arm64: pass a task parameter to unwind_frame()
[linux-2.6/btrfs-unstable.git] / include / linux / skbuff.h
blob4355129fff91b6f188136af2a499d6100f2e5bfd
1 /*
2 * Definitions for the 'struct sk_buff' memory handlers.
4 * Authors:
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_dissector.h>
38 #include <linux/splice.h>
39 #include <linux/in6.h>
40 #include <net/flow.h>
42 /* A. Checksumming of received packets by device.
44 * CHECKSUM_NONE:
46 * Device failed to checksum this packet e.g. due to lack of capabilities.
47 * The packet contains full (though not verified) checksum in packet but
48 * not in skb->csum. Thus, skb->csum is undefined in this case.
50 * CHECKSUM_UNNECESSARY:
52 * The hardware you're dealing with doesn't calculate the full checksum
53 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
54 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
55 * if their checksums are okay. skb->csum is still undefined in this case
56 * though. It is a bad option, but, unfortunately, nowadays most vendors do
57 * this. Apparently with the secret goal to sell you new devices, when you
58 * will add new protocol to your host, f.e. IPv6 8)
60 * CHECKSUM_UNNECESSARY is applicable to following protocols:
61 * TCP: IPv6 and IPv4.
62 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
63 * zero UDP checksum for either IPv4 or IPv6, the networking stack
64 * may perform further validation in this case.
65 * GRE: only if the checksum is present in the header.
66 * SCTP: indicates the CRC in SCTP header has been validated.
68 * skb->csum_level indicates the number of consecutive checksums found in
69 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
70 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
71 * and a device is able to verify the checksums for UDP (possibly zero),
72 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
73 * two. If the device were only able to verify the UDP checksum and not
74 * GRE, either because it doesn't support GRE checksum of because GRE
75 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
76 * not considered in this case).
78 * CHECKSUM_COMPLETE:
80 * This is the most generic way. The device supplied checksum of the _whole_
81 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
82 * hardware doesn't need to parse L3/L4 headers to implement this.
84 * Note: Even if device supports only some protocols, but is able to produce
85 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
87 * CHECKSUM_PARTIAL:
89 * A checksum is set up to be offloaded to a device as described in the
90 * output description for CHECKSUM_PARTIAL. This may occur on a packet
91 * received directly from another Linux OS, e.g., a virtualized Linux kernel
92 * on the same host, or it may be set in the input path in GRO or remote
93 * checksum offload. For the purposes of checksum verification, the checksum
94 * referred to by skb->csum_start + skb->csum_offset and any preceding
95 * checksums in the packet are considered verified. Any checksums in the
96 * packet that are after the checksum being offloaded are not considered to
97 * be verified.
99 * B. Checksumming on output.
101 * CHECKSUM_NONE:
103 * The skb was already checksummed by the protocol, or a checksum is not
104 * required.
106 * CHECKSUM_PARTIAL:
108 * The device is required to checksum the packet as seen by hard_start_xmit()
109 * from skb->csum_start up to the end, and to record/write the checksum at
110 * offset skb->csum_start + skb->csum_offset.
112 * The device must show its capabilities in dev->features, set up at device
113 * setup time, e.g. netdev_features.h:
115 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
116 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
117 * IPv4. Sigh. Vendors like this way for an unknown reason.
118 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
119 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
120 * NETIF_F_... - Well, you get the picture.
122 * CHECKSUM_UNNECESSARY:
124 * Normally, the device will do per protocol specific checksumming. Protocol
125 * implementations that do not want the NIC to perform the checksum
126 * calculation should use this flag in their outgoing skbs.
128 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
129 * offload. Correspondingly, the FCoE protocol driver
130 * stack should use CHECKSUM_UNNECESSARY.
132 * Any questions? No questions, good. --ANK
135 /* Don't change this without changing skb_csum_unnecessary! */
136 #define CHECKSUM_NONE 0
137 #define CHECKSUM_UNNECESSARY 1
138 #define CHECKSUM_COMPLETE 2
139 #define CHECKSUM_PARTIAL 3
141 /* Maximum value in skb->csum_level */
142 #define SKB_MAX_CSUM_LEVEL 3
144 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
145 #define SKB_WITH_OVERHEAD(X) \
146 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
147 #define SKB_MAX_ORDER(X, ORDER) \
148 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
149 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
150 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
152 /* return minimum truesize of one skb containing X bytes of data */
153 #define SKB_TRUESIZE(X) ((X) + \
154 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
155 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
157 struct net_device;
158 struct scatterlist;
159 struct pipe_inode_info;
160 struct iov_iter;
161 struct napi_struct;
163 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
164 struct nf_conntrack {
165 atomic_t use;
167 #endif
169 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
170 struct nf_bridge_info {
171 atomic_t use;
172 enum {
173 BRNF_PROTO_UNCHANGED,
174 BRNF_PROTO_8021Q,
175 BRNF_PROTO_PPPOE
176 } orig_proto:8;
177 u8 pkt_otherhost:1;
178 u8 in_prerouting:1;
179 u8 bridged_dnat:1;
180 __u16 frag_max_size;
181 struct net_device *physindev;
183 /* always valid & non-NULL from FORWARD on, for physdev match */
184 struct net_device *physoutdev;
185 union {
186 /* prerouting: detect dnat in orig/reply direction */
187 __be32 ipv4_daddr;
188 struct in6_addr ipv6_daddr;
190 /* after prerouting + nat detected: store original source
191 * mac since neigh resolution overwrites it, only used while
192 * skb is out in neigh layer.
194 char neigh_header[8];
197 #endif
199 struct sk_buff_head {
200 /* These two members must be first. */
201 struct sk_buff *next;
202 struct sk_buff *prev;
204 __u32 qlen;
205 spinlock_t lock;
208 struct sk_buff;
210 /* To allow 64K frame to be packed as single skb without frag_list we
211 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
212 * buffers which do not start on a page boundary.
214 * Since GRO uses frags we allocate at least 16 regardless of page
215 * size.
217 #if (65536/PAGE_SIZE + 1) < 16
218 #define MAX_SKB_FRAGS 16UL
219 #else
220 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
221 #endif
223 typedef struct skb_frag_struct skb_frag_t;
225 struct skb_frag_struct {
226 struct {
227 struct page *p;
228 } page;
229 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
230 __u32 page_offset;
231 __u32 size;
232 #else
233 __u16 page_offset;
234 __u16 size;
235 #endif
238 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
240 return frag->size;
243 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
245 frag->size = size;
248 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
250 frag->size += delta;
253 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
255 frag->size -= delta;
258 #define HAVE_HW_TIME_STAMP
261 * struct skb_shared_hwtstamps - hardware time stamps
262 * @hwtstamp: hardware time stamp transformed into duration
263 * since arbitrary point in time
265 * Software time stamps generated by ktime_get_real() are stored in
266 * skb->tstamp.
268 * hwtstamps can only be compared against other hwtstamps from
269 * the same device.
271 * This structure is attached to packets as part of the
272 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
274 struct skb_shared_hwtstamps {
275 ktime_t hwtstamp;
278 /* Definitions for tx_flags in struct skb_shared_info */
279 enum {
280 /* generate hardware time stamp */
281 SKBTX_HW_TSTAMP = 1 << 0,
283 /* generate software time stamp when queueing packet to NIC */
284 SKBTX_SW_TSTAMP = 1 << 1,
286 /* device driver is going to provide hardware time stamp */
287 SKBTX_IN_PROGRESS = 1 << 2,
289 /* device driver supports TX zero-copy buffers */
290 SKBTX_DEV_ZEROCOPY = 1 << 3,
292 /* generate wifi status information (where possible) */
293 SKBTX_WIFI_STATUS = 1 << 4,
295 /* This indicates at least one fragment might be overwritten
296 * (as in vmsplice(), sendfile() ...)
297 * If we need to compute a TX checksum, we'll need to copy
298 * all frags to avoid possible bad checksum
300 SKBTX_SHARED_FRAG = 1 << 5,
302 /* generate software time stamp when entering packet scheduling */
303 SKBTX_SCHED_TSTAMP = 1 << 6,
305 /* generate software timestamp on peer data acknowledgment */
306 SKBTX_ACK_TSTAMP = 1 << 7,
309 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
310 SKBTX_SCHED_TSTAMP | \
311 SKBTX_ACK_TSTAMP)
312 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
315 * The callback notifies userspace to release buffers when skb DMA is done in
316 * lower device, the skb last reference should be 0 when calling this.
317 * The zerocopy_success argument is true if zero copy transmit occurred,
318 * false on data copy or out of memory error caused by data copy attempt.
319 * The ctx field is used to track device context.
320 * The desc field is used to track userspace buffer index.
322 struct ubuf_info {
323 void (*callback)(struct ubuf_info *, bool zerocopy_success);
324 void *ctx;
325 unsigned long desc;
328 /* This data is invariant across clones and lives at
329 * the end of the header data, ie. at skb->end.
331 struct skb_shared_info {
332 unsigned char nr_frags;
333 __u8 tx_flags;
334 unsigned short gso_size;
335 /* Warning: this field is not always filled in (UFO)! */
336 unsigned short gso_segs;
337 unsigned short gso_type;
338 struct sk_buff *frag_list;
339 struct skb_shared_hwtstamps hwtstamps;
340 u32 tskey;
341 __be32 ip6_frag_id;
344 * Warning : all fields before dataref are cleared in __alloc_skb()
346 atomic_t dataref;
348 /* Intermediate layers must ensure that destructor_arg
349 * remains valid until skb destructor */
350 void * destructor_arg;
352 /* must be last field, see pskb_expand_head() */
353 skb_frag_t frags[MAX_SKB_FRAGS];
356 /* We divide dataref into two halves. The higher 16 bits hold references
357 * to the payload part of skb->data. The lower 16 bits hold references to
358 * the entire skb->data. A clone of a headerless skb holds the length of
359 * the header in skb->hdr_len.
361 * All users must obey the rule that the skb->data reference count must be
362 * greater than or equal to the payload reference count.
364 * Holding a reference to the payload part means that the user does not
365 * care about modifications to the header part of skb->data.
