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 * This is identical to the case for output below. This may occur on a packet
87 * received directly from another Linux OS, e.g., a virtualized Linux kernel
88 * on the same host. The packet can be treated in the same way as
89 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
90 * checksum must be filled in by the OS or the hardware.
92 * B. Checksumming on output.
96 * The skb was already checksummed by the protocol, or a checksum is not
101 * The device is required to checksum the packet as seen by hard_start_xmit()
102 * from skb->csum_start up to the end, and to record/write the checksum at
103 * offset skb->csum_start + skb->csum_offset.
105 * The device must show its capabilities in dev->features, set up at device
106 * setup time, e.g. netdev_features.h:
108 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
109 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
110 * IPv4. Sigh. Vendors like this way for an unknown reason.
111 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
112 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
113 * NETIF_F_... - Well, you get the picture.
115 * CHECKSUM_UNNECESSARY:
117 * Normally, the device will do per protocol specific checksumming. Protocol
118 * implementations that do not want the NIC to perform the checksum
119 * calculation should use this flag in their outgoing skbs.
121 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
122 * offload. Correspondingly, the FCoE protocol driver
123 * stack should use CHECKSUM_UNNECESSARY.
125 * Any questions? No questions, good. --ANK
128 /* Don't change this without changing skb_csum_unnecessary! */
129 #define CHECKSUM_NONE 0
130 #define CHECKSUM_UNNECESSARY 1
131 #define CHECKSUM_COMPLETE 2
132 #define CHECKSUM_PARTIAL 3
134 /* Maximum value in skb->csum_level */
135 #define SKB_MAX_CSUM_LEVEL 3
137 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
138 #define SKB_WITH_OVERHEAD(X) \
139 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
140 #define SKB_MAX_ORDER(X, ORDER) \
141 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
142 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
143 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
145 /* return minimum truesize of one skb containing X bytes of data */
146 #define SKB_TRUESIZE(X) ((X) + \
147 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
148 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
152 struct pipe_inode_info;
155 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
156 struct nf_conntrack {
161 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
162 struct nf_bridge_info {
165 struct net_device *physindev;
166 struct net_device *physoutdev;
167 unsigned long data[32 / sizeof(unsigned long)];
171 struct sk_buff_head {
172 /* These two members must be first. */
173 struct sk_buff *next;
174 struct sk_buff *prev;
182 /* To allow 64K frame to be packed as single skb without frag_list we
183 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
184 * buffers which do not start on a page boundary.
186 * Since GRO uses frags we allocate at least 16 regardless of page
189 #if (65536/PAGE_SIZE + 1) < 16
190 #define MAX_SKB_FRAGS 16UL
192 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
195 typedef struct skb_frag_struct skb_frag_t;
197 struct skb_frag_struct {
201 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
210 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
215 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
220 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
225 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
230 #define HAVE_HW_TIME_STAMP
233 * struct skb_shared_hwtstamps - hardware time stamps
234 * @hwtstamp: hardware time stamp transformed into duration
235 * since arbitrary point in time
237 * Software time stamps generated by ktime_get_real() are stored in
240 * hwtstamps can only be compared against other hwtstamps from
243 * This structure is attached to packets as part of the
244 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
246 struct skb_shared_hwtstamps {
250 /* Definitions for tx_flags in struct skb_shared_info */
252 /* generate hardware time stamp */
253 SKBTX_HW_TSTAMP = 1 << 0,
255 /* generate software time stamp when queueing packet to NIC */
256 SKBTX_SW_TSTAMP = 1 << 1,
258 /* device driver is going to provide hardware time stamp */
259 SKBTX_IN_PROGRESS = 1 << 2,
261 /* device driver supports TX zero-copy buffers */
262 SKBTX_DEV_ZEROCOPY = 1 << 3,
264 /* generate wifi status information (where possible) */
265 SKBTX_WIFI_STATUS = 1 << 4,
267 /* This indicates at least one fragment might be overwritten
268 * (as in vmsplice(), sendfile() ...)
269 * If we need to compute a TX checksum, we'll need to copy
270 * all frags to avoid possible bad checksum
272 SKBTX_SHARED_FRAG = 1 << 5,
274 /* generate software time stamp when entering packet scheduling */
275 SKBTX_SCHED_TSTAMP = 1 << 6,
277 /* generate software timestamp on peer data acknowledgment */
278 SKBTX_ACK_TSTAMP = 1 << 7,
281 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
282 SKBTX_SCHED_TSTAMP | \
284 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
287 * The callback notifies userspace to release buffers when skb DMA is done in
288 * lower device, the skb last reference should be 0 when calling this.
289 * The zerocopy_success argument is true if zero copy transmit occurred,
290 * false on data copy or out of memory error caused by data copy attempt.
291 * The ctx field is used to track device context.
292 * The desc field is used to track userspace buffer index.
295 void (*callback)(struct ubuf_info *, bool zerocopy_success);
300 /* This data is invariant across clones and lives at
301 * the end of the header data, ie. at skb->end.
303 struct skb_shared_info {
304 unsigned char nr_frags;
306 unsigned short gso_size;
307 /* Warning: this field is not always filled in (UFO)! */
308 unsigned short gso_segs;
309 unsigned short gso_type;
310 struct sk_buff *frag_list;
311 struct skb_shared_hwtstamps hwtstamps;
316 * Warning : all fields before dataref are cleared in __alloc_skb()
320 /* Intermediate layers must ensure that destructor_arg
321 * remains valid until skb destructor */
322 void * destructor_arg;
324 /* must be last field, see pskb_expand_head() */
325 skb_frag_t frags[MAX_SKB_FRAGS];
328 /* We divide dataref into two halves. The higher 16 bits hold references
329 * to the payload part of skb->data. The lower 16 bits hold references to
330 * the entire skb->data. A clone of a headerless skb holds the length of
331 * the header in skb->hdr_len.
333 * All users must obey the rule that the skb->data reference count must be
334 * greater than or equal to the payload reference count.
336 * Holding a reference to the payload part means that the user does not
337 * care about modifications to the header part of skb->data.
339 #define SKB_DATAREF_SHIFT 16
340 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
344 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
345 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
346 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
350 SKB_GSO_TCPV4 = 1 << 0,
351 SKB_GSO_UDP = 1 << 1,
353 /* This indicates the skb is from an untrusted source. */
354 SKB_GSO_DODGY = 1 << 2,
356 /* This indicates the tcp segment has CWR set. */
357 SKB_GSO_TCP_ECN = 1 << 3,
359 SKB_GSO_TCPV6 = 1 << 4,
361 SKB_GSO_FCOE = 1 << 5,
363 SKB_GSO_GRE = 1 << 6,
365 SKB_GSO_GRE_CSUM = 1 << 7,
367 SKB_GSO_IPIP = 1 << 8,
369 SKB_GSO_SIT = 1 << 9,
371 SKB_GSO_UDP_TUNNEL = 1 << 10,
373 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
375 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
378 #if BITS_PER_LONG > 32
379 #define NET_SKBUFF_DATA_USES_OFFSET 1
382 #ifdef NET_SKBUFF_DATA_USES_OFFSET
383 typedef unsigned int sk_buff_data_t;
385 typedef unsigned char *sk_buff_data_t;
389 * struct skb_mstamp - multi resolution time stamps
390 * @stamp_us: timestamp in us resolution
391 * @stamp_jiffies: timestamp in jiffies
404 * skb_mstamp_get - get current timestamp
405 * @cl: place to store timestamps
407 static inline void skb_mstamp_get(struct skb_mstamp *cl)
409 u64 val = local_clock();
411 do_div(val, NSEC_PER_USEC);
412 cl->stamp_us = (u32)val;
413 cl->stamp_jiffies = (u32)jiffies;
417 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
418 * @t1: pointer to newest sample
419 * @t0: pointer to oldest sample
421 static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
422 const struct skb_mstamp *t0)
424 s32 delta_us = t1->stamp_us - t0->stamp_us;
425 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
427 /* If delta_us is negative, this might be because interval is too big,
428 * or local_clock() drift is too big : fallback using jiffies.
