2 * Definitions for the 'struct sk_buff' memory handlers.
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/kernel.h>
18 #include <linux/kmemcheck.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/bug.h>
22 #include <linux/cache.h>
23 #include <linux/rbtree.h>
24 #include <linux/socket.h>
26 #include <linux/atomic.h>
27 #include <asm/types.h>
28 #include <linux/spinlock.h>
29 #include <linux/net.h>
30 #include <linux/textsearch.h>
31 #include <net/checksum.h>
32 #include <linux/rcupdate.h>
33 #include <linux/hrtimer.h>
34 #include <linux/dma-mapping.h>
35 #include <linux/netdev_features.h>
36 #include <linux/sched.h>
37 #include <net/flow_keys.h>
39 /* A. Checksumming of received packets by device.
43 * Device failed to checksum this packet e.g. due to lack of capabilities.
44 * The packet contains full (though not verified) checksum in packet but
45 * not in skb->csum. Thus, skb->csum is undefined in this case.
47 * CHECKSUM_UNNECESSARY:
49 * The hardware you're dealing with doesn't calculate the full checksum
50 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
51 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
52 * if their checksums are okay. skb->csum is still undefined in this case
53 * though. It is a bad option, but, unfortunately, nowadays most vendors do
54 * this. Apparently with the secret goal to sell you new devices, when you
55 * will add new protocol to your host, f.e. IPv6 8)
57 * CHECKSUM_UNNECESSARY is applicable to following protocols:
59 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
60 * zero UDP checksum for either IPv4 or IPv6, the networking stack
61 * may perform further validation in this case.
62 * GRE: only if the checksum is present in the header.
63 * SCTP: indicates the CRC in SCTP header has been validated.
65 * skb->csum_level indicates the number of consecutive checksums found in
66 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
67 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
68 * and a device is able to verify the checksums for UDP (possibly zero),
69 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
70 * two. If the device were only able to verify the UDP checksum and not
71 * GRE, either because it doesn't support GRE checksum of because GRE
72 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
73 * not considered in this case).
77 * This is the most generic way. The device supplied checksum of the _whole_
78 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
79 * hardware doesn't need to parse L3/L4 headers to implement this.
81 * Note: Even if device supports only some protocols, but is able to produce
82 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
86 * A checksum is set up to be offloaded to a device as described in the
87 * output description for CHECKSUM_PARTIAL. This may occur on a packet
88 * received directly from another Linux OS, e.g., a virtualized Linux kernel
89 * on the same host, or it may be set in the input path in GRO or remote
90 * checksum offload. For the purposes of checksum verification, the checksum
91 * referred to by skb->csum_start + skb->csum_offset and any preceding
92 * checksums in the packet are considered verified. Any checksums in the
93 * packet that are after the checksum being offloaded are not considered to
96 * B. Checksumming on output.
100 * The skb was already checksummed by the protocol, or a checksum is not
105 * The device is required to checksum the packet as seen by hard_start_xmit()
106 * from skb->csum_start up to the end, and to record/write the checksum at
107 * offset skb->csum_start + skb->csum_offset.
109 * The device must show its capabilities in dev->features, set up at device
110 * setup time, e.g. netdev_features.h:
112 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
113 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
114 * IPv4. Sigh. Vendors like this way for an unknown reason.
115 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
116 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
117 * NETIF_F_... - Well, you get the picture.
119 * CHECKSUM_UNNECESSARY:
121 * Normally, the device will do per protocol specific checksumming. Protocol
122 * implementations that do not want the NIC to perform the checksum
123 * calculation should use this flag in their outgoing skbs.
125 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
126 * offload. Correspondingly, the FCoE protocol driver
127 * stack should use CHECKSUM_UNNECESSARY.
129 * Any questions? No questions, good. --ANK
132 /* Don't change this without changing skb_csum_unnecessary! */
133 #define CHECKSUM_NONE 0
134 #define CHECKSUM_UNNECESSARY 1
135 #define CHECKSUM_COMPLETE 2
136 #define CHECKSUM_PARTIAL 3
138 /* Maximum value in skb->csum_level */
139 #define SKB_MAX_CSUM_LEVEL 3
141 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
142 #define SKB_WITH_OVERHEAD(X) \
143 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
144 #define SKB_MAX_ORDER(X, ORDER) \
145 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
146 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
147 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
149 /* return minimum truesize of one skb containing X bytes of data */
150 #define SKB_TRUESIZE(X) ((X) + \
151 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
152 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
156 struct pipe_inode_info;
160 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
161 struct nf_conntrack {
166 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
167 struct nf_bridge_info {
170 struct net_device *physindev;
171 struct net_device *physoutdev;
172 unsigned long data[32 / sizeof(unsigned long)];
176 struct sk_buff_head {
177 /* These two members must be first. */
178 struct sk_buff *next;
179 struct sk_buff *prev;
187 /* To allow 64K frame to be packed as single skb without frag_list we
188 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
189 * buffers which do not start on a page boundary.
191 * Since GRO uses frags we allocate at least 16 regardless of page
194 #if (65536/PAGE_SIZE + 1) < 16
195 #define MAX_SKB_FRAGS 16UL
197 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
200 typedef struct skb_frag_struct skb_frag_t;
202 struct skb_frag_struct {
206 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
215 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
220 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
225 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
230 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
235 #define HAVE_HW_TIME_STAMP
238 * struct skb_shared_hwtstamps - hardware time stamps
239 * @hwtstamp: hardware time stamp transformed into duration
240 * since arbitrary point in time
242 * Software time stamps generated by ktime_get_real() are stored in
245 * hwtstamps can only be compared against other hwtstamps from
248 * This structure is attached to packets as part of the
249 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
251 struct skb_shared_hwtstamps {
255 /* Definitions for tx_flags in struct skb_shared_info */
257 /* generate hardware time stamp */
258 SKBTX_HW_TSTAMP = 1 << 0,
260 /* generate software time stamp when queueing packet to NIC */
261 SKBTX_SW_TSTAMP = 1 << 1,
263 /* device driver is going to provide hardware time stamp */
264 SKBTX_IN_PROGRESS = 1 << 2,
266 /* device driver supports TX zero-copy buffers */
267 SKBTX_DEV_ZEROCOPY = 1 << 3,
269 /* generate wifi status information (where possible) */
270 SKBTX_WIFI_STATUS = 1 << 4,
272 /* This indicates at least one fragment might be overwritten
273 * (as in vmsplice(), sendfile() ...)
274 * If we need to compute a TX checksum, we'll need to copy
275 * all frags to avoid possible bad checksum
277 SKBTX_SHARED_FRAG = 1 << 5,
279 /* generate software time stamp when entering packet scheduling */
280 SKBTX_SCHED_TSTAMP = 1 << 6,
282 /* generate software timestamp on peer data acknowledgment */
283 SKBTX_ACK_TSTAMP = 1 << 7,
286 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
287 SKBTX_SCHED_TSTAMP | \
289 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
292 * The callback notifies userspace to release buffers when skb DMA is done in
293 * lower device, the skb last reference should be 0 when calling this.
294 * The zerocopy_success argument is true if zero copy transmit occurred,
295 * false on data copy or out of memory error caused by data copy attempt.
296 * The ctx field is used to track device context.
297 * The desc field is used to track userspace buffer index.
300 void (*callback)(struct ubuf_info *, bool zerocopy_success);
305 /* This data is invariant across clones and lives at
306 * the end of the header data, ie. at skb->end.
308 struct skb_shared_info {
309 unsigned char nr_frags;
311 unsigned short gso_size;
312 /* Warning: this field is not always filled in (UFO)! */
313 unsigned short gso_segs;
314 unsigned short gso_type;
315 struct sk_buff *frag_list;
316 struct skb_shared_hwtstamps hwtstamps;
321 * Warning : all fields before dataref are cleared in __alloc_skb()
325 /* Intermediate layers must ensure that destructor_arg
326 * remains valid until skb destructor */
327 void * destructor_arg;
329 /* must be last field, see pskb_expand_head() */
330 skb_frag_t frags[MAX_SKB_FRAGS];
333 /* We divide dataref into two halves. The higher 16 bits hold references
334 * to the payload part of skb->data. The lower 16 bits hold references to
335 * the entire skb->data. A clone of a headerless skb holds the length of
336 * the header in skb->hdr_len.
338 * All users must obey the rule that the skb->data reference count must be
339 * greater than or equal to the payload reference count.
341 * Holding a reference to the payload part means that the user does not
342 * care about modifications to the header part of skb->data.