367 #define SKB_DATAREF_SHIFT 16
368 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
371 enum {
372 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
373 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
374 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
377 enum {
378 SKB_GSO_TCPV4 = 1 << 0,
379 SKB_GSO_UDP = 1 << 1,
381 /* This indicates the skb is from an untrusted source. */
382 SKB_GSO_DODGY = 1 << 2,
384 /* This indicates the tcp segment has CWR set. */
385 SKB_GSO_TCP_ECN = 1 << 3,
387 SKB_GSO_TCPV6 = 1 << 4,
389 SKB_GSO_FCOE = 1 << 5,
391 SKB_GSO_GRE = 1 << 6,
393 SKB_GSO_GRE_CSUM = 1 << 7,
395 SKB_GSO_IPIP = 1 << 8,
397 SKB_GSO_SIT = 1 << 9,
399 SKB_GSO_UDP_TUNNEL = 1 << 10,
401 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
403 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
406 #if BITS_PER_LONG > 32
407 #define NET_SKBUFF_DATA_USES_OFFSET 1
408 #endif
410 #ifdef NET_SKBUFF_DATA_USES_OFFSET
411 typedef unsigned int sk_buff_data_t;
412 #else
413 typedef unsigned char *sk_buff_data_t;
414 #endif
417 * struct skb_mstamp - multi resolution time stamps
418 * @stamp_us: timestamp in us resolution
419 * @stamp_jiffies: timestamp in jiffies
421 struct skb_mstamp {
422 union {
423 u64 v64;
424 struct {
425 u32 stamp_us;
426 u32 stamp_jiffies;
432 * skb_mstamp_get - get current timestamp
433 * @cl: place to store timestamps
435 static inline void skb_mstamp_get(struct skb_mstamp *cl)
437 u64 val = local_clock();
439 do_div(val, NSEC_PER_USEC);
440 cl->stamp_us = (u32)val;
441 cl->stamp_jiffies = (u32)jiffies;
445 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
446 * @t1: pointer to newest sample
447 * @t0: pointer to oldest sample
449 static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
450 const struct skb_mstamp *t0)
452 s32 delta_us = t1->stamp_us - t0->stamp_us;
453 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
455 /* If delta_us is negative, this might be because interval is too big,
456 * or local_clock() drift is too big : fallback using jiffies.
458 if (delta_us <= 0 ||
459 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
461 delta_us = jiffies_to_usecs(delta_jiffies);
463 return delta_us;
466 static inline bool skb_mstamp_after(const struct skb_mstamp *t1,
467 const struct skb_mstamp *t0)
469 s32 diff = t1->stamp_jiffies - t0->stamp_jiffies;
471 if (!diff)
472 diff = t1->stamp_us - t0->stamp_us;
473 return diff > 0;
476 /**
477 * struct sk_buff - socket buffer
478 * @next: Next buffer in list
479 * @prev: Previous buffer in list
480 * @tstamp: Time we arrived/left
481 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
482 * @sk: Socket we are owned by
483 * @dev: Device we arrived on/are leaving by
484 * @cb: Control buffer. Free for use by every layer. Put private vars here
485 * @_skb_refdst: destination entry (with norefcount bit)
486 * @sp: the security path, used for xfrm
487 * @len: Length of actual data
488 * @data_len: Data length
489 * @mac_len: Length of link layer header
490 * @hdr_len: writable header length of cloned skb
491 * @csum: Checksum (must include start/offset pair)
492 * @csum_start: Offset from skb->head where checksumming should start
493 * @csum_offset: Offset from csum_start where checksum should be stored
494 * @priority: Packet queueing priority
495 * @ignore_df: allow local fragmentation
496 * @cloned: Head may be cloned (check refcnt to be sure)
497 * @ip_summed: Driver fed us an IP checksum
498 * @nohdr: Payload reference only, must not modify header
499 * @nfctinfo: Relationship of this skb to the connection
500 * @pkt_type: Packet class
501 * @fclone: skbuff clone status
502 * @ipvs_property: skbuff is owned by ipvs
503 * @peeked: this packet has been seen already, so stats have been
504 * done for it, don't do them again
505 * @nf_trace: netfilter packet trace flag
506 * @protocol: Packet protocol from driver
507 * @destructor: Destruct function
508 * @nfct: Associated connection, if any
509 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
510 * @skb_iif: ifindex of device we arrived on
511 * @tc_index: Traffic control index
512 * @tc_verd: traffic control verdict
513 * @hash: the packet hash
514 * @queue_mapping: Queue mapping for multiqueue devices
515 * @xmit_more: More SKBs are pending for this queue
516 * @ndisc_nodetype: router type (from link layer)
517 * @ooo_okay: allow the mapping of a socket to a queue to be changed
518 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
519 * ports.
520 * @sw_hash: indicates hash was computed in software stack
521 * @wifi_acked_valid: wifi_acked was set
522 * @wifi_acked: whether frame was acked on wifi or not
523 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
524 * @napi_id: id of the NAPI struct this skb came from
525 * @secmark: security marking
526 * @offload_fwd_mark: fwding offload mark
527 * @mark: Generic packet mark
528 * @vlan_proto: vlan encapsulation protocol
529 * @vlan_tci: vlan tag control information
530 * @inner_protocol: Protocol (encapsulation)
531 * @inner_transport_header: Inner transport layer header (encapsulation)
532 * @inner_network_header: Network layer header (encapsulation)
533 * @inner_mac_header: Link layer header (encapsulation)
534 * @transport_header: Transport layer header
535 * @network_header: Network layer header
536 * @mac_header: Link layer header
537 * @tail: Tail pointer
538 * @end: End pointer
539 * @head: Head of buffer
540 * @data: Data head pointer
541 * @truesize: Buffer size
542 * @users: User count - see {datagram,tcp}.c
545 struct sk_buff {
546 union {
547 struct {
548 /* These two members must be first. */
549 struct sk_buff *next;
550 struct sk_buff *prev;
552 union {
553 ktime_t tstamp;
554 struct skb_mstamp skb_mstamp;
557 struct rb_node rbnode; /* used in netem & tcp stack */
559 struct sock *sk;
560 struct net_device *dev;
563 * This is the control buffer. It is free to use for every
564 * layer. Please put your private variables there. If you
565 * want to keep them across layers you have to do a skb_clone()
566 * first. This is owned by whoever has the skb queued ATM.
568 char cb[48] __aligned(8);
570 unsigned long _skb_refdst;
571 void (*destructor)(struct sk_buff *skb);
572 #ifdef CONFIG_XFRM
573 struct sec_path *sp;
574 #endif
575 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
576 struct nf_conntrack *nfct;
577 #endif
578 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
579 struct nf_bridge_info *nf_bridge;
580 #endif
581 unsigned int len,
582 data_len;
583 __u16 mac_len,
584 hdr_len;
586 /* Following fields are _not_ copied in __copy_skb_header()
587 * Note that queue_mapping is here mostly to fill a hole.
589 kmemcheck_bitfield_begin(flags1);
590 __u16 queue_mapping;
591 __u8 cloned:1,
592 nohdr:1,
593 fclone:2,
594 peeked:1,
595 head_frag:1,
596 xmit_more:1;
597 /* one bit hole */
598 kmemcheck_bitfield_end(flags1);
600 /* fields enclosed in headers_start/headers_end are copied
601 * using a single memcpy() in __copy_skb_header()
603 /* private: */
604 __u32 headers_start[0];
605 /* public: */
607 /* if you move pkt_type around you also must adapt those constants */
608 #ifdef __BIG_ENDIAN_BITFIELD
609 #define PKT_TYPE_MAX (7 << 5)
610 #else
611 #define PKT_TYPE_MAX 7
612 #endif
613 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
615 __u8 __pkt_type_offset[0];
616 __u8 pkt_type:3;
617 __u8 pfmemalloc:1;
618 __u8 ignore_df:1;
619 __u8 nfctinfo:3;
621 __u8 nf_trace:1;
622 __u8 ip_summed:2;
623 __u8 ooo_okay:1;
624 __u8 l4_hash:1;
625 __u8 sw_hash:1;
626 __u8 wifi_acked_valid:1;
627 __u8 wifi_acked:1;
629 __u8 no_fcs:1;
630 /* Indicates the inner headers are valid in the skbuff. */
631 __u8 encapsulation:1;
632 __u8 encap_hdr_csum:1;
633 __u8 csum_valid:1;
634 __u8 csum_complete_sw:1;
635 __u8 csum_level:2;
636 __u8 csum_bad:1;
638 #ifdef CONFIG_IPV6_NDISC_NODETYPE
639 __u8 ndisc_nodetype:2;
640 #endif
641 __u8 ipvs_property:1;
642 __u8 inner_protocol_type:1;
643 __u8 remcsum_offload:1;
644 /* 3 or 5 bit hole */
646 #ifdef CONFIG_NET_SCHED
647 __u16 tc_index; /* traffic control index */
648 #ifdef CONFIG_NET_CLS_ACT
649 __u16 tc_verd; /* traffic control verdict */
650 #endif
651 #endif
653 union {
654 __wsum csum;
655 struct {
656 __u16 csum_start;
657 __u16 csum_offset;
660 __u32 priority;
661 int skb_iif;
662 __u32 hash;
663 __be16 vlan_proto;
664 __u16 vlan_tci;
665 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
666 union {
667 unsigned int napi_id;
668 unsigned int sender_cpu;
670 #endif
671 union {
672 #ifdef CONFIG_NETWORK_SECMARK
673 __u32 secmark;
674 #endif
675 #ifdef CONFIG_NET_SWITCHDEV
676 __u32 offload_fwd_mark;
677 #endif
680 union {
681 __u32 mark;
682 __u32 reserved_tailroom;
685 union {
686 __be16 inner_protocol;
687 __u8 inner_ipproto;
690 __u16 inner_transport_header;
691 __u16 inner_network_header;
692 __u16 inner_mac_header;
694 __be16 protocol;
695 __u16 transport_header;
696 __u16 network_header;
697 __u16 mac_header;
699 /* private: */
700 __u32 headers_end[0];
701 /* public: */
703 /* These elements must be at the end, see alloc_skb() for details. */
704 sk_buff_data_t tail;
705 sk_buff_data_t end;
706 unsigned char *head,
707 *data;
708 unsigned int truesize;
709 atomic_t users;
712 #ifdef __KERNEL__
714 * Handling routines are only of interest to the kernel
716 #include <linux/slab.h>
719 #define SKB_ALLOC_FCLONE 0x01
720 #define SKB_ALLOC_RX 0x02
721 #define SKB_ALLOC_NAPI 0x04
723 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
724 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
726 return unlikely(skb->pfmemalloc);
730 * skb might have a dst pointer attached, refcounted or not.
731 * _skb_refdst low order bit is set if refcount was _not_ taken
733 #define SKB_DST_NOREF 1UL
734 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
737 * skb_dst - returns skb dst_entry
738 * @skb: buffer
740 * Returns skb dst_entry, regardless of reference taken or not.
742 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
744 /* If refdst was not refcounted, check we still are in a
745 * rcu_read_lock section
747 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
748 !rcu_read_lock_held() &&
749 !rcu_read_lock_bh_held());
750 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
754 * skb_dst_set - sets skb dst
755 * @skb: buffer
756 * @dst: dst entry
758 * Sets skb dst, assuming a reference was taken on dst and should
759 * be released by skb_dst_drop()
761 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
763 skb->_skb_refdst = (unsigned long)dst;
767 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
768 * @skb: buffer
769 * @dst: dst entry
771 * Sets skb dst, assuming a reference was not taken on dst.
