skb_queue_splice(const struct sk_buff_head
*list
,
1365 struct sk_buff
*prev
,
1366 struct sk_buff
*next
)
1368 struct sk_buff
*first
= list
->next
;
1369 struct sk_buff
*last
= list
->prev
;
1379 * skb_queue_splice - join two skb lists, this is designed for stacks
1380 * @list: the new list to add
1381 * @head: the place to add it in the first list
1383 static inline void skb_queue_splice(const struct sk_buff_head
*list
,
1384 struct sk_buff_head
*head
)
1386 if (!skb_queue_empty(list
)) {
1387 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1388 head
->qlen
+= list
->qlen
;
1393 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1394 * @list: the new list to add
1395 * @head: the place to add it in the first list
1397 * The list at @list is reinitialised
1399 static inline void skb_queue_splice_init(struct sk_buff_head
*list
,
1400 struct sk_buff_head
*head
)
1402 if (!skb_queue_empty(list
)) {
1403 __skb_queue_splice(list
, (struct sk_buff
*) head
, head
->next
);
1404 head
->qlen
+= list
->qlen
;
1405 __skb_queue_head_init(list
);
1410 * skb_queue_splice_tail - join two skb lists, each list being a queue
1411 * @list: the new list to add
1412 * @head: the place to add it in the first list
1414 static inline void skb_queue_splice_tail(const struct sk_buff_head
*list
,
1415 struct sk_buff_head
*head
)
1417 if (!skb_queue_empty(list
)) {
1418 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1419 head
->qlen
+= list
->qlen
;
1424 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1425 * @list: the new list to add
1426 * @head: the place to add it in the first list
1428 * Each of the lists is a queue.
1429 * The list at @list is reinitialised
1431 static inline void skb_queue_splice_tail_init(struct sk_buff_head
*list
,
1432 struct sk_buff_head
*head
)
1434 if (!skb_queue_empty(list
)) {
1435 __skb_queue_splice(list
, head
->prev
, (struct sk_buff
*) head
);
1436 head
->qlen
+= list
->qlen
;
1437 __skb_queue_head_init(list
);
1442 * __skb_queue_after - queue a buffer at the list head
1443 * @list: list to use
1444 * @prev: place after this buffer
1445 * @newsk: buffer to queue
1447 * Queue a buffer int the middle of a list. This function takes no locks
1448 * and you must therefore hold required locks before calling it.
1450 * A buffer cannot be placed on two lists at the same time.
1452 static inline void __skb_queue_after(struct sk_buff_head
*list
,
1453 struct sk_buff
*prev
,
1454 struct sk_buff
*newsk
)
1456 __skb_insert(newsk
, prev
, prev
->next
, list
);
1459 void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
,
1460 struct sk_buff_head
*list
);
1462 static inline void __skb_queue_before(struct sk_buff_head
*list
,
1463 struct sk_buff
*next
,
1464 struct sk_buff
*newsk
)
1466 __skb_insert(newsk
, next
->prev
, next
, list
);
1470 * __skb_queue_head - queue a buffer at the list head
1471 * @list: list to use
1472 * @newsk: buffer to queue
1474 * Queue a buffer at the start of a list. This function takes no locks
1475 * and you must therefore hold required locks before calling it.
1477 * A buffer cannot be placed on two lists at the same time.
1479 void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1480 static inline void __skb_queue_head(struct sk_buff_head
*list
,
1481 struct sk_buff
*newsk
)
1483 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
1487 * __skb_queue_tail - queue a buffer at the list tail
1488 * @list: list to use
1489 * @newsk: buffer to queue
1491 * Queue a buffer at the end of a list. This function takes no locks
1492 * and you must therefore hold required locks before calling it.
1494 * A buffer cannot be placed on two lists at the same time.
1496 void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
1497 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
1498 struct sk_buff
*newsk
)
1500 __skb_queue_before(list
, (struct sk_buff
*)list
, newsk
);
1504 * remove sk_buff from list. _Must_ be called atomically, and with
1507 void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
1508 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
1510 struct sk_buff
*next
, *prev
;
1515 skb
->next
= skb
->prev
= NULL
;
1521 * __skb_dequeue - remove from the head of the queue
1522 * @list: list to dequeue from
1524 * Remove the head of the list. This function does not take any locks
1525 * so must be used with appropriate locks held only. The head item is
1526 * returned or %NULL if the list is empty.
1528 struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
1529 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
1531 struct sk_buff
*skb
= skb_peek(list
);
1533 __skb_unlink(skb
, list
);
1538 * __skb_dequeue_tail - remove from the tail of the queue
1539 * @list: list to dequeue from
1541 * Remove the tail of the list. This function does not take any locks
1542 * so must be used with appropriate locks held only. The tail item is
1543 * returned or %NULL if the list is empty.
1545 struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
1546 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
1548 struct sk_buff
*skb
= skb_peek_tail(list
);
1550 __skb_unlink(skb
, list
);
1555 static inline bool skb_is_nonlinear(const struct sk_buff
*skb
)
1557 return skb
->data_len
;
1560 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
1562 return skb
->len
- skb
->data_len
;
1565 static inline int skb_pagelen(const struct sk_buff
*skb
)
1569 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
1570 len
+= skb_frag_size(&skb_shinfo(skb
)->frags
[i
]);
1571 return len
+ skb_headlen(skb
);
1575 * __skb_fill_page_desc - initialise a paged fragment in an skb
1576 * @skb: buffer containing fragment to be initialised
1577 * @i: paged fragment index to initialise
1578 * @page: the page to use for this fragment
1579 * @off: the offset to the data with @page
1580 * @size: the length of the data
1582 * Initialises the @i'th fragment of @skb to point to &size bytes at
1583 * offset @off within @page.
1585 * Does not take any additional reference on the fragment.
1587 static inline void __skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1588 struct page
*page
, int off
, int size
)
1590 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
1593 * Propagate page->pfmemalloc to the skb if we can. The problem is
1594 * that not all callers have unique ownership of the page. If
1595 * pfmemalloc is set, we check the mapping as a mapping implies
1596 * page->index is set (index and pfmemalloc share space).
1597 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1598 * do not lose pfmemalloc information as the pages would not be
1599 * allocated using __GFP_MEMALLOC.
1601 frag
->page
.p
= page
;
1602 frag
->page_offset
= off
;
1603 skb_frag_size_set(frag
, size
);
1605 page
= compound_head(page
);
1606 if (page
->pfmemalloc
&& !page
->mapping
)
1607 skb
->pfmemalloc
= true;
1611 * skb_fill_page_desc - initialise a paged fragment in an skb
1612 * @skb: buffer containing fragment to be initialised
1613 * @i: paged fragment index to initialise
1614 * @page: the page to use for this fragment
1615 * @off: the offset to the data with @page
1616 * @size: the length of the data
1618 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1619 * @skb to point to @size bytes at offset @off within @page. In
1620 * addition updates @skb such that @i is the last fragment.
1622 * Does not take any additional reference on the fragment.