344 #define SKB_DATAREF_SHIFT 16
345 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
349 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
350 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
351 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
355 SKB_GSO_TCPV4 = 1 << 0,
356 SKB_GSO_UDP = 1 << 1,
358 /* This indicates the skb is from an untrusted source. */
359 SKB_GSO_DODGY = 1 << 2,
361 /* This indicates the tcp segment has CWR set. */
362 SKB_GSO_TCP_ECN = 1 << 3,
364 SKB_GSO_TCPV6 = 1 << 4,
366 SKB_GSO_FCOE = 1 << 5,
368 SKB_GSO_GRE = 1 << 6,
370 SKB_GSO_GRE_CSUM = 1 << 7,
372 SKB_GSO_IPIP = 1 << 8,
374 SKB_GSO_SIT = 1 << 9,
376 SKB_GSO_UDP_TUNNEL = 1 << 10,
378 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
380 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
383 #if BITS_PER_LONG > 32
384 #define NET_SKBUFF_DATA_USES_OFFSET 1
387 #ifdef NET_SKBUFF_DATA_USES_OFFSET
388 typedef unsigned int sk_buff_data_t;
390 typedef unsigned char *sk_buff_data_t;
394 * struct skb_mstamp - multi resolution time stamps
395 * @stamp_us: timestamp in us resolution
396 * @stamp_jiffies: timestamp in jiffies
409 * skb_mstamp_get - get current timestamp
410 * @cl: place to store timestamps
412 static inline void skb_mstamp_get(struct skb_mstamp *cl)
414 u64 val = local_clock();
416 do_div(val, NSEC_PER_USEC);
417 cl->stamp_us = (u32)val;
418 cl->stamp_jiffies = (u32)jiffies;
422 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
423 * @t1: pointer to newest sample
424 * @t0: pointer to oldest sample
426 static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
427 const struct skb_mstamp *t0)
429 s32 delta_us = t1->stamp_us - t0->stamp_us;
430 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
432 /* If delta_us is negative, this might be because interval is too big,
433 * or local_clock() drift is too big : fallback using jiffies.
436 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
438 delta_us = jiffies_to_usecs(delta_jiffies);
445 * struct sk_buff - socket buffer
446 * @next: Next buffer in list
447 * @prev: Previous buffer in list
448 * @tstamp: Time we arrived/left
449 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
450 * @sk: Socket we are owned by
451 * @dev: Device we arrived on/are leaving by
452 * @cb: Control buffer. Free for use by every layer. Put private vars here
453 * @_skb_refdst: destination entry (with norefcount bit)
454 * @sp: the security path, used for xfrm
455 * @len: Length of actual data
456 * @data_len: Data length
457 * @mac_len: Length of link layer header
458 * @hdr_len: writable header length of cloned skb
459 * @csum: Checksum (must include start/offset pair)
460 * @csum_start: Offset from skb->head where checksumming should start
461 * @csum_offset: Offset from csum_start where checksum should be stored
462 * @priority: Packet queueing priority
463 * @ignore_df: allow local fragmentation
464 * @cloned: Head may be cloned (check refcnt to be sure)
465 * @ip_summed: Driver fed us an IP checksum
466 * @nohdr: Payload reference only, must not modify header
467 * @nfctinfo: Relationship of this skb to the connection
468 * @pkt_type: Packet class
469 * @fclone: skbuff clone status
470 * @ipvs_property: skbuff is owned by ipvs
471 * @peeked: this packet has been seen already, so stats have been
472 * done for it, don't do them again
473 * @nf_trace: netfilter packet trace flag
474 * @protocol: Packet protocol from driver
475 * @destructor: Destruct function
476 * @nfct: Associated connection, if any
477 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
478 * @skb_iif: ifindex of device we arrived on
479 * @tc_index: Traffic control index
480 * @tc_verd: traffic control verdict
481 * @hash: the packet hash
482 * @queue_mapping: Queue mapping for multiqueue devices
483 * @xmit_more: More SKBs are pending for this queue
484 * @ndisc_nodetype: router type (from link layer)
485 * @ooo_okay: allow the mapping of a socket to a queue to be changed
486 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
488 * @sw_hash: indicates hash was computed in software stack
489 * @wifi_acked_valid: wifi_acked was set
490 * @wifi_acked: whether frame was acked on wifi or not
491 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
492 * @napi_id: id of the NAPI struct this skb came from
493 * @secmark: security marking
494 * @mark: Generic packet mark
495 * @dropcount: total number of sk_receive_queue overflows
496 * @vlan_proto: vlan encapsulation protocol
497 * @vlan_tci: vlan tag control information
498 * @inner_protocol: Protocol (encapsulation)
499 * @inner_transport_header: Inner transport layer header (encapsulation)
500 * @inner_network_header: Network layer header (encapsulation)
501 * @inner_mac_header: Link layer header (encapsulation)
502 * @transport_header: Transport layer header
503 * @network_header: Network layer header
504 * @mac_header: Link layer header
505 * @tail: Tail pointer
507 * @head: Head of buffer
508 * @data: Data head pointer
509 * @truesize: Buffer size
510 * @users: User count - see {datagram,tcp}.c
516 /* These two members must be first. */
517 struct sk_buff *next;
518 struct sk_buff *prev;
522 struct skb_mstamp skb_mstamp;
525 struct rb_node rbnode; /* used in netem & tcp stack */
528 struct net_device *dev;
531 * This is the control buffer. It is free to use for every
532 * layer. Please put your private variables there. If you
533 * want to keep them across layers you have to do a skb_clone()
534 * first. This is owned by whoever has the skb queued ATM.
536 char cb[48] __aligned(8);
538 unsigned long _skb_refdst;
539 void (*destructor)(struct sk_buff *skb);
543 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
544 struct nf_conntrack *nfct;
546 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
547 struct nf_bridge_info *nf_bridge;
554 /* Following fields are _not_ copied in __copy_skb_header()
555 * Note that queue_mapping is here mostly to fill a hole.
557 kmemcheck_bitfield_begin(flags1);
566 kmemcheck_bitfield_end(flags1);
568 /* fields enclosed in headers_start/headers_end are copied
569 * using a single memcpy() in __copy_skb_header()
572 __u32 headers_start[0];
575 /* if you move pkt_type around you also must adapt those constants */
576 #ifdef __BIG_ENDIAN_BITFIELD
577 #define PKT_TYPE_MAX (7 << 5)
579 #define PKT_TYPE_MAX 7
581 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
583 __u8 __pkt_type_offset[0];
594 __u8 wifi_acked_valid:1;
598 /* Indicates the inner headers are valid in the skbuff. */
599 __u8 encapsulation:1;
600 __u8 encap_hdr_csum:1;
602 __u8 csum_complete_sw:1;
606 #ifdef CONFIG_IPV6_NDISC_NODETYPE
607 __u8 ndisc_nodetype:2;
609 __u8 ipvs_property:1;
610 __u8 inner_protocol_type:1;
611 __u8 remcsum_offload:1;
612 /* 3 or 5 bit hole */
614 #ifdef CONFIG_NET_SCHED
615 __u16 tc_index; /* traffic control index */
616 #ifdef CONFIG_NET_CLS_ACT
617 __u16 tc_verd; /* traffic control verdict */
633 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
635 unsigned int napi_id;
636 unsigned int sender_cpu;
639 #ifdef CONFIG_NETWORK_SECMARK
645 __u32 reserved_tailroom;
649 __be16 inner_protocol;
653 __u16 inner_transport_header;
654 __u16 inner_network_header;
655 __u16 inner_mac_header;
658 __u16 transport_header;
659 __u16 network_header;
663 __u32 headers_end[0];
666 /* These elements must be at the end, see alloc_skb() for details. */
671 unsigned int truesize;
677 * Handling routines are only of interest to the kernel
679 #include <linux/slab.h>
682 #define SKB_ALLOC_FCLONE 0x01
683 #define SKB_ALLOC_RX 0x02
684 #define SKB_ALLOC_NAPI 0x04
686 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
687 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
689 return unlikely(skb->pfmemalloc);
693 * skb might have a dst pointer attached, refcounted or not.
694 * _skb_refdst low order bit is set if refcount was _not_ taken
696 #define SKB_DST_NOREF 1UL
697 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
700 * skb_dst - returns skb dst_entry
703 * Returns skb dst_entry, regardless of reference taken or not.
705 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
707 /* If refdst was not refcounted, check we still are in a
708 * rcu_read_lock section
710 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
711 !rcu_read_lock_held() &&
712 !rcu_read_lock_bh_held());
713 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
717 * skb_dst_set - sets skb dst
721 * Sets skb dst, assuming a reference was taken on dst and should
722 * be released by skb_dst_drop()
724 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
726 skb->_skb_refdst = (unsigned long)dst;
730 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
734 * Sets skb dst, assuming a reference was not taken on dst.
735 * If dst entry is cached, we do not take reference and dst_release
736 * will be avoided by refdst_drop. If dst entry is not cached, we take
737 * reference, so that last dst_release can destroy the dst immediately.
739 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
741 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
742 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
746 * skb_dst_is_noref - Test if skb dst isn't refcounted
749 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
751 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
754 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
756 return (struct rtable *)skb_dst(skb);
759 void kfree_skb(struct sk_buff *skb);
760 void kfree_skb_list(struct sk_buff *segs);
761 void skb_tx_error(struct sk_buff *skb);
762 void consume_skb(struct sk_buff *skb);
763 void __kfree_skb(struct sk_buff *skb);
764 extern struct kmem_cache *skbuff_head_cache;
766 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
767 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
768 bool *fragstolen, int *delta_truesize);
770 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
772 struct sk_buff *build_skb(void *data, unsigned int frag_size);
773 static inline struct sk_buff *alloc_skb(unsigned int size,
776 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
779 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
780 unsigned long data_len,
785 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
786 struct sk_buff_fclones {
795 * skb_fclone_busy - check if fclone is busy
798 * Returns true is skb is a fast clone, and its clone is not freed.