431 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
433 delta_us = jiffies_to_usecs(delta_jiffies);
440 * struct sk_buff - socket buffer
441 * @next: Next buffer in list
442 * @prev: Previous buffer in list
443 * @tstamp: Time we arrived/left
444 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
445 * @sk: Socket we are owned by
446 * @dev: Device we arrived on/are leaving by
447 * @cb: Control buffer. Free for use by every layer. Put private vars here
448 * @_skb_refdst: destination entry (with norefcount bit)
449 * @sp: the security path, used for xfrm
450 * @len: Length of actual data
451 * @data_len: Data length
452 * @mac_len: Length of link layer header
453 * @hdr_len: writable header length of cloned skb
454 * @csum: Checksum (must include start/offset pair)
455 * @csum_start: Offset from skb->head where checksumming should start
456 * @csum_offset: Offset from csum_start where checksum should be stored
457 * @priority: Packet queueing priority
458 * @ignore_df: allow local fragmentation
459 * @cloned: Head may be cloned (check refcnt to be sure)
460 * @ip_summed: Driver fed us an IP checksum
461 * @nohdr: Payload reference only, must not modify header
462 * @nfctinfo: Relationship of this skb to the connection
463 * @pkt_type: Packet class
464 * @fclone: skbuff clone status
465 * @ipvs_property: skbuff is owned by ipvs
466 * @peeked: this packet has been seen already, so stats have been
467 * done for it, don't do them again
468 * @nf_trace: netfilter packet trace flag
469 * @protocol: Packet protocol from driver
470 * @destructor: Destruct function
471 * @nfct: Associated connection, if any
472 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
473 * @skb_iif: ifindex of device we arrived on
474 * @tc_index: Traffic control index
475 * @tc_verd: traffic control verdict
476 * @hash: the packet hash
477 * @queue_mapping: Queue mapping for multiqueue devices
478 * @xmit_more: More SKBs are pending for this queue
479 * @ndisc_nodetype: router type (from link layer)
480 * @ooo_okay: allow the mapping of a socket to a queue to be changed
481 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
483 * @sw_hash: indicates hash was computed in software stack
484 * @wifi_acked_valid: wifi_acked was set
485 * @wifi_acked: whether frame was acked on wifi or not
486 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
487 * @napi_id: id of the NAPI struct this skb came from
488 * @secmark: security marking
489 * @mark: Generic packet mark
490 * @dropcount: total number of sk_receive_queue overflows
491 * @vlan_proto: vlan encapsulation protocol
492 * @vlan_tci: vlan tag control information
493 * @inner_protocol: Protocol (encapsulation)
494 * @inner_transport_header: Inner transport layer header (encapsulation)
495 * @inner_network_header: Network layer header (encapsulation)
496 * @inner_mac_header: Link layer header (encapsulation)
497 * @transport_header: Transport layer header
498 * @network_header: Network layer header
499 * @mac_header: Link layer header
500 * @tail: Tail pointer
502 * @head: Head of buffer
503 * @data: Data head pointer
504 * @truesize: Buffer size
505 * @users: User count - see {datagram,tcp}.c
511 /* These two members must be first. */
512 struct sk_buff *next;
513 struct sk_buff *prev;
517 struct skb_mstamp skb_mstamp;
520 struct rb_node rbnode; /* used in netem & tcp stack */
523 struct net_device *dev;
526 * This is the control buffer. It is free to use for every
527 * layer. Please put your private variables there. If you
528 * want to keep them across layers you have to do a skb_clone()
529 * first. This is owned by whoever has the skb queued ATM.
531 char cb[48] __aligned(8);
533 unsigned long _skb_refdst;
534 void (*destructor)(struct sk_buff *skb);
538 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
539 struct nf_conntrack *nfct;
541 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
542 struct nf_bridge_info *nf_bridge;
549 /* Following fields are _not_ copied in __copy_skb_header()
550 * Note that queue_mapping is here mostly to fill a hole.
552 kmemcheck_bitfield_begin(flags1);
561 kmemcheck_bitfield_end(flags1);
563 /* fields enclosed in headers_start/headers_end are copied
564 * using a single memcpy() in __copy_skb_header()
567 __u32 headers_start[0];
570 /* if you move pkt_type around you also must adapt those constants */
571 #ifdef __BIG_ENDIAN_BITFIELD
572 #define PKT_TYPE_MAX (7 << 5)
574 #define PKT_TYPE_MAX 7
576 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
578 __u8 __pkt_type_offset[0];
589 __u8 wifi_acked_valid:1;
593 /* Indicates the inner headers are valid in the skbuff. */
594 __u8 encapsulation:1;
595 __u8 encap_hdr_csum:1;
597 __u8 csum_complete_sw:1;
601 #ifdef CONFIG_IPV6_NDISC_NODETYPE
602 __u8 ndisc_nodetype:2;
604 __u8 ipvs_property:1;
605 __u8 inner_protocol_type:1;
606 __u8 remcsum_offload:1;
607 /* 3 or 5 bit hole */
609 #ifdef CONFIG_NET_SCHED
610 __u16 tc_index; /* traffic control index */
611 #ifdef CONFIG_NET_CLS_ACT
612 __u16 tc_verd; /* traffic control verdict */
628 #ifdef CONFIG_NET_RX_BUSY_POLL
629 unsigned int napi_id;
631 #ifdef CONFIG_NETWORK_SECMARK
637 __u32 reserved_tailroom;
641 __be16 inner_protocol;
645 __u16 inner_transport_header;
646 __u16 inner_network_header;
647 __u16 inner_mac_header;
650 __u16 transport_header;
651 __u16 network_header;
655 __u32 headers_end[0];
658 /* These elements must be at the end, see alloc_skb() for details. */
663 unsigned int truesize;
669 * Handling routines are only of interest to the kernel
671 #include <linux/slab.h>
674 #define SKB_ALLOC_FCLONE 0x01
675 #define SKB_ALLOC_RX 0x02
677 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
678 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
680 return unlikely(skb->pfmemalloc);
684 * skb might have a dst pointer attached, refcounted or not.
685 * _skb_refdst low order bit is set if refcount was _not_ taken
687 #define SKB_DST_NOREF 1UL
688 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
691 * skb_dst - returns skb dst_entry
694 * Returns skb dst_entry, regardless of reference taken or not.
696 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
698 /* If refdst was not refcounted, check we still are in a
699 * rcu_read_lock section
701 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
702 !rcu_read_lock_held() &&
703 !rcu_read_lock_bh_held());
704 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
708 * skb_dst_set - sets skb dst
712 * Sets skb dst, assuming a reference was taken on dst and should
713 * be released by skb_dst_drop()
715 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
717 skb->_skb_refdst = (unsigned long)dst;
721 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
725 * Sets skb dst, assuming a reference was not taken on dst.
726 * If dst entry is cached, we do not take reference and dst_release
727 * will be avoided by refdst_drop. If dst entry is not cached, we take
728 * reference, so that last dst_release can destroy the dst immediately.
730 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
732 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
733 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
737 * skb_dst_is_noref - Test if skb dst isn't refcounted
740 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
742 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
745 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
747 return (struct rtable *)skb_dst(skb);
750 void kfree_skb(struct sk_buff *skb);
751 void kfree_skb_list(struct sk_buff *segs);
752 void skb_tx_error(struct sk_buff *skb);
753 void consume_skb(struct sk_buff *skb);
754 void __kfree_skb(struct sk_buff *skb);
755 extern struct kmem_cache *skbuff_head_cache;
757 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
758 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
759 bool *fragstolen, int *delta_truesize);
761 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
763 struct sk_buff *build_skb(void *data, unsigned int frag_size);
764 static inline struct sk_buff *alloc_skb(unsigned int size,
767 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
770 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
771 unsigned long data_len,
776 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
777 struct sk_buff_fclones {
786 * skb_fclone_busy - check if fclone is busy
789 * Returns true is skb is a fast clone, and its clone is not freed.
790 * Some drivers call skb_orphan() in their ndo_start_xmit(),
791 * so we also check that this didnt happen.
793 static inline bool skb_fclone_busy(const struct sock *sk,
794 const struct sk_buff *skb)
796 const struct sk_buff_fclones *fclones;
798 fclones = container_of(skb, struct sk_buff_fclones, skb1);
800 return skb->fclone == SKB_FCLONE_ORIG &&
801 atomic_read(&fclones->fclone_ref) > 1 &&
802 fclones->skb2.sk == sk;
805 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
808 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
811 struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
812 static inline struct sk_buff *alloc_skb_head(gfp_t priority)
814 return __alloc_skb_head(priority, -1);
817 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
818 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
819 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
820 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
821 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
822 gfp_t gfp_mask, bool fclone);
823 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
826 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
829 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
830 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
831 unsigned int headroom);
832 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
833 int newtailroom, gfp_t priority);
834 int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
835 int offset, int len);
836 int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
838 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
839 int skb_pad(struct sk_buff *skb, int pad);
840 #define dev_kfree_skb(a) consume_skb(a)
842 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
843 int getfrag(void *from, char *to, int offset,
844 int len, int odd, struct sk_buff *skb),
845 void *from, int length);
847 struct skb_seq_state {
851 __u32 stepped_offset;
852 struct sk_buff *root_skb;
853 struct sk_buff *cur_skb;
857 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
858 unsigned int to, struct skb_seq_state *st);
859 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
860 struct skb_seq_state *st);
861 void skb_abort_seq_read(struct skb_seq_state *st);
863 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
864 unsigned int to, struct ts_config *config,
865 struct ts_state *state);
868 * Packet hash types specify the type of hash in skb_set_hash.