1624 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
1625 struct page
*page
, int off
, int size
)
1627 __skb_fill_page_desc(skb
, i
, page
, off
, size
);
1628 skb_shinfo(skb
)->nr_frags
= i
+ 1;
1631 void skb_add_rx_frag(struct sk_buff
*skb
, int i
, struct page
*page
, int off
,
1632 int size
, unsigned int truesize
);
1634 void skb_coalesce_rx_frag(struct sk_buff
*skb
, int i
, int size
,
1635 unsigned int truesize
);
1637 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1638 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1639 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1641 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1642 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1644 return skb
->head
+ skb
->tail
;
1647 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1649 skb
->tail
= skb
->data
- skb
->head
;
1652 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1654 skb_reset_tail_pointer(skb
);
1655 skb
->tail
+= offset
;
1658 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1659 static inline unsigned char *skb_tail_pointer(const struct sk_buff
*skb
)
1664 static inline void skb_reset_tail_pointer(struct sk_buff
*skb
)
1666 skb
->tail
= skb
->data
;
1669 static inline void skb_set_tail_pointer(struct sk_buff
*skb
, const int offset
)
1671 skb
->tail
= skb
->data
+ offset
;
1674 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1677 * Add data to an sk_buff
1679 unsigned char *pskb_put(struct sk_buff
*skb
, struct sk_buff
*tail
, int len
);
1680 unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
);
1681 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
1683 unsigned char *tmp
= skb_tail_pointer(skb
);
1684 SKB_LINEAR_ASSERT(skb
);
1690 unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
);
1691 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
1698 unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
);
1699 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
1702 BUG_ON(skb
->len
< skb
->data_len
);
1703 return skb
->data
+= len
;
1706 static inline unsigned char *skb_pull_inline(struct sk_buff
*skb
, unsigned int len
)
1708 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
1711 unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
1713 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1715 if (len
> skb_headlen(skb
) &&
1716 !__pskb_pull_tail(skb
, len
- skb_headlen(skb
)))
1719 return skb
->data
+= len
;
1722 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
1724 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
1727 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
1729 if (likely(len
<= skb_headlen(skb
)))
1731 if (unlikely(len
> skb
->len
))
1733 return __pskb_pull_tail(skb
, len
- skb_headlen(skb
)) != NULL
;
1737 * skb_headroom - bytes at buffer head
1738 * @skb: buffer to check
1740 * Return the number of bytes of free space at the head of an &sk_buff.
1742 static inline unsigned int skb_headroom(const struct sk_buff
*skb
)
1744 return skb
->data
- skb
->head
;
1748 * skb_tailroom - bytes at buffer end
1749 * @skb: buffer to check
1751 * Return the number of bytes of free space at the tail of an sk_buff
1753 static inline int skb_tailroom(const struct sk_buff
*skb
)
1755 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
1759 * skb_availroom - bytes at buffer end
1760 * @skb: buffer to check
1762 * Return the number of bytes of free space at the tail of an sk_buff
1763 * allocated by sk_stream_alloc()
1765 static inline int skb_availroom(const struct sk_buff
*skb
)
1767 if (skb_is_nonlinear(skb
))
1770 return skb
->end
- skb
->tail
- skb
->reserved_tailroom
;
1774 * skb_reserve - adjust headroom
1775 * @skb: buffer to alter
1776 * @len: bytes to move
1778 * Increase the headroom of an empty &sk_buff by reducing the tail
1779 * room. This is only allowed for an empty buffer.
1781 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
1787 #define ENCAP_TYPE_ETHER 0
1788 #define ENCAP_TYPE_IPPROTO 1
1790 static inline void skb_set_inner_protocol(struct sk_buff
*skb
,
1793 skb
->inner_protocol
= protocol
;
1794 skb
->inner_protocol_type
= ENCAP_TYPE_ETHER
;
1797 static inline void skb_set_inner_ipproto(struct sk_buff
*skb
,
1800 skb
->inner_ipproto
= ipproto
;
1801 skb
->inner_protocol_type
= ENCAP_TYPE_IPPROTO
;
1804 static inline void skb_reset_inner_headers(struct sk_buff
*skb
)
1806 skb
->inner_mac_header
= skb
->mac_header
;
1807 skb
->inner_network_header
= skb
->network_header
;
1808 skb
->inner_transport_header
= skb
->transport_header
;
1811 static inline void skb_reset_mac_len(struct sk_buff
*skb
)
1813 skb
->mac_len
= skb
->network_header
- skb
->mac_header
;
1816 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1819 return skb
->head
+ skb
->inner_transport_header
;
1822 static inline void skb_reset_inner_transport_header(struct sk_buff
*skb
)
1824 skb
->inner_transport_header
= skb
->data
- skb
->head
;
1827 static inline void skb_set_inner_transport_header(struct sk_buff
*skb
,
1830 skb_reset_inner_transport_header(skb
);
1831 skb
->inner_transport_header
+= offset
;
1834 static inline unsigned char *skb_inner_network_header(const struct sk_buff
*skb
)
1836 return skb
->head
+ skb
->inner_network_header
;
1839 static inline void skb_reset_inner_network_header(struct sk_buff
*skb
)
1841 skb
->inner_network_header
= skb
->data
- skb
->head
;
1844 static inline void skb_set_inner_network_header(struct sk_buff
*skb
,
1847 skb_reset_inner_network_header(skb
);
1848 skb
->inner_network_header
+= offset
;
1851 static inline unsigned char *skb_inner_mac_header(const struct sk_buff
*skb
)
1853 return skb
->head
+ skb
->inner_mac_header
;
1856 static inline void skb_reset_inner_mac_header(struct sk_buff
*skb
)
1858 skb
->inner_mac_header
= skb
->data
- skb
->head
;
1861 static inline void skb_set_inner_mac_header(struct sk_buff
*skb
,
1864 skb_reset_inner_mac_header(skb
);
1865 skb
->inner_mac_header
+= offset
;
1867 static inline bool skb_transport_header_was_set(const struct sk_buff
*skb
)
1869 return skb
->transport_header
!= (typeof(skb
->transport_header
))~0U;
1872 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
1874 return skb
->head
+ skb
->transport_header
;
1877 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
1879 skb
->transport_header
= skb
->data
- skb
->head
;
1882 static inline void skb_set_transport_header(struct sk_buff
*skb
,
1885 skb_reset_transport_header(skb
);
1886 skb
->transport_header
+= offset
;
1889 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
1891 return skb
->head
+ skb
->network_header
;
1894 static inline void skb_reset_network_header(struct sk_buff
*skb
)
1896 skb
->network_header
= skb
->data
- skb
->head
;
1899 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
1901 skb_reset_network_header(skb
);
1902 skb
->network_header
+= offset
;
1905 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
1907 return skb
->head
+ skb
->mac_header
;
1910 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
1912 return skb
->mac_header
!= (typeof(skb
->mac_header
))~0U;
1915 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1917 skb
->mac_header
= skb
->data
- skb
->head
;
1920 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1922 skb_reset_mac_header(skb
);
1923 skb
->mac_header
+= offset
;
1926 static inline void skb_pop_mac_header(struct sk_buff
*skb
)
1928 skb
->mac_header
= skb
->network_header
;
1931 static inline void skb_probe_transport_header(struct sk_buff
*skb
,
1932 const int offset_hint
)
1934 struct flow_keys keys
;
1936 if (skb_transport_header_was_set(skb
))
1938 else if (skb_flow_dissect(skb
, &keys
))
1939 skb_set_transport_header(skb
, keys
.thoff
);
1941 skb_set_transport_header(skb
, offset_hint
);
1944 static inline void skb_mac_header_rebuild(struct sk_buff
*skb
)
1946 if (skb_mac_header_was_set(skb
)) {
1947 const unsigned char *old_mac
= skb_mac_header(skb
);
1949 skb_set_mac_header(skb
, -skb
->mac_len
);
1950 memmove(skb_mac_header(skb
), old_mac
, skb
->mac_len
);
1954 static inline int skb_checksum_start_offset(const struct sk_buff
*skb
)
1956 return skb
->csum_start
- skb_headroom(skb
);
1959 static inline int skb_transport_offset(const struct sk_buff
*skb
)
1961 return skb_transport_header(skb
) - skb
->data
;
1964 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
1966 return skb
->transport_header
- skb
->network_header
;
1969 static inline u32
skb_inner_network_header_len(const struct sk_buff
*skb
)
1971 return skb
->inner_transport_header
- skb
->inner_network_header
;
1974 static inline int skb_network_offset(const struct sk_buff
*skb
)
1976 return skb_network_header(skb
) - skb
->data
;
1979 static inline int skb_inner_network_offset(const struct sk_buff
*skb
)
1981 return skb_inner_network_header(skb
) - skb
->data
;
1984 static inline int pskb_network_may_pull(struct sk_buff
*skb
, unsigned int len
)
1986 return pskb_may_pull(skb
, skb_network_offset(skb
) + len
);
1990 * CPUs often take a performance hit when accessing unaligned memory
1991 * locations. The actual performance hit varies, it can be small if the
1992 * hardware handles it or large if we have to take an exception and fix it
1995 * Since an ethernet header is 14 bytes network drivers often end up with
1996 * the IP header at an unaligned offset. The IP header can be aligned by
1997 * shifting the start of the packet by 2 bytes. Drivers should do this
2000 * skb_reserve(skb, NET_IP_ALIGN);
2002 * The downside to this alignment of the IP header is that the DMA is now
2003 * unaligned. On some architectures the cost of an unaligned DMA is high
2004 * and this cost outweighs the gains made by aligning the IP header.
