]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/skbuff.h
net: do not pretend FRAGLIST support
[mirror_ubuntu-bionic-kernel.git] / include / linux / skbuff.h
CommitLineData
1da177e4
LT
1/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
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.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
1da177e4 17#include <linux/kernel.h>
fe55f6d5 18#include <linux/kmemcheck.h>
1da177e4
LT
19#include <linux/compiler.h>
20#include <linux/time.h>
187f1882 21#include <linux/bug.h>
1da177e4
LT
22#include <linux/cache.h>
23
60063497 24#include <linux/atomic.h>
1da177e4
LT
25#include <asm/types.h>
26#include <linux/spinlock.h>
1da177e4 27#include <linux/net.h>
3fc7e8a6 28#include <linux/textsearch.h>
1da177e4 29#include <net/checksum.h>
a80958f4 30#include <linux/rcupdate.h>
97fc2f08 31#include <linux/dmaengine.h>
b7aa0bf7 32#include <linux/hrtimer.h>
131ea667 33#include <linux/dma-mapping.h>
c8f44aff 34#include <linux/netdev_features.h>
5203cd28 35#include <net/flow_keys.h>
1da177e4 36
60476372 37/* Don't change this without changing skb_csum_unnecessary! */
1da177e4 38#define CHECKSUM_NONE 0
60476372
HX
39#define CHECKSUM_UNNECESSARY 1
40#define CHECKSUM_COMPLETE 2
41#define CHECKSUM_PARTIAL 3
1da177e4
LT
42
43#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
44 ~(SMP_CACHE_BYTES - 1))
fc910a27 45#define SKB_WITH_OVERHEAD(X) \
deea84b0 46 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
47#define SKB_MAX_ORDER(X, ORDER) \
48 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
49#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
50#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
51
87fb4b7b
ED
52/* return minimum truesize of one skb containing X bytes of data */
53#define SKB_TRUESIZE(X) ((X) + \
54 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
55 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
56
1da177e4
LT
57/* A. Checksumming of received packets by device.
58 *
59 * NONE: device failed to checksum this packet.
60 * skb->csum is undefined.
61 *
62 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
63 * skb->csum is undefined.
64 * It is bad option, but, unfortunately, many of vendors do this.
65 * Apparently with secret goal to sell you new device, when you
66 * will add new protocol to your host. F.e. IPv6. 8)
67 *
84fa7933 68 * COMPLETE: the most generic way. Device supplied checksum of _all_
1da177e4
LT
69 * the packet as seen by netif_rx in skb->csum.
70 * NOTE: Even if device supports only some protocols, but
84fa7933 71 * is able to produce some skb->csum, it MUST use COMPLETE,
1da177e4
LT
72 * not UNNECESSARY.
73 *
c6c6e3e0
HX
74 * PARTIAL: identical to the case for output below. This may occur
75 * on a packet received directly from another Linux OS, e.g.,
76 * a virtualised Linux kernel on the same host. The packet can
77 * be treated in the same way as UNNECESSARY except that on
78 * output (i.e., forwarding) the checksum must be filled in
79 * by the OS or the hardware.
80 *
1da177e4
LT
81 * B. Checksumming on output.
82 *
83 * NONE: skb is checksummed by protocol or csum is not required.
84 *
84fa7933 85 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
c6c6e3e0
HX
86 * from skb->csum_start to the end and to record the checksum
87 * at skb->csum_start + skb->csum_offset.
1da177e4
LT
88 *
89 * Device must show its capabilities in dev->features, set
90 * at device setup time.
91 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
92 * everything.
1da177e4
LT
93 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
94 * TCP/UDP over IPv4. Sigh. Vendors like this
95 * way by an unknown reason. Though, see comment above
96 * about CHECKSUM_UNNECESSARY. 8)
c6c6e3e0 97 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
1da177e4 98 *
3af79302
YZ
99 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
100 * that do not want net to perform the checksum calculation should use
101 * this flag in their outgoing skbs.
102 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
103 * offload. Correspondingly, the FCoE protocol driver
104 * stack should use CHECKSUM_UNNECESSARY.
105 *
1da177e4
LT
106 * Any questions? No questions, good. --ANK
107 */
108
1da177e4 109struct net_device;
716ea3a7 110struct scatterlist;
9c55e01c 111struct pipe_inode_info;
1da177e4 112
5f79e0f9 113#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
114struct nf_conntrack {
115 atomic_t use;
1da177e4 116};
5f79e0f9 117#endif
1da177e4
LT
118
119#ifdef CONFIG_BRIDGE_NETFILTER
120struct nf_bridge_info {
bf1ac5ca
ED
121 atomic_t use;
122 unsigned int mask;
123 struct net_device *physindev;
124 struct net_device *physoutdev;
125 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
126};
127#endif
128
1da177e4
LT
129struct sk_buff_head {
130 /* These two members must be first. */
131 struct sk_buff *next;
132 struct sk_buff *prev;
133
134 __u32 qlen;
135 spinlock_t lock;
136};
137
138struct sk_buff;
139
9d4dde52
IC
140/* To allow 64K frame to be packed as single skb without frag_list we
141 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
142 * buffers which do not start on a page boundary.
143 *
144 * Since GRO uses frags we allocate at least 16 regardless of page
145 * size.
a715dea3 146 */
9d4dde52 147#if (65536/PAGE_SIZE + 1) < 16
eec00954 148#define MAX_SKB_FRAGS 16UL
a715dea3 149#else
9d4dde52 150#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 151#endif
1da177e4
LT
152
153typedef struct skb_frag_struct skb_frag_t;
154
155struct skb_frag_struct {
a8605c60
IC
156 struct {
157 struct page *p;
158 } page;
cb4dfe56 159#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
160 __u32 page_offset;
161 __u32 size;
cb4dfe56
ED
162#else
163 __u16 page_offset;
164 __u16 size;
165#endif
1da177e4
LT
166};
167
9e903e08
ED
168static inline unsigned int skb_frag_size(const skb_frag_t *frag)
169{
170 return frag->size;
171}
172
173static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
174{
175 frag->size = size;
176}
177
178static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
179{
180 frag->size += delta;
181}
182
183static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
184{
185 frag->size -= delta;
186}
187
ac45f602
PO
188#define HAVE_HW_TIME_STAMP
189
190/**
d3a21be8 191 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
192 * @hwtstamp: hardware time stamp transformed into duration
193 * since arbitrary point in time
194 * @syststamp: hwtstamp transformed to system time base
195 *
196 * Software time stamps generated by ktime_get_real() are stored in
197 * skb->tstamp. The relation between the different kinds of time
198 * stamps is as follows:
199 *
200 * syststamp and tstamp can be compared against each other in
201 * arbitrary combinations. The accuracy of a
202 * syststamp/tstamp/"syststamp from other device" comparison is
203 * limited by the accuracy of the transformation into system time
204 * base. This depends on the device driver and its underlying
205 * hardware.
206 *
207 * hwtstamps can only be compared against other hwtstamps from
208 * the same device.
209 *
210 * This structure is attached to packets as part of the
211 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
212 */
213struct skb_shared_hwtstamps {
214 ktime_t hwtstamp;
215 ktime_t syststamp;
216};
217
2244d07b
OH
218/* Definitions for tx_flags in struct skb_shared_info */
219enum {
220 /* generate hardware time stamp */
221 SKBTX_HW_TSTAMP = 1 << 0,
222
223 /* generate software time stamp */
224 SKBTX_SW_TSTAMP = 1 << 1,
225
226 /* device driver is going to provide hardware time stamp */
227 SKBTX_IN_PROGRESS = 1 << 2,
228
a6686f2f 229 /* device driver supports TX zero-copy buffers */
62b1a8ab 230 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
231
232 /* generate wifi status information (where possible) */
62b1a8ab 233 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
234
235 /* This indicates at least one fragment might be overwritten
236 * (as in vmsplice(), sendfile() ...)
237 * If we need to compute a TX checksum, we'll need to copy
238 * all frags to avoid possible bad checksum
239 */
240 SKBTX_SHARED_FRAG = 1 << 5,
a6686f2f
SM
241};
242
243/*
244 * The callback notifies userspace to release buffers when skb DMA is done in
245 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
246 * The zerocopy_success argument is true if zero copy transmit occurred,
247 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
248 * The ctx field is used to track device context.
249 * The desc field is used to track userspace buffer index.
a6686f2f
SM
250 */
251struct ubuf_info {
e19d6763 252 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 253 void *ctx;
a6686f2f 254 unsigned long desc;
ac45f602
PO
255};
256
1da177e4
LT
257/* This data is invariant across clones and lives at
258 * the end of the header data, ie. at skb->end.
259 */
260struct skb_shared_info {
9f42f126
IC
261 unsigned char nr_frags;
262 __u8 tx_flags;
7967168c
HX
263 unsigned short gso_size;
264 /* Warning: this field is not always filled in (UFO)! */
265 unsigned short gso_segs;
266 unsigned short gso_type;
1da177e4 267 struct sk_buff *frag_list;
ac45f602 268 struct skb_shared_hwtstamps hwtstamps;
9f42f126 269 __be32 ip6_frag_id;
ec7d2f2c
ED
270
271 /*
272 * Warning : all fields before dataref are cleared in __alloc_skb()
273 */
274 atomic_t dataref;
275
69e3c75f
JB
276 /* Intermediate layers must ensure that destructor_arg
277 * remains valid until skb destructor */
278 void * destructor_arg;
a6686f2f 279
fed66381
ED
280 /* must be last field, see pskb_expand_head() */
281 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
282};
283
284/* We divide dataref into two halves. The higher 16 bits hold references
285 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
286 * the entire skb->data. A clone of a headerless skb holds the length of
287 * the header in skb->hdr_len.
1da177e4
LT
288 *
289 * All users must obey the rule that the skb->data reference count must be
290 * greater than or equal to the payload reference count.
291 *
292 * Holding a reference to the payload part means that the user does not
293 * care about modifications to the header part of skb->data.