799 * Some drivers call skb_orphan() in their ndo_start_xmit(),
800 * so we also check that this didnt happen.
802 static inline bool skb_fclone_busy(const struct sock *sk,
803 const struct sk_buff *skb)
805 const struct sk_buff_fclones *fclones;
807 fclones = container_of(skb, struct sk_buff_fclones, skb1);
809 return skb->fclone == SKB_FCLONE_ORIG &&
810 atomic_read(&fclones->fclone_ref) > 1 &&
811 fclones->skb2.sk == sk;
814 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
817 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
820 struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
821 static inline struct sk_buff *alloc_skb_head(gfp_t priority)
823 return __alloc_skb_head(priority, -1);
826 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
827 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
828 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
829 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
830 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
831 gfp_t gfp_mask, bool fclone);
832 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
835 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
838 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
839 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
840 unsigned int headroom);
841 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
842 int newtailroom, gfp_t priority);
843 int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
844 int offset, int len);
845 int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
847 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
848 int skb_pad(struct sk_buff *skb, int pad);
849 #define dev_kfree_skb(a) consume_skb(a)
851 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
852 int getfrag(void *from, char *to, int offset,
853 int len, int odd, struct sk_buff *skb),
854 void *from, int length);
856 struct skb_seq_state {
860 __u32 stepped_offset;
861 struct sk_buff *root_skb;
862 struct sk_buff *cur_skb;
866 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
867 unsigned int to, struct skb_seq_state *st);
868 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
869 struct skb_seq_state *st);
870 void skb_abort_seq_read(struct skb_seq_state *st);
872 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
873 unsigned int to, struct ts_config *config,
874 struct ts_state *state);
877 * Packet hash types specify the type of hash in skb_set_hash.
879 * Hash types refer to the protocol layer addresses which are used to
880 * construct a packet's hash. The hashes are used to differentiate or identify
881 * flows of the protocol layer for the hash type. Hash types are either
882 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
884 * Properties of hashes:
886 * 1) Two packets in different flows have different hash values
887 * 2) Two packets in the same flow should have the same hash value
889 * A hash at a higher layer is considered to be more specific. A driver should
890 * set the most specific hash possible.
892 * A driver cannot indicate a more specific hash than the layer at which a hash
893 * was computed. For instance an L3 hash cannot be set as an L4 hash.
895 * A driver may indicate a hash level which is less specific than the
896 * actual layer the hash was computed on. For instance, a hash computed
897 * at L4 may be considered an L3 hash. This should only be done if the
898 * driver can't unambiguously determine that the HW computed the hash at
899 * the higher layer. Note that the "should" in the second property above
902 enum pkt_hash_types {
903 PKT_HASH_TYPE_NONE, /* Undefined type */
904 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
905 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
906 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
910 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
912 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
917 void __skb_get_hash(struct sk_buff *skb);
918 static inline __u32 skb_get_hash(struct sk_buff *skb)
920 if (!skb->l4_hash && !skb->sw_hash)
926 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
931 static inline void skb_clear_hash(struct sk_buff *skb)
938 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
944 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
946 to->hash = from->hash;
947 to->sw_hash = from->sw_hash;
948 to->l4_hash = from->l4_hash;
951 static inline void skb_sender_cpu_clear(struct sk_buff *skb)
958 #ifdef NET_SKBUFF_DATA_USES_OFFSET
959 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
961 return skb->head + skb->end;
964 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
969 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
974 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
976 return skb->end - skb->head;
981 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
983 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
985 return &skb_shinfo(skb)->hwtstamps;
989 * skb_queue_empty - check if a queue is empty
992 * Returns true if the queue is empty, false otherwise.
994 static inline int skb_queue_empty(const struct sk_buff_head *list)
996 return list->next == (const struct sk_buff *) list;
1000 * skb_queue_is_last - check if skb is the last entry in the queue
1004 * Returns true if @skb is the last buffer on the list.
1006 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1007 const struct sk_buff *skb)
1009 return skb->next == (const struct sk_buff *) list;
1013 * skb_queue_is_first - check if skb is the first entry in the queue
1017 * Returns true if @skb is the first buffer on the list.
1019 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1020 const struct sk_buff *skb)
1022 return skb->prev == (const struct sk_buff *) list;
1026 * skb_queue_next - return the next packet in the queue
1028 * @skb: current buffer
1030 * Return the next packet in @list after @skb. It is only valid to
1031 * call this if skb_queue_is_last() evaluates to false.
1033 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1034 const struct sk_buff *skb)
1036 /* This BUG_ON may seem severe, but if we just return then we
1037 * are going to dereference garbage.
1039 BUG_ON(skb_queue_is_last(list, skb));
1044 * skb_queue_prev - return the prev packet in the queue
1046 * @skb: current buffer
1048 * Return the prev packet in @list before @skb. It is only valid to
1049 * call this if skb_queue_is_first() evaluates to false.
1051 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1052 const struct sk_buff *skb)
1054 /* This BUG_ON may seem severe, but if we just return then we
1055 * are going to dereference garbage.
1057 BUG_ON(skb_queue_is_first(list, skb));
1062 * skb_get - reference buffer
1063 * @skb: buffer to reference
1065 * Makes another reference to a socket buffer and returns a pointer
1068 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1070 atomic_inc(&skb->users);
1075 * If users == 1, we are the only owner and are can avoid redundant
1080 * skb_cloned - is the buffer a clone
1081 * @skb: buffer to check
1083 * Returns true if the buffer was generated with skb_clone() and is
1084 * one of multiple shared copies of the buffer. Cloned buffers are
1085 * shared data so must not be written to under normal circumstances.
1087 static inline int skb_cloned(const struct sk_buff *skb)
1089 return skb->cloned &&
1090 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1093 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1095 might_sleep_if(pri & __GFP_WAIT);
1097 if (skb_cloned(skb))
1098 return pskb_expand_head(skb, 0, 0, pri);
1104 * skb_header_cloned - is the header a clone
1105 * @skb: buffer to check
1107 * Returns true if modifying the header part of the buffer requires
1108 * the data to be copied.
1110 static inline int skb_header_cloned(const struct sk_buff *skb)
1117 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1118 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1119 return dataref != 1;
1123 * skb_header_release - release reference to header
1124 * @skb: buffer to operate on
1126 * Drop a reference to the header part of the buffer. This is done
1127 * by acquiring a payload reference. You must not read from the header
1128 * part of skb->data after this.
1129 * Note : Check if you can use __skb_header_release() instead.
1131 static inline void skb_header_release(struct sk_buff *skb)
1135 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1139 * __skb_header_release - release reference to header
1140 * @skb: buffer to operate on
1142 * Variant of skb_header_release() assuming skb is private to caller.
1143 * We can avoid one atomic operation.
1145 static inline void __skb_header_release(struct sk_buff *skb)
1148 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1153 * skb_shared - is the buffer shared
1154 * @skb: buffer to check
1156 * Returns true if more than one person has a reference to this
1159 static inline int skb_shared(const struct sk_buff *skb)
1161 return atomic_read(&skb->users) != 1;
1165 * skb_share_check - check if buffer is shared and if so clone it
1166 * @skb: buffer to check
1167 * @pri: priority for memory allocation
1169 * If the buffer is shared the buffer is cloned and the old copy
1170 * drops a reference. A new clone with a single reference is returned.
1171 * If the buffer is not shared the original buffer is returned. When
1172 * being called from interrupt status or with spinlocks held pri must
1175 * NULL is returned on a memory allocation failure.
1177 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1179 might_sleep_if(pri & __GFP_WAIT);
1180 if (skb_shared(skb)) {
1181 struct sk_buff *nskb = skb_clone(skb, pri);
1193 * Copy shared buffers into a new sk_buff. We effectively do COW on
1194 * packets to handle cases where we have a local reader and forward
1195 * and a couple of other messy ones. The normal one is tcpdumping
1196 * a packet thats being forwarded.
1200 * skb_unshare - make a copy of a shared buffer
1201 * @skb: buffer to check
1202 * @pri: priority for memory allocation
1204 * If the socket buffer is a clone then this function creates a new
1205 * copy of the data, drops a reference count on the old copy and returns
1206 * the new copy with the reference count at 1. If the buffer is not a clone
1207 * the original buffer is returned. When called with a spinlock held or
1208 * from interrupt state @pri must be %GFP_ATOMIC
1210 * %NULL is returned on a memory allocation failure.
1212 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1215 might_sleep_if(pri & __GFP_WAIT);
1216 if (skb_cloned(skb)) {
1217 struct sk_buff *nskb = skb_copy(skb, pri);
1219 /* Free our shared copy */
1230 * skb_peek - peek at the head of an &sk_buff_head
1231 * @list_: list to peek at
1233 * Peek an &sk_buff. Unlike most other operations you _MUST_
1234 * be careful with this one. A peek leaves the buffer on the
1235 * list and someone else may run off with it. You must hold
1236 * the appropriate locks or have a private queue to do this.