772 * If dst entry is cached, we do not take reference and dst_release
773 * will be avoided by refdst_drop. If dst entry is not cached, we take
774 * reference, so that last dst_release can destroy the dst immediately.
776 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
778 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
779 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
783 * skb_dst_is_noref - Test if skb dst isn't refcounted
784 * @skb: buffer
786 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
788 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
791 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
793 return (struct rtable *)skb_dst(skb);
796 void kfree_skb(struct sk_buff *skb);
797 void kfree_skb_list(struct sk_buff *segs);
798 void skb_tx_error(struct sk_buff *skb);
799 void consume_skb(struct sk_buff *skb);
800 void __kfree_skb(struct sk_buff *skb);
801 extern struct kmem_cache *skbuff_head_cache;
803 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
804 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
805 bool *fragstolen, int *delta_truesize);
807 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
808 int node);
809 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
810 struct sk_buff *build_skb(void *data, unsigned int frag_size);
811 static inline struct sk_buff *alloc_skb(unsigned int size,
812 gfp_t priority)
814 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
817 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
818 unsigned long data_len,
819 int max_page_order,
820 int *errcode,
821 gfp_t gfp_mask);
823 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
824 struct sk_buff_fclones {
825 struct sk_buff skb1;
827 struct sk_buff skb2;
829 atomic_t fclone_ref;
833 * skb_fclone_busy - check if fclone is busy
834 * @skb: buffer
836 * Returns true is skb is a fast clone, and its clone is not freed.
837 * Some drivers call skb_orphan() in their ndo_start_xmit(),
838 * so we also check that this didnt happen.
840 static inline bool skb_fclone_busy(const struct sock *sk,
841 const struct sk_buff *skb)
843 const struct sk_buff_fclones *fclones;
845 fclones = container_of(skb, struct sk_buff_fclones, skb1);
847 return skb->fclone == SKB_FCLONE_ORIG &&
848 atomic_read(&fclones->fclone_ref) > 1 &&
849 fclones->skb2.sk == sk;
852 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
853 gfp_t priority)
855 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
858 struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
859 static inline struct sk_buff *alloc_skb_head(gfp_t priority)
861 return __alloc_skb_head(priority, -1);
864 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
865 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
866 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
867 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
868 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
869 gfp_t gfp_mask, bool fclone);
870 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
871 gfp_t gfp_mask)
873 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
876 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
877 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
878 unsigned int headroom);
879 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
880 int newtailroom, gfp_t priority);
881 int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
882 int offset, int len);
883 int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
884 int len);
885 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
886 int skb_pad(struct sk_buff *skb, int pad);
887 #define dev_kfree_skb(a) consume_skb(a)
889 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
890 int getfrag(void *from, char *to, int offset,
891 int len, int odd, struct sk_buff *skb),
892 void *from, int length);
894 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
895 int offset, size_t size);
897 struct skb_seq_state {
898 __u32 lower_offset;
899 __u32 upper_offset;
900 __u32 frag_idx;
901 __u32 stepped_offset;
902 struct sk_buff *root_skb;
903 struct sk_buff *cur_skb;
904 __u8 *frag_data;
907 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
908 unsigned int to, struct skb_seq_state *st);
909 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
910 struct skb_seq_state *st);
911 void skb_abort_seq_read(struct skb_seq_state *st);
913 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
914 unsigned int to, struct ts_config *config);
917 * Packet hash types specify the type of hash in skb_set_hash.
919 * Hash types refer to the protocol layer addresses which are used to
920 * construct a packet's hash. The hashes are used to differentiate or identify
921 * flows of the protocol layer for the hash type. Hash types are either
922 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
924 * Properties of hashes:
926 * 1) Two packets in different flows have different hash values
927 * 2) Two packets in the same flow should have the same hash value
929 * A hash at a higher layer is considered to be more specific. A driver should
930 * set the most specific hash possible.
932 * A driver cannot indicate a more specific hash than the layer at which a hash
933 * was computed. For instance an L3 hash cannot be set as an L4 hash.
935 * A driver may indicate a hash level which is less specific than the
936 * actual layer the hash was computed on. For instance, a hash computed
937 * at L4 may be considered an L3 hash. This should only be done if the
938 * driver can't unambiguously determine that the HW computed the hash at
939 * the higher layer. Note that the "should" in the second property above
940 * permits this.
942 enum pkt_hash_types {
943 PKT_HASH_TYPE_NONE, /* Undefined type */
944 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
945 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
946 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
949 static inline void skb_clear_hash(struct sk_buff *skb)
951 skb->hash = 0;
952 skb->sw_hash = 0;
953 skb->l4_hash = 0;
956 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
958 if (!skb->l4_hash)
959 skb_clear_hash(skb);
962 static inline void
963 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
965 skb->l4_hash = is_l4;
966 skb->sw_hash = is_sw;
967 skb->hash = hash;
970 static inline void
971 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
973 /* Used by drivers to set hash from HW */
974 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
977 static inline void
978 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
980 __skb_set_hash(skb, hash, true, is_l4);
983 void __skb_get_hash(struct sk_buff *skb);
984 u32 skb_get_poff(const struct sk_buff *skb);
985 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
986 const struct flow_keys *keys, int hlen);
987 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
988 void *data, int hlen_proto);
990 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
991 int thoff, u8 ip_proto)
993 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
996 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
997 const struct flow_dissector_key *key,
998 unsigned int key_count);
1000 bool __skb_flow_dissect(const struct sk_buff *skb,
1001 struct flow_dissector *flow_dissector,
1002 void *target_container,
1003 void *data, __be16 proto, int nhoff, int hlen,
1004 unsigned int flags);
1006 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1007 struct flow_dissector *flow_dissector,
1008 void *target_container, unsigned int flags)
1010 return __skb_flow_dissect(skb, flow_dissector, target_container,
1011 NULL, 0, 0, 0, flags);
1014 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1015 struct flow_keys *flow,
1016 unsigned int flags)
1018 memset(flow, 0, sizeof(*flow));
1019 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
1020 NULL, 0, 0, 0, flags);
1023 static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1024 void *data, __be16 proto,
1025 int nhoff, int hlen,
1026 unsigned int flags)
1028 memset(flow, 0, sizeof(*flow));
1029 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
1030 data, proto, nhoff, hlen, flags);
1033 static inline __u32 skb_get_hash(struct sk_buff *skb)
1035 if (!skb->l4_hash && !skb->sw_hash)
1036 __skb_get_hash(skb);
1038 return skb->hash;
1041 __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
1043 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1045 if (!skb->l4_hash && !skb->sw_hash) {
1046 struct flow_keys keys;
1047 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1049 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1052 return skb->hash;
1055 __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
1057 static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
1059 if (!skb->l4_hash && !skb->sw_hash) {
1060 struct flow_keys keys;
1061 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
1063 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1066 return skb->hash;
1069 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1071 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1073 return skb->hash;
1076 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1078 to->hash = from->hash;
1079 to->sw_hash = from->sw_hash;
1080 to->l4_hash = from->l4_hash;
1083 static inline void skb_sender_cpu_clear(struct sk_buff *skb)
1085 #ifdef CONFIG_XPS
1086 skb->sender_cpu = 0;
1087 #endif
1090 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1091 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1093 return skb->head + skb->end;
1096 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1098 return skb->end;
1100 #else
1101 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1103 return skb->end;
1106 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1108 return skb->end - skb->head;
1110 #endif
1112 /* Internal */
1113 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1115 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1117 return &skb_shinfo(skb)->hwtstamps;
1121 * skb_queue_empty - check if a queue is empty
1122 * @list: queue head
1124 * Returns true if the queue is empty, false otherwise.
1126 static inline int skb_queue_empty(const struct sk_buff_head *list)
1128 return list->next == (const struct sk_buff *) list;
1132 * skb_queue_is_last - check if skb is the last entry in the queue
1133 * @list: queue head
1134 * @skb: buffer
1136 * Returns true if @skb is the last buffer on the list.
1138 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1139 const struct sk_buff *skb)
1141 return skb->next == (const struct sk_buff *) list;
1145 * skb_queue_is_first - check if skb is the first entry in the queue
1146 * @list: queue head
1147 * @skb: buffer
1149 * Returns true if @skb is the first buffer on the list.
1151 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1152 const struct sk_buff *skb)
1154 return skb->prev == (const struct sk_buff *) list;
1158 * skb_queue_next - return the next packet in the queue
1159 * @list: queue head
1160 * @skb: current buffer
1162 * Return the next packet in @list after @skb. It is only valid to
1163 * call this if skb_queue_is_last() evaluates to false.
1165 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1166 const struct sk_buff *skb)
1168 /* This BUG_ON may seem severe, but if we just return then we
1169 * are going to dereference garbage.
1171 BUG_ON(skb_queue_is_last(list, skb));
1172 return skb->next;
1176 * skb_queue_prev - return the prev packet in the queue
1177 * @list: queue head
1178 * @skb: current buffer
1180 * Return the prev packet in @list before @skb. It is only valid to
1181 * call this if skb_queue_is_first() evaluates to false.
1183 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1184 const struct sk_buff *skb)
1186 /* This BUG_ON may seem severe, but if we just return then we
1187 * are going to dereference garbage.
1189 BUG_ON(skb_queue_is_first(list, skb));
1190 return skb->prev;
1194 * skb_get - reference buffer
1195 * @skb: buffer to reference
1197 * Makes another reference to a socket buffer and returns a pointer
1198 * to the buffer.
1200 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1202 atomic_inc(&skb->users);
1203 return skb;
1207 * If users == 1, we are the only owner and are can avoid redundant
1208 * atomic change.
1212 * skb_cloned - is the buffer a clone
1213 * @skb: buffer to check
1215 * Returns true if the buffer was generated with skb_clone() and is
1216 * one of multiple shared copies of the buffer. Cloned buffers are
1217 * shared data so must not be written to under normal circumstances.
1219 static inline int skb_cloned(const struct sk_buff *skb)
1221 return skb->cloned &&
1222 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1225 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1227 might_sleep_if(gfpflags_allow_blocking(pri));
1229 if (skb_cloned(skb))
1230 return pskb_expand_head(skb, 0, 0, pri);
1232 return 0;
1236 * skb_header_cloned - is the header a clone
1237 * @skb: buffer to check
1239 * Returns true if modifying the header part of the buffer requires
1240 * the data to be copied.
1242 static inline int skb_header_cloned(const struct sk_buff *skb)
1244 int dataref;
1246 if (!skb->cloned)
1247 return 0;
1249 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1250 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1251 return dataref != 1;
1255 * skb_header_release - release reference to header
1256 * @skb: buffer to operate on
1258 * Drop a reference to the header part of the buffer. This is done
1259 * by acquiring a payload reference. You must not read from the header
1260 * part of skb->data after this.