294 */
295#define SKB_DATAREF_SHIFT 16
296#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
297
d179cd12
DM
298
299enum {
300 SKB_FCLONE_UNAVAILABLE,
301 SKB_FCLONE_ORIG,
302 SKB_FCLONE_CLONE,
303};
304
7967168c
HX
305enum {
306 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 307 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
308
309 /* This indicates the skb is from an untrusted source. */
310 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
311
312 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
313 SKB_GSO_TCP_ECN = 1 << 3,
314
315 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
316
317 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
318
319 SKB_GSO_GRE = 1 << 6,
73136267 320
cb32f511 321 SKB_GSO_IPIP = 1 << 7,
0d89d203 322
61c1db7f 323 SKB_GSO_SIT = 1 << 8,
cb32f511 324
61c1db7f
ED
325 SKB_GSO_UDP_TUNNEL = 1 << 9,
326
327 SKB_GSO_MPLS = 1 << 10,
7967168c
HX
328};
329
2e07fa9c
ACM
330#if BITS_PER_LONG > 32
331#define NET_SKBUFF_DATA_USES_OFFSET 1
332#endif
333
334#ifdef NET_SKBUFF_DATA_USES_OFFSET
335typedef unsigned int sk_buff_data_t;
336#else
337typedef unsigned char *sk_buff_data_t;
338#endif
339
1da177e4
LT
340/**
341 * struct sk_buff - socket buffer
342 * @next: Next buffer in list
343 * @prev: Previous buffer in list
325ed823 344 * @tstamp: Time we arrived
d84e0bd7 345 * @sk: Socket we are owned by
1da177e4 346 * @dev: Device we arrived on/are leaving by
d84e0bd7 347 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 348 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 349 * @sp: the security path, used for xfrm
1da177e4
LT
350 * @len: Length of actual data
351 * @data_len: Data length
352 * @mac_len: Length of link layer header
334a8132 353 * @hdr_len: writable header length of cloned skb
663ead3b
HX
354 * @csum: Checksum (must include start/offset pair)
355 * @csum_start: Offset from skb->head where checksumming should start
356 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 357 * @priority: Packet queueing priority
67be2dd1 358 * @local_df: allow local fragmentation
1da177e4 359 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 360 * @ip_summed: Driver fed us an IP checksum
1da177e4 361 * @nohdr: Payload reference only, must not modify header
d84e0bd7 362 * @nfctinfo: Relationship of this skb to the connection
1da177e4 363 * @pkt_type: Packet class
c83c2486 364 * @fclone: skbuff clone status
c83c2486 365 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
366 * @peeked: this packet has been seen already, so stats have been
367 * done for it, don't do them again
ba9dda3a 368 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
369 * @protocol: Packet protocol from driver
370 * @destructor: Destruct function
371 * @nfct: Associated connection, if any
1da177e4 372 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 373 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
374 * @tc_index: Traffic control index
375 * @tc_verd: traffic control verdict
d84e0bd7
DB
376 * @rxhash: the packet hash computed on receive
377 * @queue_mapping: Queue mapping for multiqueue devices
553a5672 378 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 379 * @ooo_okay: allow the mapping of a socket to a queue to be changed
4ca2462e
CG
380 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
381 * ports.
6e3e939f
JB
382 * @wifi_acked_valid: wifi_acked was set
383 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 384 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
f4b8ea78
RD
385 * @dma_cookie: a cookie to one of several possible DMA operations
386 * done by skb DMA functions
06021292 387 * @napi_id: id of the NAPI struct this skb came from
984bc16c 388 * @secmark: security marking
d84e0bd7
DB
389 * @mark: Generic packet mark
390 * @dropcount: total number of sk_receive_queue overflows
86a9bad3 391 * @vlan_proto: vlan encapsulation protocol
6aa895b0 392 * @vlan_tci: vlan tag control information
0d89d203 393 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
394 * @inner_transport_header: Inner transport layer header (encapsulation)
395 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 396 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
397 * @transport_header: Transport layer header
398 * @network_header: Network layer header
399 * @mac_header: Link layer header
400 * @tail: Tail pointer
401 * @end: End pointer
402 * @head: Head of buffer
403 * @data: Data head pointer
404 * @truesize: Buffer size
405 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
406 */
407
408struct sk_buff {
409 /* These two members must be first. */
410 struct sk_buff *next;
411 struct sk_buff *prev;
412
b7aa0bf7 413 ktime_t tstamp;
da3f5cf1
FF
414
415 struct sock *sk;
1da177e4 416 struct net_device *dev;
1da177e4 417
1da177e4
LT
418 /*
419 * This is the control buffer. It is free to use for every
420 * layer. Please put your private variables there. If you
421 * want to keep them across layers you have to do a skb_clone()
422 * first. This is owned by whoever has the skb queued ATM.
423 */
da3f5cf1 424 char cb[48] __aligned(8);
1da177e4 425
7fee226a 426 unsigned long _skb_refdst;
da3f5cf1
FF
427#ifdef CONFIG_XFRM
428 struct sec_path *sp;
429#endif
1da177e4 430 unsigned int len,
334a8132
PM
431 data_len;
432 __u16 mac_len,
433 hdr_len;
ff1dcadb
AV
434 union {
435 __wsum csum;
663ead3b
HX
436 struct {
437 __u16 csum_start;
438 __u16 csum_offset;
439 };
ff1dcadb 440 };
1da177e4 441 __u32 priority;
fe55f6d5 442 kmemcheck_bitfield_begin(flags1);
1cbb3380
TG
443 __u8 local_df:1,
444 cloned:1,
445 ip_summed:2,
6869c4d8
HW
446 nohdr:1,
447 nfctinfo:3;
d179cd12 448 __u8 pkt_type:3,
b84f4cc9 449 fclone:2,
ba9dda3a 450 ipvs_property:1,
a59322be 451 peeked:1,
ba9dda3a 452 nf_trace:1;
fe55f6d5 453 kmemcheck_bitfield_end(flags1);
4ab408de 454 __be16 protocol;
1da177e4
LT
455
456 void (*destructor)(struct sk_buff *skb);
9fb9cbb1 457#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 458 struct nf_conntrack *nfct;
2fc72c7b 459#endif
1da177e4
LT
460#ifdef CONFIG_BRIDGE_NETFILTER
461 struct nf_bridge_info *nf_bridge;
462#endif
f25f4e44 463
8964be4a 464 int skb_iif;
4031ae6e
AD
465
466 __u32 rxhash;
467
86a9bad3 468 __be16 vlan_proto;
4031ae6e
AD
469 __u16 vlan_tci;
470
1da177e4 471#ifdef CONFIG_NET_SCHED
b6b99eb5 472 __u16 tc_index; /* traffic control index */
1da177e4 473#ifdef CONFIG_NET_CLS_ACT
b6b99eb5 474 __u16 tc_verd; /* traffic control verdict */
1da177e4 475#endif
1da177e4 476#endif
fe55f6d5 477
0a14842f 478 __u16 queue_mapping;
fe55f6d5 479 kmemcheck_bitfield_begin(flags2);
de357cc0 480#ifdef CONFIG_IPV6_NDISC_NODETYPE
8a4eb573 481 __u8 ndisc_nodetype:2;
d0f09804 482#endif
c93bdd0e 483 __u8 pfmemalloc:1;
3853b584 484 __u8 ooo_okay:1;
bdeab991 485 __u8 l4_rxhash:1;
6e3e939f
JB
486 __u8 wifi_acked_valid:1;
487 __u8 wifi_acked:1;
3bdc0eba 488 __u8 no_fcs:1;
d3836f21 489 __u8 head_frag:1;
6a674e9c
JG
490 /* Encapsulation protocol and NIC drivers should use
491 * this flag to indicate to each other if the skb contains
492 * encapsulated packet or not and maybe use the inner packet
493 * headers if needed
494 */
495 __u8 encapsulation:1;
45906723 496 /* 6/8 bit hole (depending on ndisc_nodetype presence) */
fe55f6d5
VN
497 kmemcheck_bitfield_end(flags2);
498
e0d1095a 499#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
06021292
ET
500 union {
501 unsigned int napi_id;
502 dma_cookie_t dma_cookie;
503 };
97fc2f08 504#endif
984bc16c
JM
505#ifdef CONFIG_NETWORK_SECMARK
506 __u32 secmark;
507#endif
3b885787
NH
508 union {
509 __u32 mark;
510 __u32 dropcount;
16fad69c 511 __u32 reserved_tailroom;
3b885787 512 };
1da177e4 513
0d89d203 514 __be16 inner_protocol;
1a37e412
SH
515 __u16 inner_transport_header;
516 __u16 inner_network_header;
517 __u16 inner_mac_header;
518 __u16 transport_header;
519 __u16 network_header;
520 __u16 mac_header;
1da177e4 521 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 522 sk_buff_data_t tail;
4305b541 523 sk_buff_data_t end;
1da177e4 524 unsigned char *head,
4305b541 525 *data;
27a884dc
ACM
526 unsigned int truesize;
527 atomic_t users;
1da177e4
LT
528};
529
530#ifdef __KERNEL__
531/*
532 * Handling routines are only of interest to the kernel
533 */
534#include <linux/slab.h>
535
1da177e4 536
c93bdd0e
MG
537#define SKB_ALLOC_FCLONE 0x01
538#define SKB_ALLOC_RX 0x02
539
540/* Returns true if the skb was allocated from PFMEMALLOC reserves */
541static inline bool skb_pfmemalloc(const struct sk_buff *skb)
542{
543 return unlikely(skb->pfmemalloc);
544}
545
7fee226a
ED
546/*
547 * skb might have a dst pointer attached, refcounted or not.
548 * _skb_refdst low order bit is set if refcount was _not_ taken
549 */
550#define SKB_DST_NOREF 1UL
551#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
552
553/**
554 * skb_dst - returns skb dst_entry
555 * @skb: buffer
556 *
557 * Returns skb dst_entry, regardless of reference taken or not.
558 */
adf30907
ED
559static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
560{
7fee226a
ED
561 /* If refdst was not refcounted, check we still are in a
562 * rcu_read_lock section
563 */
564 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
565 !rcu_read_lock_held() &&
566 !rcu_read_lock_bh_held());
567 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
568}
569
7fee226a
ED
570/**
571 * skb_dst_set - sets skb dst
572 * @skb: buffer
573 * @dst: dst entry
574 *
575 * Sets skb dst, assuming a reference was taken on dst and should
576 * be released by skb_dst_drop()
577 */
adf30907
ED
578static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
579{
7fee226a
ED
580 skb->_skb_refdst = (unsigned long)dst;
581}
582
7965bd4d
JP
583void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
584 bool force);
932bc4d7
JA
585
586/**
587 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
588 * @skb: buffer
589 * @dst: dst entry
590 *
591 * Sets skb dst, assuming a reference was not taken on dst.
592 * If dst entry is cached, we do not take reference and dst_release
593 * will be avoided by refdst_drop. If dst entry is not cached, we take
594 * reference, so that last dst_release can destroy the dst immediately.
595 */
596static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
597{
598 __skb_dst_set_noref(skb, dst, false);
599}
600
601/**
602 * skb_dst_set_noref_force - sets skb dst, without taking reference
603 * @skb: buffer
604 * @dst: dst entry
605 *
606 * Sets skb dst, assuming a reference was not taken on dst.