870 * Hash types refer to the protocol layer addresses which are used to
871 * construct a packet's hash. The hashes are used to differentiate or identify
872 * flows of the protocol layer for the hash type. Hash types are either
873 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
875 * Properties of hashes:
877 * 1) Two packets in different flows have different hash values
878 * 2) Two packets in the same flow should have the same hash value
880 * A hash at a higher layer is considered to be more specific. A driver should
881 * set the most specific hash possible.
883 * A driver cannot indicate a more specific hash than the layer at which a hash
884 * was computed. For instance an L3 hash cannot be set as an L4 hash.
886 * A driver may indicate a hash level which is less specific than the
887 * actual layer the hash was computed on. For instance, a hash computed
888 * at L4 may be considered an L3 hash. This should only be done if the
889 * driver can't unambiguously determine that the HW computed the hash at
890 * the higher layer. Note that the "should" in the second property above
893 enum pkt_hash_types {
894 PKT_HASH_TYPE_NONE, /* Undefined type */
895 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
896 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
897 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
901 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
903 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
908 void __skb_get_hash(struct sk_buff *skb);
909 static inline __u32 skb_get_hash(struct sk_buff *skb)
911 if (!skb->l4_hash && !skb->sw_hash)
917 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
922 static inline void skb_clear_hash(struct sk_buff *skb)
929 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
935 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
937 to->hash = from->hash;
938 to->sw_hash = from->sw_hash;
939 to->l4_hash = from->l4_hash;
942 #ifdef NET_SKBUFF_DATA_USES_OFFSET
943 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
945 return skb->head + skb->end;
948 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
953 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
958 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
960 return skb->end - skb->head;
965 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
967 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
969 return &skb_shinfo(skb)->hwtstamps;
973 * skb_queue_empty - check if a queue is empty
976 * Returns true if the queue is empty, false otherwise.
978 static inline int skb_queue_empty(const struct sk_buff_head *list)
980 return list->next == (const struct sk_buff *) list;
984 * skb_queue_is_last - check if skb is the last entry in the queue
988 * Returns true if @skb is the last buffer on the list.
990 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
991 const struct sk_buff *skb)
993 return skb->next == (const struct sk_buff *) list;
997 * skb_queue_is_first - check if skb is the first entry in the queue
1001 * Returns true if @skb is the first buffer on the list.
1003 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1004 const struct sk_buff *skb)
1006 return skb->prev == (const struct sk_buff *) list;
1010 * skb_queue_next - return the next packet in the queue
1012 * @skb: current buffer
1014 * Return the next packet in @list after @skb. It is only valid to
1015 * call this if skb_queue_is_last() evaluates to false.
1017 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1018 const struct sk_buff *skb)
1020 /* This BUG_ON may seem severe, but if we just return then we
1021 * are going to dereference garbage.
1023 BUG_ON(skb_queue_is_last(list, skb));
1028 * skb_queue_prev - return the prev packet in the queue
1030 * @skb: current buffer
1032 * Return the prev packet in @list before @skb. It is only valid to
1033 * call this if skb_queue_is_first() evaluates to false.
1035 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1036 const struct sk_buff *skb)
1038 /* This BUG_ON may seem severe, but if we just return then we
1039 * are going to dereference garbage.
1041 BUG_ON(skb_queue_is_first(list, skb));
1046 * skb_get - reference buffer
1047 * @skb: buffer to reference
1049 * Makes another reference to a socket buffer and returns a pointer
1052 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1054 atomic_inc(&skb->users);
1059 * If users == 1, we are the only owner and are can avoid redundant
1064 * skb_cloned - is the buffer a clone
1065 * @skb: buffer to check
1067 * Returns true if the buffer was generated with skb_clone() and is
1068 * one of multiple shared copies of the buffer. Cloned buffers are
1069 * shared data so must not be written to under normal circumstances.
1071 static inline int skb_cloned(const struct sk_buff *skb)
1073 return skb->cloned &&
1074 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1077 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1079 might_sleep_if(pri & __GFP_WAIT);
1081 if (skb_cloned(skb))
1082 return pskb_expand_head(skb, 0, 0, pri);
1088 * skb_header_cloned - is the header a clone
1089 * @skb: buffer to check
1091 * Returns true if modifying the header part of the buffer requires
1092 * the data to be copied.
1094 static inline int skb_header_cloned(const struct sk_buff *skb)
1101 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1102 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1103 return dataref != 1;
1107 * skb_header_release - release reference to header
1108 * @skb: buffer to operate on
1110 * Drop a reference to the header part of the buffer. This is done
1111 * by acquiring a payload reference. You must not read from the header
1112 * part of skb->data after this.
1113 * Note : Check if you can use __skb_header_release() instead.
1115 static inline void skb_header_release(struct sk_buff *skb)
1119 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1123 * __skb_header_release - release reference to header
1124 * @skb: buffer to operate on
1126 * Variant of skb_header_release() assuming skb is private to caller.
1127 * We can avoid one atomic operation.
1129 static inline void __skb_header_release(struct sk_buff *skb)
1132 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1137 * skb_shared - is the buffer shared
1138 * @skb: buffer to check
1140 * Returns true if more than one person has a reference to this
1143 static inline int skb_shared(const struct sk_buff *skb)
1145 return atomic_read(&skb->users) != 1;
1149 * skb_share_check - check if buffer is shared and if so clone it
1150 * @skb: buffer to check
1151 * @pri: priority for memory allocation
1153 * If the buffer is shared the buffer is cloned and the old copy
1154 * drops a reference. A new clone with a single reference is returned.
1155 * If the buffer is not shared the original buffer is returned. When
1156 * being called from interrupt status or with spinlocks held pri must
1159 * NULL is returned on a memory allocation failure.
1161 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1163 might_sleep_if(pri & __GFP_WAIT);
1164 if (skb_shared(skb)) {
1165 struct sk_buff *nskb = skb_clone(skb, pri);
1177 * Copy shared buffers into a new sk_buff. We effectively do COW on
1178 * packets to handle cases where we have a local reader and forward
1179 * and a couple of other messy ones. The normal one is tcpdumping
1180 * a packet thats being forwarded.
1184 * skb_unshare - make a copy of a shared buffer
1185 * @skb: buffer to check
1186 * @pri: priority for memory allocation
1188 * If the socket buffer is a clone then this function creates a new
1189 * copy of the data, drops a reference count on the old copy and returns
1190 * the new copy with the reference count at 1. If the buffer is not a clone
1191 * the original buffer is returned. When called with a spinlock held or
1192 * from interrupt state @pri must be %GFP_ATOMIC
1194 * %NULL is returned on a memory allocation failure.
1196 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1199 might_sleep_if(pri & __GFP_WAIT);
1200 if (skb_cloned(skb)) {
1201 struct sk_buff *nskb = skb_copy(skb, pri);
1203 /* Free our shared copy */
1214 * skb_peek - peek at the head of an &sk_buff_head
1215 * @list_: list to peek at
1217 * Peek an &sk_buff. Unlike most other operations you _MUST_
1218 * be careful with this one. A peek leaves the buffer on the
1219 * list and someone else may run off with it. You must hold
1220 * the appropriate locks or have a private queue to do this.
1222 * Returns %NULL for an empty list or a pointer to the head element.
1223 * The reference count is not incremented and the reference is therefore
1224 * volatile. Use with caution.
1226 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1228 struct sk_buff *skb = list_->next;
1230 if (skb == (struct sk_buff *)list_)
1236 * skb_peek_next - peek skb following the given one from a queue
1237 * @skb: skb to start from
1238 * @list_: list to peek at
1240 * Returns %NULL when the end of the list is met or a pointer to the
1241 * next element. The reference count is not incremented and the
1242 * reference is therefore volatile. Use with caution.
1244 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1245 const struct sk_buff_head *list_)
1247 struct sk_buff *next = skb->next;
1249 if (next == (struct sk_buff *)list_)
1255 * skb_peek_tail - peek at the tail of an &sk_buff_head
1256 * @list_: list to peek at
1258 * Peek an &sk_buff. Unlike most other operations you _MUST_
1259 * be careful with this one. A peek leaves the buffer on the
1260 * list and someone else may run off with it. You must hold
1261 * the appropriate locks or have a private queue to do this.
1263 * Returns %NULL for an empty list or a pointer to the tail element.
1264 * The reference count is not incremented and the reference is therefore
1265 * volatile. Use with caution.
1267 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1269 struct sk_buff *skb = list_->prev;
1271 if (skb == (struct sk_buff *)list_)
1278 * skb_queue_len - get queue length
1279 * @list_: list to measure
1281 * Return the length of an &sk_buff queue.