1261 * Note : Check if you can use __skb_header_release() instead.
1263 static inline void skb_header_release(struct sk_buff *skb)
1265 BUG_ON(skb->nohdr);
1266 skb->nohdr = 1;
1267 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1271 * __skb_header_release - release reference to header
1272 * @skb: buffer to operate on
1274 * Variant of skb_header_release() assuming skb is private to caller.
1275 * We can avoid one atomic operation.
1277 static inline void __skb_header_release(struct sk_buff *skb)
1279 skb->nohdr = 1;
1280 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1285 * skb_shared - is the buffer shared
1286 * @skb: buffer to check
1288 * Returns true if more than one person has a reference to this
1289 * buffer.
1291 static inline int skb_shared(const struct sk_buff *skb)
1293 return atomic_read(&skb->users) != 1;
1297 * skb_share_check - check if buffer is shared and if so clone it
1298 * @skb: buffer to check
1299 * @pri: priority for memory allocation
1301 * If the buffer is shared the buffer is cloned and the old copy
1302 * drops a reference. A new clone with a single reference is returned.
1303 * If the buffer is not shared the original buffer is returned. When
1304 * being called from interrupt status or with spinlocks held pri must
1305 * be GFP_ATOMIC.
1307 * NULL is returned on a memory allocation failure.
1309 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1311 might_sleep_if(gfpflags_allow_blocking(pri));
1312 if (skb_shared(skb)) {
1313 struct sk_buff *nskb = skb_clone(skb, pri);
1315 if (likely(nskb))
1316 consume_skb(skb);
1317 else
1318 kfree_skb(skb);
1319 skb = nskb;
1321 return skb;
1325 * Copy shared buffers into a new sk_buff. We effectively do COW on
1326 * packets to handle cases where we have a local reader and forward
1327 * and a couple of other messy ones. The normal one is tcpdumping
1328 * a packet thats being forwarded.
1332 * skb_unshare - make a copy of a shared buffer
1333 * @skb: buffer to check
1334 * @pri: priority for memory allocation
1336 * If the socket buffer is a clone then this function creates a new
1337 * copy of the data, drops a reference count on the old copy and returns
1338 * the new copy with the reference count at 1. If the buffer is not a clone
1339 * the original buffer is returned. When called with a spinlock held or
1340 * from interrupt state @pri must be %GFP_ATOMIC
1342 * %NULL is returned on a memory allocation failure.
1344 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1345 gfp_t pri)
1347 might_sleep_if(gfpflags_allow_blocking(pri));
1348 if (skb_cloned(skb)) {
1349 struct sk_buff *nskb = skb_copy(skb, pri);
1351 /* Free our shared copy */
1352 if (likely(nskb))
1353 consume_skb(skb);
1354 else
1355 kfree_skb(skb);
1356 skb = nskb;
1358 return skb;
1362 * skb_peek - peek at the head of an &sk_buff_head
1363 * @list_: list to peek at
1365 * Peek an &sk_buff. Unlike most other operations you _MUST_
1366 * be careful with this one. A peek leaves the buffer on the
1367 * list and someone else may run off with it. You must hold
1368 * the appropriate locks or have a private queue to do this.
1370 * Returns %NULL for an empty list or a pointer to the head element.
1371 * The reference count is not incremented and the reference is therefore
1372 * volatile. Use with caution.
1374 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1376 struct sk_buff *skb = list_->next;
1378 if (skb == (struct sk_buff *)list_)
1379 skb = NULL;
1380 return skb;
1384 * skb_peek_next - peek skb following the given one from a queue
1385 * @skb: skb to start from
1386 * @list_: list to peek at
1388 * Returns %NULL when the end of the list is met or a pointer to the
1389 * next element. The reference count is not incremented and the
1390 * reference is therefore volatile. Use with caution.
1392 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1393 const struct sk_buff_head *list_)
1395 struct sk_buff *next = skb->next;
1397 if (next == (struct sk_buff *)list_)
1398 next = NULL;
1399 return next;
1403 * skb_peek_tail - peek at the tail of an &sk_buff_head
1404 * @list_: list to peek at
1406 * Peek an &sk_buff. Unlike most other operations you _MUST_
1407 * be careful with this one. A peek leaves the buffer on the
1408 * list and someone else may run off with it. You must hold
1409 * the appropriate locks or have a private queue to do this.
1411 * Returns %NULL for an empty list or a pointer to the tail element.
1412 * The reference count is not incremented and the reference is therefore
1413 * volatile. Use with caution.
1415 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1417 struct sk_buff *skb = list_->prev;
1419 if (skb == (struct sk_buff *)list_)
1420 skb = NULL;
1421 return skb;
1426 * skb_queue_len - get queue length
1427 * @list_: list to measure
1429 * Return the length of an &sk_buff queue.
1431 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1433 return list_->qlen;
1437 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1438 * @list: queue to initialize
1440 * This initializes only the list and queue length aspects of
1441 * an sk_buff_head object. This allows to initialize the list
1442 * aspects of an sk_buff_head without reinitializing things like
1443 * the spinlock. It can also be used for on-stack sk_buff_head
1444 * objects where the spinlock is known to not be used.
1446 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1448 list->prev = list->next = (struct sk_buff *)list;
1449 list->qlen = 0;
1453 * This function creates a split out lock class for each invocation;
1454 * this is needed for now since a whole lot of users of the skb-queue
1455 * infrastructure in drivers have different locking usage (in hardirq)
1456 * than the networking core (in softirq only). In the long run either the
1457 * network layer or drivers should need annotation to consolidate the
1458 * main types of usage into 3 classes.
1460 static inline void skb_queue_head_init(struct sk_buff_head *list)
1462 spin_lock_init(&list->lock);
1463 __skb_queue_head_init(list);
1466 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1467 struct lock_class_key *class)
1469 skb_queue_head_init(list);
1470 lockdep_set_class(&list->lock, class);
1474 * Insert an sk_buff on a list.
1476 * The "__skb_xxxx()" functions are the non-atomic ones that
1477 * can only be called with interrupts disabled.
1479 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1480 struct sk_buff_head *list);
1481 static inline void __skb_insert(struct sk_buff *newsk,
1482 struct sk_buff *prev, struct sk_buff *next,
1483 struct sk_buff_head *list)
1485 newsk->next = next;
1486 newsk->prev = prev;
1487 next->prev = prev->next = newsk;
1488 list->qlen++;
1491 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1492 struct sk_buff *prev,
1493 struct sk_buff *next)
1495 struct sk_buff *first = list->next;
1496 struct sk_buff *last = list->prev;
1498 first->prev = prev;
1499 prev->next = first;
1501 last->next = next;
1502 next->prev = last;
1506 * skb_queue_splice - join two skb lists, this is designed for stacks
1507 * @list: the new list to add
1508 * @head: the place to add it in the first list
1510 static inline void skb_queue_splice(const struct sk_buff_head *list,
1511 struct sk_buff_head *head)
1513 if (!skb_queue_empty(list)) {
1514 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1515 head->qlen += list->qlen;
1520 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1521 * @list: the new list to add
1522 * @head: the place to add it in the first list
1524 * The list at @list is reinitialised
1526 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1527 struct sk_buff_head *head)
1529 if (!skb_queue_empty(list)) {
1530 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1531 head->qlen += list->qlen;
1532 __skb_queue_head_init(list);
1537 * skb_queue_splice_tail - join two skb lists, each list being a queue
1538 * @list: the new list to add
1539 * @head: the place to add it in the first list
1541 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1542 struct sk_buff_head *head)
1544 if (!skb_queue_empty(list)) {
1545 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1546 head->qlen += list->qlen;
1551 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1552 * @list: the new list to add
1553 * @head: the place to add it in the first list
1555 * Each of the lists is a queue.
1556 * The list at @list is reinitialised
1558 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1559 struct sk_buff_head *head)
1561 if (!skb_queue_empty(list)) {
1562 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1563 head->qlen += list->qlen;
1564 __skb_queue_head_init(list);
1569 * __skb_queue_after - queue a buffer at the list head
1570 * @list: list to use
1571 * @prev: place after this buffer
1572 * @newsk: buffer to queue
1574 * Queue a buffer int the middle of a list. This function takes no locks
1575 * and you must therefore hold required locks before calling it.
1577 * A buffer cannot be placed on two lists at the same time.
1579 static inline void __skb_queue_after(struct sk_buff_head *list,
1580 struct sk_buff *prev,
1581 struct sk_buff *newsk)
1583 __skb_insert(newsk, prev, prev->next, list);
1586 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1587 struct sk_buff_head *list);
1589 static inline void __skb_queue_before(struct sk_buff_head *list,
1590 struct sk_buff *next,
1591 struct sk_buff *newsk)
1593 __skb_insert(newsk, next->prev, next, list);
1597 * __skb_queue_head - queue a buffer at the list head
1598 * @list: list to use
1599 * @newsk: buffer to queue
1601 * Queue a buffer at the start of a list. This function takes no locks
1602 * and you must therefore hold required locks before calling it.
1604 * A buffer cannot be placed on two lists at the same time.
1606 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1607 static inline void __skb_queue_head(struct sk_buff_head *list,
1608 struct sk_buff *newsk)
1610 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1614 * __skb_queue_tail - queue a buffer at the list tail
1615 * @list: list to use
1616 * @newsk: buffer to queue
1618 * Queue a buffer at the end of a list. This function takes no locks
1619 * and you must therefore hold required locks before calling it.
1621 * A buffer cannot be placed on two lists at the same time.
1623 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1624 static inline void __skb_queue_tail(struct sk_buff_head *list,
1625 struct sk_buff *newsk)
1627 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1631 * remove sk_buff from list. _Must_ be called atomically, and with
1632 * the list known..
1634 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1635 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1637 struct sk_buff *next, *prev;
1639 list->qlen--;
1640 next = skb->next;
1641 prev = skb->prev;
1642 skb->next = skb->prev = NULL;
1643 next->prev = prev;
1644 prev->next = next;
1648 * __skb_dequeue - remove from the head of the queue
1649 * @list: list to dequeue from
1651 * Remove the head of the list. This function does not take any locks
1652 * so must be used with appropriate locks held only. The head item is
1653 * returned or %NULL if the list is empty.
1655 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1656 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1658 struct sk_buff *skb = skb_peek(list);
1659 if (skb)
1660 __skb_unlink(skb, list);
1661 return skb;
1665 * __skb_dequeue_tail - remove from the tail of the queue
1666 * @list: list to dequeue from
1668 * Remove the tail of the list. This function does not take any locks
1669 * so must be used with appropriate locks held only. The tail item is
1670 * returned or %NULL if the list is empty.