607 * No reference is taken and no dst_release will be called. While for
608 * cached dsts deferred reclaim is a basic feature, for entries that are
609 * not cached it is caller's job to guarantee that last dst_release for
610 * provided dst happens when nobody uses it, eg. after a RCU grace period.
611 */
612static inline void skb_dst_set_noref_force(struct sk_buff *skb,
613 struct dst_entry *dst)
614{
615 __skb_dst_set_noref(skb, dst, true);
616}
7fee226a
ED
617
618/**
25985edc 619 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
620 * @skb: buffer
621 */
622static inline bool skb_dst_is_noref(const struct sk_buff *skb)
623{
624 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
625}
626
511c3f92
ED
627static inline struct rtable *skb_rtable(const struct sk_buff *skb)
628{
adf30907 629 return (struct rtable *)skb_dst(skb);
511c3f92
ED
630}
631
7965bd4d
JP
632void kfree_skb(struct sk_buff *skb);
633void kfree_skb_list(struct sk_buff *segs);
634void skb_tx_error(struct sk_buff *skb);
635void consume_skb(struct sk_buff *skb);
636void __kfree_skb(struct sk_buff *skb);
d7e8883c 637extern struct kmem_cache *skbuff_head_cache;
bad43ca8 638
7965bd4d
JP
639void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
640bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
641 bool *fragstolen, int *delta_truesize);
bad43ca8 642
7965bd4d
JP
643struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
644 int node);
645struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 646static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 647 gfp_t priority)
d179cd12 648{
564824b0 649 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
650}
651
652static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 653 gfp_t priority)
d179cd12 654{
c93bdd0e 655 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
656}
657
7965bd4d 658struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
659static inline struct sk_buff *alloc_skb_head(gfp_t priority)
660{
661 return __alloc_skb_head(priority, -1);
662}
663
7965bd4d
JP
664struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
665int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
666struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
667struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
668struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
669
670int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
671struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
672 unsigned int headroom);
673struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
674 int newtailroom, gfp_t priority);
675int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
676 int len);
677int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
678int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 679#define dev_kfree_skb(a) consume_skb(a)
1da177e4 680
7965bd4d
JP
681int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
682 int getfrag(void *from, char *to, int offset,
683 int len, int odd, struct sk_buff *skb),
684 void *from, int length);
e89e9cf5 685
d94d9fee 686struct skb_seq_state {
677e90ed
TG
687 __u32 lower_offset;
688 __u32 upper_offset;
689 __u32 frag_idx;
690 __u32 stepped_offset;
691 struct sk_buff *root_skb;
692 struct sk_buff *cur_skb;
693 __u8 *frag_data;
694};
695
7965bd4d
JP
696void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
697 unsigned int to, struct skb_seq_state *st);
698unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
699 struct skb_seq_state *st);
700void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 701
7965bd4d
JP
702unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
703 unsigned int to, struct ts_config *config,
704 struct ts_state *state);
3fc7e8a6 705
7965bd4d 706void __skb_get_rxhash(struct sk_buff *skb);
bfb564e7
KK
707static inline __u32 skb_get_rxhash(struct sk_buff *skb)
708{
ecd5cf5d 709 if (!skb->l4_rxhash)
bdeab991 710 __skb_get_rxhash(skb);
bfb564e7
KK
711
712 return skb->rxhash;
713}
714
4305b541
ACM
715#ifdef NET_SKBUFF_DATA_USES_OFFSET
716static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
717{
718 return skb->head + skb->end;
719}
ec47ea82
AD
720
721static inline unsigned int skb_end_offset(const struct sk_buff *skb)
722{
723 return skb->end;
724}
4305b541
ACM
725#else
726static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
727{
728 return skb->end;
729}
ec47ea82
AD
730
731static inline unsigned int skb_end_offset(const struct sk_buff *skb)
732{
733 return skb->end - skb->head;
734}
4305b541
ACM
735#endif
736
1da177e4 737/* Internal */
4305b541 738#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 739
ac45f602
PO
740static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
741{
742 return &skb_shinfo(skb)->hwtstamps;
743}
744
1da177e4
LT
745/**
746 * skb_queue_empty - check if a queue is empty
747 * @list: queue head
748 *
749 * Returns true if the queue is empty, false otherwise.
750 */
751static inline int skb_queue_empty(const struct sk_buff_head *list)
752{
753 return list->next == (struct sk_buff *)list;
754}
755
fc7ebb21
DM
756/**
757 * skb_queue_is_last - check if skb is the last entry in the queue
758 * @list: queue head
759 * @skb: buffer
760 *
761 * Returns true if @skb is the last buffer on the list.
762 */
763static inline bool skb_queue_is_last(const struct sk_buff_head *list,
764 const struct sk_buff *skb)
765{
a02cec21 766 return skb->next == (struct sk_buff *)list;
fc7ebb21
DM
767}
768
832d11c5
IJ
769/**
770 * skb_queue_is_first - check if skb is the first entry in the queue
771 * @list: queue head
772 * @skb: buffer
773 *
774 * Returns true if @skb is the first buffer on the list.
775 */
776static inline bool skb_queue_is_first(const struct sk_buff_head *list,
777 const struct sk_buff *skb)
778{
a02cec21 779 return skb->prev == (struct sk_buff *)list;
832d11c5
IJ
780}
781
249c8b42
DM
782/**
783 * skb_queue_next - return the next packet in the queue
784 * @list: queue head
785 * @skb: current buffer
786 *
787 * Return the next packet in @list after @skb. It is only valid to
788 * call this if skb_queue_is_last() evaluates to false.
789 */
790static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
791 const struct sk_buff *skb)
792{
793 /* This BUG_ON may seem severe, but if we just return then we
794 * are going to dereference garbage.
795 */
796 BUG_ON(skb_queue_is_last(list, skb));
797 return skb->next;
798}
799
832d11c5
IJ
800/**
801 * skb_queue_prev - return the prev packet in the queue
802 * @list: queue head
803 * @skb: current buffer
804 *
805 * Return the prev packet in @list before @skb. It is only valid to
806 * call this if skb_queue_is_first() evaluates to false.
807 */
808static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
809 const struct sk_buff *skb)
810{
811 /* This BUG_ON may seem severe, but if we just return then we
812 * are going to dereference garbage.
813 */
814 BUG_ON(skb_queue_is_first(list, skb));
815 return skb->prev;
816}
817
1da177e4
LT
818/**
819 * skb_get - reference buffer
820 * @skb: buffer to reference
821 *
822 * Makes another reference to a socket buffer and returns a pointer
823 * to the buffer.
824 */
825static inline struct sk_buff *skb_get(struct sk_buff *skb)
826{
827 atomic_inc(&skb->users);
828 return skb;
829}
830
831/*
832 * If users == 1, we are the only owner and are can avoid redundant
833 * atomic change.
834 */
835
1da177e4
LT
836/**
837 * skb_cloned - is the buffer a clone
838 * @skb: buffer to check
839 *
840 * Returns true if the buffer was generated with skb_clone() and is
841 * one of multiple shared copies of the buffer. Cloned buffers are
842 * shared data so must not be written to under normal circumstances.
843 */
844static inline int skb_cloned(const struct sk_buff *skb)
845{
846 return skb->cloned &&
847 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
848}
849
14bbd6a5
PS
850static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
851{
852 might_sleep_if(pri & __GFP_WAIT);
853
854 if (skb_cloned(skb))
855 return pskb_expand_head(skb, 0, 0, pri);
856
857 return 0;
858}
859
1da177e4
LT
860/**
861 * skb_header_cloned - is the header a clone
862 * @skb: buffer to check
863 *
864 * Returns true if modifying the header part of the buffer requires
865 * the data to be copied.
866 */
867static inline int skb_header_cloned(const struct sk_buff *skb)
868{
869 int dataref;
870
871 if (!skb->cloned)
872 return 0;
873
874 dataref = atomic_read(&skb_shinfo(skb)->dataref);
875 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
876 return dataref != 1;
877}
878
879/**
880 * skb_header_release - release reference to header
881 * @skb: buffer to operate on
882 *
883 * Drop a reference to the header part of the buffer. This is done
884 * by acquiring a payload reference. You must not read from the header
885 * part of skb->data after this.
886 */
887static inline void skb_header_release(struct sk_buff *skb)
888{
889 BUG_ON(skb->nohdr);
890 skb->nohdr = 1;
891 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
892}
893
894/**
895 * skb_shared - is the buffer shared
896 * @skb: buffer to check
897 *
898 * Returns true if more than one person has a reference to this
899 * buffer.
900 */
901static inline int skb_shared(const struct sk_buff *skb)
902{
903 return atomic_read(&skb->users) != 1;
904}
905
906/**
907 * skb_share_check - check if buffer is shared and if so clone it
908 * @skb: buffer to check
909 * @pri: priority for memory allocation
910 *
911 * If the buffer is shared the buffer is cloned and the old copy
912 * drops a reference. A new clone with a single reference is returned.
913 * If the buffer is not shared the original buffer is returned. When
914 * being called from interrupt status or with spinlocks held pri must
915 * be GFP_ATOMIC.
916 *
917 * NULL is returned on a memory allocation failure.
918 */
47061bc4 919static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
920{
921 might_sleep_if(pri & __GFP_WAIT);
922 if (skb_shared(skb)) {
923 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
924
925 if (likely(nskb))
926 consume_skb(skb);
927 else
928 kfree_skb(skb);
1da177e4
LT
929 skb = nskb;
930 }
931 return skb;
932}
933
934/*
935 * Copy shared buffers into a new sk_buff. We effectively do COW on
936 * packets to handle cases where we have a local reader and forward
937 * and a couple of other messy ones. The normal one is tcpdumping
938 * a packet thats being forwarded.
939 */
940
941/**
942 * skb_unshare - make a copy of a shared buffer
943 * @skb: buffer to check
944 * @pri: priority for memory allocation
945 *
946 * If the socket buffer is a clone then this function creates a new
947 * copy of the data, drops a reference count on the old copy and returns
948 * the new copy with the reference count at 1. If the buffer is not a clone
949 * the original buffer is returned. When called with a spinlock held or
950 * from interrupt state @pri must be %GFP_ATOMIC
951 *
952 * %NULL is returned on a memory allocation failure.
953 */
e2bf521d 954static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 955 gfp_t pri)
1da177e4
LT
956{
957 might_sleep_if(pri & __GFP_WAIT);
958 if (skb_cloned(skb)) {
959 struct sk_buff *nskb = skb_copy(skb, pri);
960 kfree_skb(skb); /* Free our shared copy */
961 skb = nskb;
962 }
963 return skb;
964}
965
966/**
1a5778aa 967 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
968 * @list_: list to peek at
969 *
970 * Peek an &sk_buff. Unlike most other operations you _MUST_
971 * be careful with this one. A peek leaves the buffer on the
972 * list and someone else may run off with it. You must hold
973 * the appropriate locks or have a private queue to do this.