1238 * Returns %NULL for an empty list or a pointer to the head element.
1239 * The reference count is not incremented and the reference is therefore
1240 * volatile. Use with caution.
1242 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1244 struct sk_buff *skb = list_->next;
1246 if (skb == (struct sk_buff *)list_)
1252 * skb_peek_next - peek skb following the given one from a queue
1253 * @skb: skb to start from
1254 * @list_: list to peek at
1256 * Returns %NULL when the end of the list is met or a pointer to the
1257 * next element. The reference count is not incremented and the
1258 * reference is therefore volatile. Use with caution.
1260 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1261 const struct sk_buff_head *list_)
1263 struct sk_buff *next = skb->next;
1265 if (next == (struct sk_buff *)list_)
1271 * skb_peek_tail - peek at the tail of an &sk_buff_head
1272 * @list_: list to peek at
1274 * Peek an &sk_buff. Unlike most other operations you _MUST_
1275 * be careful with this one. A peek leaves the buffer on the
1276 * list and someone else may run off with it. You must hold
1277 * the appropriate locks or have a private queue to do this.
1279 * Returns %NULL for an empty list or a pointer to the tail element.
1280 * The reference count is not incremented and the reference is therefore
1281 * volatile. Use with caution.
1283 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1285 struct sk_buff *skb = list_->prev;
1287 if (skb == (struct sk_buff *)list_)
1294 * skb_queue_len - get queue length
1295 * @list_: list to measure
1297 * Return the length of an &sk_buff queue.
1299 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1305 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1306 * @list: queue to initialize
1308 * This initializes only the list and queue length aspects of
1309 * an sk_buff_head object. This allows to initialize the list
1310 * aspects of an sk_buff_head without reinitializing things like
1311 * the spinlock. It can also be used for on-stack sk_buff_head
1312 * objects where the spinlock is known to not be used.
1314 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1316 list->prev = list->next = (struct sk_buff *)list;
1321 * This function creates a split out lock class for each invocation;
1322 * this is needed for now since a whole lot of users of the skb-queue
1323 * infrastructure in drivers have different locking usage (in hardirq)
1324 * than the networking core (in softirq only). In the long run either the
1325 * network layer or drivers should need annotation to consolidate the
1326 * main types of usage into 3 classes.
1328 static inline void skb_queue_head_init(struct sk_buff_head *list)
1330 spin_lock_init(&list->lock);
1331 __skb_queue_head_init(list);
1334 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1335 struct lock_class_key *class)
1337 skb_queue_head_init(list);
1338 lockdep_set_class(&list->lock, class);
1342 * Insert an sk_buff on a list.
1344 * The "__skb_xxxx()" functions are the non-atomic ones that
1345 * can only be called with interrupts disabled.
1347 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1348 struct sk_buff_head *list);
1349 static inline void __skb_insert(struct sk_buff *newsk,
1350 struct sk_buff *prev, struct sk_buff *next,
1351 struct sk_buff_head *list)
1355 next->prev = prev->next = newsk;
1359 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1360 struct sk_buff *prev,
1361 struct sk_buff *next)
1363 struct sk_buff *first = list->next;
1364 struct sk_buff *last = list->prev;
1374 * skb_queue_splice - join two skb lists, this is designed for stacks
1375 * @list: the new list to add
1376 * @head: the place to add it in the first list
1378 static inline void skb_queue_splice(const 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;
1388 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1389 * @list: the new list to add
1390 * @head: the place to add it in the first list
1392 * The list at @list is reinitialised
1394 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1395 struct sk_buff_head *head)
1397 if (!skb_queue_empty(list)) {
1398 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1399 head->qlen += list->qlen;
1400 __skb_queue_head_init(list);
1405 * skb_queue_splice_tail - join two skb lists, each list being a queue
1406 * @list: the new list to add
1407 * @head: the place to add it in the first list
1409 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1410 struct sk_buff_head *head)
1412 if (!skb_queue_empty(list)) {
1413 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1414 head->qlen += list->qlen;
1419 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1420 * @list: the new list to add
1421 * @head: the place to add it in the first list
1423 * Each of the lists is a queue.
1424 * The list at @list is reinitialised
1426 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1427 struct sk_buff_head *head)
1429 if (!skb_queue_empty(list)) {
1430 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1431 head->qlen += list->qlen;
1432 __skb_queue_head_init(list);
1437 * __skb_queue_after - queue a buffer at the list head
1438 * @list: list to use
1439 * @prev: place after this buffer
1440 * @newsk: buffer to queue
1442 * Queue a buffer int the middle of a list. This function takes no locks
1443 * and you must therefore hold required locks before calling it.
1445 * A buffer cannot be placed on two lists at the same time.
1447 static inline void __skb_queue_after(struct sk_buff_head *list,
1448 struct sk_buff *prev,
1449 struct sk_buff *newsk)
1451 __skb_insert(newsk, prev, prev->next, list);
1454 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1455 struct sk_buff_head *list);
1457 static inline void __skb_queue_before(struct sk_buff_head *list,
1458 struct sk_buff *next,
1459 struct sk_buff *newsk)
1461 __skb_insert(newsk, next->prev, next, list);
1465 * __skb_queue_head - queue a buffer at the list head
1466 * @list: list to use
1467 * @newsk: buffer to queue
1469 * Queue a buffer at the start of a list. This function takes no locks
1470 * and you must therefore hold required locks before calling it.
1472 * A buffer cannot be placed on two lists at the same time.
1474 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1475 static inline void __skb_queue_head(struct sk_buff_head *list,
1476 struct sk_buff *newsk)
1478 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1482 * __skb_queue_tail - queue a buffer at the list tail
1483 * @list: list to use
1484 * @newsk: buffer to queue
1486 * Queue a buffer at the end of a list. This function takes no locks
1487 * and you must therefore hold required locks before calling it.
1489 * A buffer cannot be placed on two lists at the same time.
1491 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1492 static inline void __skb_queue_tail(struct sk_buff_head *list,
1493 struct sk_buff *newsk)
1495 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1499 * remove sk_buff from list. _Must_ be called atomically, and with
1502 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1503 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1505 struct sk_buff *next, *prev;
1510 skb->next = skb->prev = NULL;
1516 * __skb_dequeue - remove from the head of the queue
1517 * @list: list to dequeue from
1519 * Remove the head of the list. This function does not take any locks
1520 * so must be used with appropriate locks held only. The head item is
1521 * returned or %NULL if the list is empty.
1523 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1524 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1526 struct sk_buff *skb = skb_peek(list);
1528 __skb_unlink(skb, list);
1533 * __skb_dequeue_tail - remove from the tail of the queue
1534 * @list: list to dequeue from
1536 * Remove the tail of the list. This function does not take any locks
1537 * so must be used with appropriate locks held only. The tail item is
1538 * returned or %NULL if the list is empty.
1540 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1541 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1543 struct sk_buff *skb = skb_peek_tail(list);
1545 __skb_unlink(skb, list);
1550 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1552 return skb->data_len;
1555 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1557 return skb->len - skb->data_len;
1560 static inline int skb_pagelen(const struct sk_buff *skb)
1564 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1565 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1566 return len + skb_headlen(skb);
1570 * __skb_fill_page_desc - initialise a paged fragment in an skb
1571 * @skb: buffer containing fragment to be initialised
1572 * @i: paged fragment index to initialise
1573 * @page: the page to use for this fragment
1574 * @off: the offset to the data with @page
1575 * @size: the length of the data
1577 * Initialises the @i'th fragment of @skb to point to &size bytes at
1578 * offset @off within @page.
1580 * Does not take any additional reference on the fragment.
1582 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1583 struct page *page, int off, int size)
1585 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1588 * Propagate page->pfmemalloc to the skb if we can. The problem is
1589 * that not all callers have unique ownership of the page. If
1590 * pfmemalloc is set, we check the mapping as a mapping implies
1591 * page->index is set (index and pfmemalloc share space).
1592 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1593 * do not lose pfmemalloc information as the pages would not be
1594 * allocated using __GFP_MEMALLOC.
1596 frag->page.p = page;
1597 frag->page_offset = off;
1598 skb_frag_size_set(frag, size);
1600 page = compound_head(page);
1601 if (page->pfmemalloc && !page->mapping)
1602 skb->pfmemalloc = true;
1606 * skb_fill_page_desc - initialise a paged fragment in an skb
1607 * @skb: buffer containing fragment to be initialised
1608 * @i: paged fragment index to initialise
1609 * @page: the page to use for this fragment
1610 * @off: the offset to the data with @page
1611 * @size: the length of the data
1613 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1614 * @skb to point to @size bytes at offset @off within @page. In
1615 * addition updates @skb such that @i is the last fragment.
1617 * Does not take any additional reference on the fragment.