2006 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2009 #ifndef NET_IP_ALIGN
2010 #define NET_IP_ALIGN 2
2014 * The networking layer reserves some headroom in skb data (via
2015 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2016 * the header has to grow. In the default case, if the header has to grow
2017 * 32 bytes or less we avoid the reallocation.
2019 * Unfortunately this headroom changes the DMA alignment of the resulting
2020 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2021 * on some architectures. An architecture can override this value,
2022 * perhaps setting it to a cacheline in size (since that will maintain
2023 * cacheline alignment of the DMA). It must be a power of 2.
2025 * Various parts of the networking layer expect at least 32 bytes of
2026 * headroom, you should not reduce this.
2028 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2029 * to reduce average number of cache lines per packet.
2030 * get_rps_cpus() for example only access one 64 bytes aligned block :
2031 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2034 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2037 int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
2039 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
2041 if (unlikely(skb_is_nonlinear(skb
))) {
2046 skb_set_tail_pointer(skb
, len
);
2049 void skb_trim(struct sk_buff
*skb
, unsigned int len
);
2051 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2054 return ___pskb_trim(skb
, len
);
2055 __skb_trim(skb
, len
);
2059 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
2061 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
2065 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2066 * @skb: buffer to alter
2069 * This is identical to pskb_trim except that the caller knows that
2070 * the skb is not cloned so we should never get an error due to out-
2073 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
2075 int err
= pskb_trim(skb
, len
);
2080 * skb_orphan - orphan a buffer
2081 * @skb: buffer to orphan
2083 * If a buffer currently has an owner then we call the owner's
2084 * destructor function and make the @skb unowned. The buffer continues
2085 * to exist but is no longer charged to its former owner.
2087 static inline void skb_orphan(struct sk_buff
*skb
)
2089 if (skb
->destructor
) {
2090 skb
->destructor(skb
);
2091 skb
->destructor
= NULL
;
2099 * skb_orphan_frags - orphan the frags contained in a buffer
2100 * @skb: buffer to orphan frags from
2101 * @gfp_mask: allocation mask for replacement pages
2103 * For each frag in the SKB which needs a destructor (i.e. has an
2104 * owner) create a copy of that frag and release the original
2105 * page by calling the destructor.
2107 static inline int skb_orphan_frags(struct sk_buff
*skb
, gfp_t gfp_mask
)
2109 if (likely(!(skb_shinfo(skb
)->tx_flags
& SKBTX_DEV_ZEROCOPY
)))
2111 return skb_copy_ubufs(skb
, gfp_mask
);
2115 * __skb_queue_purge - empty a list
2116 * @list: list to empty
2118 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2119 * the list and one reference dropped. This function does not take the
2120 * list lock and the caller must hold the relevant locks to use it.
2122 void skb_queue_purge(struct sk_buff_head
*list
);
2123 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
2125 struct sk_buff
*skb
;
2126 while ((skb
= __skb_dequeue(list
)) != NULL
)
2130 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2131 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2132 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2134 void *netdev_alloc_frag(unsigned int fragsz
);
2136 struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
, unsigned int length
,
2140 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2141 * @dev: network device to receive on
2142 * @length: length to allocate
2144 * Allocate a new &sk_buff and assign it a usage count of one. The
2145 * buffer has unspecified headroom built in. Users should allocate
2146 * the headroom they think they need without accounting for the
2147 * built in space. The built in space is used for optimisations.
2149 * %NULL is returned if there is no free memory. Although this function
2150 * allocates memory it can be called from an interrupt.
2152 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
2153 unsigned int length
)
2155 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
2158 /* legacy helper around __netdev_alloc_skb() */
2159 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
2162 return __netdev_alloc_skb(NULL
, length
, gfp_mask
);
2165 /* legacy helper around netdev_alloc_skb() */
2166 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
2168 return netdev_alloc_skb(NULL
, length
);
2172 static inline struct sk_buff
*__netdev_alloc_skb_ip_align(struct net_device
*dev
,
2173 unsigned int length
, gfp_t gfp
)
2175 struct sk_buff
*skb
= __netdev_alloc_skb(dev
, length
+ NET_IP_ALIGN
, gfp
);
2177 if (NET_IP_ALIGN
&& skb
)
2178 skb_reserve(skb
, NET_IP_ALIGN
);
2182 static inline struct sk_buff
*netdev_alloc_skb_ip_align(struct net_device
*dev
,
2183 unsigned int length
)
2185 return __netdev_alloc_skb_ip_align(dev
, length
, GFP_ATOMIC
);
2188 void *napi_alloc_frag(unsigned int fragsz
);
2189 struct sk_buff
*__napi_alloc_skb(struct napi_struct
*napi
,
2190 unsigned int length
, gfp_t gfp_mask
);
2191 static inline struct sk_buff
*napi_alloc_skb(struct napi_struct
*napi
,
2192 unsigned int length
)
2194 return __napi_alloc_skb(napi
, length
, GFP_ATOMIC
);
2198 * __dev_alloc_pages - allocate page for network Rx
2199 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2200 * @order: size of the allocation
2202 * Allocate a new page.
2204 * %NULL is returned if there is no free memory.
2206 static inline struct page
*__dev_alloc_pages(gfp_t gfp_mask
,
2209 /* This piece of code contains several assumptions.
2210 * 1. This is for device Rx, therefor a cold page is preferred.
2211 * 2. The expectation is the user wants a compound page.