974 *
975 * Returns %NULL for an empty list or a pointer to the head element.
976 * The reference count is not incremented and the reference is therefore
977 * volatile. Use with caution.
978 */
05bdd2f1 979static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 980{
18d07000
ED
981 struct sk_buff *skb = list_->next;
982
983 if (skb == (struct sk_buff *)list_)
984 skb = NULL;
985 return skb;
1da177e4
LT
986}
987
da5ef6e5
PE
988/**
989 * skb_peek_next - peek skb following the given one from a queue
990 * @skb: skb to start from
991 * @list_: list to peek at
992 *
993 * Returns %NULL when the end of the list is met or a pointer to the
994 * next element. The reference count is not incremented and the
995 * reference is therefore volatile. Use with caution.
996 */
997static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
998 const struct sk_buff_head *list_)
999{
1000 struct sk_buff *next = skb->next;
18d07000 1001
da5ef6e5
PE
1002 if (next == (struct sk_buff *)list_)
1003 next = NULL;
1004 return next;
1005}
1006
1da177e4 1007/**
1a5778aa 1008 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1009 * @list_: list to peek at
1010 *
1011 * Peek an &sk_buff. Unlike most other operations you _MUST_
1012 * be careful with this one. A peek leaves the buffer on the
1013 * list and someone else may run off with it. You must hold
1014 * the appropriate locks or have a private queue to do this.
1015 *
1016 * Returns %NULL for an empty list or a pointer to the tail element.
1017 * The reference count is not incremented and the reference is therefore
1018 * volatile. Use with caution.
1019 */
05bdd2f1 1020static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1021{
18d07000
ED
1022 struct sk_buff *skb = list_->prev;
1023
1024 if (skb == (struct sk_buff *)list_)
1025 skb = NULL;
1026 return skb;
1027
1da177e4
LT
1028}
1029
1030/**
1031 * skb_queue_len - get queue length
1032 * @list_: list to measure
1033 *
1034 * Return the length of an &sk_buff queue.
1035 */
1036static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1037{
1038 return list_->qlen;
1039}
1040
67fed459
DM
1041/**
1042 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1043 * @list: queue to initialize
1044 *
1045 * This initializes only the list and queue length aspects of
1046 * an sk_buff_head object. This allows to initialize the list
1047 * aspects of an sk_buff_head without reinitializing things like
1048 * the spinlock. It can also be used for on-stack sk_buff_head
1049 * objects where the spinlock is known to not be used.
1050 */
1051static inline void __skb_queue_head_init(struct sk_buff_head *list)
1052{
1053 list->prev = list->next = (struct sk_buff *)list;
1054 list->qlen = 0;
1055}
1056
76f10ad0
AV
1057/*
1058 * This function creates a split out lock class for each invocation;
1059 * this is needed for now since a whole lot of users of the skb-queue
1060 * infrastructure in drivers have different locking usage (in hardirq)
1061 * than the networking core (in softirq only). In the long run either the
1062 * network layer or drivers should need annotation to consolidate the
1063 * main types of usage into 3 classes.
1064 */
1da177e4
LT
1065static inline void skb_queue_head_init(struct sk_buff_head *list)
1066{
1067 spin_lock_init(&list->lock);
67fed459 1068 __skb_queue_head_init(list);
1da177e4
LT
1069}
1070
c2ecba71
PE
1071static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1072 struct lock_class_key *class)
1073{
1074 skb_queue_head_init(list);
1075 lockdep_set_class(&list->lock, class);
1076}
1077
1da177e4 1078/*
bf299275 1079 * Insert an sk_buff on a list.
1da177e4
LT
1080 *
1081 * The "__skb_xxxx()" functions are the non-atomic ones that
1082 * can only be called with interrupts disabled.
1083 */
7965bd4d
JP
1084void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1085 struct sk_buff_head *list);
bf299275
GR
1086static inline void __skb_insert(struct sk_buff *newsk,
1087 struct sk_buff *prev, struct sk_buff *next,
1088 struct sk_buff_head *list)
1089{
1090 newsk->next = next;
1091 newsk->prev = prev;
1092 next->prev = prev->next = newsk;
1093 list->qlen++;
1094}
1da177e4 1095
67fed459
DM
1096static inline void __skb_queue_splice(const struct sk_buff_head *list,
1097 struct sk_buff *prev,
1098 struct sk_buff *next)
1099{
1100 struct sk_buff *first = list->next;
1101 struct sk_buff *last = list->prev;
1102
1103 first->prev = prev;
1104 prev->next = first;
1105
1106 last->next = next;
1107 next->prev = last;
1108}
1109
1110/**
1111 * skb_queue_splice - join two skb lists, this is designed for stacks
1112 * @list: the new list to add
1113 * @head: the place to add it in the first list
1114 */
1115static inline void skb_queue_splice(const struct sk_buff_head *list,
1116 struct sk_buff_head *head)
1117{
1118 if (!skb_queue_empty(list)) {
1119 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1120 head->qlen += list->qlen;
67fed459
DM
1121 }
1122}
1123
1124/**
d9619496 1125 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1126 * @list: the new list to add
1127 * @head: the place to add it in the first list
1128 *
1129 * The list at @list is reinitialised
1130 */
1131static inline void skb_queue_splice_init(struct sk_buff_head *list,
1132 struct sk_buff_head *head)
1133{
1134 if (!skb_queue_empty(list)) {
1135 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1136 head->qlen += list->qlen;
67fed459
DM
1137 __skb_queue_head_init(list);
1138 }
1139}
1140
1141/**
1142 * skb_queue_splice_tail - join two skb lists, each list being a queue
1143 * @list: the new list to add
1144 * @head: the place to add it in the first list
1145 */
1146static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1147 struct sk_buff_head *head)
1148{
1149 if (!skb_queue_empty(list)) {
1150 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1151 head->qlen += list->qlen;
67fed459
DM
1152 }
1153}
1154
1155/**
d9619496 1156 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1157 * @list: the new list to add
1158 * @head: the place to add it in the first list
1159 *
1160 * Each of the lists is a queue.
1161 * The list at @list is reinitialised
1162 */
1163static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1164 struct sk_buff_head *head)
1165{
1166 if (!skb_queue_empty(list)) {
1167 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1168 head->qlen += list->qlen;
67fed459
DM
1169 __skb_queue_head_init(list);
1170 }
1171}
1172
1da177e4 1173/**
300ce174 1174 * __skb_queue_after - queue a buffer at the list head
1da177e4 1175 * @list: list to use
300ce174 1176 * @prev: place after this buffer
1da177e4
LT
1177 * @newsk: buffer to queue
1178 *
300ce174 1179 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1180 * and you must therefore hold required locks before calling it.
1181 *
1182 * A buffer cannot be placed on two lists at the same time.
1183 */
300ce174
SH
1184static inline void __skb_queue_after(struct sk_buff_head *list,
1185 struct sk_buff *prev,
1186 struct sk_buff *newsk)
1da177e4 1187{
bf299275 1188 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1189}
1190
7965bd4d
JP
1191void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1192 struct sk_buff_head *list);
7de6c033 1193
f5572855
GR
1194static inline void __skb_queue_before(struct sk_buff_head *list,
1195 struct sk_buff *next,
1196 struct sk_buff *newsk)
1197{
1198 __skb_insert(newsk, next->prev, next, list);
1199}
1200
300ce174
SH
1201/**
1202 * __skb_queue_head - queue a buffer at the list head
1203 * @list: list to use
1204 * @newsk: buffer to queue
1205 *
1206 * Queue a buffer at the start of a list. This function takes no locks
1207 * and you must therefore hold required locks before calling it.
1208 *
1209 * A buffer cannot be placed on two lists at the same time.
1210 */
7965bd4d 1211void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1212static inline void __skb_queue_head(struct sk_buff_head *list,
1213 struct sk_buff *newsk)
1214{
1215 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1216}
1217
1da177e4
LT
1218/**
1219 * __skb_queue_tail - queue a buffer at the list tail
1220 * @list: list to use
1221 * @newsk: buffer to queue
1222 *
1223 * Queue a buffer at the end of a list. This function takes no locks
1224 * and you must therefore hold required locks before calling it.
1225 *
1226 * A buffer cannot be placed on two lists at the same time.
1227 */
7965bd4d 1228void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1229static inline void __skb_queue_tail(struct sk_buff_head *list,
1230 struct sk_buff *newsk)
1231{
f5572855 1232 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1233}
1234
1da177e4
LT
1235/*
1236 * remove sk_buff from list. _Must_ be called atomically, and with
1237 * the list known..
1238 */
7965bd4d 1239void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1240static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1241{
1242 struct sk_buff *next, *prev;
1243
1244 list->qlen--;
1245 next = skb->next;
1246 prev = skb->prev;
1247 skb->next = skb->prev = NULL;
1da177e4
LT
1248 next->prev = prev;
1249 prev->next = next;
1250}
1251
f525c06d
GR
1252/**
1253 * __skb_dequeue - remove from the head of the queue
1254 * @list: list to dequeue from
1255 *
1256 * Remove the head of the list. This function does not take any locks
1257 * so must be used with appropriate locks held only. The head item is
1258 * returned or %NULL if the list is empty.
1259 */
7965bd4d 1260struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1261static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1262{
1263 struct sk_buff *skb = skb_peek(list);
1264 if (skb)
1265 __skb_unlink(skb, list);
1266 return skb;
1267}
1da177e4
LT
1268
1269/**
1270 * __skb_dequeue_tail - remove from the tail of the queue
1271 * @list: list to dequeue from
1272 *
1273 * Remove the tail of the list. This function does not take any locks
1274 * so must be used with appropriate locks held only. The tail item is
1275 * returned or %NULL if the list is empty.
1276 */
7965bd4d 1277struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1278static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1279{
1280 struct sk_buff *skb = skb_peek_tail(list);
1281 if (skb)
1282 __skb_unlink(skb, list);
1283 return skb;
1284}
1285
1286
bdcc0924 1287static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1288{
1289 return skb->data_len;
1290}
1291
1292static inline unsigned int skb_headlen(const struct sk_buff *skb)
1293{
1294 return skb->len - skb->data_len;
1295}
1296
1297static inline int skb_pagelen(const struct sk_buff *skb)
1298{
1299 int i, len = 0;
1300
1301 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1302 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1303 return len + skb_headlen(skb);
1304}
1305
131ea667
IC
1306/**
1307 * __skb_fill_page_desc - initialise a paged fragment in an skb
1308 * @skb: buffer containing fragment to be initialised
1309 * @i: paged fragment index to initialise
1310 * @page: the page to use for this fragment
1311 * @off: the offset to the data with @page
1312 * @size: the length of the data
1313 *
1314 * Initialises the @i'th fragment of @skb to point to &size bytes at
1315 * offset @off within @page.