1619 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1620 struct page *page, int off, int size)
1622 __skb_fill_page_desc(skb, i, page, off, size);
1623 skb_shinfo(skb)->nr_frags = i + 1;
1626 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1627 int size, unsigned int truesize);
1629 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1630 unsigned int truesize);
1632 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1633 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1634 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1636 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1637 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1639 return skb->head + skb->tail;
1642 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1644 skb->tail = skb->data - skb->head;
1647 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1649 skb_reset_tail_pointer(skb);
1650 skb->tail += offset;
1653 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1654 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1659 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1661 skb->tail = skb->data;
1664 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1666 skb->tail = skb->data + offset;
1669 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1672 * Add data to an sk_buff
1674 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1675 unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1676 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1678 unsigned char *tmp = skb_tail_pointer(skb);
1679 SKB_LINEAR_ASSERT(skb);
1685 unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1686 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1693 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1694 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1697 BUG_ON(skb->len < skb->data_len);
1698 return skb->data += len;
1701 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1703 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1706 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1708 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1710 if (len > skb_headlen(skb) &&
1711 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1714 return skb->data += len;
1717 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1719 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1722 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1724 if (likely(len <= skb_headlen(skb)))
1726 if (unlikely(len > skb->len))
1728 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1732 * skb_headroom - bytes at buffer head
1733 * @skb: buffer to check
1735 * Return the number of bytes of free space at the head of an &sk_buff.
1737 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1739 return skb->data - skb->head;
1743 * skb_tailroom - bytes at buffer end
1744 * @skb: buffer to check
1746 * Return the number of bytes of free space at the tail of an sk_buff
1748 static inline int skb_tailroom(const struct sk_buff *skb)
1750 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1754 * skb_availroom - bytes at buffer end
1755 * @skb: buffer to check
1757 * Return the number of bytes of free space at the tail of an sk_buff
1758 * allocated by sk_stream_alloc()
1760 static inline int skb_availroom(const struct sk_buff *skb)
1762 if (skb_is_nonlinear(skb))
1765 return skb->end - skb->tail - skb->reserved_tailroom;
1769 * skb_reserve - adjust headroom
1770 * @skb: buffer to alter
1771 * @len: bytes to move
1773 * Increase the headroom of an empty &sk_buff by reducing the tail
1774 * room. This is only allowed for an empty buffer.
1776 static inline void skb_reserve(struct sk_buff *skb, int len)
1782 #define ENCAP_TYPE_ETHER 0
1783 #define ENCAP_TYPE_IPPROTO 1
1785 static inline void skb_set_inner_protocol(struct sk_buff *skb,
1788 skb->inner_protocol = protocol;
1789 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1792 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1795 skb->inner_ipproto = ipproto;
1796 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1799 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1801 skb->inner_mac_header = skb->mac_header;
1802 skb->inner_network_header = skb->network_header;
1803 skb->inner_transport_header = skb->transport_header;
1806 static inline void skb_reset_mac_len(struct sk_buff *skb)
1808 skb->mac_len = skb->network_header - skb->mac_header;
1811 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1814 return skb->head + skb->inner_transport_header;
1817 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1819 skb->inner_transport_header = skb->data - skb->head;
1822 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1825 skb_reset_inner_transport_header(skb);
1826 skb->inner_transport_header += offset;
1829 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1831 return skb->head + skb->inner_network_header;
1834 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1836 skb->inner_network_header = skb->data - skb->head;
1839 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1842 skb_reset_inner_network_header(skb);
1843 skb->inner_network_header += offset;
1846 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1848 return skb->head + skb->inner_mac_header;
1851 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1853 skb->inner_mac_header = skb->data - skb->head;
1856 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1859 skb_reset_inner_mac_header(skb);
1860 skb->inner_mac_header += offset;
1862 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1864 return skb->transport_header != (typeof(skb->transport_header))~0U;
1867 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1869 return skb->head + skb->transport_header;
1872 static inline void skb_reset_transport_header(struct sk_buff *skb)
1874 skb->transport_header = skb->data - skb->head;
1877 static inline void skb_set_transport_header(struct sk_buff *skb,
1880 skb_reset_transport_header(skb);
1881 skb->transport_header += offset;
1884 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1886 return skb->head + skb->network_header;
1889 static inline void skb_reset_network_header(struct sk_buff *skb)
1891 skb->network_header = skb->data - skb->head;
1894 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1896 skb_reset_network_header(skb);
1897 skb->network_header += offset;
1900 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1902 return skb->head + skb->mac_header;
1905 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1907 return skb->mac_header != (typeof(skb->mac_header))~0U;
1910 static inline void skb_reset_mac_header(struct sk_buff *skb)
1912 skb->mac_header = skb->data - skb->head;
1915 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1917 skb_reset_mac_header(skb);
1918 skb->mac_header += offset;
1921 static inline void skb_pop_mac_header(struct sk_buff *skb)
1923 skb->mac_header = skb->network_header;
1926 static inline void skb_probe_transport_header(struct sk_buff *skb,
1927 const int offset_hint)
1929 struct flow_keys keys;
1931 if (skb_transport_header_was_set(skb))
1933 else if (skb_flow_dissect(skb, &keys))
1934 skb_set_transport_header(skb, keys.thoff);
1936 skb_set_transport_header(skb, offset_hint);
1939 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1941 if (skb_mac_header_was_set(skb)) {
1942 const unsigned char *old_mac = skb_mac_header(skb);
1944 skb_set_mac_header(skb, -skb->mac_len);
1945 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1949 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1951 return skb->csum_start - skb_headroom(skb);
1954 static inline int skb_transport_offset(const struct sk_buff *skb)
1956 return skb_transport_header(skb) - skb->data;
1959 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1961 return skb->transport_header - skb->network_header;
1964 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1966 return skb->inner_transport_header - skb->inner_network_header;
1969 static inline int skb_network_offset(const struct sk_buff *skb)
1971 return skb_network_header(skb) - skb->data;
1974 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1976 return skb_inner_network_header(skb) - skb->data;
1979 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1981 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1985 * CPUs often take a performance hit when accessing unaligned memory
1986 * locations. The actual performance hit varies, it can be small if the
1987 * hardware handles it or large if we have to take an exception and fix it
1990 * Since an ethernet header is 14 bytes network drivers often end up with
1991 * the IP header at an unaligned offset. The IP header can be aligned by
1992 * shifting the start of the packet by 2 bytes. Drivers should do this
1995 * skb_reserve(skb, NET_IP_ALIGN);
1997 * The downside to this alignment of the IP header is that the DMA is now
1998 * unaligned. On some architectures the cost of an unaligned DMA is high
1999 * and this cost outweighs the gains made by aligning the IP header.
2001 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2004 #ifndef NET_IP_ALIGN
2005 #define NET_IP_ALIGN 2
2009 * The networking layer reserves some headroom in skb data (via
2010 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2011 * the header has to grow. In the default case, if the header has to grow
2012 * 32 bytes or less we avoid the reallocation.
2014 * Unfortunately this headroom changes the DMA alignment of the resulting
2015 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2016 * on some architectures. An architecture can override this value,
2017 * perhaps setting it to a cacheline in size (since that will maintain
2018 * cacheline alignment of the DMA). It must be a power of 2.
2020 * Various parts of the networking layer expect at least 32 bytes of
2021 * headroom, you should not reduce this.
2023 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2024 * to reduce average number of cache lines per packet.
2025 * get_rps_cpus() for example only access one 64 bytes aligned block :
2026 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2029 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2032 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2034 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2036 if (unlikely(skb_is_nonlinear(skb))) {
2041 skb_set_tail_pointer(skb, len);
2044 void skb_trim(struct sk_buff *skb, unsigned int len);
2046 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2049 return ___pskb_trim(skb, len);
2050 __skb_trim(skb, len);
2054 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2056 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2060 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2061 * @skb: buffer to alter
2064 * This is identical to pskb_trim except that the caller knows that
2065 * the skb is not cloned so we should never get an error due to out-
2068 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2070 int err = pskb_trim(skb, len);
2075 * skb_orphan - orphan a buffer
2076 * @skb: buffer to orphan
2078 * If a buffer currently has an owner then we call the owner's
2079 * destructor function and make the @skb unowned. The buffer continues
2080 * to exist but is no longer charged to its former owner.
2082 static inline void skb_orphan(struct sk_buff *skb)
2084 if (skb->destructor) {
2085 skb->destructor(skb);
2086 skb->destructor = NULL;
2094 * skb_orphan_frags - orphan the frags contained in a buffer
2095 * @skb: buffer to orphan frags from
2096 * @gfp_mask: allocation mask for replacement pages
2098 * For each frag in the SKB which needs a destructor (i.e. has an
2099 * owner) create a copy of that frag and release the original
2100 * page by calling the destructor.
2102 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2104 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2106 return skb_copy_ubufs(skb, gfp_mask);
2110 * __skb_queue_purge - empty a list
2111 * @list: list to empty
2113 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2114 * the list and one reference dropped. This function does not take the
2115 * list lock and the caller must hold the relevant locks to use it.
2117 void skb_queue_purge(struct sk_buff_head *list);
2118 static inline void __skb_queue_purge(struct sk_buff_head *list)
2120 struct sk_buff *skb;
2121 while ((skb = __skb_dequeue(list)) != NULL)
2125 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2126 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2127 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2129 void *netdev_alloc_frag(unsigned int fragsz);
2131 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2135 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2136 * @dev: network device to receive on
2137 * @length: length to allocate
2139 * Allocate a new &sk_buff and assign it a usage count of one. The
2140 * buffer has unspecified headroom built in. Users should allocate
2141 * the headroom they think they need without accounting for the
2142 * built in space. The built in space is used for optimisations.