1672 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1673 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1675 struct sk_buff *skb = skb_peek_tail(list);
1676 if (skb)
1677 __skb_unlink(skb, list);
1678 return skb;
1682 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1684 return skb->data_len;
1687 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1689 return skb->len - skb->data_len;
1692 static inline int skb_pagelen(const struct sk_buff *skb)
1694 int i, len = 0;
1696 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1697 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1698 return len + skb_headlen(skb);
1702 * __skb_fill_page_desc - initialise a paged fragment in an skb
1703 * @skb: buffer containing fragment to be initialised
1704 * @i: paged fragment index to initialise
1705 * @page: the page to use for this fragment
1706 * @off: the offset to the data with @page
1707 * @size: the length of the data
1709 * Initialises the @i'th fragment of @skb to point to &size bytes at
1710 * offset @off within @page.
1712 * Does not take any additional reference on the fragment.
1714 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1715 struct page *page, int off, int size)
1717 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1720 * Propagate page pfmemalloc to the skb if we can. The problem is
1721 * that not all callers have unique ownership of the page but rely
1722 * on page_is_pfmemalloc doing the right thing(tm).
1724 frag->page.p = page;
1725 frag->page_offset = off;
1726 skb_frag_size_set(frag, size);
1728 page = compound_head(page);
1729 if (page_is_pfmemalloc(page))
1730 skb->pfmemalloc = true;
1734 * skb_fill_page_desc - initialise a paged fragment in an skb
1735 * @skb: buffer containing fragment to be initialised
1736 * @i: paged fragment index to initialise
1737 * @page: the page to use for this fragment
1738 * @off: the offset to the data with @page
1739 * @size: the length of the data
1741 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1742 * @skb to point to @size bytes at offset @off within @page. In
1743 * addition updates @skb such that @i is the last fragment.
1745 * Does not take any additional reference on the fragment.
1747 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1748 struct page *page, int off, int size)
1750 __skb_fill_page_desc(skb, i, page, off, size);
1751 skb_shinfo(skb)->nr_frags = i + 1;
1754 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1755 int size, unsigned int truesize);
1757 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1758 unsigned int truesize);
1760 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1761 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1762 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1764 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1765 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1767 return skb->head + skb->tail;
1770 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1772 skb->tail = skb->data - skb->head;
1775 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1777 skb_reset_tail_pointer(skb);
1778 skb->tail += offset;
1781 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1782 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1784 return skb->tail;
1787 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1789 skb->tail = skb->data;
1792 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1794 skb->tail = skb->data + offset;
1797 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1800 * Add data to an sk_buff
1802 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1803 unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1804 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1806 unsigned char *tmp = skb_tail_pointer(skb);
1807 SKB_LINEAR_ASSERT(skb);
1808 skb->tail += len;
1809 skb->len += len;
1810 return tmp;
1813 unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1814 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1816 skb->data -= len;
1817 skb->len += len;
1818 return skb->data;
1821 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1822 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1824 skb->len -= len;
1825 BUG_ON(skb->len < skb->data_len);
1826 return skb->data += len;
1829 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1831 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1834 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1836 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1838 if (len > skb_headlen(skb) &&
1839 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1840 return NULL;
1841 skb->len -= len;
1842 return skb->data += len;
1845 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1847 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1850 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1852 if (likely(len <= skb_headlen(skb)))
1853 return 1;
1854 if (unlikely(len > skb->len))
1855 return 0;
1856 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1860 * skb_headroom - bytes at buffer head
1861 * @skb: buffer to check
1863 * Return the number of bytes of free space at the head of an &sk_buff.
1865 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1867 return skb->data - skb->head;
1871 * skb_tailroom - bytes at buffer end
1872 * @skb: buffer to check
1874 * Return the number of bytes of free space at the tail of an sk_buff
1876 static inline int skb_tailroom(const struct sk_buff *skb)
1878 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1882 * skb_availroom - bytes at buffer end
1883 * @skb: buffer to check
1885 * Return the number of bytes of free space at the tail of an sk_buff
1886 * allocated by sk_stream_alloc()
1888 static inline int skb_availroom(const struct sk_buff *skb)
1890 if (skb_is_nonlinear(skb))
1891 return 0;
1893 return skb->end - skb->tail - skb->reserved_tailroom;
1897 * skb_reserve - adjust headroom
1898 * @skb: buffer to alter
1899 * @len: bytes to move
1901 * Increase the headroom of an empty &sk_buff by reducing the tail
1902 * room. This is only allowed for an empty buffer.
1904 static inline void skb_reserve(struct sk_buff *skb, int len)
1906 skb->data += len;
1907 skb->tail += len;
1910 #define ENCAP_TYPE_ETHER 0
1911 #define ENCAP_TYPE_IPPROTO 1
1913 static inline void skb_set_inner_protocol(struct sk_buff *skb,
1914 __be16 protocol)
1916 skb->inner_protocol = protocol;
1917 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1920 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1921 __u8 ipproto)
1923 skb->inner_ipproto = ipproto;
1924 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1927 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1929 skb->inner_mac_header = skb->mac_header;
1930 skb->inner_network_header = skb->network_header;
1931 skb->inner_transport_header = skb->transport_header;
1934 static inline void skb_reset_mac_len(struct sk_buff *skb)
1936 skb->mac_len = skb->network_header - skb->mac_header;
1939 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1940 *skb)
1942 return skb->head + skb->inner_transport_header;
1945 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1947 skb->inner_transport_header = skb->data - skb->head;
1950 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1951 const int offset)
1953 skb_reset_inner_transport_header(skb);
1954 skb->inner_transport_header += offset;
1957 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1959 return skb->head + skb->inner_network_header;
1962 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1964 skb->inner_network_header = skb->data - skb->head;
1967 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1968 const int offset)
1970 skb_reset_inner_network_header(skb);
1971 skb->inner_network_header += offset;
1974 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1976 return skb->head + skb->inner_mac_header;
1979 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1981 skb->inner_mac_header = skb->data - skb->head;
1984 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1985 const int offset)
1987 skb_reset_inner_mac_header(skb);
1988 skb->inner_mac_header += offset;
1990 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1992 return skb->transport_header != (typeof(skb->transport_header))~0U;
1995 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1997 return skb->head + skb->transport_header;
2000 static inline void skb_reset_transport_header(struct sk_buff *skb)
2002 skb->transport_header = skb->data - skb->head;
2005 static inline void skb_set_transport_header(struct sk_buff *skb,
2006 const int offset)
2008 skb_reset_transport_header(skb);
2009 skb->transport_header += offset;
2012 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2014 return skb->head + skb->network_header;
2017 static inline void skb_reset_network_header(struct sk_buff *skb)
2019 skb->network_header = skb->data - skb->head;
2022 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2024 skb_reset_network_header(skb);
2025 skb->network_header += offset;
2028 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2030 return skb->head + skb->mac_header;
2033 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2035 return skb->mac_header != (typeof(skb->mac_header))~0U;
2038 static inline void skb_reset_mac_header(struct sk_buff *skb)
2040 skb->mac_header = skb->data - skb->head;
2043 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2045 skb_reset_mac_header(skb);
2046 skb->mac_header += offset;
2049 static inline void skb_pop_mac_header(struct sk_buff *skb)
2051 skb->mac_header = skb->network_header;
2054 static inline void skb_probe_transport_header(struct sk_buff *skb,
2055 const int offset_hint)
2057 struct flow_keys keys;
2059 if (skb_transport_header_was_set(skb))
2060 return;
2061 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
2062 skb_set_transport_header(skb, keys.control.thoff);
2063 else
2064 skb_set_transport_header(skb, offset_hint);
2067 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2069 if (skb_mac_header_was_set(skb)) {
2070 const unsigned char *old_mac = skb_mac_header(skb);
2072 skb_set_mac_header(skb, -skb->mac_len);
2073 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2077 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2079 return skb->csum_start - skb_headroom(skb);
2082 static inline int skb_transport_offset(const struct sk_buff *skb)
2084 return skb_transport_header(skb) - skb->data;
2087 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2089 return skb->transport_header - skb->network_header;
2092 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2094 return skb->inner_transport_header - skb->inner_network_header;
2097 static inline int skb_network_offset(const struct sk_buff *skb)
2099 return skb_network_header(skb) - skb->data;
2102 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2104 return skb_inner_network_header(skb) - skb->data;
2107 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2109 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2113 * CPUs often take a performance hit when accessing unaligned memory
2114 * locations. The actual performance hit varies, it can be small if the
2115 * hardware handles it or large if we have to take an exception and fix it
2116 * in software.
2118 * Since an ethernet header is 14 bytes network drivers often end up with
2119 * the IP header at an unaligned offset. The IP header can be aligned by
2120 * shifting the start of the packet by 2 bytes. Drivers should do this
2121 * with:
2123 * skb_reserve(skb, NET_IP_ALIGN);
2125 * The downside to this alignment of the IP header is that the DMA is now
2126 * unaligned. On some architectures the cost of an unaligned DMA is high
2127 * and this cost outweighs the gains made by aligning the IP header.
2129 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2130 * to be overridden.
2132 #ifndef NET_IP_ALIGN
2133 #define NET_IP_ALIGN 2
2134 #endif
2137 * The networking layer reserves some headroom in skb data (via
2138 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2139 * the header has to grow. In the default case, if the header has to grow
2140 * 32 bytes or less we avoid the reallocation.
2142 * Unfortunately this headroom changes the DMA alignment of the resulting
2143 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2144 * on some architectures. An architecture can override this value,
2145 * perhaps setting it to a cacheline in size (since that will maintain
2146 * cacheline alignment of the DMA). It must be a power of 2.
2148 * Various parts of the networking layer expect at least 32 bytes of
2149 * headroom, you should not reduce this.
2151 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2152 * to reduce average number of cache lines per packet.
2153 * get_rps_cpus() for example only access one 64 bytes aligned block :
2154 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2156 #ifndef NET_SKB_PAD
2157 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2158 #endif
2160 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2162 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2164 if (unlikely(skb_is_nonlinear(skb))) {
2165 WARN_ON(1);
2166 return;
2168 skb->len = len;
2169 skb_set_tail_pointer(skb, len);
2172 void skb_trim(struct sk_buff *skb, unsigned int len);
2174 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2176 if (skb->data_len)
2177 return ___pskb_trim(skb, len);
2178 __skb_trim(skb, len);
2179 return 0;
2182 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2184 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2188 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2189 * @skb: buffer to alter
2190 * @len: new length
2192 * This is identical to pskb_trim except that the caller knows that
2193 * the skb is not cloned so we should never get an error due to out-
2194 * of-memory.
2196 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2198 int err = pskb_trim(skb, len);
2199 BUG_ON(err);
2203 * skb_orphan - orphan a buffer
2204 * @skb: buffer to orphan
2206 * If a buffer currently has an owner then we call the owner's
2207 * destructor function and make the @skb unowned. The buffer continues
2208 * to exist but is no longer charged to its former owner.