1316 *
1317 * Does not take any additional reference on the fragment.
1318 */
1319static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1320 struct page *page, int off, int size)
1da177e4
LT
1321{
1322 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1323
c48a11c7
MG
1324 /*
1325 * Propagate page->pfmemalloc to the skb if we can. The problem is
1326 * that not all callers have unique ownership of the page. If
1327 * pfmemalloc is set, we check the mapping as a mapping implies
1328 * page->index is set (index and pfmemalloc share space).
1329 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1330 * do not lose pfmemalloc information as the pages would not be
1331 * allocated using __GFP_MEMALLOC.
1332 */
a8605c60 1333 frag->page.p = page;
1da177e4 1334 frag->page_offset = off;
9e903e08 1335 skb_frag_size_set(frag, size);
cca7af38
PE
1336
1337 page = compound_head(page);
1338 if (page->pfmemalloc && !page->mapping)
1339 skb->pfmemalloc = true;
131ea667
IC
1340}
1341
1342/**
1343 * skb_fill_page_desc - initialise a paged fragment in an skb
1344 * @skb: buffer containing fragment to be initialised
1345 * @i: paged fragment index to initialise
1346 * @page: the page to use for this fragment
1347 * @off: the offset to the data with @page
1348 * @size: the length of the data
1349 *
1350 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1351 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1352 * addition updates @skb such that @i is the last fragment.
1353 *
1354 * Does not take any additional reference on the fragment.
1355 */
1356static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1357 struct page *page, int off, int size)
1358{
1359 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1360 skb_shinfo(skb)->nr_frags = i + 1;
1361}
1362
7965bd4d
JP
1363void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1364 int size, unsigned int truesize);
654bed16 1365
f8e617e1
JW
1366void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1367 unsigned int truesize);
1368
1da177e4 1369#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1370#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1371#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1372
27a884dc
ACM
1373#ifdef NET_SKBUFF_DATA_USES_OFFSET
1374static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1375{
1376 return skb->head + skb->tail;
1377}
1378
1379static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1380{
1381 skb->tail = skb->data - skb->head;
1382}
1383
1384static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1385{
1386 skb_reset_tail_pointer(skb);
1387 skb->tail += offset;
1388}
7cc46190 1389
27a884dc
ACM
1390#else /* NET_SKBUFF_DATA_USES_OFFSET */
1391static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1392{
1393 return skb->tail;
1394}
1395
1396static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1397{
1398 skb->tail = skb->data;
1399}
1400
1401static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1402{
1403 skb->tail = skb->data + offset;
1404}
4305b541 1405
27a884dc
ACM
1406#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1407
1da177e4
LT
1408/*
1409 * Add data to an sk_buff
1410 */
0c7ddf36 1411unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1412unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1413static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1414{
27a884dc 1415 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1416 SKB_LINEAR_ASSERT(skb);
1417 skb->tail += len;
1418 skb->len += len;
1419 return tmp;
1420}
1421
7965bd4d 1422unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1423static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1424{
1425 skb->data -= len;
1426 skb->len += len;
1427 return skb->data;
1428}
1429
7965bd4d 1430unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1431static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1432{
1433 skb->len -= len;
1434 BUG_ON(skb->len < skb->data_len);
1435 return skb->data += len;
1436}
1437
47d29646
DM
1438static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1439{
1440 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1441}
1442
7965bd4d 1443unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1444
1445static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1446{
1447 if (len > skb_headlen(skb) &&
987c402a 1448 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1449 return NULL;
1450 skb->len -= len;
1451 return skb->data += len;
1452}
1453
1454static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1455{
1456 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1457}
1458
1459static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1460{
1461 if (likely(len <= skb_headlen(skb)))
1462 return 1;
1463 if (unlikely(len > skb->len))
1464 return 0;
987c402a 1465 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1466}
1467
1468/**
1469 * skb_headroom - bytes at buffer head
1470 * @skb: buffer to check
1471 *
1472 * Return the number of bytes of free space at the head of an &sk_buff.
1473 */
c2636b4d 1474static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1475{
1476 return skb->data - skb->head;
1477}
1478
1479/**
1480 * skb_tailroom - bytes at buffer end
1481 * @skb: buffer to check
1482 *
1483 * Return the number of bytes of free space at the tail of an sk_buff
1484 */
1485static inline int skb_tailroom(const struct sk_buff *skb)
1486{
4305b541 1487 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1488}
1489
a21d4572
ED
1490/**
1491 * skb_availroom - bytes at buffer end
1492 * @skb: buffer to check
1493 *
1494 * Return the number of bytes of free space at the tail of an sk_buff
1495 * allocated by sk_stream_alloc()
1496 */
1497static inline int skb_availroom(const struct sk_buff *skb)
1498{
16fad69c
ED
1499 if (skb_is_nonlinear(skb))
1500 return 0;
1501
1502 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1503}
1504
1da177e4
LT
1505/**
1506 * skb_reserve - adjust headroom
1507 * @skb: buffer to alter
1508 * @len: bytes to move
1509 *
1510 * Increase the headroom of an empty &sk_buff by reducing the tail
1511 * room. This is only allowed for an empty buffer.
1512 */
8243126c 1513static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1514{
1515 skb->data += len;
1516 skb->tail += len;
1517}
1518
6a674e9c
JG
1519static inline void skb_reset_inner_headers(struct sk_buff *skb)
1520{
aefbd2b3 1521 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1522 skb->inner_network_header = skb->network_header;
1523 skb->inner_transport_header = skb->transport_header;
1524}
1525
0b5c9db1
JP
1526static inline void skb_reset_mac_len(struct sk_buff *skb)
1527{
1528 skb->mac_len = skb->network_header - skb->mac_header;
1529}
1530
6a674e9c
JG
1531static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1532 *skb)
1533{
1534 return skb->head + skb->inner_transport_header;
1535}
1536
1537static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1538{
1539 skb->inner_transport_header = skb->data - skb->head;
1540}
1541
1542static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1543 const int offset)
1544{
1545 skb_reset_inner_transport_header(skb);
1546 skb->inner_transport_header += offset;
1547}
1548
1549static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1550{
1551 return skb->head + skb->inner_network_header;
1552}
1553
1554static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1555{
1556 skb->inner_network_header = skb->data - skb->head;
1557}
1558
1559static inline void skb_set_inner_network_header(struct sk_buff *skb,
1560 const int offset)
1561{
1562 skb_reset_inner_network_header(skb);
1563 skb->inner_network_header += offset;
1564}
1565
aefbd2b3
PS
1566static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1567{
1568 return skb->head + skb->inner_mac_header;
1569}
1570
1571static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1572{
1573 skb->inner_mac_header = skb->data - skb->head;
1574}
1575
1576static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1577 const int offset)
1578{
1579 skb_reset_inner_mac_header(skb);
1580 skb->inner_mac_header += offset;
1581}
fda55eca
ED
1582static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1583{
35d04610 1584 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1585}
1586
9c70220b
ACM
1587static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1588{
2e07fa9c 1589 return skb->head + skb->transport_header;
9c70220b
ACM
1590}
1591
badff6d0
ACM
1592static inline void skb_reset_transport_header(struct sk_buff *skb)
1593{
2e07fa9c 1594 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1595}
1596
967b05f6
ACM
1597static inline void skb_set_transport_header(struct sk_buff *skb,
1598 const int offset)
1599{
2e07fa9c
ACM
1600 skb_reset_transport_header(skb);
1601 skb->transport_header += offset;
ea2ae17d
ACM
1602}
1603
d56f90a7
ACM
1604static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1605{
2e07fa9c 1606 return skb->head + skb->network_header;
d56f90a7
ACM
1607}
1608
c1d2bbe1
ACM
1609static inline void skb_reset_network_header(struct sk_buff *skb)
1610{
2e07fa9c 1611 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1612}
1613
c14d2450
ACM
1614static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1615{
2e07fa9c
ACM
1616 skb_reset_network_header(skb);
1617 skb->network_header += offset;
c14d2450
ACM
1618}
1619
2e07fa9c 1620static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1621{
2e07fa9c 1622 return skb->head + skb->mac_header;
bbe735e4
ACM
1623}
1624
2e07fa9c 1625static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1626{
35d04610 1627 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1628}
1629
1630static inline void skb_reset_mac_header(struct sk_buff *skb)
1631{
1632 skb->mac_header = skb->data - skb->head;
1633}
1634
1635static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1636{
1637 skb_reset_mac_header(skb);
1638 skb->mac_header += offset;
1639}
1640
fbbdb8f0
YX
1641static inline void skb_probe_transport_header(struct sk_buff *skb,
1642 const int offset_hint)
1643{
1644 struct flow_keys keys;
1645
1646 if (skb_transport_header_was_set(skb))
1647 return;
1648 else if (skb_flow_dissect(skb, &keys))
1649 skb_set_transport_header(skb, keys.thoff);
1650 else
1651 skb_set_transport_header(skb, offset_hint);
1652}
1653
03606895
ED
1654static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1655{
1656 if (skb_mac_header_was_set(skb)) {
1657 const unsigned char *old_mac = skb_mac_header(skb);
1658
1659 skb_set_mac_header(skb, -skb->mac_len);
1660 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1661 }
1662}
1663
04fb451e
MM
1664static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1665{
1666 return skb->csum_start - skb_headroom(skb);
1667}
1668
2e07fa9c
ACM
1669static inline int skb_transport_offset(const struct sk_buff *skb)
1670{
1671 return skb_transport_header(skb) - skb->data;
1672}
1673
1674static inline u32 skb_network_header_len(const struct sk_buff *skb)
1675{
1676 return skb->transport_header - skb->network_header;
1677}
1678
6a674e9c
JG
1679static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1680{
1681 return skb->inner_transport_header - skb->inner_network_header;
1682}
1683
2e07fa9c
ACM
1684static inline int skb_network_offset(const struct sk_buff *skb)
1685{
1686 return skb_network_header(skb) - skb->data;
1687}
48d49d0c 1688
6a674e9c
JG
1689static inline int skb_inner_network_offset(const struct sk_buff *skb)
1690{
1691 return skb_inner_network_header(skb) - skb->data;
1692}
1693
f9599ce1
CG
1694static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1695{
1696 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1697}
1698
1da177e4
LT
1699/*
1700 * CPUs often take a performance hit when accessing unaligned memory
1701 * locations. The actual performance hit varies, it can be small if the
1702 * hardware handles it or large if we have to take an exception and fix it
1703 * in software.