1283 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1289 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1290 * @list: queue to initialize
1292 * This initializes only the list and queue length aspects of
1293 * an sk_buff_head object. This allows to initialize the list
1294 * aspects of an sk_buff_head without reinitializing things like
1295 * the spinlock. It can also be used for on-stack sk_buff_head
1296 * objects where the spinlock is known to not be used.
1298 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1300 list->prev = list->next = (struct sk_buff *)list;
1305 * This function creates a split out lock class for each invocation;
1306 * this is needed for now since a whole lot of users of the skb-queue
1307 * infrastructure in drivers have different locking usage (in hardirq)
1308 * than the networking core (in softirq only). In the long run either the
1309 * network layer or drivers should need annotation to consolidate the
1310 * main types of usage into 3 classes.
1312 static inline void skb_queue_head_init(struct sk_buff_head *list)
1314 spin_lock_init(&list->lock);
1315 __skb_queue_head_init(list);
1318 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1319 struct lock_class_key *class)
1321 skb_queue_head_init(list);
1322 lockdep_set_class(&list->lock, class);
1326 * Insert an sk_buff on a list.
1328 * The "__skb_xxxx()" functions are the non-atomic ones that
1329 * can only be called with interrupts disabled.
1331 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1332 struct sk_buff_head *list);
1333 static inline void __skb_insert(struct sk_buff *newsk,
1334 struct sk_buff *prev, struct sk_buff *next,
1335 struct sk_buff_head *list)
1339 next->prev = prev->next = newsk;
1343 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1344 struct sk_buff *prev,
1345 struct sk_buff *next)
1347 struct sk_buff *first = list->next;
1348 struct sk_buff *last = list->prev;
1358 * skb_queue_splice - join two skb lists, this is designed for stacks
1359 * @list: the new list to add
1360 * @head: the place to add it in the first list
1362 static inline void skb_queue_splice(const struct sk_buff_head *list,
1363 struct sk_buff_head *head)
1365 if (!skb_queue_empty(list)) {
1366 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1367 head->qlen += list->qlen;
1372 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1373 * @list: the new list to add
1374 * @head: the place to add it in the first list
1376 * The list at @list is reinitialised
1378 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1379 struct sk_buff_head *head)
1381 if (!skb_queue_empty(list)) {
1382 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1383 head->qlen += list->qlen;
1384 __skb_queue_head_init(list);
1389 * skb_queue_splice_tail - join two skb lists, each list being a queue
1390 * @list: the new list to add
1391 * @head: the place to add it in the first list
1393 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1394 struct sk_buff_head *head)
1396 if (!skb_queue_empty(list)) {
1397 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1398 head->qlen += list->qlen;
1403 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1404 * @list: the new list to add
1405 * @head: the place to add it in the first list
1407 * Each of the lists is a queue.
1408 * The list at @list is reinitialised
1410 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1411 struct sk_buff_head *head)
1413 if (!skb_queue_empty(list)) {
1414 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1415 head->qlen += list->qlen;
1416 __skb_queue_head_init(list);
1421 * __skb_queue_after - queue a buffer at the list head
1422 * @list: list to use
1423 * @prev: place after this buffer
1424 * @newsk: buffer to queue
1426 * Queue a buffer int the middle of a list. This function takes no locks
1427 * and you must therefore hold required locks before calling it.
1429 * A buffer cannot be placed on two lists at the same time.
1431 static inline void __skb_queue_after(struct sk_buff_head *list,
1432 struct sk_buff *prev,
1433 struct sk_buff *newsk)
1435 __skb_insert(newsk, prev, prev->next, list);
1438 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1439 struct sk_buff_head *list);
1441 static inline void __skb_queue_before(struct sk_buff_head *list,
1442 struct sk_buff *next,
1443 struct sk_buff *newsk)
1445 __skb_insert(newsk, next->prev, next, list);
1449 * __skb_queue_head - queue a buffer at the list head
1450 * @list: list to use
1451 * @newsk: buffer to queue
1453 * Queue a buffer at the start of a list. This function takes no locks
1454 * and you must therefore hold required locks before calling it.
1456 * A buffer cannot be placed on two lists at the same time.
1458 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1459 static inline void __skb_queue_head(struct sk_buff_head *list,
1460 struct sk_buff *newsk)
1462 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1466 * __skb_queue_tail - queue a buffer at the list tail
1467 * @list: list to use
1468 * @newsk: buffer to queue
1470 * Queue a buffer at the end of a list. This function takes no locks
1471 * and you must therefore hold required locks before calling it.
1473 * A buffer cannot be placed on two lists at the same time.
1475 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1476 static inline void __skb_queue_tail(struct sk_buff_head *list,
1477 struct sk_buff *newsk)
1479 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1483 * remove sk_buff from list. _Must_ be called atomically, and with
1486 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1487 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1489 struct sk_buff *next, *prev;
1494 skb->next = skb->prev = NULL;
1500 * __skb_dequeue - remove from the head of the queue
1501 * @list: list to dequeue from
1503 * Remove the head of the list. This function does not take any locks
1504 * so must be used with appropriate locks held only. The head item is
1505 * returned or %NULL if the list is empty.
1507 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1508 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1510 struct sk_buff *skb = skb_peek(list);
1512 __skb_unlink(skb, list);
1517 * __skb_dequeue_tail - remove from the tail of the queue
1518 * @list: list to dequeue from
1520 * Remove the tail of the list. This function does not take any locks
1521 * so must be used with appropriate locks held only. The tail item is
1522 * returned or %NULL if the list is empty.
1524 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1525 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1527 struct sk_buff *skb = skb_peek_tail(list);
1529 __skb_unlink(skb, list);
1534 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1536 return skb->data_len;
1539 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1541 return skb->len - skb->data_len;
1544 static inline int skb_pagelen(const struct sk_buff *skb)
1548 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1549 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1550 return len + skb_headlen(skb);
1554 * __skb_fill_page_desc - initialise a paged fragment in an skb
1555 * @skb: buffer containing fragment to be initialised
1556 * @i: paged fragment index to initialise
1557 * @page: the page to use for this fragment
1558 * @off: the offset to the data with @page
1559 * @size: the length of the data
1561 * Initialises the @i'th fragment of @skb to point to &size bytes at
1562 * offset @off within @page.
1564 * Does not take any additional reference on the fragment.
1566 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1567 struct page *page, int off, int size)
1569 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1572 * Propagate page->pfmemalloc to the skb if we can. The problem is
1573 * that not all callers have unique ownership of the page. If
1574 * pfmemalloc is set, we check the mapping as a mapping implies
1575 * page->index is set (index and pfmemalloc share space).
1576 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1577 * do not lose pfmemalloc information as the pages would not be
1578 * allocated using __GFP_MEMALLOC.
1580 frag->page.p = page;
1581 frag->page_offset = off;
1582 skb_frag_size_set(frag, size);
1584 page = compound_head(page);
1585 if (page->pfmemalloc && !page->mapping)
1586 skb->pfmemalloc = true;
1590 * skb_fill_page_desc - initialise a paged fragment in an skb
1591 * @skb: buffer containing fragment to be initialised
1592 * @i: paged fragment index to initialise
1593 * @page: the page to use for this fragment
1594 * @off: the offset to the data with @page
1595 * @size: the length of the data
1597 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1598 * @skb to point to @size bytes at offset @off within @page. In
1599 * addition updates @skb such that @i is the last fragment.
1601 * Does not take any additional reference on the fragment.
1603 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1604 struct page *page, int off, int size)
1606 __skb_fill_page_desc(skb, i, page, off, size);
1607 skb_shinfo(skb)->nr_frags = i + 1;
1610 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1611 int size, unsigned int truesize);
1613 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1614 unsigned int truesize);
1616 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1617 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1618 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1620 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1621 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1623 return skb->head + skb->tail;
1626 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1628 skb->tail = skb->data - skb->head;
1631 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1633 skb_reset_tail_pointer(skb);
1634 skb->tail += offset;
1637 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1638 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1643 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1645 skb->tail = skb->data;
1648 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1650 skb->tail = skb->data + offset;
1653 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1656 * Add data to an sk_buff
1658 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1659 unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1660 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1662 unsigned char *tmp = skb_tail_pointer(skb);
1663 SKB_LINEAR_ASSERT(skb);
1669 unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1670 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1677 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1678 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1681 BUG_ON(skb->len < skb->data_len);
1682 return skb->data += len;
1685 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1687 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1690 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1692 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1694 if (len > skb_headlen(skb) &&
1695 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1698 return skb->data += len;
1701 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1703 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1706 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1708 if (likely(len <= skb_headlen(skb)))
1710 if (unlikely(len > skb->len))
1712 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1716 * skb_headroom - bytes at buffer head
1717 * @skb: buffer to check
1719 * Return the number of bytes of free space at the head of an &sk_buff.