2144 * %NULL is returned if there is no free memory. Although this function
2145 * allocates memory it can be called from an interrupt.
2147 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2148 unsigned int length)
2150 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2153 /* legacy helper around __netdev_alloc_skb() */
2154 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2157 return __netdev_alloc_skb(NULL, length, gfp_mask);
2160 /* legacy helper around netdev_alloc_skb() */
2161 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2163 return netdev_alloc_skb(NULL, length);
2167 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2168 unsigned int length, gfp_t gfp)
2170 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2172 if (NET_IP_ALIGN && skb)
2173 skb_reserve(skb, NET_IP_ALIGN);
2177 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2178 unsigned int length)
2180 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2183 void *napi_alloc_frag(unsigned int fragsz);
2184 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2185 unsigned int length, gfp_t gfp_mask);
2186 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2187 unsigned int length)
2189 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2193 * __dev_alloc_pages - allocate page for network Rx
2194 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2195 * @order: size of the allocation
2197 * Allocate a new page.
2199 * %NULL is returned if there is no free memory.
2201 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2204 /* This piece of code contains several assumptions.
2205 * 1. This is for device Rx, therefor a cold page is preferred.
2206 * 2. The expectation is the user wants a compound page.
2207 * 3. If requesting a order 0 page it will not be compound
2208 * due to the check to see if order has a value in prep_new_page
2209 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2210 * code in gfp_to_alloc_flags that should be enforcing this.
2212 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2214 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2217 static inline struct page *dev_alloc_pages(unsigned int order)
2219 return __dev_alloc_pages(GFP_ATOMIC, order);
2223 * __dev_alloc_page - allocate a page for network Rx
2224 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2226 * Allocate a new page.
2228 * %NULL is returned if there is no free memory.
2230 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2232 return __dev_alloc_pages(gfp_mask, 0);
2235 static inline struct page *dev_alloc_page(void)
2237 return __dev_alloc_page(GFP_ATOMIC);
2241 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2242 * @page: The page that was allocated from skb_alloc_page
2243 * @skb: The skb that may need pfmemalloc set
2245 static inline void skb_propagate_pfmemalloc(struct page *page,
2246 struct sk_buff *skb)
2248 if (page && page->pfmemalloc)
2249 skb->pfmemalloc = true;
2253 * skb_frag_page - retrieve the page referred to by a paged fragment
2254 * @frag: the paged fragment
2256 * Returns the &struct page associated with @frag.
2258 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2260 return frag->page.p;
2264 * __skb_frag_ref - take an addition reference on a paged fragment.
2265 * @frag: the paged fragment
2267 * Takes an additional reference on the paged fragment @frag.
2269 static inline void __skb_frag_ref(skb_frag_t *frag)
2271 get_page(skb_frag_page(frag));
2275 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2277 * @f: the fragment offset.
2279 * Takes an additional reference on the @f'th paged fragment of @skb.
2281 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2283 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2287 * __skb_frag_unref - release a reference on a paged fragment.
2288 * @frag: the paged fragment
2290 * Releases a reference on the paged fragment @frag.
2292 static inline void __skb_frag_unref(skb_frag_t *frag)
2294 put_page(skb_frag_page(frag));
2298 * skb_frag_unref - release a reference on a paged fragment of an skb.
2300 * @f: the fragment offset
2302 * Releases a reference on the @f'th paged fragment of @skb.
2304 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2306 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2310 * skb_frag_address - gets the address of the data contained in a paged fragment
2311 * @frag: the paged fragment buffer
2313 * Returns the address of the data within @frag. The page must already
2316 static inline void *skb_frag_address(const skb_frag_t *frag)
2318 return page_address(skb_frag_page(frag)) + frag->page_offset;
2322 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2323 * @frag: the paged fragment buffer
2325 * Returns the address of the data within @frag. Checks that the page
2326 * is mapped and returns %NULL otherwise.
2328 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2330 void *ptr = page_address(skb_frag_page(frag));
2334 return ptr + frag->page_offset;
2338 * __skb_frag_set_page - sets the page contained in a paged fragment
2339 * @frag: the paged fragment
2340 * @page: the page to set
2342 * Sets the fragment @frag to contain @page.
2344 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2346 frag->page.p = page;
2350 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2352 * @f: the fragment offset
2353 * @page: the page to set
2355 * Sets the @f'th fragment of @skb to contain @page.
2357 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2360 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2363 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2366 * skb_frag_dma_map - maps a paged fragment via the DMA API
2367 * @dev: the device to map the fragment to
2368 * @frag: the paged fragment to map
2369 * @offset: the offset within the fragment (starting at the
2370 * fragment's own offset)
2371 * @size: the number of bytes to map
2372 * @dir: the direction of the mapping (%PCI_DMA_*)
2374 * Maps the page associated with @frag to @device.
2376 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2377 const skb_frag_t *frag,
2378 size_t offset, size_t size,
2379 enum dma_data_direction dir)
2381 return dma_map_page(dev, skb_frag_page(frag),
2382 frag->page_offset + offset, size, dir);
2385 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2388 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2392 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2395 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2400 * skb_clone_writable - is the header of a clone writable
2401 * @skb: buffer to check
2402 * @len: length up to which to write
2404 * Returns true if modifying the header part of the cloned buffer
2405 * does not requires the data to be copied.
2407 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2409 return !skb_header_cloned(skb) &&
2410 skb_headroom(skb) + len <= skb->hdr_len;
2413 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2418 if (headroom > skb_headroom(skb))
2419 delta = headroom - skb_headroom(skb);
2421 if (delta || cloned)
2422 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2428 * skb_cow - copy header of skb when it is required
2429 * @skb: buffer to cow
2430 * @headroom: needed headroom
2432 * If the skb passed lacks sufficient headroom or its data part
2433 * is shared, data is reallocated. If reallocation fails, an error
2434 * is returned and original skb is not changed.
2436 * The result is skb with writable area skb->head...skb->tail
2437 * and at least @headroom of space at head.
2439 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2441 return __skb_cow(skb, headroom, skb_cloned(skb));
2445 * skb_cow_head - skb_cow but only making the head writable
2446 * @skb: buffer to cow
2447 * @headroom: needed headroom
2449 * This function is identical to skb_cow except that we replace the
2450 * skb_cloned check by skb_header_cloned. It should be used when
2451 * you only need to push on some header and do not need to modify
2454 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2456 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2460 * skb_padto - pad an skbuff up to a minimal size
2461 * @skb: buffer to pad
2462 * @len: minimal length
2464 * Pads up a buffer to ensure the trailing bytes exist and are
2465 * blanked. If the buffer already contains sufficient data it
2466 * is untouched. Otherwise it is extended. Returns zero on
2467 * success. The skb is freed on error.
2469 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2471 unsigned int size = skb->len;
2472 if (likely(size >= len))
2474 return skb_pad(skb, len - size);
2478 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2479 * @skb: buffer to pad
2480 * @len: minimal length
2482 * Pads up a buffer to ensure the trailing bytes exist and are
2483 * blanked. If the buffer already contains sufficient data it
2484 * is untouched. Otherwise it is extended. Returns zero on
2485 * success. The skb is freed on error.
2487 static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2489 unsigned int size = skb->len;
2491 if (unlikely(size < len)) {
2493 if (skb_pad(skb, len))
2495 __skb_put(skb, len);
2500 static inline int skb_add_data(struct sk_buff *skb,
2501 struct iov_iter *from, int copy)
2503 const int off = skb->len;
2505 if (skb->ip_summed == CHECKSUM_NONE) {
2507 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2508 &csum, from) == copy) {
2509 skb->csum = csum_block_add(skb->csum, csum, off);
2512 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
2515 __skb_trim(skb, off);
2519 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2520 const struct page *page, int off)
2523 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2525 return page == skb_frag_page(frag) &&
2526 off == frag->page_offset + skb_frag_size(frag);
2531 static inline int __skb_linearize(struct sk_buff *skb)
2533 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2537 * skb_linearize - convert paged skb to linear one
2538 * @skb: buffer to linarize
2540 * If there is no free memory -ENOMEM is returned, otherwise zero
2541 * is returned and the old skb data released.
2543 static inline int skb_linearize(struct sk_buff *skb)
2545 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2549 * skb_has_shared_frag - can any frag be overwritten
2550 * @skb: buffer to test
2552 * Return true if the skb has at least one frag that might be modified
2553 * by an external entity (as in vmsplice()/sendfile())
2555 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2557 return skb_is_nonlinear(skb) &&
2558 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2562 * skb_linearize_cow - make sure skb is linear and writable
2563 * @skb: buffer to process
2565 * If there is no free memory -ENOMEM is returned, otherwise zero
2566 * is returned and the old skb data released.