1704 *
1705 * Since an ethernet header is 14 bytes network drivers often end up with
1706 * the IP header at an unaligned offset. The IP header can be aligned by
1707 * shifting the start of the packet by 2 bytes. Drivers should do this
1708 * with:
1709 *
8660c124 1710 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1711 *
1712 * The downside to this alignment of the IP header is that the DMA is now
1713 * unaligned. On some architectures the cost of an unaligned DMA is high
1714 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1715 *
1da177e4
LT
1716 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1717 * to be overridden.
1718 */
1719#ifndef NET_IP_ALIGN
1720#define NET_IP_ALIGN 2
1721#endif
1722
025be81e
AB
1723/*
1724 * The networking layer reserves some headroom in skb data (via
1725 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1726 * the header has to grow. In the default case, if the header has to grow
d6301d3d 1727 * 32 bytes or less we avoid the reallocation.
025be81e
AB
1728 *
1729 * Unfortunately this headroom changes the DMA alignment of the resulting
1730 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1731 * on some architectures. An architecture can override this value,
1732 * perhaps setting it to a cacheline in size (since that will maintain
1733 * cacheline alignment of the DMA). It must be a power of 2.
1734 *
d6301d3d 1735 * Various parts of the networking layer expect at least 32 bytes of
025be81e 1736 * headroom, you should not reduce this.
5933dd2f
ED
1737 *
1738 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1739 * to reduce average number of cache lines per packet.
1740 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 1741 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
1742 */
1743#ifndef NET_SKB_PAD
5933dd2f 1744#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
1745#endif
1746
7965bd4d 1747int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1748
1749static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1750{
c4264f27 1751 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
1752 WARN_ON(1);
1753 return;
1754 }
27a884dc
ACM
1755 skb->len = len;
1756 skb_set_tail_pointer(skb, len);
1da177e4
LT
1757}
1758
7965bd4d 1759void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1760
1761static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1762{
3cc0e873
HX
1763 if (skb->data_len)
1764 return ___pskb_trim(skb, len);
1765 __skb_trim(skb, len);
1766 return 0;
1da177e4
LT
1767}
1768
1769static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1770{
1771 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1772}
1773
e9fa4f7b
HX
1774/**
1775 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1776 * @skb: buffer to alter
1777 * @len: new length
1778 *
1779 * This is identical to pskb_trim except that the caller knows that
1780 * the skb is not cloned so we should never get an error due to out-
1781 * of-memory.
1782 */
1783static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1784{
1785 int err = pskb_trim(skb, len);
1786 BUG_ON(err);
1787}
1788
1da177e4
LT
1789/**
1790 * skb_orphan - orphan a buffer
1791 * @skb: buffer to orphan
1792 *
1793 * If a buffer currently has an owner then we call the owner's
1794 * destructor function and make the @skb unowned. The buffer continues
1795 * to exist but is no longer charged to its former owner.
1796 */
1797static inline void skb_orphan(struct sk_buff *skb)
1798{
c34a7612 1799 if (skb->destructor) {
1da177e4 1800 skb->destructor(skb);
c34a7612
ED
1801 skb->destructor = NULL;
1802 skb->sk = NULL;
376c7311
ED
1803 } else {
1804 BUG_ON(skb->sk);
c34a7612 1805 }
1da177e4
LT
1806}
1807
a353e0ce
MT
1808/**
1809 * skb_orphan_frags - orphan the frags contained in a buffer
1810 * @skb: buffer to orphan frags from
1811 * @gfp_mask: allocation mask for replacement pages
1812 *
1813 * For each frag in the SKB which needs a destructor (i.e. has an
1814 * owner) create a copy of that frag and release the original
1815 * page by calling the destructor.
1816 */
1817static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1818{
1819 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1820 return 0;
1821 return skb_copy_ubufs(skb, gfp_mask);
1822}
1823
1da177e4
LT
1824/**
1825 * __skb_queue_purge - empty a list
1826 * @list: list to empty
1827 *
1828 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1829 * the list and one reference dropped. This function does not take the
1830 * list lock and the caller must hold the relevant locks to use it.
1831 */
7965bd4d 1832void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
1833static inline void __skb_queue_purge(struct sk_buff_head *list)
1834{
1835 struct sk_buff *skb;
1836 while ((skb = __skb_dequeue(list)) != NULL)
1837 kfree_skb(skb);
1838}
1839
e5e67305
AD
1840#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1841#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1842#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1843
7965bd4d 1844void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 1845
7965bd4d
JP
1846struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1847 gfp_t gfp_mask);
8af27456
CH
1848
1849/**
1850 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1851 * @dev: network device to receive on
1852 * @length: length to allocate
1853 *
1854 * Allocate a new &sk_buff and assign it a usage count of one. The
1855 * buffer has unspecified headroom built in. Users should allocate
1856 * the headroom they think they need without accounting for the
1857 * built in space. The built in space is used for optimisations.
1858 *
1859 * %NULL is returned if there is no free memory. Although this function
1860 * allocates memory it can be called from an interrupt.
1861 */
1862static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 1863 unsigned int length)
8af27456
CH
1864{
1865 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1866}
1867
6f532612
ED
1868/* legacy helper around __netdev_alloc_skb() */
1869static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1870 gfp_t gfp_mask)
1871{
1872 return __netdev_alloc_skb(NULL, length, gfp_mask);
1873}
1874
1875/* legacy helper around netdev_alloc_skb() */
1876static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1877{
1878 return netdev_alloc_skb(NULL, length);
1879}
1880
1881
4915a0de
ED
1882static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1883 unsigned int length, gfp_t gfp)
61321bbd 1884{
4915a0de 1885 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
1886
1887 if (NET_IP_ALIGN && skb)
1888 skb_reserve(skb, NET_IP_ALIGN);
1889 return skb;
1890}
1891
4915a0de
ED
1892static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1893 unsigned int length)
1894{
1895 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1896}
1897
bc6fc9fa
FF
1898/**
1899 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
1900 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1901 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1902 * @order: size of the allocation
1903 *
1904 * Allocate a new page.
1905 *
1906 * %NULL is returned if there is no free memory.
1907*/
1908static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1909 struct sk_buff *skb,
1910 unsigned int order)
1911{
1912 struct page *page;
1913
1914 gfp_mask |= __GFP_COLD;
1915
1916 if (!(gfp_mask & __GFP_NOMEMALLOC))
1917 gfp_mask |= __GFP_MEMALLOC;
1918
1919 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
1920 if (skb && page && page->pfmemalloc)
1921 skb->pfmemalloc = true;
1922
1923 return page;
1924}
1925
1926/**
1927 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
1928 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1929 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1930 *
1931 * Allocate a new page.
1932 *
1933 * %NULL is returned if there is no free memory.
1934 */
1935static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
1936 struct sk_buff *skb)
1937{
1938 return __skb_alloc_pages(gfp_mask, skb, 0);
1939}
1940
1941/**
1942 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
1943 * @page: The page that was allocated from skb_alloc_page
1944 * @skb: The skb that may need pfmemalloc set
1945 */
1946static inline void skb_propagate_pfmemalloc(struct page *page,
1947 struct sk_buff *skb)
1948{
1949 if (page && page->pfmemalloc)
1950 skb->pfmemalloc = true;
1951}
1952
131ea667
IC
1953/**
1954 * skb_frag_page - retrieve the page refered to by a paged fragment
1955 * @frag: the paged fragment
1956 *
1957 * Returns the &struct page associated with @frag.
1958 */
1959static inline struct page *skb_frag_page(const skb_frag_t *frag)
1960{
a8605c60 1961 return frag->page.p;
131ea667
IC
1962}
1963
1964/**
1965 * __skb_frag_ref - take an addition reference on a paged fragment.
1966 * @frag: the paged fragment
1967 *
1968 * Takes an additional reference on the paged fragment @frag.
1969 */
1970static inline void __skb_frag_ref(skb_frag_t *frag)
1971{
1972 get_page(skb_frag_page(frag));
1973}
1974
1975/**
1976 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1977 * @skb: the buffer
1978 * @f: the fragment offset.
1979 *
1980 * Takes an additional reference on the @f'th paged fragment of @skb.
1981 */
1982static inline void skb_frag_ref(struct sk_buff *skb, int f)
1983{
1984 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1985}
1986
1987/**
1988 * __skb_frag_unref - release a reference on a paged fragment.
1989 * @frag: the paged fragment
1990 *
1991 * Releases a reference on the paged fragment @frag.
1992 */
1993static inline void __skb_frag_unref(skb_frag_t *frag)
1994{
1995 put_page(skb_frag_page(frag));
1996}
1997
1998/**
1999 * skb_frag_unref - release a reference on a paged fragment of an skb.
2000 * @skb: the buffer
2001 * @f: the fragment offset
2002 *
2003 * Releases a reference on the @f'th paged fragment of @skb.
2004 */
2005static inline void skb_frag_unref(struct sk_buff *skb, int f)
2006{
2007 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2008}
2009
2010/**
2011 * skb_frag_address - gets the address of the data contained in a paged fragment
2012 * @frag: the paged fragment buffer
2013 *
2014 * Returns the address of the data within @frag. The page must already
2015 * be mapped.
2016 */
2017static inline void *skb_frag_address(const skb_frag_t *frag)
2018{
2019 return page_address(skb_frag_page(frag)) + frag->page_offset;
2020}
2021
2022/**
2023 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2024 * @frag: the paged fragment buffer
2025 *
2026 * Returns the address of the data within @frag. Checks that the page
2027 * is mapped and returns %NULL otherwise.
2028 */
2029static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2030{
2031 void *ptr = page_address(skb_frag_page(frag));
2032 if (unlikely(!ptr))
2033 return NULL;
2034
2035 return ptr + frag->page_offset;
2036}
2037
2038/**
2039 * __skb_frag_set_page - sets the page contained in a paged fragment
2040 * @frag: the paged fragment
2041 * @page: the page to set
2042 *
2043 * Sets the fragment @frag to contain @page.
2044 */
2045static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2046{
a8605c60 2047 frag->page.p = page;
131ea667
IC
2048}
2049
2050/**
2051 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2052 * @skb: the buffer
2053 * @f: the fragment offset
2054 * @page: the page to set
2055 *
2056 * Sets the @f'th fragment of @skb to contain @page.
2057 */
2058static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2059 struct page *page)
2060{
2061 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2062}
2063
400dfd3a
ED
2064bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2065
131ea667
IC
2066/**
2067 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2068 * @dev: the device to map the fragment to
131ea667
IC
2069 * @frag: the paged fragment to map
2070 * @offset: the offset within the fragment (starting at the
2071 * fragment's own offset)
2072 * @size: the number of bytes to map
f83347df 2073 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2074 *
2075 * Maps the page associated with @frag to @device.