1721 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1723 return skb->data - skb->head;
1727 * skb_tailroom - bytes at buffer end
1728 * @skb: buffer to check
1730 * Return the number of bytes of free space at the tail of an sk_buff
1732 static inline int skb_tailroom(const struct sk_buff *skb)
1734 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1738 * skb_availroom - bytes at buffer end
1739 * @skb: buffer to check
1741 * Return the number of bytes of free space at the tail of an sk_buff
1742 * allocated by sk_stream_alloc()
1744 static inline int skb_availroom(const struct sk_buff *skb)
1746 if (skb_is_nonlinear(skb))
1749 return skb->end - skb->tail - skb->reserved_tailroom;
1753 * skb_reserve - adjust headroom
1754 * @skb: buffer to alter
1755 * @len: bytes to move
1757 * Increase the headroom of an empty &sk_buff by reducing the tail
1758 * room. This is only allowed for an empty buffer.
1760 static inline void skb_reserve(struct sk_buff *skb, int len)
1766 #define ENCAP_TYPE_ETHER 0
1767 #define ENCAP_TYPE_IPPROTO 1
1769 static inline void skb_set_inner_protocol(struct sk_buff *skb,
1772 skb->inner_protocol = protocol;
1773 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1776 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1779 skb->inner_ipproto = ipproto;
1780 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1783 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1785 skb->inner_mac_header = skb->mac_header;
1786 skb->inner_network_header = skb->network_header;
1787 skb->inner_transport_header = skb->transport_header;
1790 static inline void skb_reset_mac_len(struct sk_buff *skb)
1792 skb->mac_len = skb->network_header - skb->mac_header;
1795 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1798 return skb->head + skb->inner_transport_header;
1801 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1803 skb->inner_transport_header = skb->data - skb->head;
1806 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1809 skb_reset_inner_transport_header(skb);
1810 skb->inner_transport_header += offset;
1813 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1815 return skb->head + skb->inner_network_header;
1818 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1820 skb->inner_network_header = skb->data - skb->head;
1823 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1826 skb_reset_inner_network_header(skb);
1827 skb->inner_network_header += offset;
1830 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1832 return skb->head + skb->inner_mac_header;
1835 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1837 skb->inner_mac_header = skb->data - skb->head;
1840 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1843 skb_reset_inner_mac_header(skb);
1844 skb->inner_mac_header += offset;
1846 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1848 return skb->transport_header != (typeof(skb->transport_header))~0U;
1851 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1853 return skb->head + skb->transport_header;
1856 static inline void skb_reset_transport_header(struct sk_buff *skb)
1858 skb->transport_header = skb->data - skb->head;
1861 static inline void skb_set_transport_header(struct sk_buff *skb,
1864 skb_reset_transport_header(skb);
1865 skb->transport_header += offset;
1868 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1870 return skb->head + skb->network_header;
1873 static inline void skb_reset_network_header(struct sk_buff *skb)
1875 skb->network_header = skb->data - skb->head;
1878 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1880 skb_reset_network_header(skb);
1881 skb->network_header += offset;
1884 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1886 return skb->head + skb->mac_header;
1889 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1891 return skb->mac_header != (typeof(skb->mac_header))~0U;
1894 static inline void skb_reset_mac_header(struct sk_buff *skb)
1896 skb->mac_header = skb->data - skb->head;
1899 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1901 skb_reset_mac_header(skb);
1902 skb->mac_header += offset;
1905 static inline void skb_pop_mac_header(struct sk_buff *skb)
1907 skb->mac_header = skb->network_header;
1910 static inline void skb_probe_transport_header(struct sk_buff *skb,
1911 const int offset_hint)
1913 struct flow_keys keys;
1915 if (skb_transport_header_was_set(skb))
1917 else if (skb_flow_dissect(skb, &keys))
1918 skb_set_transport_header(skb, keys.thoff);
1920 skb_set_transport_header(skb, offset_hint);
1923 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1925 if (skb_mac_header_was_set(skb)) {
1926 const unsigned char *old_mac = skb_mac_header(skb);
1928 skb_set_mac_header(skb, -skb->mac_len);
1929 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1933 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1935 return skb->csum_start - skb_headroom(skb);
1938 static inline int skb_transport_offset(const struct sk_buff *skb)
1940 return skb_transport_header(skb) - skb->data;
1943 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1945 return skb->transport_header - skb->network_header;
1948 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1950 return skb->inner_transport_header - skb->inner_network_header;
1953 static inline int skb_network_offset(const struct sk_buff *skb)
1955 return skb_network_header(skb) - skb->data;
1958 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1960 return skb_inner_network_header(skb) - skb->data;
1963 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1965 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1969 * CPUs often take a performance hit when accessing unaligned memory
1970 * locations. The actual performance hit varies, it can be small if the
1971 * hardware handles it or large if we have to take an exception and fix it
1974 * Since an ethernet header is 14 bytes network drivers often end up with
1975 * the IP header at an unaligned offset. The IP header can be aligned by
1976 * shifting the start of the packet by 2 bytes. Drivers should do this
1979 * skb_reserve(skb, NET_IP_ALIGN);
1981 * The downside to this alignment of the IP header is that the DMA is now
1982 * unaligned. On some architectures the cost of an unaligned DMA is high
1983 * and this cost outweighs the gains made by aligning the IP header.
1985 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1988 #ifndef NET_IP_ALIGN
1989 #define NET_IP_ALIGN 2
1993 * The networking layer reserves some headroom in skb data (via
1994 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1995 * the header has to grow. In the default case, if the header has to grow
1996 * 32 bytes or less we avoid the reallocation.
1998 * Unfortunately this headroom changes the DMA alignment of the resulting
1999 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2000 * on some architectures. An architecture can override this value,
2001 * perhaps setting it to a cacheline in size (since that will maintain
2002 * cacheline alignment of the DMA). It must be a power of 2.
2004 * Various parts of the networking layer expect at least 32 bytes of
2005 * headroom, you should not reduce this.
2007 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2008 * to reduce average number of cache lines per packet.
2009 * get_rps_cpus() for example only access one 64 bytes aligned block :
2010 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2013 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2016 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2018 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2020 if (unlikely(skb_is_nonlinear(skb))) {
2025 skb_set_tail_pointer(skb, len);
2028 void skb_trim(struct sk_buff *skb, unsigned int len);
2030 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2033 return ___pskb_trim(skb, len);
2034 __skb_trim(skb, len);
2038 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2040 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2044 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2045 * @skb: buffer to alter
2048 * This is identical to pskb_trim except that the caller knows that
2049 * the skb is not cloned so we should never get an error due to out-
2052 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2054 int err = pskb_trim(skb, len);
2059 * skb_orphan - orphan a buffer
2060 * @skb: buffer to orphan
2062 * If a buffer currently has an owner then we call the owner's
2063 * destructor function and make the @skb unowned. The buffer continues
2064 * to exist but is no longer charged to its former owner.
2066 static inline void skb_orphan(struct sk_buff *skb)
2068 if (skb->destructor) {
2069 skb->destructor(skb);
2070 skb->destructor = NULL;
2078 * skb_orphan_frags - orphan the frags contained in a buffer
2079 * @skb: buffer to orphan frags from
2080 * @gfp_mask: allocation mask for replacement pages
2082 * For each frag in the SKB which needs a destructor (i.e. has an
2083 * owner) create a copy of that frag and release the original
2084 * page by calling the destructor.
2086 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2088 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2090 return skb_copy_ubufs(skb, gfp_mask);
2094 * __skb_queue_purge - empty a list
2095 * @list: list to empty
2097 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2098 * the list and one reference dropped. This function does not take the
2099 * list lock and the caller must hold the relevant locks to use it.
2101 void skb_queue_purge(struct sk_buff_head *list);
2102 static inline void __skb_queue_purge(struct sk_buff_head *list)
2104 struct sk_buff *skb;
2105 while ((skb = __skb_dequeue(list)) != NULL)
2109 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2110 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2111 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2113 void *netdev_alloc_frag(unsigned int fragsz);
2115 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2119 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2120 * @dev: network device to receive on
2121 * @length: length to allocate
2123 * Allocate a new &sk_buff and assign it a usage count of one. The
2124 * buffer has unspecified headroom built in. Users should allocate
2125 * the headroom they think they need without accounting for the
2126 * built in space. The built in space is used for optimisations.
2128 * %NULL is returned if there is no free memory. Although this function
2129 * allocates memory it can be called from an interrupt.