2210 static inline void skb_orphan(struct sk_buff *skb)
2212 if (skb->destructor) {
2213 skb->destructor(skb);
2214 skb->destructor = NULL;
2215 skb->sk = NULL;
2216 } else {
2217 BUG_ON(skb->sk);
2222 * skb_orphan_frags - orphan the frags contained in a buffer
2223 * @skb: buffer to orphan frags from
2224 * @gfp_mask: allocation mask for replacement pages
2226 * For each frag in the SKB which needs a destructor (i.e. has an
2227 * owner) create a copy of that frag and release the original
2228 * page by calling the destructor.
2230 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2232 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2233 return 0;
2234 return skb_copy_ubufs(skb, gfp_mask);
2238 * __skb_queue_purge - empty a list
2239 * @list: list to empty
2241 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2242 * the list and one reference dropped. This function does not take the
2243 * list lock and the caller must hold the relevant locks to use it.
2245 void skb_queue_purge(struct sk_buff_head *list);
2246 static inline void __skb_queue_purge(struct sk_buff_head *list)
2248 struct sk_buff *skb;
2249 while ((skb = __skb_dequeue(list)) != NULL)
2250 kfree_skb(skb);
2253 void *netdev_alloc_frag(unsigned int fragsz);
2255 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2256 gfp_t gfp_mask);
2259 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2260 * @dev: network device to receive on
2261 * @length: length to allocate
2263 * Allocate a new &sk_buff and assign it a usage count of one. The
2264 * buffer has unspecified headroom built in. Users should allocate
2265 * the headroom they think they need without accounting for the
2266 * built in space. The built in space is used for optimisations.
2268 * %NULL is returned if there is no free memory. Although this function
2269 * allocates memory it can be called from an interrupt.
2271 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2272 unsigned int length)
2274 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2277 /* legacy helper around __netdev_alloc_skb() */
2278 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2279 gfp_t gfp_mask)
2281 return __netdev_alloc_skb(NULL, length, gfp_mask);
2284 /* legacy helper around netdev_alloc_skb() */
2285 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2287 return netdev_alloc_skb(NULL, length);
2291 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2292 unsigned int length, gfp_t gfp)
2294 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2296 if (NET_IP_ALIGN && skb)
2297 skb_reserve(skb, NET_IP_ALIGN);
2298 return skb;
2301 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2302 unsigned int length)
2304 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2307 static inline void skb_free_frag(void *addr)
2309 __free_page_frag(addr);
2312 void *napi_alloc_frag(unsigned int fragsz);
2313 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2314 unsigned int length, gfp_t gfp_mask);
2315 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2316 unsigned int length)
2318 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2322 * __dev_alloc_pages - allocate page for network Rx
2323 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2324 * @order: size of the allocation
2326 * Allocate a new page.
2328 * %NULL is returned if there is no free memory.
2330 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2331 unsigned int order)
2333 /* This piece of code contains several assumptions.
2334 * 1. This is for device Rx, therefor a cold page is preferred.
2335 * 2. The expectation is the user wants a compound page.
2336 * 3. If requesting a order 0 page it will not be compound
2337 * due to the check to see if order has a value in prep_new_page
2338 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2339 * code in gfp_to_alloc_flags that should be enforcing this.
2341 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2343 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2346 static inline struct page *dev_alloc_pages(unsigned int order)
2348 return __dev_alloc_pages(GFP_ATOMIC, order);
2352 * __dev_alloc_page - allocate a page for network Rx
2353 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2355 * Allocate a new page.
2357 * %NULL is returned if there is no free memory.
2359 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2361 return __dev_alloc_pages(gfp_mask, 0);
2364 static inline struct page *dev_alloc_page(void)
2366 return __dev_alloc_page(GFP_ATOMIC);
2370 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2371 * @page: The page that was allocated from skb_alloc_page
2372 * @skb: The skb that may need pfmemalloc set
2374 static inline void skb_propagate_pfmemalloc(struct page *page,
2375 struct sk_buff *skb)
2377 if (page_is_pfmemalloc(page))
2378 skb->pfmemalloc = true;
2382 * skb_frag_page - retrieve the page referred to by a paged fragment
2383 * @frag: the paged fragment
2385 * Returns the &struct page associated with @frag.
2387 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2389 return frag->page.p;
2393 * __skb_frag_ref - take an addition reference on a paged fragment.
2394 * @frag: the paged fragment
2396 * Takes an additional reference on the paged fragment @frag.
2398 static inline void __skb_frag_ref(skb_frag_t *frag)
2400 get_page(skb_frag_page(frag));
2404 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2405 * @skb: the buffer
2406 * @f: the fragment offset.
2408 * Takes an additional reference on the @f'th paged fragment of @skb.
2410 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2412 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2416 * __skb_frag_unref - release a reference on a paged fragment.
2417 * @frag: the paged fragment
2419 * Releases a reference on the paged fragment @frag.
2421 static inline void __skb_frag_unref(skb_frag_t *frag)
2423 put_page(skb_frag_page(frag));
2427 * skb_frag_unref - release a reference on a paged fragment of an skb.
2428 * @skb: the buffer
2429 * @f: the fragment offset
2431 * Releases a reference on the @f'th paged fragment of @skb.
2433 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2435 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2439 * skb_frag_address - gets the address of the data contained in a paged fragment
2440 * @frag: the paged fragment buffer
2442 * Returns the address of the data within @frag. The page must already
2443 * be mapped.
2445 static inline void *skb_frag_address(const skb_frag_t *frag)
2447 return page_address(skb_frag_page(frag)) + frag->page_offset;
2451 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2452 * @frag: the paged fragment buffer
2454 * Returns the address of the data within @frag. Checks that the page
2455 * is mapped and returns %NULL otherwise.
2457 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2459 void *ptr = page_address(skb_frag_page(frag));
2460 if (unlikely(!ptr))
2461 return NULL;
2463 return ptr + frag->page_offset;
2467 * __skb_frag_set_page - sets the page contained in a paged fragment
2468 * @frag: the paged fragment
2469 * @page: the page to set
2471 * Sets the fragment @frag to contain @page.
2473 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2475 frag->page.p = page;
2479 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2480 * @skb: the buffer
2481 * @f: the fragment offset
2482 * @page: the page to set
2484 * Sets the @f'th fragment of @skb to contain @page.
2486 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2487 struct page *page)
2489 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2492 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2495 * skb_frag_dma_map - maps a paged fragment via the DMA API
2496 * @dev: the device to map the fragment to
2497 * @frag: the paged fragment to map
2498 * @offset: the offset within the fragment (starting at the
2499 * fragment's own offset)
2500 * @size: the number of bytes to map
2501 * @dir: the direction of the mapping (%PCI_DMA_*)
2503 * Maps the page associated with @frag to @device.
2505 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2506 const skb_frag_t *frag,
2507 size_t offset, size_t size,
2508 enum dma_data_direction dir)
2510 return dma_map_page(dev, skb_frag_page(frag),
2511 frag->page_offset + offset, size, dir);
2514 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2515 gfp_t gfp_mask)
2517 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2521 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2522 gfp_t gfp_mask)
2524 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2529 * skb_clone_writable - is the header of a clone writable
2530 * @skb: buffer to check
2531 * @len: length up to which to write
2533 * Returns true if modifying the header part of the cloned buffer
2534 * does not requires the data to be copied.
2536 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2538 return !skb_header_cloned(skb) &&
2539 skb_headroom(skb) + len <= skb->hdr_len;
2542 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2543 int cloned)
2545 int delta = 0;
2547 if (headroom > skb_headroom(skb))
2548 delta = headroom - skb_headroom(skb);
2550 if (delta || cloned)
2551 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2552 GFP_ATOMIC);
2553 return 0;
2557 * skb_cow - copy header of skb when it is required
2558 * @skb: buffer to cow
2559 * @headroom: needed headroom
2561 * If the skb passed lacks sufficient headroom or its data part
2562 * is shared, data is reallocated. If reallocation fails, an error
2563 * is returned and original skb is not changed.
2565 * The result is skb with writable area skb->head...skb->tail
2566 * and at least @headroom of space at head.
2568 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2570 return __skb_cow(skb, headroom, skb_cloned(skb));
2574 * skb_cow_head - skb_cow but only making the head writable
2575 * @skb: buffer to cow
2576 * @headroom: needed headroom
2578 * This function is identical to skb_cow except that we replace the
2579 * skb_cloned check by skb_header_cloned. It should be used when
2580 * you only need to push on some header and do not need to modify
2581 * the data.
2583 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2585 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2589 * skb_padto - pad an skbuff up to a minimal size
2590 * @skb: buffer to pad
2591 * @len: minimal length
2593 * Pads up a buffer to ensure the trailing bytes exist and are
2594 * blanked. If the buffer already contains sufficient data it
2595 * is untouched. Otherwise it is extended. Returns zero on
2596 * success. The skb is freed on error.
2598 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2600 unsigned int size = skb->len;
2601 if (likely(size >= len))
2602 return 0;
2603 return skb_pad(skb, len - size);
2607 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2608 * @skb: buffer to pad
2609 * @len: minimal length
2611 * Pads up a buffer to ensure the trailing bytes exist and are
2612 * blanked. If the buffer already contains sufficient data it
2613 * is untouched. Otherwise it is extended. Returns zero on
2614 * success. The skb is freed on error.
2616 static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2618 unsigned int size = skb->len;
2620 if (unlikely(size < len)) {
2621 len -= size;
2622 if (skb_pad(skb, len))
2623 return -ENOMEM;
2624 __skb_put(skb, len);
2626 return 0;
2629 static inline int skb_add_data(struct sk_buff *skb,
2630 struct iov_iter *from, int copy)
2632 const int off = skb->len;
2634 if (skb->ip_summed == CHECKSUM_NONE) {
2635 __wsum csum = 0;
2636 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2637 &csum, from) == copy) {
2638 skb->csum = csum_block_add(skb->csum, csum, off);
2639 return 0;
2641 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
2642 return 0;
2644 __skb_trim(skb, off);
2645 return -EFAULT;
2648 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2649 const struct page *page, int off)
2651 if (i) {
2652 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2654 return page == skb_frag_page(frag) &&
2655 off == frag->page_offset + skb_frag_size(frag);
2657 return false;
2660 static inline int __skb_linearize(struct sk_buff *skb)
2662 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2666 * skb_linearize - convert paged skb to linear one
2667 * @skb: buffer to linarize
2669 * If there is no free memory -ENOMEM is returned, otherwise zero
2670 * is returned and the old skb data released.