2212 * 3. If requesting a order 0 page it will not be compound
2213 * due to the check to see if order has a value in prep_new_page
2214 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2215 * code in gfp_to_alloc_flags that should be enforcing this.
2217 gfp_mask
|= __GFP_COLD
| __GFP_COMP
| __GFP_MEMALLOC
;
2219 return alloc_pages_node(NUMA_NO_NODE
, gfp_mask
, order
);
2222 static inline struct page
*dev_alloc_pages(unsigned int order
)
2224 return __dev_alloc_pages(GFP_ATOMIC
, order
);
2228 * __dev_alloc_page - allocate a page for network Rx
2229 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2231 * Allocate a new page.
2233 * %NULL is returned if there is no free memory.
2235 static inline struct page
*__dev_alloc_page(gfp_t gfp_mask
)
2237 return __dev_alloc_pages(gfp_mask
, 0);
2240 static inline struct page
*dev_alloc_page(void)
2242 return __dev_alloc_page(GFP_ATOMIC
);
2246 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2247 * @page: The page that was allocated from skb_alloc_page
2248 * @skb: The skb that may need pfmemalloc set
2250 static inline void skb_propagate_pfmemalloc(struct page
*page
,
2251 struct sk_buff
*skb
)
2253 if (page
&& page
->pfmemalloc
)
2254 skb
->pfmemalloc
= true;
2258 * skb_frag_page - retrieve the page referred to by a paged fragment
2259 * @frag: the paged fragment
2261 * Returns the &struct page associated with @frag.
2263 static inline struct page
*skb_frag_page(const skb_frag_t
*frag
)
2265 return frag
->page
.p
;
2269 * __skb_frag_ref - take an addition reference on a paged fragment.
2270 * @frag: the paged fragment
2272 * Takes an additional reference on the paged fragment @frag.
2274 static inline void __skb_frag_ref(skb_frag_t
*frag
)
2276 get_page(skb_frag_page(frag
));
2280 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2282 * @f: the fragment offset.
2284 * Takes an additional reference on the @f'th paged fragment of @skb.
2286 static inline void skb_frag_ref(struct sk_buff
*skb
, int f
)
2288 __skb_frag_ref(&skb_shinfo(skb
)->frags
[f
]);
2292 * __skb_frag_unref - release a reference on a paged fragment.
2293 * @frag: the paged fragment
2295 * Releases a reference on the paged fragment @frag.
2297 static inline void __skb_frag_unref(skb_frag_t
*frag
)
2299 put_page(skb_frag_page(frag
));
2303 * skb_frag_unref - release a reference on a paged fragment of an skb.
2305 * @f: the fragment offset
2307 * Releases a reference on the @f'th paged fragment of @skb.
2309 static inline void skb_frag_unref(struct sk_buff
*skb
, int f
)
2311 __skb_frag_unref(&skb_shinfo(skb
)->frags
[f
]);
2315 * skb_frag_address - gets the address of the data contained in a paged fragment
2316 * @frag: the paged fragment buffer
2318 * Returns the address of the data within @frag. The page must already
2321 static inline void *skb_frag_address(const skb_frag_t
*frag
)
2323 return page_address(skb_frag_page(frag
)) + frag
->page_offset
;
2327 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2328 * @frag: the paged fragment buffer
2330 * Returns the address of the data within @frag. Checks that the page
2331 * is mapped and returns %NULL otherwise.
2333 static inline void *skb_frag_address_safe(const skb_frag_t
*frag
)
2335 void *ptr
= page_address(skb_frag_page(frag
));
2339 return ptr
+ frag
->page_offset
;
2343 * __skb_frag_set_page - sets the page contained in a paged fragment
2344 * @frag: the paged fragment
2345 * @page: the page to set
2347 * Sets the fragment @frag to contain @page.
2349 static inline void __skb_frag_set_page(skb_frag_t
*frag
, struct page
*page
)
2351 frag
->page
.p
= page
;
2355 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2357 * @f: the fragment offset
2358 * @page: the page to set
2360 * Sets the @f'th fragment of @skb to contain @page.
2362 static inline void skb_frag_set_page(struct sk_buff
*skb
, int f
,
2365 __skb_frag_set_page(&skb_shinfo(skb
)->frags
[f
], page
);
2368 bool skb_page_frag_refill(unsigned int sz
, struct page_frag
*pfrag
, gfp_t prio
);
2371 * skb_frag_dma_map - maps a paged fragment via the DMA API
2372 * @dev: the device to map the fragment to
2373 * @frag: the paged fragment to map
2374 * @offset: the offset within the fragment (starting at the
2375 * fragment's own offset)
2376 * @size: the number of bytes to map
2377 * @dir: the direction of the mapping (%PCI_DMA_*)
2379 * Maps the page associated with @frag to @device.
2381 static inline dma_addr_t
skb_frag_dma_map(struct device
*dev
,
2382 const skb_frag_t
*frag
,
2383 size_t offset
, size_t size
,
2384 enum dma_data_direction dir
)
2386 return dma_map_page(dev
, skb_frag_page(frag
),
2387 frag
->page_offset
+ offset
, size
, dir
);
2390 static inline struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
2393 return __pskb_copy(skb
, skb_headroom(skb
), gfp_mask
);
2397 static inline struct sk_buff
*pskb_copy_for_clone(struct sk_buff
*skb
,
2400 return __pskb_copy_fclone(skb
, skb_headroom(skb
), gfp_mask
, true);
2405 * skb_clone_writable - is the header of a clone writable
2406 * @skb: buffer to check
2407 * @len: length up to which to write
2409 * Returns true if modifying the header part of the cloned buffer
2410 * does not requires the data to be copied.
2412 static inline int skb_clone_writable(const struct sk_buff
*skb
, unsigned int len
)
2414 return !skb_header_cloned(skb
) &&
2415 skb_headroom(skb
) + len
<= skb
->hdr_len
;
2418 static inline int __skb_cow(struct sk_buff
*skb
, unsigned int headroom
,
2423 if (headroom
> skb_headroom(skb
))
2424 delta
= headroom
- skb_headroom(skb
);
2426 if (delta
|| cloned
)
2427 return pskb_expand_head(skb
, ALIGN(delta
, NET_SKB_PAD
), 0,
2433 * skb_cow - copy header of skb when it is required
2434 * @skb: buffer to cow
2435 * @headroom: needed headroom
2437 * If the skb passed lacks sufficient headroom or its data part
2438 * is shared, data is reallocated. If reallocation fails, an error
2439 * is returned and original skb is not changed.
2441 * The result is skb with writable area skb->head...skb->tail
2442 * and at least @headroom of space at head.
2444 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
2446 return __skb_cow(skb
, headroom
, skb_cloned(skb
));
2450 * skb_cow_head - skb_cow but only making the head writable
2451 * @skb: buffer to cow
2452 * @headroom: needed headroom
2454 * This function is identical to skb_cow except that we replace the
2455 * skb_cloned check by skb_header_cloned. It should be used when
2456 * you only need to push on some header and do not need to modify
2459 static inline int skb_cow_head(struct sk_buff
*skb
, unsigned int headroom
)
2461 return __skb_cow(skb
, headroom
, skb_header_cloned(skb
));
2465 * skb_padto - pad an skbuff up to a minimal size
2466 * @skb: buffer to pad
2467 * @len: minimal length
2469 * Pads up a buffer to ensure the trailing bytes exist and are
2470 * blanked. If the buffer already contains sufficient data it
2471 * is untouched. Otherwise it is extended. Returns zero on
2472 * success. The skb is freed on error.