2568 static inline int skb_linearize_cow(struct sk_buff *skb)
2570 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2571 __skb_linearize(skb) : 0;
2575 * skb_postpull_rcsum - update checksum for received skb after pull
2576 * @skb: buffer to update
2577 * @start: start of data before pull
2578 * @len: length of data pulled
2580 * After doing a pull on a received packet, you need to call this to
2581 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2582 * CHECKSUM_NONE so that it can be recomputed from scratch.
2585 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2586 const void *start, unsigned int len)
2588 if (skb->ip_summed == CHECKSUM_COMPLETE)
2589 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2592 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2595 * pskb_trim_rcsum - trim received skb and update checksum
2596 * @skb: buffer to trim
2599 * This is exactly the same as pskb_trim except that it ensures the
2600 * checksum of received packets are still valid after the operation.
2603 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2605 if (likely(len >= skb->len))
2607 if (skb->ip_summed == CHECKSUM_COMPLETE)
2608 skb->ip_summed = CHECKSUM_NONE;
2609 return __pskb_trim(skb, len);
2612 #define skb_queue_walk(queue, skb) \
2613 for (skb = (queue)->next; \
2614 skb != (struct sk_buff *)(queue); \
2617 #define skb_queue_walk_safe(queue, skb, tmp) \
2618 for (skb = (queue)->next, tmp = skb->next; \
2619 skb != (struct sk_buff *)(queue); \
2620 skb = tmp, tmp = skb->next)
2622 #define skb_queue_walk_from(queue, skb) \
2623 for (; skb != (struct sk_buff *)(queue); \
2626 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2627 for (tmp = skb->next; \
2628 skb != (struct sk_buff *)(queue); \
2629 skb = tmp, tmp = skb->next)
2631 #define skb_queue_reverse_walk(queue, skb) \
2632 for (skb = (queue)->prev; \
2633 skb != (struct sk_buff *)(queue); \
2636 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2637 for (skb = (queue)->prev, tmp = skb->prev; \
2638 skb != (struct sk_buff *)(queue); \
2639 skb = tmp, tmp = skb->prev)
2641 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2642 for (tmp = skb->prev; \
2643 skb != (struct sk_buff *)(queue); \
2644 skb = tmp, tmp = skb->prev)
2646 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2648 return skb_shinfo(skb)->frag_list != NULL;
2651 static inline void skb_frag_list_init(struct sk_buff *skb)
2653 skb_shinfo(skb)->frag_list = NULL;
2656 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2658 frag->next = skb_shinfo(skb)->frag_list;
2659 skb_shinfo(skb)->frag_list = frag;
2662 #define skb_walk_frags(skb, iter) \
2663 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2665 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2666 int *peeked, int *off, int *err);
2667 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2669 unsigned int datagram_poll(struct file *file, struct socket *sock,
2670 struct poll_table_struct *wait);
2671 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2672 struct iov_iter *to, int size);
2673 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2674 struct msghdr *msg, int size)
2676 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
2678 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2679 struct msghdr *msg);
2680 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2681 struct iov_iter *from, int len);
2682 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
2683 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2684 void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2685 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
2686 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2687 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2688 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2689 int len, __wsum csum);
2690 int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2691 struct pipe_inode_info *pipe, unsigned int len,
2692 unsigned int flags);
2693 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2694 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2695 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2697 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2698 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2699 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
2700 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
2701 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
2702 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
2703 int skb_ensure_writable(struct sk_buff *skb, int write_len);
2704 int skb_vlan_pop(struct sk_buff *skb);
2705 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
2707 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2709 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
2712 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2714 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
2717 struct skb_checksum_ops {
2718 __wsum (*update)(const void *mem, int len, __wsum wsum);
2719 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2722 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2723 __wsum csum, const struct skb_checksum_ops *ops);
2724 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2727 static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2728 int len, void *data, int hlen, void *buffer)
2730 if (hlen - offset >= len)
2731 return data + offset;
2734 skb_copy_bits(skb, offset, buffer, len) < 0)
2740 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2741 int len, void *buffer)
2743 return __skb_header_pointer(skb, offset, len, skb->data,
2744 skb_headlen(skb), buffer);
2748 * skb_needs_linearize - check if we need to linearize a given skb
2749 * depending on the given device features.
2750 * @skb: socket buffer to check
2751 * @features: net device features
2753 * Returns true if either:
2754 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2755 * 2. skb is fragmented and the device does not support SG.
2757 static inline bool skb_needs_linearize(struct sk_buff *skb,
2758 netdev_features_t features)
2760 return skb_is_nonlinear(skb) &&
2761 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2762 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2765 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2767 const unsigned int len)
2769 memcpy(to, skb->data, len);
2772 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2773 const int offset, void *to,
2774 const unsigned int len)
2776 memcpy(to, skb->data + offset, len);
2779 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2781 const unsigned int len)
2783 memcpy(skb->data, from, len);
2786 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2789 const unsigned int len)
2791 memcpy(skb->data + offset, from, len);
2794 void skb_init(void);
2796 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2802 * skb_get_timestamp - get timestamp from a skb
2803 * @skb: skb to get stamp from
2804 * @stamp: pointer to struct timeval to store stamp in
2806 * Timestamps are stored in the skb as offsets to a base timestamp.
2807 * This function converts the offset back to a struct timeval and stores
2810 static inline void skb_get_timestamp(const struct sk_buff *skb,
2811 struct timeval *stamp)
2813 *stamp = ktime_to_timeval(skb->tstamp);
2816 static inline void skb_get_timestampns(const struct sk_buff *skb,
2817 struct timespec *stamp)
2819 *stamp = ktime_to_timespec(skb->tstamp);
2822 static inline void __net_timestamp(struct sk_buff *skb)
2824 skb->tstamp = ktime_get_real();
2827 static inline ktime_t net_timedelta(ktime_t t)
2829 return ktime_sub(ktime_get_real(), t);
2832 static inline ktime_t net_invalid_timestamp(void)
2834 return ktime_set(0, 0);
2837 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2839 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2841 void skb_clone_tx_timestamp(struct sk_buff *skb);
2842 bool skb_defer_rx_timestamp(struct sk_buff *skb);
2844 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2846 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2850 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2855 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2858 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2860 * PHY drivers may accept clones of transmitted packets for
2861 * timestamping via their phy_driver.txtstamp method. These drivers
2862 * must call this function to return the skb back to the stack, with
2863 * or without a timestamp.
2865 * @skb: clone of the the original outgoing packet
2866 * @hwtstamps: hardware time stamps, may be NULL if not available
2869 void skb_complete_tx_timestamp(struct sk_buff *skb,
2870 struct skb_shared_hwtstamps *hwtstamps);
2872 void __skb_tstamp_tx(struct sk_buff *orig_skb,
2873 struct skb_shared_hwtstamps *hwtstamps,
2874 struct sock *sk, int tstype);
2877 * skb_tstamp_tx - queue clone of skb with send time stamps
2878 * @orig_skb: the original outgoing packet
2879 * @hwtstamps: hardware time stamps, may be NULL if not available
2881 * If the skb has a socket associated, then this function clones the
2882 * skb (thus sharing the actual data and optional structures), stores
2883 * the optional hardware time stamping information (if non NULL) or
2884 * generates a software time stamp (otherwise), then queues the clone
2885 * to the error queue of the socket. Errors are silently ignored.
2887 void skb_tstamp_tx(struct sk_buff *orig_skb,
2888 struct skb_shared_hwtstamps *hwtstamps);
2890 static inline void sw_tx_timestamp(struct sk_buff *skb)
2892 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2893 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2894 skb_tstamp_tx(skb, NULL);
2898 * skb_tx_timestamp() - Driver hook for transmit timestamping
2900 * Ethernet MAC Drivers should call this function in their hard_xmit()
2901 * function immediately before giving the sk_buff to the MAC hardware.
2903 * Specifically, one should make absolutely sure that this function is
2904 * called before TX completion of this packet can trigger. Otherwise
2905 * the packet could potentially already be freed.
2907 * @skb: A socket buffer.
2909 static inline void skb_tx_timestamp(struct sk_buff *skb)
2911 skb_clone_tx_timestamp(skb);
2912 sw_tx_timestamp(skb);
2916 * skb_complete_wifi_ack - deliver skb with wifi status
2918 * @skb: the original outgoing packet
2919 * @acked: ack status
2922 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2924 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2925 __sum16 __skb_checksum_complete(struct sk_buff *skb);
2927 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2929 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
2931 (skb->ip_summed == CHECKSUM_PARTIAL &&
2932 skb_checksum_start_offset(skb) >= 0));
2936 * skb_checksum_complete - Calculate checksum of an entire packet
2937 * @skb: packet to process
2939 * This function calculates the checksum over the entire packet plus
2940 * the value of skb->csum. The latter can be used to supply the
2941 * checksum of a pseudo header as used by TCP/UDP. It returns the
2944 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2945 * this function can be used to verify that checksum on received
2946 * packets. In that case the function should return zero if the
2947 * checksum is correct. In particular, this function will return zero
2948 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2949 * hardware has already verified the correctness of the checksum.
2951 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2953 return skb_csum_unnecessary(skb) ?