2672 static inline int skb_linearize(struct sk_buff *skb)
2674 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2678 * skb_has_shared_frag - can any frag be overwritten
2679 * @skb: buffer to test
2681 * Return true if the skb has at least one frag that might be modified
2682 * by an external entity (as in vmsplice()/sendfile())
2684 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2686 return skb_is_nonlinear(skb) &&
2687 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2691 * skb_linearize_cow - make sure skb is linear and writable
2692 * @skb: buffer to process
2694 * If there is no free memory -ENOMEM is returned, otherwise zero
2695 * is returned and the old skb data released.
2697 static inline int skb_linearize_cow(struct sk_buff *skb)
2699 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2700 __skb_linearize(skb) : 0;
2704 * skb_postpull_rcsum - update checksum for received skb after pull
2705 * @skb: buffer to update
2706 * @start: start of data before pull
2707 * @len: length of data pulled
2709 * After doing a pull on a received packet, you need to call this to
2710 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2711 * CHECKSUM_NONE so that it can be recomputed from scratch.
2714 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2715 const void *start, unsigned int len)
2717 if (skb->ip_summed == CHECKSUM_COMPLETE)
2718 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2719 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2720 skb_checksum_start_offset(skb) < 0)
2721 skb->ip_summed = CHECKSUM_NONE;
2724 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2727 * pskb_trim_rcsum - trim received skb and update checksum
2728 * @skb: buffer to trim
2729 * @len: new length
2731 * This is exactly the same as pskb_trim except that it ensures the
2732 * checksum of received packets are still valid after the operation.
2735 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2737 if (likely(len >= skb->len))
2738 return 0;
2739 if (skb->ip_summed == CHECKSUM_COMPLETE)
2740 skb->ip_summed = CHECKSUM_NONE;
2741 return __pskb_trim(skb, len);
2744 #define skb_queue_walk(queue, skb) \
2745 for (skb = (queue)->next; \
2746 skb != (struct sk_buff *)(queue); \
2747 skb = skb->next)
2749 #define skb_queue_walk_safe(queue, skb, tmp) \
2750 for (skb = (queue)->next, tmp = skb->next; \
2751 skb != (struct sk_buff *)(queue); \
2752 skb = tmp, tmp = skb->next)
2754 #define skb_queue_walk_from(queue, skb) \
2755 for (; skb != (struct sk_buff *)(queue); \
2756 skb = skb->next)
2758 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2759 for (tmp = skb->next; \
2760 skb != (struct sk_buff *)(queue); \
2761 skb = tmp, tmp = skb->next)
2763 #define skb_queue_reverse_walk(queue, skb) \
2764 for (skb = (queue)->prev; \
2765 skb != (struct sk_buff *)(queue); \
2766 skb = skb->prev)
2768 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2769 for (skb = (queue)->prev, tmp = skb->prev; \
2770 skb != (struct sk_buff *)(queue); \
2771 skb = tmp, tmp = skb->prev)
2773 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2774 for (tmp = skb->prev; \
2775 skb != (struct sk_buff *)(queue); \
2776 skb = tmp, tmp = skb->prev)
2778 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2780 return skb_shinfo(skb)->frag_list != NULL;
2783 static inline void skb_frag_list_init(struct sk_buff *skb)
2785 skb_shinfo(skb)->frag_list = NULL;
2788 #define skb_walk_frags(skb, iter) \
2789 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2791 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2792 int *peeked, int *off, int *err);
2793 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2794 int *err);
2795 unsigned int datagram_poll(struct file *file, struct socket *sock,
2796 struct poll_table_struct *wait);
2797 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2798 struct iov_iter *to, int size);
2799 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2800 struct msghdr *msg, int size)
2802 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
2804 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2805 struct msghdr *msg);
2806 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2807 struct iov_iter *from, int len);
2808 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
2809 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2810 void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2811 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
2812 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2813 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2814 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2815 int len, __wsum csum);
2816 ssize_t skb_socket_splice(struct sock *sk,
2817 struct pipe_inode_info *pipe,
2818 struct splice_pipe_desc *spd);
2819 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
2820 struct pipe_inode_info *pipe, unsigned int len,
2821 unsigned int flags,
2822 ssize_t (*splice_cb)(struct sock *,
2823 struct pipe_inode_info *,
2824 struct splice_pipe_desc *));
2825 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2826 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2827 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2828 int len, int hlen);
2829 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2830 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2831 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
2832 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
2833 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
2834 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
2835 int skb_ensure_writable(struct sk_buff *skb, int write_len);
2836 int skb_vlan_pop(struct sk_buff *skb);
2837 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
2839 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2841 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
2844 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2846 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
2849 struct skb_checksum_ops {
2850 __wsum (*update)(const void *mem, int len, __wsum wsum);
2851 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2854 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2855 __wsum csum, const struct skb_checksum_ops *ops);
2856 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2857 __wsum csum);
2859 static inline void * __must_check
2860 __skb_header_pointer(const struct sk_buff *skb, int offset,
2861 int len, void *data, int hlen, void *buffer)
2863 if (hlen - offset >= len)
2864 return data + offset;
2866 if (!skb ||
2867 skb_copy_bits(skb, offset, buffer, len) < 0)
2868 return NULL;
2870 return buffer;
2873 static inline void * __must_check
2874 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
2876 return __skb_header_pointer(skb, offset, len, skb->data,
2877 skb_headlen(skb), buffer);
2881 * skb_needs_linearize - check if we need to linearize a given skb
2882 * depending on the given device features.
2883 * @skb: socket buffer to check
2884 * @features: net device features
2886 * Returns true if either:
2887 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2888 * 2. skb is fragmented and the device does not support SG.
2890 static inline bool skb_needs_linearize(struct sk_buff *skb,
2891 netdev_features_t features)
2893 return skb_is_nonlinear(skb) &&
2894 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2895 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2898 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2899 void *to,
2900 const unsigned int len)
2902 memcpy(to, skb->data, len);
2905 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2906 const int offset, void *to,
2907 const unsigned int len)
2909 memcpy(to, skb->data + offset, len);
2912 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2913 const void *from,
2914 const unsigned int len)
2916 memcpy(skb->data, from, len);
2919 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2920 const int offset,
2921 const void *from,
2922 const unsigned int len)
2924 memcpy(skb->data + offset, from, len);
2927 void skb_init(void);
2929 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2931 return skb->tstamp;
2935 * skb_get_timestamp - get timestamp from a skb
2936 * @skb: skb to get stamp from
2937 * @stamp: pointer to struct timeval to store stamp in
2939 * Timestamps are stored in the skb as offsets to a base timestamp.
2940 * This function converts the offset back to a struct timeval and stores
2941 * it in stamp.
2943 static inline void skb_get_timestamp(const struct sk_buff *skb,
2944 struct timeval *stamp)
2946 *stamp = ktime_to_timeval(skb->tstamp);
2949 static inline void skb_get_timestampns(const struct sk_buff *skb,
2950 struct timespec *stamp)
2952 *stamp = ktime_to_timespec(skb->tstamp);
2955 static inline void __net_timestamp(struct sk_buff *skb)
2957 skb->tstamp = ktime_get_real();
2960 static inline ktime_t net_timedelta(ktime_t t)
2962 return ktime_sub(ktime_get_real(), t);
2965 static inline ktime_t net_invalid_timestamp(void)
2967 return ktime_set(0, 0);
2970 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2972 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2974 void skb_clone_tx_timestamp(struct sk_buff *skb);
2975 bool skb_defer_rx_timestamp(struct sk_buff *skb);
2977 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2979 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2983 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2985 return false;
2988 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2991 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2993 * PHY drivers may accept clones of transmitted packets for
2994 * timestamping via their phy_driver.txtstamp method. These drivers
2995 * must call this function to return the skb back to the stack with a
2996 * timestamp.
2998 * @skb: clone of the the original outgoing packet
2999 * @hwtstamps: hardware time stamps
3002 void skb_complete_tx_timestamp(struct sk_buff *skb,
3003 struct skb_shared_hwtstamps *hwtstamps);
3005 void __skb_tstamp_tx(struct sk_buff *orig_skb,
3006 struct skb_shared_hwtstamps *hwtstamps,
3007 struct sock *sk, int tstype);
3010 * skb_tstamp_tx - queue clone of skb with send time stamps
3011 * @orig_skb: the original outgoing packet
3012 * @hwtstamps: hardware time stamps, may be NULL if not available
3014 * If the skb has a socket associated, then this function clones the
3015 * skb (thus sharing the actual data and optional structures), stores
3016 * the optional hardware time stamping information (if non NULL) or
3017 * generates a software time stamp (otherwise), then queues the clone
3018 * to the error queue of the socket. Errors are silently ignored.
3020 void skb_tstamp_tx(struct sk_buff *orig_skb,
3021 struct skb_shared_hwtstamps *hwtstamps);
3023 static inline void sw_tx_timestamp(struct sk_buff *skb)
3025 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3026 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
3027 skb_tstamp_tx(skb, NULL);
3031 * skb_tx_timestamp() - Driver hook for transmit timestamping
3033 * Ethernet MAC Drivers should call this function in their hard_xmit()
3034 * function immediately before giving the sk_buff to the MAC hardware.
3036 * Specifically, one should make absolutely sure that this function is
3037 * called before TX completion of this packet can trigger. Otherwise
3038 * the packet could potentially already be freed.
3040 * @skb: A socket buffer.
3042 static inline void skb_tx_timestamp(struct sk_buff *skb)
3044 skb_clone_tx_timestamp(skb);
3045 sw_tx_timestamp(skb);
3049 * skb_complete_wifi_ack - deliver skb with wifi status
3051 * @skb: the original outgoing packet
3052 * @acked: ack status
3055 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3057 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3058 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3060 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3062 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3063 skb->csum_valid ||
3064 (skb->ip_summed == CHECKSUM_PARTIAL &&
3065 skb_checksum_start_offset(skb) >= 0));
3069 * skb_checksum_complete - Calculate checksum of an entire packet
3070 * @skb: packet to process
3072 * This function calculates the checksum over the entire packet plus
3073 * the value of skb->csum. The latter can be used to supply the
3074 * checksum of a pseudo header as used by TCP/UDP. It returns the
3075 * checksum.
3077 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3078 * this function can be used to verify that checksum on received
3079 * packets. In that case the function should return zero if the
3080 * checksum is correct. In particular, this function will return zero
3081 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3082 * hardware has already verified the correctness of the checksum.
3084 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
3086 return skb_csum_unnecessary(skb) ?
3087 0 : __skb_checksum_complete(skb);
3090 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3092 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3093 if (skb->csum_level == 0)
3094 skb->ip_summed = CHECKSUM_NONE;
3095 else
3096 skb->csum_level--;
3100 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3102 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3103 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3104 skb->csum_level++;
3105 } else if (skb->ip_summed == CHECKSUM_NONE) {
3106 skb->ip_summed = CHECKSUM_UNNECESSARY;
3107 skb->csum_level = 0;
3111 static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3113 /* Mark current checksum as bad (typically called from GRO
3114 * path). In the case that ip_summed is CHECKSUM_NONE
3115 * this must be the first checksum encountered in the packet.