2474 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
2476 unsigned int size
= skb
->len
;
2477 if (likely(size
>= len
))
2479 return skb_pad(skb
, len
- size
);
2483 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2484 * @skb: buffer to pad
2485 * @len: minimal length
2487 * Pads up a buffer to ensure the trailing bytes exist and are
2488 * blanked. If the buffer already contains sufficient data it
2489 * is untouched. Otherwise it is extended. Returns zero on
2490 * success. The skb is freed on error.
2492 static inline int skb_put_padto(struct sk_buff
*skb
, unsigned int len
)
2494 unsigned int size
= skb
->len
;
2496 if (unlikely(size
< len
)) {
2498 if (skb_pad(skb
, len
))
2500 __skb_put(skb
, len
);
2505 static inline int skb_add_data(struct sk_buff
*skb
,
2506 struct iov_iter
*from
, int copy
)
2508 const int off
= skb
->len
;
2510 if (skb
->ip_summed
== CHECKSUM_NONE
) {
2512 if (csum_and_copy_from_iter(skb_put(skb
, copy
), copy
,
2513 &csum
, from
) == copy
) {
2514 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
2517 } else if (copy_from_iter(skb_put(skb
, copy
), copy
, from
) == copy
)
2520 __skb_trim(skb
, off
);
2524 static inline bool skb_can_coalesce(struct sk_buff
*skb
, int i
,
2525 const struct page
*page
, int off
)
2528 const struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
2530 return page
== skb_frag_page(frag
) &&
2531 off
== frag
->page_offset
+ skb_frag_size(frag
);
2536 static inline int __skb_linearize(struct sk_buff
*skb
)
2538 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
2542 * skb_linearize - convert paged skb to linear one
2543 * @skb: buffer to linarize
2545 * If there is no free memory -ENOMEM is returned, otherwise zero
2546 * is returned and the old skb data released.
2548 static inline int skb_linearize(struct sk_buff
*skb
)
2550 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
2554 * skb_has_shared_frag - can any frag be overwritten
2555 * @skb: buffer to test
2557 * Return true if the skb has at least one frag that might be modified
2558 * by an external entity (as in vmsplice()/sendfile())
2560 static inline bool skb_has_shared_frag(const struct sk_buff
*skb
)
2562 return skb_is_nonlinear(skb
) &&
2563 skb_shinfo(skb
)->tx_flags
& SKBTX_SHARED_FRAG
;
2567 * skb_linearize_cow - make sure skb is linear and writable
2568 * @skb: buffer to process
2570 * If there is no free memory -ENOMEM is returned, otherwise zero
2571 * is returned and the old skb data released.
2573 static inline int skb_linearize_cow(struct sk_buff
*skb
)
2575 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
2576 __skb_linearize(skb
) : 0;
2580 * skb_postpull_rcsum - update checksum for received skb after pull
2581 * @skb: buffer to update
2582 * @start: start of data before pull
2583 * @len: length of data pulled
2585 * After doing a pull on a received packet, you need to call this to
2586 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2587 * CHECKSUM_NONE so that it can be recomputed from scratch.
2590 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
2591 const void *start
, unsigned int len
)
2593 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2594 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
2597 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
2600 * pskb_trim_rcsum - trim received skb and update checksum
2601 * @skb: buffer to trim
2604 * This is exactly the same as pskb_trim except that it ensures the
2605 * checksum of received packets are still valid after the operation.
2608 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
2610 if (likely(len
>= skb
->len
))
2612 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
2613 skb
->ip_summed
= CHECKSUM_NONE
;
2614 return __pskb_trim(skb
, len
);
2617 #define skb_queue_walk(queue, skb) \
2618 for (skb = (queue)->next; \
2619 skb != (struct sk_buff *)(queue); \
2622 #define skb_queue_walk_safe(queue, skb, tmp) \
2623 for (skb = (queue)->next, tmp = skb->next; \
2624 skb != (struct sk_buff *)(queue); \
2625 skb = tmp, tmp = skb->next)
2627 #define skb_queue_walk_from(queue, skb) \
2628 for (; skb != (struct sk_buff *)(queue); \
2631 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2632 for (tmp = skb->next; \
2633 skb != (struct sk_buff *)(queue); \
2634 skb = tmp, tmp = skb->next)
2636 #define skb_queue_reverse_walk(queue, skb) \
2637 for (skb = (queue)->prev; \
2638 skb != (struct sk_buff *)(queue); \
2641 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2642 for (skb = (queue)->prev, tmp = skb->prev; \
2643 skb != (struct sk_buff *)(queue); \
2644 skb = tmp, tmp = skb->prev)
2646 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2647 for (tmp = skb->prev; \
2648 skb != (struct sk_buff *)(queue); \
2649 skb = tmp, tmp = skb->prev)
2651 static inline bool skb_has_frag_list(const struct sk_buff
*skb
)
2653 return skb_shinfo(skb
)->frag_list
!= NULL
;
2656 static inline void skb_frag_list_init(struct sk_buff
*skb
)
2658 skb_shinfo(skb
)->frag_list
= NULL
;
2661 static inline void skb_frag_add_head(struct sk_buff
*skb
, struct sk_buff
*frag
)
2663 frag
->next
= skb_shinfo(skb
)->frag_list
;
2664 skb_shinfo(skb
)->frag_list
= frag
;
2667 #define skb_walk_frags(skb, iter) \
2668 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2670 struct sk_buff
*__skb_recv_datagram(struct sock
*sk
, unsigned flags
,
2671 int *peeked
, int *off
, int *err
);
2672 struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
, int noblock
,
2674 unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
2675 struct poll_table_struct
*wait
);
2676 int skb_copy_datagram_iter(const struct sk_buff
*from
, int offset
,
2677 struct iov_iter
*to
, int size
);
2678 static inline int skb_copy_datagram_msg(const struct sk_buff
*from
, int offset
,
2679 struct msghdr
*msg
, int size
)
2681 return skb_copy_datagram_iter(from
, offset
, &msg
->msg_iter
, size
);
2683 int skb_copy_and_csum_datagram_msg(struct sk_buff
*skb
, int hlen
,
2684 struct msghdr
*msg
);
2685 int skb_copy_datagram_from_iter(struct sk_buff
*skb
, int offset
,
2686 struct iov_iter
*from
, int len
);
2687 int zerocopy_sg_from_iter(struct sk_buff
*skb
, struct iov_iter
*frm
);
2688 void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
2689 void skb_free_datagram_locked(struct sock
*sk
, struct sk_buff
*skb
);
2690 int skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
, unsigned int flags
);
2691 int skb_copy_bits(const struct sk_buff
*skb
, int offset
, void *to
, int len
);
2692 int skb_store_bits(struct sk_buff
*skb
, int offset
, const void *from
, int len
);
2693 __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
, int offset
, u8
*to
,
2694 int len
, __wsum csum
);
2695 int skb_splice_bits(struct sk_buff
*skb
, unsigned int offset
,
2696 struct pipe_inode_info
*pipe
, unsigned int len
,
2697 unsigned int flags
);
2698 void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
2699 unsigned int skb_zerocopy_headlen(const struct sk_buff
*from
);
2700 int skb_zerocopy(struct sk_buff
*to
, struct sk_buff
*from
,
2702 void skb_split(struct sk_buff
*skb
, struct sk_buff
*skb1
, const u32 len
);
2703 int skb_shift(struct sk_buff
*tgt
, struct sk_buff
*skb
, int shiftlen
);
2704 void skb_scrub_packet(struct sk_buff
*skb
, bool xnet
);
2705 unsigned int skb_gso_transport_seglen(const struct sk_buff
*skb
);
2706 struct sk_buff
*skb_segment(struct sk_buff
*skb
, netdev_features_t features
);
2707 struct sk_buff
*skb_vlan_untag(struct sk_buff
*skb
);
2708 int skb_ensure_writable(struct sk_buff
*skb
, int write_len
);
2709 int skb_vlan_pop(struct sk_buff
*skb
);
2710 int skb_vlan_push(struct sk_buff
*skb
, __be16 vlan_proto
, u16 vlan_tci
);
2712 static inline int memcpy_from_msg(void *data
, struct msghdr
*msg
, int len
)
2714 return copy_from_iter(data
, len
, &msg
->msg_iter
) == len
? 0 : -EFAULT
;
2717 static inline int memcpy_to_msg(struct msghdr
*msg
, void *data
, int len
)
2719 return copy_to_iter(data
, len
, &msg
->msg_iter
) == len
? 0 : -EFAULT
;
2722 struct skb_checksum_ops
{
2723 __wsum (*update
)(const void *mem
, int len
, __wsum wsum
);
2724 __wsum (*combine
)(__wsum csum
, __wsum csum2
, int offset
, int len
);
2727 __wsum
__skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
2728 __wsum csum
, const struct skb_checksum_ops
*ops
);
2729 __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
, int len
,
2732 static inline void *__skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2733 int len
, void *data
, int hlen
, void *buffer
)
2735 if (hlen
- offset
>= len
)
2736 return data
+ offset
;
2739 skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
2745 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
2746 int len
, void *buffer
)
2748 return __skb_header_pointer(skb
, offset
, len
, skb
->data
,
2749 skb_headlen(skb
), buffer
);
2753 * skb_needs_linearize - check if we need to linearize a given skb
2754 * depending on the given device features.