2131 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2132 unsigned int length)
2134 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2137 /* legacy helper around __netdev_alloc_skb() */
2138 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2141 return __netdev_alloc_skb(NULL, length, gfp_mask);
2144 /* legacy helper around netdev_alloc_skb() */
2145 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2147 return netdev_alloc_skb(NULL, length);
2151 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2152 unsigned int length, gfp_t gfp)
2154 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2156 if (NET_IP_ALIGN && skb)
2157 skb_reserve(skb, NET_IP_ALIGN);
2161 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2162 unsigned int length)
2164 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2168 * __dev_alloc_pages - allocate page for network Rx
2169 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2170 * @order: size of the allocation
2172 * Allocate a new page.
2174 * %NULL is returned if there is no free memory.
2176 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2179 /* This piece of code contains several assumptions.
2180 * 1. This is for device Rx, therefor a cold page is preferred.
2181 * 2. The expectation is the user wants a compound page.
2182 * 3. If requesting a order 0 page it will not be compound
2183 * due to the check to see if order has a value in prep_new_page
2184 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2185 * code in gfp_to_alloc_flags that should be enforcing this.
2187 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2189 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2192 static inline struct page *dev_alloc_pages(unsigned int order)
2194 return __dev_alloc_pages(GFP_ATOMIC, order);
2198 * __dev_alloc_page - allocate a page for network Rx
2199 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2201 * Allocate a new page.
2203 * %NULL is returned if there is no free memory.
2205 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2207 return __dev_alloc_pages(gfp_mask, 0);
2210 static inline struct page *dev_alloc_page(void)
2212 return __dev_alloc_page(GFP_ATOMIC);
2216 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2217 * @page: The page that was allocated from skb_alloc_page
2218 * @skb: The skb that may need pfmemalloc set
2220 static inline void skb_propagate_pfmemalloc(struct page *page,
2221 struct sk_buff *skb)
2223 if (page && page->pfmemalloc)
2224 skb->pfmemalloc = true;
2228 * skb_frag_page - retrieve the page referred to by a paged fragment
2229 * @frag: the paged fragment
2231 * Returns the &struct page associated with @frag.
2233 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2235 return frag->page.p;
2239 * __skb_frag_ref - take an addition reference on a paged fragment.
2240 * @frag: the paged fragment
2242 * Takes an additional reference on the paged fragment @frag.
2244 static inline void __skb_frag_ref(skb_frag_t *frag)
2246 get_page(skb_frag_page(frag));
2250 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2252 * @f: the fragment offset.
2254 * Takes an additional reference on the @f'th paged fragment of @skb.
2256 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2258 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2262 * __skb_frag_unref - release a reference on a paged fragment.
2263 * @frag: the paged fragment
2265 * Releases a reference on the paged fragment @frag.
2267 static inline void __skb_frag_unref(skb_frag_t *frag)
2269 put_page(skb_frag_page(frag));
2273 * skb_frag_unref - release a reference on a paged fragment of an skb.
2275 * @f: the fragment offset
2277 * Releases a reference on the @f'th paged fragment of @skb.
2279 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2281 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2285 * skb_frag_address - gets the address of the data contained in a paged fragment
2286 * @frag: the paged fragment buffer
2288 * Returns the address of the data within @frag. The page must already
2291 static inline void *skb_frag_address(const skb_frag_t *frag)
2293 return page_address(skb_frag_page(frag)) + frag->page_offset;
2297 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2298 * @frag: the paged fragment buffer
2300 * Returns the address of the data within @frag. Checks that the page
2301 * is mapped and returns %NULL otherwise.
2303 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2305 void *ptr = page_address(skb_frag_page(frag));
2309 return ptr + frag->page_offset;
2313 * __skb_frag_set_page - sets the page contained in a paged fragment
2314 * @frag: the paged fragment
2315 * @page: the page to set
2317 * Sets the fragment @frag to contain @page.
2319 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2321 frag->page.p = page;
2325 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2327 * @f: the fragment offset
2328 * @page: the page to set
2330 * Sets the @f'th fragment of @skb to contain @page.
2332 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2335 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2338 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2341 * skb_frag_dma_map - maps a paged fragment via the DMA API
2342 * @dev: the device to map the fragment to
2343 * @frag: the paged fragment to map
2344 * @offset: the offset within the fragment (starting at the
2345 * fragment's own offset)
2346 * @size: the number of bytes to map
2347 * @dir: the direction of the mapping (%PCI_DMA_*)
2349 * Maps the page associated with @frag to @device.
2351 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2352 const skb_frag_t *frag,
2353 size_t offset, size_t size,
2354 enum dma_data_direction dir)
2356 return dma_map_page(dev, skb_frag_page(frag),
2357 frag->page_offset + offset, size, dir);
2360 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2363 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2367 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2370 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2375 * skb_clone_writable - is the header of a clone writable
2376 * @skb: buffer to check
2377 * @len: length up to which to write
2379 * Returns true if modifying the header part of the cloned buffer
2380 * does not requires the data to be copied.
2382 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2384 return !skb_header_cloned(skb) &&
2385 skb_headroom(skb) + len <= skb->hdr_len;
2388 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2393 if (headroom > skb_headroom(skb))
2394 delta = headroom - skb_headroom(skb);
2396 if (delta || cloned)
2397 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2403 * skb_cow - copy header of skb when it is required
2404 * @skb: buffer to cow
2405 * @headroom: needed headroom
2407 * If the skb passed lacks sufficient headroom or its data part
2408 * is shared, data is reallocated. If reallocation fails, an error
2409 * is returned and original skb is not changed.
2411 * The result is skb with writable area skb->head...skb->tail
2412 * and at least @headroom of space at head.
2414 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2416 return __skb_cow(skb, headroom, skb_cloned(skb));
2420 * skb_cow_head - skb_cow but only making the head writable
2421 * @skb: buffer to cow
2422 * @headroom: needed headroom
2424 * This function is identical to skb_cow except that we replace the
2425 * skb_cloned check by skb_header_cloned. It should be used when
2426 * you only need to push on some header and do not need to modify
2429 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2431 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2435 * skb_padto - pad an skbuff up to a minimal size
2436 * @skb: buffer to pad
2437 * @len: minimal length
2439 * Pads up a buffer to ensure the trailing bytes exist and are
2440 * blanked. If the buffer already contains sufficient data it
2441 * is untouched. Otherwise it is extended. Returns zero on
2442 * success. The skb is freed on error.
2444 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2446 unsigned int size = skb->len;
2447 if (likely(size >= len))
2449 return skb_pad(skb, len - size);
2453 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2454 * @skb: buffer to pad
2455 * @len: minimal length
2457 * Pads up a buffer to ensure the trailing bytes exist and are
2458 * blanked. If the buffer already contains sufficient data it
2459 * is untouched. Otherwise it is extended. Returns zero on
2460 * success. The skb is freed on error.
2462 static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2464 unsigned int size = skb->len;
2466 if (unlikely(size < len)) {
2468 if (skb_pad(skb, len))
2470 __skb_put(skb, len);
2475 static inline int skb_add_data(struct sk_buff *skb,
2476 char __user *from, int copy)
2478 const int off = skb->len;
2480 if (skb->ip_summed == CHECKSUM_NONE) {
2482 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
2485 skb->csum = csum_block_add(skb->csum, csum, off);
2488 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2491 __skb_trim(skb, off);
2495 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2496 const struct page *page, int off)
2499 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2501 return page == skb_frag_page(frag) &&
2502 off == frag->page_offset + skb_frag_size(frag);
2507 static inline int __skb_linearize(struct sk_buff *skb)
2509 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2513 * skb_linearize - convert paged skb to linear one
2514 * @skb: buffer to linarize
2516 * If there is no free memory -ENOMEM is returned, otherwise zero
2517 * is returned and the old skb data released.
2519 static inline int skb_linearize(struct sk_buff *skb)
2521 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2525 * skb_has_shared_frag - can any frag be overwritten
2526 * @skb: buffer to test
2528 * Return true if the skb has at least one frag that might be modified
2529 * by an external entity (as in vmsplice()/sendfile())
2531 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2533 return skb_is_nonlinear(skb) &&
2534 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2538 * skb_linearize_cow - make sure skb is linear and writable
2539 * @skb: buffer to process
2541 * If there is no free memory -ENOMEM is returned, otherwise zero
2542 * is returned and the old skb data released.
2544 static inline int skb_linearize_cow(struct sk_buff *skb)
2546 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2547 __skb_linearize(skb) : 0;
2551 * skb_postpull_rcsum - update checksum for received skb after pull
2552 * @skb: buffer to update
2553 * @start: start of data before pull
2554 * @len: length of data pulled
2556 * After doing a pull on a received packet, you need to call this to
2557 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2558 * CHECKSUM_NONE so that it can be recomputed from scratch.
2561 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2562 const void *start, unsigned int len)
2564 if (skb->ip_summed == CHECKSUM_COMPLETE)
2565 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2568 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2571 * pskb_trim_rcsum - trim received skb and update checksum
2572 * @skb: buffer to trim
2575 * This is exactly the same as pskb_trim except that it ensures the
2576 * checksum of received packets are still valid after the operation.