2954 0 : __skb_checksum_complete(skb);
2957 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2959 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2960 if (skb->csum_level == 0)
2961 skb->ip_summed = CHECKSUM_NONE;
2967 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2969 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2970 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2972 } else if (skb->ip_summed == CHECKSUM_NONE) {
2973 skb->ip_summed = CHECKSUM_UNNECESSARY;
2974 skb->csum_level = 0;
2978 static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2980 /* Mark current checksum as bad (typically called from GRO
2981 * path). In the case that ip_summed is CHECKSUM_NONE
2982 * this must be the first checksum encountered in the packet.
2983 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2984 * checksum after the last one validated. For UDP, a zero
2985 * checksum can not be marked as bad.
2988 if (skb->ip_summed == CHECKSUM_NONE ||
2989 skb->ip_summed == CHECKSUM_UNNECESSARY)
2993 /* Check if we need to perform checksum complete validation.
2995 * Returns true if checksum complete is needed, false otherwise
2996 * (either checksum is unnecessary or zero checksum is allowed).
2998 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3002 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3003 skb->csum_valid = 1;
3004 __skb_decr_checksum_unnecessary(skb);
3011 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3014 #define CHECKSUM_BREAK 76
3016 /* Validate (init) checksum based on checksum complete.
3019 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3020 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3021 * checksum is stored in skb->csum for use in __skb_checksum_complete
3022 * non-zero: value of invalid checksum
3025 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3029 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3030 if (!csum_fold(csum_add(psum, skb->csum))) {
3031 skb->csum_valid = 1;
3034 } else if (skb->csum_bad) {
3035 /* ip_summed == CHECKSUM_NONE in this case */
3041 if (complete || skb->len <= CHECKSUM_BREAK) {
3044 csum = __skb_checksum_complete(skb);
3045 skb->csum_valid = !csum;
3052 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3057 /* Perform checksum validate (init). Note that this is a macro since we only
3058 * want to calculate the pseudo header which is an input function if necessary.
3059 * First we try to validate without any computation (checksum unnecessary) and
3060 * then calculate based on checksum complete calling the function to compute
3064 * 0: checksum is validated or try to in skb_checksum_complete
3065 * non-zero: value of invalid checksum
3067 #define __skb_checksum_validate(skb, proto, complete, \
3068 zero_okay, check, compute_pseudo) \
3070 __sum16 __ret = 0; \
3071 skb->csum_valid = 0; \
3072 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3073 __ret = __skb_checksum_validate_complete(skb, \
3074 complete, compute_pseudo(skb, proto)); \
3078 #define skb_checksum_init(skb, proto, compute_pseudo) \
3079 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3081 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3082 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3084 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3085 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3087 #define skb_checksum_validate_zero_check(skb, proto, check, \
3089 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3091 #define skb_checksum_simple_validate(skb) \
3092 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3094 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3096 return (skb->ip_summed == CHECKSUM_NONE &&
3097 skb->csum_valid && !skb->csum_bad);
3100 static inline void __skb_checksum_convert(struct sk_buff *skb,
3101 __sum16 check, __wsum pseudo)
3103 skb->csum = ~pseudo;
3104 skb->ip_summed = CHECKSUM_COMPLETE;
3107 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3109 if (__skb_checksum_convert_check(skb)) \
3110 __skb_checksum_convert(skb, check, \
3111 compute_pseudo(skb, proto)); \
3114 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3115 u16 start, u16 offset)
3117 skb->ip_summed = CHECKSUM_PARTIAL;
3118 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3119 skb->csum_offset = offset - start;
3122 /* Update skbuf and packet to reflect the remote checksum offload operation.
3123 * When called, ptr indicates the starting point for skb->csum when
3124 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3125 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3127 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
3128 int start, int offset, bool nopartial)
3133 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3137 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3138 __skb_checksum_complete(skb);
3139 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3142 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3144 /* Adjust skb->csum since we changed the packet */
3145 skb->csum = csum_add(skb->csum, delta);
3148 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3149 void nf_conntrack_destroy(struct nf_conntrack *nfct);
3150 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3152 if (nfct && atomic_dec_and_test(&nfct->use))
3153 nf_conntrack_destroy(nfct);
3155 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3158 atomic_inc(&nfct->use);
3161 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3162 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3164 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3167 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3170 atomic_inc(&nf_bridge->use);
3172 #endif /* CONFIG_BRIDGE_NETFILTER */
3173 static inline void nf_reset(struct sk_buff *skb)
3175 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3176 nf_conntrack_put(skb->nfct);
3179 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3180 nf_bridge_put(skb->nf_bridge);
3181 skb->nf_bridge = NULL;
3185 static inline void nf_reset_trace(struct sk_buff *skb)
3187 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3192 /* Note: This doesn't put any conntrack and bridge info in dst. */
3193 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3196 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3197 dst->nfct = src->nfct;
3198 nf_conntrack_get(src->nfct);
3200 dst->nfctinfo = src->nfctinfo;
3202 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3203 dst->nf_bridge = src->nf_bridge;
3204 nf_bridge_get(src->nf_bridge);
3206 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3208 dst->nf_trace = src->nf_trace;
3212 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3214 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3215 nf_conntrack_put(dst->nfct);
3217 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3218 nf_bridge_put(dst->nf_bridge);
3220 __nf_copy(dst, src, true);
3223 #ifdef CONFIG_NETWORK_SECMARK
3224 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3226 to->secmark = from->secmark;
3229 static inline void skb_init_secmark(struct sk_buff *skb)
3234 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3237 static inline void skb_init_secmark(struct sk_buff *skb)
3241 static inline bool skb_irq_freeable(const struct sk_buff *skb)
3243 return !skb->destructor &&
3244 #if IS_ENABLED(CONFIG_XFRM)
3247 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3250 !skb->_skb_refdst &&
3251 !skb_has_frag_list(skb);
3254 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3256 skb->queue_mapping = queue_mapping;
3259 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
3261 return skb->queue_mapping;
3264 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3266 to->queue_mapping = from->queue_mapping;
3269 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3271 skb->queue_mapping = rx_queue + 1;
3274 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
3276 return skb->queue_mapping - 1;
3279 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
3281 return skb->queue_mapping != 0;
3284 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3285 unsigned int num_tx_queues);
3287 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3296 /* Keeps track of mac header offset relative to skb->head.
3297 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3298 * For non-tunnel skb it points to skb_mac_header() and for
3299 * tunnel skb it points to outer mac header.
3300 * Keeps track of level of encapsulation of network headers.
3307 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3309 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3311 return (skb_mac_header(inner_skb) - inner_skb->head) -
3312 SKB_GSO_CB(inner_skb)->mac_offset;
3315 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3317 int new_headroom, headroom;
3320 headroom = skb_headroom(skb);
3321 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3325 new_headroom = skb_headroom(skb);
3326 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3330 /* Compute the checksum for a gso segment. First compute the checksum value
3331 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3332 * then add in skb->csum (checksum from csum_start to end of packet).
3333 * skb->csum and csum_start are then updated to reflect the checksum of the
3334 * resultant packet starting from the transport header-- the resultant checksum
3335 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3338 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3340 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3341 skb_transport_offset(skb);
3344 csum = csum_fold(csum_partial(skb_transport_header(skb),
3347 SKB_GSO_CB(skb)->csum_start -= plen;
3352 static inline bool skb_is_gso(const struct sk_buff *skb)
3354 return skb_shinfo(skb)->gso_size;
3357 /* Note: Should be called only if skb_is_gso(skb) is true */
3358 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
3360 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3363 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
3365 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3367 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3368 * wanted then gso_type will be set. */
3369 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3371 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3372 unlikely(shinfo->gso_type == 0)) {
3373 __skb_warn_lro_forwarding(skb);
3379 static inline void skb_forward_csum(struct sk_buff *skb)
3381 /* Unfortunately we don't support this one. Any brave souls? */
3382 if (skb->ip_summed == CHECKSUM_COMPLETE)
3383 skb->ip_summed = CHECKSUM_NONE;
3387 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3388 * @skb: skb to check
3390 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3391 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3392 * use this helper, to document places where we make this assertion.
3394 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
3397 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3401 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
3403 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3405 u32 skb_get_poff(const struct sk_buff *skb);
3406 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3407 const struct flow_keys *keys, int hlen);
3410 * skb_head_is_locked - Determine if the skb->head is locked down
3411 * @skb: skb to check
3413 * The head on skbs build around a head frag can be removed if they are
3414 * not cloned. This function returns true if the skb head is locked down
3415 * due to either being allocated via kmalloc, or by being a clone with
3416 * multiple references to the head.
3418 static inline bool skb_head_is_locked(const struct sk_buff *skb)
3420 return !skb->head_frag || skb_cloned(skb);
3424 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3428 * skb_gso_network_seglen is used to determine the real size of the
3429 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3431 * The MAC/L2 header is not accounted for.
3433 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3435 unsigned int hdr_len = skb_transport_header(skb) -
3436 skb_network_header(skb);
3437 return hdr_len + skb_gso_transport_seglen(skb);
3439 #endif /* __KERNEL__ */
3440 #endif /* _LINUX_SKBUFF_H */