2755 * @skb: socket buffer to check
2756 * @features: net device features
2758 * Returns true if either:
2759 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2760 * 2. skb is fragmented and the device does not support SG.
2762 static inline bool skb_needs_linearize(struct sk_buff
*skb
,
2763 netdev_features_t features
)
2765 return skb_is_nonlinear(skb
) &&
2766 ((skb_has_frag_list(skb
) && !(features
& NETIF_F_FRAGLIST
)) ||
2767 (skb_shinfo(skb
)->nr_frags
&& !(features
& NETIF_F_SG
)));
2770 static inline void skb_copy_from_linear_data(const struct sk_buff
*skb
,
2772 const unsigned int len
)
2774 memcpy(to
, skb
->data
, len
);
2777 static inline void skb_copy_from_linear_data_offset(const struct sk_buff
*skb
,
2778 const int offset
, void *to
,
2779 const unsigned int len
)
2781 memcpy(to
, skb
->data
+ offset
, len
);
2784 static inline void skb_copy_to_linear_data(struct sk_buff
*skb
,
2786 const unsigned int len
)
2788 memcpy(skb
->data
, from
, len
);
2791 static inline void skb_copy_to_linear_data_offset(struct sk_buff
*skb
,
2794 const unsigned int len
)
2796 memcpy(skb
->data
+ offset
, from
, len
);
2799 void skb_init(void);
2801 static inline ktime_t
skb_get_ktime(const struct sk_buff
*skb
)
2807 * skb_get_timestamp - get timestamp from a skb
2808 * @skb: skb to get stamp from
2809 * @stamp: pointer to struct timeval to store stamp in
2811 * Timestamps are stored in the skb as offsets to a base timestamp.
2812 * This function converts the offset back to a struct timeval and stores
2815 static inline void skb_get_timestamp(const struct sk_buff
*skb
,
2816 struct timeval
*stamp
)
2818 *stamp
= ktime_to_timeval(skb
->tstamp
);
2821 static inline void skb_get_timestampns(const struct sk_buff
*skb
,
2822 struct timespec
*stamp
)
2824 *stamp
= ktime_to_timespec(skb
->tstamp
);
2827 static inline void __net_timestamp(struct sk_buff
*skb
)
2829 skb
->tstamp
= ktime_get_real();
2832 static inline ktime_t
net_timedelta(ktime_t t
)
2834 return ktime_sub(ktime_get_real(), t
);
2837 static inline ktime_t
net_invalid_timestamp(void)
2839 return ktime_set(0, 0);
2842 struct sk_buff
*skb_clone_sk(struct sk_buff
*skb
);
2844 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2846 void skb_clone_tx_timestamp(struct sk_buff
*skb
);
2847 bool skb_defer_rx_timestamp(struct sk_buff
*skb
);
2849 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2851 static inline void skb_clone_tx_timestamp(struct sk_buff
*skb
)
2855 static inline bool skb_defer_rx_timestamp(struct sk_buff
*skb
)
2860 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2863 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2865 * PHY drivers may accept clones of transmitted packets for
2866 * timestamping via their phy_driver.txtstamp method. These drivers
2867 * must call this function to return the skb back to the stack, with
2868 * or without a timestamp.
2870 * @skb: clone of the the original outgoing packet
2871 * @hwtstamps: hardware time stamps, may be NULL if not available
2874 void skb_complete_tx_timestamp(struct sk_buff
*skb
,
2875 struct skb_shared_hwtstamps
*hwtstamps
);
2877 void __skb_tstamp_tx(struct sk_buff
*orig_skb
,
2878 struct skb_shared_hwtstamps
*hwtstamps
,
2879 struct sock
*sk
, int tstype
);
2882 * skb_tstamp_tx - queue clone of skb with send time stamps
2883 * @orig_skb: the original outgoing packet
2884 * @hwtstamps: hardware time stamps, may be NULL if not available
2886 * If the skb has a socket associated, then this function clones the
2887 * skb (thus sharing the actual data and optional structures), stores
2888 * the optional hardware time stamping information (if non NULL) or
2889 * generates a software time stamp (otherwise), then queues the clone
2890 * to the error queue of the socket. Errors are silently ignored.
2892 void skb_tstamp_tx(struct sk_buff
*orig_skb
,
2893 struct skb_shared_hwtstamps
*hwtstamps
);
2895 static inline void sw_tx_timestamp(struct sk_buff
*skb
)
2897 if (skb_shinfo(skb
)->tx_flags
& SKBTX_SW_TSTAMP
&&
2898 !(skb_shinfo(skb
)->tx_flags
& SKBTX_IN_PROGRESS
))
2899 skb_tstamp_tx(skb
, NULL
);
2903 * skb_tx_timestamp() - Driver hook for transmit timestamping
2905 * Ethernet MAC Drivers should call this function in their hard_xmit()
2906 * function immediately before giving the sk_buff to the MAC hardware.
2908 * Specifically, one should make absolutely sure that this function is
2909 * called before TX completion of this packet can trigger. Otherwise
2910 * the packet could potentially already be freed.
2912 * @skb: A socket buffer.