2579 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2581 if (likely(len >= skb->len))
2583 if (skb->ip_summed == CHECKSUM_COMPLETE)
2584 skb->ip_summed = CHECKSUM_NONE;
2585 return __pskb_trim(skb, len);
2588 #define skb_queue_walk(queue, skb) \
2589 for (skb = (queue)->next; \
2590 skb != (struct sk_buff *)(queue); \
2593 #define skb_queue_walk_safe(queue, skb, tmp) \
2594 for (skb = (queue)->next, tmp = skb->next; \
2595 skb != (struct sk_buff *)(queue); \
2596 skb = tmp, tmp = skb->next)
2598 #define skb_queue_walk_from(queue, skb) \
2599 for (; skb != (struct sk_buff *)(queue); \
2602 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2603 for (tmp = skb->next; \
2604 skb != (struct sk_buff *)(queue); \
2605 skb = tmp, tmp = skb->next)
2607 #define skb_queue_reverse_walk(queue, skb) \
2608 for (skb = (queue)->prev; \
2609 skb != (struct sk_buff *)(queue); \
2612 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2613 for (skb = (queue)->prev, tmp = skb->prev; \
2614 skb != (struct sk_buff *)(queue); \
2615 skb = tmp, tmp = skb->prev)
2617 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2618 for (tmp = skb->prev; \
2619 skb != (struct sk_buff *)(queue); \
2620 skb = tmp, tmp = skb->prev)
2622 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2624 return skb_shinfo(skb)->frag_list != NULL;
2627 static inline void skb_frag_list_init(struct sk_buff *skb)
2629 skb_shinfo(skb)->frag_list = NULL;
2632 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2634 frag->next = skb_shinfo(skb)->frag_list;
2635 skb_shinfo(skb)->frag_list = frag;
2638 #define skb_walk_frags(skb, iter) \
2639 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2641 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2642 int *peeked, int *off, int *err);
2643 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2645 unsigned int datagram_poll(struct file *file, struct socket *sock,
2646 struct poll_table_struct *wait);
2647 int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2648 struct iovec *to, int size);
2649 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2650 struct iov_iter *to, int size);
2651 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2652 struct msghdr *msg, int size)
2654 /* XXX: stripping const */
2655 return skb_copy_datagram_iovec(from, offset, (struct iovec *)msg->msg_iter.iov, size);
2657 int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2659 static inline int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2662 /* XXX: stripping const */
2663 return skb_copy_and_csum_datagram_iovec(skb, hlen, (struct iovec *)msg->msg_iter.iov);
2665 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2666 struct iov_iter *from, int len);
2667 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
2668 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2669 void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2670 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
2671 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2672 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2673 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2674 int len, __wsum csum);
2675 int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2676 struct pipe_inode_info *pipe, unsigned int len,
2677 unsigned int flags);
2678 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2679 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2680 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2682 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2683 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2684 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
2685 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
2686 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
2687 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
2688 int skb_ensure_writable(struct sk_buff *skb, int write_len);
2689 int skb_vlan_pop(struct sk_buff *skb);
2690 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
2692 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2694 /* XXX: stripping const */
2695 return memcpy_fromiovec(data, (struct iovec *)msg->msg_iter.iov, len);
2698 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2700 /* XXX: stripping const */
2701 return memcpy_toiovec((struct iovec *)msg->msg_iter.iov, data, len);
2704 struct skb_checksum_ops {
2705 __wsum (*update)(const void *mem, int len, __wsum wsum);
2706 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2709 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2710 __wsum csum, const struct skb_checksum_ops *ops);
2711 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2714 static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2715 int len, void *data, int hlen, void *buffer)
2717 if (hlen - offset >= len)
2718 return data + offset;
2721 skb_copy_bits(skb, offset, buffer, len) < 0)
2727 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2728 int len, void *buffer)
2730 return __skb_header_pointer(skb, offset, len, skb->data,
2731 skb_headlen(skb), buffer);
2735 * skb_needs_linearize - check if we need to linearize a given skb
2736 * depending on the given device features.
2737 * @skb: socket buffer to check
2738 * @features: net device features
2740 * Returns true if either:
2741 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2742 * 2. skb is fragmented and the device does not support SG.
2744 static inline bool skb_needs_linearize(struct sk_buff *skb,
2745 netdev_features_t features)
2747 return skb_is_nonlinear(skb) &&
2748 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2749 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2752 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2754 const unsigned int len)
2756 memcpy(to, skb->data, len);
2759 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2760 const int offset, void *to,
2761 const unsigned int len)
2763 memcpy(to, skb->data + offset, len);
2766 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2768 const unsigned int len)
2770 memcpy(skb->data, from, len);
2773 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2776 const unsigned int len)
2778 memcpy(skb->data + offset, from, len);
2781 void skb_init(void);
2783 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2789 * skb_get_timestamp - get timestamp from a skb
2790 * @skb: skb to get stamp from
2791 * @stamp: pointer to struct timeval to store stamp in
2793 * Timestamps are stored in the skb as offsets to a base timestamp.
2794 * This function converts the offset back to a struct timeval and stores
2797 static inline void skb_get_timestamp(const struct sk_buff *skb,
2798 struct timeval *stamp)
2800 *stamp = ktime_to_timeval(skb->tstamp);
2803 static inline void skb_get_timestampns(const struct sk_buff *skb,
2804 struct timespec *stamp)
2806 *stamp = ktime_to_timespec(skb->tstamp);
2809 static inline void __net_timestamp(struct sk_buff *skb)
2811 skb->tstamp = ktime_get_real();
2814 static inline ktime_t net_timedelta(ktime_t t)
2816 return ktime_sub(ktime_get_real(), t);
2819 static inline ktime_t net_invalid_timestamp(void)
2821 return ktime_set(0, 0);
2824 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2826 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2828 void skb_clone_tx_timestamp(struct sk_buff *skb);
2829 bool skb_defer_rx_timestamp(struct sk_buff *skb);
2831 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2833 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2837 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2842 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2845 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2847 * PHY drivers may accept clones of transmitted packets for
2848 * timestamping via their phy_driver.txtstamp method. These drivers
2849 * must call this function to return the skb back to the stack, with
2850 * or without a timestamp.
2852 * @skb: clone of the the original outgoing packet
2853 * @hwtstamps: hardware time stamps, may be NULL if not available
2856 void skb_complete_tx_timestamp(struct sk_buff *skb,
2857 struct skb_shared_hwtstamps *hwtstamps);
2859 void __skb_tstamp_tx(struct sk_buff *orig_skb,
2860 struct skb_shared_hwtstamps *hwtstamps,
2861 struct sock *sk, int tstype);
2864 * skb_tstamp_tx - queue clone of skb with send time stamps
2865 * @orig_skb: the original outgoing packet
2866 * @hwtstamps: hardware time stamps, may be NULL if not available
2868 * If the skb has a socket associated, then this function clones the
2869 * skb (thus sharing the actual data and optional structures), stores
2870 * the optional hardware time stamping information (if non NULL) or
2871 * generates a software time stamp (otherwise), then queues the clone
2872 * to the error queue of the socket. Errors are silently ignored.
2874 void skb_tstamp_tx(struct sk_buff *orig_skb,
2875 struct skb_shared_hwtstamps *hwtstamps);
2877 static inline void sw_tx_timestamp(struct sk_buff *skb)
2879 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2880 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2881 skb_tstamp_tx(skb, NULL);
2885 * skb_tx_timestamp() - Driver hook for transmit timestamping
2887 * Ethernet MAC Drivers should call this function in their hard_xmit()
2888 * function immediately before giving the sk_buff to the MAC hardware.
2890 * Specifically, one should make absolutely sure that this function is
2891 * called before TX completion of this packet can trigger. Otherwise
2892 * the packet could potentially already be freed.
2894 * @skb: A socket buffer.
2896 static inline void skb_tx_timestamp(struct sk_buff *skb)
2898 skb_clone_tx_timestamp(skb);
2899 sw_tx_timestamp(skb);
2903 * skb_complete_wifi_ack - deliver skb with wifi status
2905 * @skb: the original outgoing packet
2906 * @acked: ack status
2909 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2911 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2912 __sum16 __skb_checksum_complete(struct sk_buff *skb);
2914 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2916 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
2920 * skb_checksum_complete - Calculate checksum of an entire packet
2921 * @skb: packet to process
2923 * This function calculates the checksum over the entire packet plus
2924 * the value of skb->csum. The latter can be used to supply the
2925 * checksum of a pseudo header as used by TCP/UDP. It returns the
2928 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2929 * this function can be used to verify that checksum on received
2930 * packets. In that case the function should return zero if the
2931 * checksum is correct. In particular, this function will return zero
2932 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2933 * hardware has already verified the correctness of the checksum.