2076 */
2077static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2078 const skb_frag_t *frag,
2079 size_t offset, size_t size,
2080 enum dma_data_direction dir)
2081{
2082 return dma_map_page(dev, skb_frag_page(frag),
2083 frag->page_offset + offset, size, dir);
2084}
2085
117632e6
ED
2086static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2087 gfp_t gfp_mask)
2088{
2089 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2090}
2091
334a8132
PM
2092/**
2093 * skb_clone_writable - is the header of a clone writable
2094 * @skb: buffer to check
2095 * @len: length up to which to write
2096 *
2097 * Returns true if modifying the header part of the cloned buffer
2098 * does not requires the data to be copied.
2099 */
05bdd2f1 2100static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2101{
2102 return !skb_header_cloned(skb) &&
2103 skb_headroom(skb) + len <= skb->hdr_len;
2104}
2105
d9cc2048
HX
2106static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2107 int cloned)
2108{
2109 int delta = 0;
2110
d9cc2048
HX
2111 if (headroom > skb_headroom(skb))
2112 delta = headroom - skb_headroom(skb);
2113
2114 if (delta || cloned)
2115 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2116 GFP_ATOMIC);
2117 return 0;
2118}
2119
1da177e4
LT
2120/**
2121 * skb_cow - copy header of skb when it is required
2122 * @skb: buffer to cow
2123 * @headroom: needed headroom
2124 *
2125 * If the skb passed lacks sufficient headroom or its data part
2126 * is shared, data is reallocated. If reallocation fails, an error
2127 * is returned and original skb is not changed.
2128 *
2129 * The result is skb with writable area skb->head...skb->tail
2130 * and at least @headroom of space at head.
2131 */
2132static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2133{
d9cc2048
HX
2134 return __skb_cow(skb, headroom, skb_cloned(skb));
2135}
1da177e4 2136
d9cc2048
HX
2137/**
2138 * skb_cow_head - skb_cow but only making the head writable
2139 * @skb: buffer to cow
2140 * @headroom: needed headroom
2141 *
2142 * This function is identical to skb_cow except that we replace the
2143 * skb_cloned check by skb_header_cloned. It should be used when
2144 * you only need to push on some header and do not need to modify
2145 * the data.
2146 */
2147static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2148{
2149 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2150}
2151
2152/**
2153 * skb_padto - pad an skbuff up to a minimal size
2154 * @skb: buffer to pad
2155 * @len: minimal length
2156 *
2157 * Pads up a buffer to ensure the trailing bytes exist and are
2158 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2159 * is untouched. Otherwise it is extended. Returns zero on
2160 * success. The skb is freed on error.
1da177e4
LT
2161 */
2162
5b057c6b 2163static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2164{
2165 unsigned int size = skb->len;
2166 if (likely(size >= len))
5b057c6b 2167 return 0;
987c402a 2168 return skb_pad(skb, len - size);
1da177e4
LT
2169}
2170
2171static inline int skb_add_data(struct sk_buff *skb,
2172 char __user *from, int copy)
2173{
2174 const int off = skb->len;
2175
2176 if (skb->ip_summed == CHECKSUM_NONE) {
2177 int err = 0;
5084205f 2178 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2179 copy, 0, &err);
2180 if (!err) {
2181 skb->csum = csum_block_add(skb->csum, csum, off);
2182 return 0;
2183 }
2184 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2185 return 0;
2186
2187 __skb_trim(skb, off);
2188 return -EFAULT;
2189}
2190
38ba0a65
ED
2191static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2192 const struct page *page, int off)
1da177e4
LT
2193{
2194 if (i) {
9e903e08 2195 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2196
ea2ab693 2197 return page == skb_frag_page(frag) &&
9e903e08 2198 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2199 }
38ba0a65 2200 return false;
1da177e4
LT
2201}
2202
364c6bad
HX
2203static inline int __skb_linearize(struct sk_buff *skb)
2204{
2205 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2206}
2207
1da177e4
LT
2208/**
2209 * skb_linearize - convert paged skb to linear one
2210 * @skb: buffer to linarize
1da177e4
LT
2211 *
2212 * If there is no free memory -ENOMEM is returned, otherwise zero
2213 * is returned and the old skb data released.
2214 */
364c6bad
HX
2215static inline int skb_linearize(struct sk_buff *skb)
2216{
2217 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2218}
2219
cef401de
ED
2220/**
2221 * skb_has_shared_frag - can any frag be overwritten
2222 * @skb: buffer to test
2223 *
2224 * Return true if the skb has at least one frag that might be modified
2225 * by an external entity (as in vmsplice()/sendfile())
2226 */
2227static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2228{
c9af6db4
PS
2229 return skb_is_nonlinear(skb) &&
2230 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2231}
2232
364c6bad
HX
2233/**
2234 * skb_linearize_cow - make sure skb is linear and writable
2235 * @skb: buffer to process
2236 *
2237 * If there is no free memory -ENOMEM is returned, otherwise zero
2238 * is returned and the old skb data released.
2239 */
2240static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2241{
364c6bad
HX
2242 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2243 __skb_linearize(skb) : 0;
1da177e4
LT
2244}
2245
2246/**
2247 * skb_postpull_rcsum - update checksum for received skb after pull
2248 * @skb: buffer to update
2249 * @start: start of data before pull
2250 * @len: length of data pulled
2251 *
2252 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2253 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2254 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2255 */
2256
2257static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2258 const void *start, unsigned int len)
1da177e4 2259{
84fa7933 2260 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2261 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2262}
2263
cbb042f9
HX
2264unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2265
1da177e4
LT
2266#define skb_queue_walk(queue, skb) \
2267 for (skb = (queue)->next; \
a1e4891f 2268 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2269 skb = skb->next)
2270
46f8914e
JC
2271#define skb_queue_walk_safe(queue, skb, tmp) \
2272 for (skb = (queue)->next, tmp = skb->next; \
2273 skb != (struct sk_buff *)(queue); \
2274 skb = tmp, tmp = skb->next)
2275
1164f52a 2276#define skb_queue_walk_from(queue, skb) \
a1e4891f 2277 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2278 skb = skb->next)
2279
2280#define skb_queue_walk_from_safe(queue, skb, tmp) \
2281 for (tmp = skb->next; \
2282 skb != (struct sk_buff *)(queue); \
2283 skb = tmp, tmp = skb->next)
2284
300ce174
SH
2285#define skb_queue_reverse_walk(queue, skb) \
2286 for (skb = (queue)->prev; \
a1e4891f 2287 skb != (struct sk_buff *)(queue); \
300ce174
SH
2288 skb = skb->prev)
2289
686a2955
DM
2290#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2291 for (skb = (queue)->prev, tmp = skb->prev; \
2292 skb != (struct sk_buff *)(queue); \
2293 skb = tmp, tmp = skb->prev)
2294
2295#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2296 for (tmp = skb->prev; \
2297 skb != (struct sk_buff *)(queue); \
2298 skb = tmp, tmp = skb->prev)
1da177e4 2299
21dc3301 2300static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2301{
2302 return skb_shinfo(skb)->frag_list != NULL;
2303}
2304
2305static inline void skb_frag_list_init(struct sk_buff *skb)
2306{
2307 skb_shinfo(skb)->frag_list = NULL;
2308}
2309
2310static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2311{
2312 frag->next = skb_shinfo(skb)->frag_list;
2313 skb_shinfo(skb)->frag_list = frag;
2314}
2315
2316#define skb_walk_frags(skb, iter) \
2317 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2318
7965bd4d
JP
2319struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2320 int *peeked, int *off, int *err);
2321struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2322 int *err);
2323unsigned int datagram_poll(struct file *file, struct socket *sock,
2324 struct poll_table_struct *wait);
2325int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2326 struct iovec *to, int size);
2327int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2328 struct iovec *iov);
2329int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2330 const struct iovec *from, int from_offset,
2331 int len);
2332int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2333 int offset, size_t count);
2334int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2335 const struct iovec *to, int to_offset,
2336 int size);
2337void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2338void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2339int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2340int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2341int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2342__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2343 int len, __wsum csum);
2344int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2345 struct pipe_inode_info *pipe, unsigned int len,
2346 unsigned int flags);
2347void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2348void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2349int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2350void skb_scrub_packet(struct sk_buff *skb, bool xnet);
7965bd4d 2351struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
20380731 2352
2817a336
DB
2353struct skb_checksum_ops {
2354 __wsum (*update)(const void *mem, int len, __wsum wsum);
2355 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2356};
2357
2358__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2359 __wsum csum, const struct skb_checksum_ops *ops);
2360__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2361 __wsum csum);
2362
018c5bba
DM
2363/**
2364 * pskb_trim_rcsum - trim received skb and update checksum
2365 * @skb: buffer to trim
2366 * @len: new length
2367 *
2368 * This is exactly the same as pskb_trim except that it ensures the
2369 * checksum of received packets are still valid after the operation.
2370 */
2371
2372static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2373{
2374 if (likely(len >= skb->len))
2375 return 0;
2376 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2377 __wsum adj = skb_checksum(skb, len, skb->len - len, 0);
2378
2379 skb->csum = csum_sub(skb->csum, adj);
2380 }
2381 return __pskb_trim(skb, len);
2382}
2383
1da177e4
LT
2384static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2385 int len, void *buffer)
2386{
2387 int hlen = skb_headlen(skb);
2388
55820ee2 2389 if (hlen - offset >= len)
1da177e4
LT
2390 return skb->data + offset;
2391
2392 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2393 return NULL;
2394
2395 return buffer;
2396}
2397
d626f62b
ACM
2398static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2399 void *to,
2400 const unsigned int len)
2401{
2402 memcpy(to, skb->data, len);
2403}
2404
2405static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2406 const int offset, void *to,
2407 const unsigned int len)
2408{
2409 memcpy(to, skb->data + offset, len);
2410}
2411
27d7ff46
ACM
2412static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2413 const void *from,
2414 const unsigned int len)
2415{
2416 memcpy(skb->data, from, len);
2417}
2418
2419static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2420 const int offset,
2421 const void *from,
2422 const unsigned int len)
2423{
2424 memcpy(skb->data + offset, from, len);
2425}
2426
7965bd4d 2427void skb_init(void);
1da177e4 2428
ac45f602
PO
2429static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2430{
2431 return skb->tstamp;
2432}
2433
a61bbcf2
PM
2434/**
2435 * skb_get_timestamp - get timestamp from a skb
2436 * @skb: skb to get stamp from
2437 * @stamp: pointer to struct timeval to store stamp in
2438 *
2439 * Timestamps are stored in the skb as offsets to a base timestamp.
2440 * This function converts the offset back to a struct timeval and stores
2441 * it in stamp.