2914 static inline void skb_tx_timestamp(struct sk_buff
*skb
)
2916 skb_clone_tx_timestamp(skb
);
2917 sw_tx_timestamp(skb
);
2921 * skb_complete_wifi_ack - deliver skb with wifi status
2923 * @skb: the original outgoing packet
2924 * @acked: ack status
2927 void skb_complete_wifi_ack(struct sk_buff
*skb
, bool acked
);
2929 __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
2930 __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
2932 static inline int skb_csum_unnecessary(const struct sk_buff
*skb
)
2934 return ((skb
->ip_summed
== CHECKSUM_UNNECESSARY
) ||
2936 (skb
->ip_summed
== CHECKSUM_PARTIAL
&&
2937 skb_checksum_start_offset(skb
) >= 0));
2941 * skb_checksum_complete - Calculate checksum of an entire packet
2942 * @skb: packet to process
2944 * This function calculates the checksum over the entire packet plus
2945 * the value of skb->csum. The latter can be used to supply the
2946 * checksum of a pseudo header as used by TCP/UDP. It returns the
2949 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2950 * this function can be used to verify that checksum on received
2951 * packets. In that case the function should return zero if the
2952 * checksum is correct. In particular, this function will return zero
2953 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2954 * hardware has already verified the correctness of the checksum.
2956 static inline __sum16
skb_checksum_complete(struct sk_buff
*skb
)
2958 return skb_csum_unnecessary(skb
) ?
2959 0 : __skb_checksum_complete(skb
);
2962 static inline void __skb_decr_checksum_unnecessary(struct sk_buff
*skb
)
2964 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
2965 if (skb
->csum_level
== 0)
2966 skb
->ip_summed
= CHECKSUM_NONE
;
2972 static inline void __skb_incr_checksum_unnecessary(struct sk_buff
*skb
)
2974 if (skb
->ip_summed
== CHECKSUM_UNNECESSARY
) {
2975 if (skb
->csum_level
< SKB_MAX_CSUM_LEVEL
)
2977 } else if (skb
->ip_summed
== CHECKSUM_NONE
) {
2978 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
2979 skb
->csum_level
= 0;
2983 static inline void __skb_mark_checksum_bad(struct sk_buff
*skb
)
2985 /* Mark current checksum as bad (typically called from GRO
2986 * path). In the case that ip_summed is CHECKSUM_NONE
2987 * this must be the first checksum encountered in the packet.
2988 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2989 * checksum after the last one validated. For UDP, a zero
2990 * checksum can not be marked as bad.
2993 if (skb
->ip_summed
== CHECKSUM_NONE
||
2994 skb
->ip_summed
== CHECKSUM_UNNECESSARY
)
2998 /* Check if we need to perform checksum complete validation.
3000 * Returns true if checksum complete is needed, false otherwise
3001 * (either checksum is unnecessary or zero checksum is allowed).
3003 static inline bool __skb_checksum_validate_needed(struct sk_buff
*skb
,
3007 if (skb_csum_unnecessary(skb
) || (zero_okay
&& !check
)) {
3008 skb
->csum_valid
= 1;
3009 __skb_decr_checksum_unnecessary(skb
);
3016 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3019 #define CHECKSUM_BREAK 76
3021 /* Unset checksum-complete
3023 * Unset checksum complete can be done when packet is being modified
3024 * (uncompressed for instance) and checksum-complete value is
3027 static inline void skb_checksum_complete_unset(struct sk_buff
*skb
)
3029 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3030 skb
->ip_summed
= CHECKSUM_NONE
;
3033 /* Validate (init) checksum based on checksum complete.
3036 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3037 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3038 * checksum is stored in skb->csum for use in __skb_checksum_complete
3039 * non-zero: value of invalid checksum
3042 static inline __sum16
__skb_checksum_validate_complete(struct sk_buff
*skb
,
3046 if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
3047 if (!csum_fold(csum_add(psum
, skb
->csum
))) {
3048 skb
->csum_valid
= 1;
3051 } else if (skb
->csum_bad
) {
3052 /* ip_summed == CHECKSUM_NONE in this case */
3058 if (complete
|| skb
->len
<= CHECKSUM_BREAK
) {
3061 csum
= __skb_checksum_complete(skb
);
3062 skb
->csum_valid
= !csum
;
3069 static inline __wsum
null_compute_pseudo(struct sk_buff
*skb
, int proto
)
3074 /* Perform checksum validate (init). Note that this is a macro since we only
3075 * want to calculate the pseudo header which is an input function if necessary.
3076 * First we try to validate without any computation (checksum unnecessary) and
3077 * then calculate based on checksum complete calling the function to compute
3081 * 0: checksum is validated or try to in skb_checksum_complete
3082 * non-zero: value of invalid checksum
3084 #define __skb_checksum_validate(skb, proto, complete, \
3085 zero_okay, check, compute_pseudo) \
3087 __sum16 __ret = 0; \
3088 skb->csum_valid = 0; \
3089 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3090 __ret = __skb_checksum_validate_complete(skb, \
3091 complete, compute_pseudo(skb, proto)); \
3095 #define skb_checksum_init(skb, proto, compute_pseudo) \
3096 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3098 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3099 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3101 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3102 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3104 #define skb_checksum_validate_zero_check(skb, proto, check, \
3106 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3108 #define skb_checksum_simple_validate(skb) \
3109 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3111 static inline bool __skb_checksum_convert_check(struct sk_buff
*skb
)
3113 return (skb
->ip_summed
== CHECKSUM_NONE
&&
3114 skb
->csum_valid
&& !skb
->csum_bad
);
3117 static inline void __skb_checksum_convert(struct sk_buff
*skb
,
3118 __sum16 check
, __wsum pseudo
)
3120 skb
->csum
= ~pseudo
;
3121 skb
->ip_summed
= CHECKSUM_COMPLETE
;
3124 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3126 if (__skb_checksum_convert_check(skb)) \
3127 __skb_checksum_convert(skb, check, \
3128 compute_pseudo(skb, proto)); \
3131 static inline void skb_remcsum_adjust_partial(struct sk_buff
*skb
, void *ptr
,
3132 u16 start
, u16 offset
)
3134 skb
->ip_summed
= CHECKSUM_PARTIAL
;
3135 skb
->csum_start
= ((unsigned char *)ptr
+ start
) - skb
->head
;
3136 skb
->csum_offset
= offset
- start
;
3139 /* Update skbuf and packet to reflect the remote checksum offload operation.