3116 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3117 * checksum after the last one validated. For UDP, a zero
3118 * checksum can not be marked as bad.
3121 if (skb->ip_summed == CHECKSUM_NONE ||
3122 skb->ip_summed == CHECKSUM_UNNECESSARY)
3123 skb->csum_bad = 1;
3126 /* Check if we need to perform checksum complete validation.
3128 * Returns true if checksum complete is needed, false otherwise
3129 * (either checksum is unnecessary or zero checksum is allowed).
3131 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3132 bool zero_okay,
3133 __sum16 check)
3135 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3136 skb->csum_valid = 1;
3137 __skb_decr_checksum_unnecessary(skb);
3138 return false;
3141 return true;
3144 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3145 * in checksum_init.
3147 #define CHECKSUM_BREAK 76
3149 /* Unset checksum-complete
3151 * Unset checksum complete can be done when packet is being modified
3152 * (uncompressed for instance) and checksum-complete value is
3153 * invalidated.
3155 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3157 if (skb->ip_summed == CHECKSUM_COMPLETE)
3158 skb->ip_summed = CHECKSUM_NONE;
3161 /* Validate (init) checksum based on checksum complete.
3163 * Return values:
3164 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3165 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3166 * checksum is stored in skb->csum for use in __skb_checksum_complete
3167 * non-zero: value of invalid checksum
3170 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3171 bool complete,
3172 __wsum psum)
3174 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3175 if (!csum_fold(csum_add(psum, skb->csum))) {
3176 skb->csum_valid = 1;
3177 return 0;
3179 } else if (skb->csum_bad) {
3180 /* ip_summed == CHECKSUM_NONE in this case */
3181 return (__force __sum16)1;
3184 skb->csum = psum;
3186 if (complete || skb->len <= CHECKSUM_BREAK) {
3187 __sum16 csum;
3189 csum = __skb_checksum_complete(skb);
3190 skb->csum_valid = !csum;
3191 return csum;
3194 return 0;
3197 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3199 return 0;
3202 /* Perform checksum validate (init). Note that this is a macro since we only
3203 * want to calculate the pseudo header which is an input function if necessary.
3204 * First we try to validate without any computation (checksum unnecessary) and
3205 * then calculate based on checksum complete calling the function to compute
3206 * pseudo header.
3208 * Return values:
3209 * 0: checksum is validated or try to in skb_checksum_complete
3210 * non-zero: value of invalid checksum
3212 #define __skb_checksum_validate(skb, proto, complete, \
3213 zero_okay, check, compute_pseudo) \
3214 ({ \
3215 __sum16 __ret = 0; \
3216 skb->csum_valid = 0; \
3217 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3218 __ret = __skb_checksum_validate_complete(skb, \
3219 complete, compute_pseudo(skb, proto)); \
3220 __ret; \
3223 #define skb_checksum_init(skb, proto, compute_pseudo) \
3224 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3226 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3227 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3229 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3230 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3232 #define skb_checksum_validate_zero_check(skb, proto, check, \
3233 compute_pseudo) \
3234 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3236 #define skb_checksum_simple_validate(skb) \
3237 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3239 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3241 return (skb->ip_summed == CHECKSUM_NONE &&
3242 skb->csum_valid && !skb->csum_bad);
3245 static inline void __skb_checksum_convert(struct sk_buff *skb,
3246 __sum16 check, __wsum pseudo)
3248 skb->csum = ~pseudo;
3249 skb->ip_summed = CHECKSUM_COMPLETE;
3252 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3253 do { \
3254 if (__skb_checksum_convert_check(skb)) \
3255 __skb_checksum_convert(skb, check, \
3256 compute_pseudo(skb, proto)); \
3257 } while (0)
3259 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3260 u16 start, u16 offset)
3262 skb->ip_summed = CHECKSUM_PARTIAL;
3263 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3264 skb->csum_offset = offset - start;
3267 /* Update skbuf and packet to reflect the remote checksum offload operation.
3268 * When called, ptr indicates the starting point for skb->csum when
3269 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3270 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3272 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
3273 int start, int offset, bool nopartial)
3275 __wsum delta;
3277 if (!nopartial) {
3278 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3279 return;
3282 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3283 __skb_checksum_complete(skb);
3284 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3287 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3289 /* Adjust skb->csum since we changed the packet */
3290 skb->csum = csum_add(skb->csum, delta);
3293 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3294 void nf_conntrack_destroy(struct nf_conntrack *nfct);
3295 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3297 if (nfct && atomic_dec_and_test(&nfct->use))
3298 nf_conntrack_destroy(nfct);
3300 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3302 if (nfct)
3303 atomic_inc(&nfct->use);
3305 #endif
3306 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3307 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3309 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3310 kfree(nf_bridge);
3312 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3314 if (nf_bridge)
3315 atomic_inc(&nf_bridge->use);
3317 #endif /* CONFIG_BRIDGE_NETFILTER */
3318 static inline void nf_reset(struct sk_buff *skb)
3320 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3321 nf_conntrack_put(skb->nfct);
3322 skb->nfct = NULL;
3323 #endif
3324 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3325 nf_bridge_put(skb->nf_bridge);
3326 skb->nf_bridge = NULL;
3327 #endif
3330 static inline void nf_reset_trace(struct sk_buff *skb)
3332 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3333 skb->nf_trace = 0;
3334 #endif
3337 /* Note: This doesn't put any conntrack and bridge info in dst. */
3338 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3339 bool copy)
3341 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3342 dst->nfct = src->nfct;
3343 nf_conntrack_get(src->nfct);
3344 if (copy)
3345 dst->nfctinfo = src->nfctinfo;
3346 #endif
3347 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3348 dst->nf_bridge = src->nf_bridge;
3349 nf_bridge_get(src->nf_bridge);
3350 #endif
3351 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3352 if (copy)
3353 dst->nf_trace = src->nf_trace;
3354 #endif
3357 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3359 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3360 nf_conntrack_put(dst->nfct);
3361 #endif
3362 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3363 nf_bridge_put(dst->nf_bridge);
3364 #endif
3365 __nf_copy(dst, src, true);
3368 #ifdef CONFIG_NETWORK_SECMARK
3369 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3371 to->secmark = from->secmark;
3374 static inline void skb_init_secmark(struct sk_buff *skb)
3376 skb->secmark = 0;
3378 #else
3379 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3382 static inline void skb_init_secmark(struct sk_buff *skb)
3384 #endif
3386 static inline bool skb_irq_freeable(const struct sk_buff *skb)
3388 return !skb->destructor &&
3389 #if IS_ENABLED(CONFIG_XFRM)
3390 !skb->sp &&
3391 #endif
3392 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3393 !skb->nfct &&
3394 #endif
3395 !skb->_skb_refdst &&
3396 !skb_has_frag_list(skb);
3399 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3401 skb->queue_mapping = queue_mapping;
3404 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
3406 return skb->queue_mapping;
3409 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3411 to->queue_mapping = from->queue_mapping;
3414 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3416 skb->queue_mapping = rx_queue + 1;
3419 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
3421 return skb->queue_mapping - 1;
3424 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
3426 return skb->queue_mapping != 0;
3429 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3431 #ifdef CONFIG_XFRM
3432 return skb->sp;
3433 #else
3434 return NULL;
3435 #endif
3438 /* Keeps track of mac header offset relative to skb->head.
3439 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3440 * For non-tunnel skb it points to skb_mac_header() and for
3441 * tunnel skb it points to outer mac header.
3442 * Keeps track of level of encapsulation of network headers.
3444 struct skb_gso_cb {
3445 int mac_offset;
3446 int encap_level;
3447 __u16 csum_start;
3449 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3451 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3453 return (skb_mac_header(inner_skb) - inner_skb->head) -
3454 SKB_GSO_CB(inner_skb)->mac_offset;
3457 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3459 int new_headroom, headroom;
3460 int ret;
3462 headroom = skb_headroom(skb);
3463 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3464 if (ret)
3465 return ret;
3467 new_headroom = skb_headroom(skb);
3468 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3469 return 0;
3472 /* Compute the checksum for a gso segment. First compute the checksum value
3473 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3474 * then add in skb->csum (checksum from csum_start to end of packet).
3475 * skb->csum and csum_start are then updated to reflect the checksum of the
3476 * resultant packet starting from the transport header-- the resultant checksum
3477 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3478 * header.
3480 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3482 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3483 skb_transport_offset(skb);
3484 __wsum partial;
3486 partial = csum_partial(skb_transport_header(skb), plen, skb->csum);
3487 skb->csum = res;
3488 SKB_GSO_CB(skb)->csum_start -= plen;
3490 return csum_fold(partial);
3493 static inline bool skb_is_gso(const struct sk_buff *skb)
3495 return skb_shinfo(skb)->gso_size;
3498 /* Note: Should be called only if skb_is_gso(skb) is true */
3499 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
3501 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3504 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
3506 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3508 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3509 * wanted then gso_type will be set. */
3510 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3512 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3513 unlikely(shinfo->gso_type == 0)) {
3514 __skb_warn_lro_forwarding(skb);
3515 return true;
3517 return false;
3520 static inline void skb_forward_csum(struct sk_buff *skb)
3522 /* Unfortunately we don't support this one. Any brave souls? */
3523 if (skb->ip_summed == CHECKSUM_COMPLETE)
3524 skb->ip_summed = CHECKSUM_NONE;
3528 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3529 * @skb: skb to check
3531 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3532 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3533 * use this helper, to document places where we make this assertion.
3535 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
3537 #ifdef DEBUG
3538 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3539 #endif
3542 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
3544 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3545 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3546 unsigned int transport_len,
3547 __sum16(*skb_chkf)(struct sk_buff *skb));
3550 * skb_head_is_locked - Determine if the skb->head is locked down
3551 * @skb: skb to check
3553 * The head on skbs build around a head frag can be removed if they are
3554 * not cloned. This function returns true if the skb head is locked down
3555 * due to either being allocated via kmalloc, or by being a clone with
3556 * multiple references to the head.
3558 static inline bool skb_head_is_locked(const struct sk_buff *skb)
3560 return !skb->head_frag || skb_cloned(skb);
3564 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3566 * @skb: GSO skb
3568 * skb_gso_network_seglen is used to determine the real size of the
3569 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3571 * The MAC/L2 header is not accounted for.
3573 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3575 unsigned int hdr_len = skb_transport_header(skb) -
3576 skb_network_header(skb);
3577 return hdr_len + skb_gso_transport_seglen(skb);
3580 #endif /* __KERNEL__ */
3581 #endif /* _LINUX_SKBUFF_H */