2935 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2937 return skb_csum_unnecessary(skb) ?
2938 0 : __skb_checksum_complete(skb);
2941 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2943 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2944 if (skb->csum_level == 0)
2945 skb->ip_summed = CHECKSUM_NONE;
2951 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2953 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2954 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2956 } else if (skb->ip_summed == CHECKSUM_NONE) {
2957 skb->ip_summed = CHECKSUM_UNNECESSARY;
2958 skb->csum_level = 0;
2962 static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2964 /* Mark current checksum as bad (typically called from GRO
2965 * path). In the case that ip_summed is CHECKSUM_NONE
2966 * this must be the first checksum encountered in the packet.
2967 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2968 * checksum after the last one validated. For UDP, a zero
2969 * checksum can not be marked as bad.
2972 if (skb->ip_summed == CHECKSUM_NONE ||
2973 skb->ip_summed == CHECKSUM_UNNECESSARY)
2977 /* Check if we need to perform checksum complete validation.
2979 * Returns true if checksum complete is needed, false otherwise
2980 * (either checksum is unnecessary or zero checksum is allowed).
2982 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2986 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2987 skb->csum_valid = 1;
2988 __skb_decr_checksum_unnecessary(skb);
2995 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2998 #define CHECKSUM_BREAK 76
3000 /* Validate (init) checksum based on checksum complete.
3003 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3004 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3005 * checksum is stored in skb->csum for use in __skb_checksum_complete
3006 * non-zero: value of invalid checksum
3009 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3013 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3014 if (!csum_fold(csum_add(psum, skb->csum))) {
3015 skb->csum_valid = 1;
3018 } else if (skb->csum_bad) {
3019 /* ip_summed == CHECKSUM_NONE in this case */
3025 if (complete || skb->len <= CHECKSUM_BREAK) {
3028 csum = __skb_checksum_complete(skb);
3029 skb->csum_valid = !csum;
3036 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3041 /* Perform checksum validate (init). Note that this is a macro since we only
3042 * want to calculate the pseudo header which is an input function if necessary.
3043 * First we try to validate without any computation (checksum unnecessary) and
3044 * then calculate based on checksum complete calling the function to compute
3048 * 0: checksum is validated or try to in skb_checksum_complete
3049 * non-zero: value of invalid checksum
3051 #define __skb_checksum_validate(skb, proto, complete, \
3052 zero_okay, check, compute_pseudo) \
3054 __sum16 __ret = 0; \
3055 skb->csum_valid = 0; \
3056 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3057 __ret = __skb_checksum_validate_complete(skb, \
3058 complete, compute_pseudo(skb, proto)); \
3062 #define skb_checksum_init(skb, proto, compute_pseudo) \
3063 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3065 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3066 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3068 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3069 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3071 #define skb_checksum_validate_zero_check(skb, proto, check, \
3073 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
3075 #define skb_checksum_simple_validate(skb) \
3076 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3078 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3080 return (skb->ip_summed == CHECKSUM_NONE &&
3081 skb->csum_valid && !skb->csum_bad);
3084 static inline void __skb_checksum_convert(struct sk_buff *skb,
3085 __sum16 check, __wsum pseudo)
3087 skb->csum = ~pseudo;
3088 skb->ip_summed = CHECKSUM_COMPLETE;
3091 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3093 if (__skb_checksum_convert_check(skb)) \
3094 __skb_checksum_convert(skb, check, \
3095 compute_pseudo(skb, proto)); \
3098 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3099 void nf_conntrack_destroy(struct nf_conntrack *nfct);
3100 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3102 if (nfct && atomic_dec_and_test(&nfct->use))
3103 nf_conntrack_destroy(nfct);
3105 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3108 atomic_inc(&nfct->use);
3111 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3112 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3114 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3117 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3120 atomic_inc(&nf_bridge->use);
3122 #endif /* CONFIG_BRIDGE_NETFILTER */
3123 static inline void nf_reset(struct sk_buff *skb)
3125 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3126 nf_conntrack_put(skb->nfct);
3129 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3130 nf_bridge_put(skb->nf_bridge);
3131 skb->nf_bridge = NULL;
3135 static inline void nf_reset_trace(struct sk_buff *skb)
3137 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3142 /* Note: This doesn't put any conntrack and bridge info in dst. */
3143 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3146 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3147 dst->nfct = src->nfct;
3148 nf_conntrack_get(src->nfct);
3150 dst->nfctinfo = src->nfctinfo;
3152 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3153 dst->nf_bridge = src->nf_bridge;
3154 nf_bridge_get(src->nf_bridge);
3156 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3158 dst->nf_trace = src->nf_trace;
3162 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3164 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3165 nf_conntrack_put(dst->nfct);
3167 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3168 nf_bridge_put(dst->nf_bridge);
3170 __nf_copy(dst, src, true);
3173 #ifdef CONFIG_NETWORK_SECMARK
3174 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3176 to->secmark = from->secmark;
3179 static inline void skb_init_secmark(struct sk_buff *skb)
3184 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3187 static inline void skb_init_secmark(struct sk_buff *skb)
3191 static inline bool skb_irq_freeable(const struct sk_buff *skb)
3193 return !skb->destructor &&
3194 #if IS_ENABLED(CONFIG_XFRM)
3197 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3200 !skb->_skb_refdst &&
3201 !skb_has_frag_list(skb);
3204 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3206 skb->queue_mapping = queue_mapping;
3209 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
3211 return skb->queue_mapping;
3214 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3216 to->queue_mapping = from->queue_mapping;
3219 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3221 skb->queue_mapping = rx_queue + 1;
3224 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
3226 return skb->queue_mapping - 1;
3229 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
3231 return skb->queue_mapping != 0;
3234 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3235 unsigned int num_tx_queues);
3237 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3246 /* Keeps track of mac header offset relative to skb->head.
3247 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3248 * For non-tunnel skb it points to skb_mac_header() and for
3249 * tunnel skb it points to outer mac header.
3250 * Keeps track of level of encapsulation of network headers.
3257 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3259 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3261 return (skb_mac_header(inner_skb) - inner_skb->head) -
3262 SKB_GSO_CB(inner_skb)->mac_offset;
3265 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3267 int new_headroom, headroom;
3270 headroom = skb_headroom(skb);
3271 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3275 new_headroom = skb_headroom(skb);
3276 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3280 /* Compute the checksum for a gso segment. First compute the checksum value
3281 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3282 * then add in skb->csum (checksum from csum_start to end of packet).
3283 * skb->csum and csum_start are then updated to reflect the checksum of the
3284 * resultant packet starting from the transport header-- the resultant checksum
3285 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3288 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3290 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3291 skb_transport_offset(skb);
3294 csum = csum_fold(csum_partial(skb_transport_header(skb),
3297 SKB_GSO_CB(skb)->csum_start -= plen;
3302 static inline bool skb_is_gso(const struct sk_buff *skb)
3304 return skb_shinfo(skb)->gso_size;
3307 /* Note: Should be called only if skb_is_gso(skb) is true */
3308 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
3310 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3313 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
3315 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3317 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3318 * wanted then gso_type will be set. */
3319 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3321 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3322 unlikely(shinfo->gso_type == 0)) {
3323 __skb_warn_lro_forwarding(skb);
3329 static inline void skb_forward_csum(struct sk_buff *skb)
3331 /* Unfortunately we don't support this one. Any brave souls? */
3332 if (skb->ip_summed == CHECKSUM_COMPLETE)
3333 skb->ip_summed = CHECKSUM_NONE;
3337 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3338 * @skb: skb to check
3340 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3341 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3342 * use this helper, to document places where we make this assertion.
3344 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
3347 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3351 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
3353 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3355 u32 skb_get_poff(const struct sk_buff *skb);
3356 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3357 const struct flow_keys *keys, int hlen);
3360 * skb_head_is_locked - Determine if the skb->head is locked down
3361 * @skb: skb to check
3363 * The head on skbs build around a head frag can be removed if they are
3364 * not cloned. This function returns true if the skb head is locked down
3365 * due to either being allocated via kmalloc, or by being a clone with
3366 * multiple references to the head.
3368 static inline bool skb_head_is_locked(const struct sk_buff *skb)
3370 return !skb->head_frag || skb_cloned(skb);
3374 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3378 * skb_gso_network_seglen is used to determine the real size of the
3379 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3381 * The MAC/L2 header is not accounted for.
3383 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3385 unsigned int hdr_len = skb_transport_header(skb) -
3386 skb_network_header(skb);
3387 return hdr_len + skb_gso_transport_seglen(skb);
3389 #endif /* __KERNEL__ */
3390 #endif /* _LINUX_SKBUFF_H */