3140 * When called, ptr indicates the starting point for skb->csum when
3141 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3142 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3144 static inline void skb_remcsum_process(struct sk_buff
*skb
, void *ptr
,
3145 int start
, int offset
, bool nopartial
)
3150 skb_remcsum_adjust_partial(skb
, ptr
, start
, offset
);
3154 if (unlikely(skb
->ip_summed
!= CHECKSUM_COMPLETE
)) {
3155 __skb_checksum_complete(skb
);
3156 skb_postpull_rcsum(skb
, skb
->data
, ptr
- (void *)skb
->data
);
3159 delta
= remcsum_adjust(ptr
, skb
->csum
, start
, offset
);
3161 /* Adjust skb->csum since we changed the packet */
3162 skb
->csum
= csum_add(skb
->csum
, delta
);
3165 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3166 void nf_conntrack_destroy(struct nf_conntrack
*nfct
);
3167 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
3169 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
3170 nf_conntrack_destroy(nfct
);
3172 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
3175 atomic_inc(&nfct
->use
);
3178 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3179 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
3181 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
3184 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
3187 atomic_inc(&nf_bridge
->use
);
3189 #endif /* CONFIG_BRIDGE_NETFILTER */
3190 static inline void nf_reset(struct sk_buff
*skb
)
3192 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3193 nf_conntrack_put(skb
->nfct
);
3196 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3197 nf_bridge_put(skb
->nf_bridge
);
3198 skb
->nf_bridge
= NULL
;
3202 static inline void nf_reset_trace(struct sk_buff
*skb
)
3204 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3209 /* Note: This doesn't put any conntrack and bridge info in dst. */
3210 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
,
3213 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3214 dst
->nfct
= src
->nfct
;
3215 nf_conntrack_get(src
->nfct
);
3217 dst
->nfctinfo
= src
->nfctinfo
;
3219 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3220 dst
->nf_bridge
= src
->nf_bridge
;
3221 nf_bridge_get(src
->nf_bridge
);
3223 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3225 dst
->nf_trace
= src
->nf_trace
;
3229 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
3231 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3232 nf_conntrack_put(dst
->nfct
);
3234 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3235 nf_bridge_put(dst
->nf_bridge
);
3237 __nf_copy(dst
, src
, true);
3240 #ifdef CONFIG_NETWORK_SECMARK
3241 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3243 to
->secmark
= from
->secmark
;
3246 static inline void skb_init_secmark(struct sk_buff
*skb
)
3251 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
3254 static inline void skb_init_secmark(struct sk_buff
*skb
)
3258 static inline bool skb_irq_freeable(const struct sk_buff
*skb
)
3260 return !skb
->destructor
&&
3261 #if IS_ENABLED(CONFIG_XFRM)
3264 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3267 !skb
->_skb_refdst
&&
3268 !skb_has_frag_list(skb
);
3271 static inline void skb_set_queue_mapping(struct sk_buff
*skb
, u16 queue_mapping
)
3273 skb
->queue_mapping
= queue_mapping
;
3276 static inline u16
skb_get_queue_mapping(const struct sk_buff
*skb
)
3278 return skb
->queue_mapping
;
3281 static inline void skb_copy_queue_mapping(struct sk_buff
*to
, const struct sk_buff
*from
)
3283 to
->queue_mapping
= from
->queue_mapping
;
3286 static inline void skb_record_rx_queue(struct sk_buff
*skb
, u16 rx_queue
)
3288 skb
->queue_mapping
= rx_queue
+ 1;
3291 static inline u16
skb_get_rx_queue(const struct sk_buff
*skb
)
3293 return skb
->queue_mapping
- 1;
3296 static inline bool skb_rx_queue_recorded(const struct sk_buff
*skb
)
3298 return skb
->queue_mapping
!= 0;
3301 u16
__skb_tx_hash(const struct net_device
*dev
, struct sk_buff
*skb
,
3302 unsigned int num_tx_queues
);
3304 static inline struct sec_path
*skb_sec_path(struct sk_buff
*skb
)
3313 /* Keeps track of mac header offset relative to skb->head.
3314 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3315 * For non-tunnel skb it points to skb_mac_header() and for
3316 * tunnel skb it points to outer mac header.
3317 * Keeps track of level of encapsulation of network headers.
3324 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3326 static inline int skb_tnl_header_len(const struct sk_buff
*inner_skb
)
3328 return (skb_mac_header(inner_skb
) - inner_skb
->head
) -
3329 SKB_GSO_CB(inner_skb
)->mac_offset
;
3332 static inline int gso_pskb_expand_head(struct sk_buff
*skb
, int extra
)
3334 int new_headroom
, headroom
;
3337 headroom
= skb_headroom(skb
);
3338 ret
= pskb_expand_head(skb
, extra
, 0, GFP_ATOMIC
);
3342 new_headroom
= skb_headroom(skb
);
3343 SKB_GSO_CB(skb
)->mac_offset
+= (new_headroom
- headroom
);
3347 /* Compute the checksum for a gso segment. First compute the checksum value
3348 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3349 * then add in skb->csum (checksum from csum_start to end of packet).
3350 * skb->csum and csum_start are then updated to reflect the checksum of the
3351 * resultant packet starting from the transport header-- the resultant checksum
3352 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3355 static inline __sum16
gso_make_checksum(struct sk_buff
*skb
, __wsum res
)
3357 int plen
= SKB_GSO_CB(skb
)->csum_start
- skb_headroom(skb
) -
3358 skb_transport_offset(skb
);
3361 csum
= csum_fold(csum_partial(skb_transport_header(skb
),
3364 SKB_GSO_CB(skb
)->csum_start
-= plen
;
3369 static inline bool skb_is_gso(const struct sk_buff
*skb
)
3371 return skb_shinfo(skb
)->gso_size
;
3374 /* Note: Should be called only if skb_is_gso(skb) is true */
3375 static inline bool skb_is_gso_v6(const struct sk_buff
*skb
)
3377 return skb_shinfo(skb
)->gso_type
& SKB_GSO_TCPV6
;
3380 void __skb_warn_lro_forwarding(const struct sk_buff
*skb
);
3382 static inline bool skb_warn_if_lro(const struct sk_buff
*skb
)
3384 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3385 * wanted then gso_type will be set. */
3386 const struct skb_shared_info
*shinfo
= skb_shinfo(skb
);
3388 if (skb_is_nonlinear(skb
) && shinfo
->gso_size
!= 0 &&
3389 unlikely(shinfo
->gso_type
== 0)) {
3390 __skb_warn_lro_forwarding(skb
);
3396 static inline void skb_forward_csum(struct sk_buff
*skb
)
3398 /* Unfortunately we don't support this one. Any brave souls? */
3399 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
3400 skb
->ip_summed
= CHECKSUM_NONE
;
3404 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3405 * @skb: skb to check
3407 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3408 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3409 * use this helper, to document places where we make this assertion.
3411 static inline void skb_checksum_none_assert(const struct sk_buff
*skb
)
3414 BUG_ON(skb
->ip_summed
!= CHECKSUM_NONE
);
3418 bool skb_partial_csum_set(struct sk_buff
*skb
, u16 start
, u16 off
);
3420 int skb_checksum_setup(struct sk_buff
*skb
, bool recalculate
);
3422 u32
skb_get_poff(const struct sk_buff
*skb
);
3423 u32
__skb_get_poff(const struct sk_buff
*skb
, void *data
,
3424 const struct flow_keys
*keys
, int hlen
);
3427 * skb_head_is_locked - Determine if the skb->head is locked down
3428 * @skb: skb to check
3430 * The head on skbs build around a head frag can be removed if they are
3431 * not cloned. This function returns true if the skb head is locked down
3432 * due to either being allocated via kmalloc, or by being a clone with
3433 * multiple references to the head.
3435 static inline bool skb_head_is_locked(const struct sk_buff
*skb
)
3437 return !skb
->head_frag
|| skb_cloned(skb
);
3441 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3445 * skb_gso_network_seglen is used to determine the real size of the
3446 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3448 * The MAC/L2 header is not accounted for.
3450 static inline unsigned int skb_gso_network_seglen(const struct sk_buff
*skb
)
3452 unsigned int hdr_len
= skb_transport_header(skb
) -
3453 skb_network_header(skb
);
3454 return hdr_len
+ skb_gso_transport_seglen(skb
);
3456 #endif /* __KERNEL__ */
3457 #endif /* _LINUX_SKBUFF_H */