2442 */
ac45f602
PO
2443static inline void skb_get_timestamp(const struct sk_buff *skb,
2444 struct timeval *stamp)
a61bbcf2 2445{
b7aa0bf7 2446 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2447}
2448
ac45f602
PO
2449static inline void skb_get_timestampns(const struct sk_buff *skb,
2450 struct timespec *stamp)
2451{
2452 *stamp = ktime_to_timespec(skb->tstamp);
2453}
2454
b7aa0bf7 2455static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2456{
b7aa0bf7 2457 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2458}
2459
164891aa
SH
2460static inline ktime_t net_timedelta(ktime_t t)
2461{
2462 return ktime_sub(ktime_get_real(), t);
2463}
2464
b9ce204f
IJ
2465static inline ktime_t net_invalid_timestamp(void)
2466{
2467 return ktime_set(0, 0);
2468}
a61bbcf2 2469
7965bd4d 2470void skb_timestamping_init(void);
c1f19b51
RC
2471
2472#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2473
7965bd4d
JP
2474void skb_clone_tx_timestamp(struct sk_buff *skb);
2475bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2476
2477#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2478
2479static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2480{
2481}
2482
2483static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2484{
2485 return false;
2486}
2487
2488#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2489
2490/**
2491 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2492 *
da92b194
RC
2493 * PHY drivers may accept clones of transmitted packets for
2494 * timestamping via their phy_driver.txtstamp method. These drivers
2495 * must call this function to return the skb back to the stack, with
2496 * or without a timestamp.
2497 *
c1f19b51 2498 * @skb: clone of the the original outgoing packet
da92b194 2499 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2500 *
2501 */
2502void skb_complete_tx_timestamp(struct sk_buff *skb,
2503 struct skb_shared_hwtstamps *hwtstamps);
2504
ac45f602
PO
2505/**
2506 * skb_tstamp_tx - queue clone of skb with send time stamps
2507 * @orig_skb: the original outgoing packet
2508 * @hwtstamps: hardware time stamps, may be NULL if not available
2509 *
2510 * If the skb has a socket associated, then this function clones the
2511 * skb (thus sharing the actual data and optional structures), stores
2512 * the optional hardware time stamping information (if non NULL) or
2513 * generates a software time stamp (otherwise), then queues the clone
2514 * to the error queue of the socket. Errors are silently ignored.
2515 */
7965bd4d
JP
2516void skb_tstamp_tx(struct sk_buff *orig_skb,
2517 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2518
4507a715
RC
2519static inline void sw_tx_timestamp(struct sk_buff *skb)
2520{
2244d07b
OH
2521 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2522 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2523 skb_tstamp_tx(skb, NULL);
2524}
2525
2526/**
2527 * skb_tx_timestamp() - Driver hook for transmit timestamping
2528 *
2529 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2530 * function immediately before giving the sk_buff to the MAC hardware.
4507a715
RC
2531 *
2532 * @skb: A socket buffer.
2533 */
2534static inline void skb_tx_timestamp(struct sk_buff *skb)
2535{
c1f19b51 2536 skb_clone_tx_timestamp(skb);
4507a715
RC
2537 sw_tx_timestamp(skb);
2538}
2539
6e3e939f
JB
2540/**
2541 * skb_complete_wifi_ack - deliver skb with wifi status
2542 *
2543 * @skb: the original outgoing packet
2544 * @acked: ack status
2545 *
2546 */
2547void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2548
7965bd4d
JP
2549__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2550__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2551
60476372
HX
2552static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2553{
2554 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2555}
2556
fb286bb2
HX
2557/**
2558 * skb_checksum_complete - Calculate checksum of an entire packet
2559 * @skb: packet to process
2560 *
2561 * This function calculates the checksum over the entire packet plus
2562 * the value of skb->csum. The latter can be used to supply the
2563 * checksum of a pseudo header as used by TCP/UDP. It returns the
2564 * checksum.
2565 *
2566 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2567 * this function can be used to verify that checksum on received
2568 * packets. In that case the function should return zero if the
2569 * checksum is correct. In particular, this function will return zero
2570 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2571 * hardware has already verified the correctness of the checksum.
2572 */
4381ca3c 2573static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2574{
60476372
HX
2575 return skb_csum_unnecessary(skb) ?
2576 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2577}
2578
5f79e0f9 2579#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 2580void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
2581static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2582{
2583 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 2584 nf_conntrack_destroy(nfct);
1da177e4
LT
2585}
2586static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2587{
2588 if (nfct)
2589 atomic_inc(&nfct->use);
2590}
2fc72c7b 2591#endif
1da177e4
LT
2592#ifdef CONFIG_BRIDGE_NETFILTER
2593static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2594{
2595 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2596 kfree(nf_bridge);
2597}
2598static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2599{
2600 if (nf_bridge)
2601 atomic_inc(&nf_bridge->use);
2602}
2603#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
2604static inline void nf_reset(struct sk_buff *skb)
2605{
5f79e0f9 2606#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
2607 nf_conntrack_put(skb->nfct);
2608 skb->nfct = NULL;
2fc72c7b 2609#endif
a193a4ab
PM
2610#ifdef CONFIG_BRIDGE_NETFILTER
2611 nf_bridge_put(skb->nf_bridge);
2612 skb->nf_bridge = NULL;
2613#endif
2614}
2615
124dff01
PM
2616static inline void nf_reset_trace(struct sk_buff *skb)
2617{
130549fe
G
2618#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
2619 skb->nf_trace = 0;
2620#endif
a193a4ab
PM
2621}
2622
edda553c
YK
2623/* Note: This doesn't put any conntrack and bridge info in dst. */
2624static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2625{
5f79e0f9 2626#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
2627 dst->nfct = src->nfct;
2628 nf_conntrack_get(src->nfct);
2629 dst->nfctinfo = src->nfctinfo;
2fc72c7b 2630#endif
edda553c
YK
2631#ifdef CONFIG_BRIDGE_NETFILTER
2632 dst->nf_bridge = src->nf_bridge;
2633 nf_bridge_get(src->nf_bridge);
2634#endif
2635}
2636
e7ac05f3
YK
2637static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2638{
e7ac05f3 2639#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 2640 nf_conntrack_put(dst->nfct);
2fc72c7b 2641#endif
e7ac05f3
YK
2642#ifdef CONFIG_BRIDGE_NETFILTER
2643 nf_bridge_put(dst->nf_bridge);
2644#endif
2645 __nf_copy(dst, src);
2646}
2647
984bc16c
JM
2648#ifdef CONFIG_NETWORK_SECMARK
2649static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2650{
2651 to->secmark = from->secmark;
2652}
2653
2654static inline void skb_init_secmark(struct sk_buff *skb)
2655{
2656 skb->secmark = 0;
2657}
2658#else
2659static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2660{ }
2661
2662static inline void skb_init_secmark(struct sk_buff *skb)
2663{ }
2664#endif
2665
f25f4e44
PWJ
2666static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2667{
f25f4e44 2668 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
2669}
2670
9247744e 2671static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 2672{
4e3ab47a 2673 return skb->queue_mapping;
4e3ab47a
PE
2674}
2675
f25f4e44
PWJ
2676static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2677{
f25f4e44 2678 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
2679}
2680
d5a9e24a
DM
2681static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2682{
2683 skb->queue_mapping = rx_queue + 1;
2684}
2685
9247744e 2686static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
2687{
2688 return skb->queue_mapping - 1;
2689}
2690
9247744e 2691static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 2692{
a02cec21 2693 return skb->queue_mapping != 0;
d5a9e24a
DM
2694}
2695
7965bd4d
JP
2696u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2697 unsigned int num_tx_queues);
9247744e 2698
def8b4fa
AD
2699static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2700{
0b3d8e08 2701#ifdef CONFIG_XFRM
def8b4fa 2702 return skb->sp;
def8b4fa 2703#else
def8b4fa 2704 return NULL;
def8b4fa 2705#endif
0b3d8e08 2706}
def8b4fa 2707
68c33163
PS
2708/* Keeps track of mac header offset relative to skb->head.
2709 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2710 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
2711 * tunnel skb it points to outer mac header.
2712 * Keeps track of level of encapsulation of network headers.
2713 */
68c33163 2714struct skb_gso_cb {
3347c960
ED
2715 int mac_offset;
2716 int encap_level;
68c33163
PS
2717};
2718#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2719
2720static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2721{
2722 return (skb_mac_header(inner_skb) - inner_skb->head) -
2723 SKB_GSO_CB(inner_skb)->mac_offset;
2724}
2725
1e2bd517
PS
2726static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
2727{
2728 int new_headroom, headroom;
2729 int ret;
2730
2731 headroom = skb_headroom(skb);
2732 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
2733 if (ret)
2734 return ret;
2735
2736 new_headroom = skb_headroom(skb);
2737 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
2738 return 0;
2739}
2740
bdcc0924 2741static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
2742{
2743 return skb_shinfo(skb)->gso_size;
2744}
2745
36a8f39e 2746/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 2747static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
2748{
2749 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2750}
2751
7965bd4d 2752void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
2753
2754static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2755{
2756 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2757 * wanted then gso_type will be set. */
05bdd2f1
ED
2758 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2759
b78462eb
AD
2760 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2761 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
2762 __skb_warn_lro_forwarding(skb);
2763 return true;
2764 }
2765 return false;
2766}
2767
35fc92a9
HX
2768static inline void skb_forward_csum(struct sk_buff *skb)
2769{
2770 /* Unfortunately we don't support this one. Any brave souls? */
2771 if (skb->ip_summed == CHECKSUM_COMPLETE)
2772 skb->ip_summed = CHECKSUM_NONE;
2773}
2774
bc8acf2c
ED
2775/**
2776 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2777 * @skb: skb to check
2778 *
2779 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2780 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2781 * use this helper, to document places where we make this assertion.
2782 */
05bdd2f1 2783static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
2784{
2785#ifdef DEBUG
2786 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2787#endif
2788}
2789
f35d9d8a 2790bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 2791
f77668dc
DB
2792u32 __skb_get_poff(const struct sk_buff *skb);
2793
3a7c1ee4
AD
2794/**
2795 * skb_head_is_locked - Determine if the skb->head is locked down
2796 * @skb: skb to check
2797 *
2798 * The head on skbs build around a head frag can be removed if they are
2799 * not cloned. This function returns true if the skb head is locked down
2800 * due to either being allocated via kmalloc, or by being a clone with
2801 * multiple references to the head.
2802 */
2803static inline bool skb_head_is_locked(const struct sk_buff *skb)
2804{
2805 return !skb->head_frag || skb_cloned(skb);
2806}
1da177e4
LT
2807#endif /* __KERNEL__ */
2808#endif /* _LINUX_SKBUFF_H */