netfilter: nf_tables: fix nf_trace always-on with XT_TRACE=n
[deliverable/linux.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
78ea85f1
DB
37/* A. Checksumming of received packets by device.
38 *
39 * CHECKSUM_NONE:
40 *
41 * Device failed to checksum this packet e.g. due to lack of capabilities.
42 * The packet contains full (though not verified) checksum in packet but
43 * not in skb->csum. Thus, skb->csum is undefined in this case.
44 *
45 * CHECKSUM_UNNECESSARY:
46 *
47 * The hardware you're dealing with doesn't calculate the full checksum
48 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
49 * for specific protocols e.g. TCP/UDP/SCTP, then, for such packets it will
50 * set CHECKSUM_UNNECESSARY if their checksums are okay. skb->csum is still
51 * undefined in this case though. It is a bad option, but, unfortunately,
52 * nowadays most vendors do this. Apparently with the secret goal to sell
53 * you new devices, when you will add new protocol to your host, f.e. IPv6 8)
54 *
55 * CHECKSUM_COMPLETE:
56 *
57 * This is the most generic way. The device supplied checksum of the _whole_
58 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
59 * hardware doesn't need to parse L3/L4 headers to implement this.
60 *
61 * Note: Even if device supports only some protocols, but is able to produce
62 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
63 *
64 * CHECKSUM_PARTIAL:
65 *
66 * This is identical to the case for output below. This may occur on a packet
67 * received directly from another Linux OS, e.g., a virtualized Linux kernel
68 * on the same host. The packet can be treated in the same way as
69 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
70 * checksum must be filled in by the OS or the hardware.
71 *
72 * B. Checksumming on output.
73 *
74 * CHECKSUM_NONE:
75 *
76 * The skb was already checksummed by the protocol, or a checksum is not
77 * required.
78 *
79 * CHECKSUM_PARTIAL:
80 *
81 * The device is required to checksum the packet as seen by hard_start_xmit()
82 * from skb->csum_start up to the end, and to record/write the checksum at
83 * offset skb->csum_start + skb->csum_offset.
84 *
85 * The device must show its capabilities in dev->features, set up at device
86 * setup time, e.g. netdev_features.h:
87 *
88 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
89 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
90 * IPv4. Sigh. Vendors like this way for an unknown reason.
91 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
92 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
93 * NETIF_F_... - Well, you get the picture.
94 *
95 * CHECKSUM_UNNECESSARY:
96 *
97 * Normally, the device will do per protocol specific checksumming. Protocol
98 * implementations that do not want the NIC to perform the checksum
99 * calculation should use this flag in their outgoing skbs.
100 *
101 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
102 * offload. Correspondingly, the FCoE protocol driver
103 * stack should use CHECKSUM_UNNECESSARY.
104 *
105 * Any questions? No questions, good. --ANK
106 */
107
60476372 108/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
109#define CHECKSUM_NONE 0
110#define CHECKSUM_UNNECESSARY 1
111#define CHECKSUM_COMPLETE 2
112#define CHECKSUM_PARTIAL 3
1da177e4
LT
113
114#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
115 ~(SMP_CACHE_BYTES - 1))
fc910a27 116#define SKB_WITH_OVERHEAD(X) \
deea84b0 117 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
118#define SKB_MAX_ORDER(X, ORDER) \
119 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
120#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
121#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
122
87fb4b7b
ED
123/* return minimum truesize of one skb containing X bytes of data */
124#define SKB_TRUESIZE(X) ((X) + \
125 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
126 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
127
1da177e4 128struct net_device;
716ea3a7 129struct scatterlist;
9c55e01c 130struct pipe_inode_info;
1da177e4 131
5f79e0f9 132#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
133struct nf_conntrack {
134 atomic_t use;
1da177e4 135};
5f79e0f9 136#endif
1da177e4
LT
137
138#ifdef CONFIG_BRIDGE_NETFILTER
139struct nf_bridge_info {
bf1ac5ca
ED
140 atomic_t use;
141 unsigned int mask;
142 struct net_device *physindev;
143 struct net_device *physoutdev;
144 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
145};
146#endif
147
1da177e4
LT
148struct sk_buff_head {
149 /* These two members must be first. */
150 struct sk_buff *next;
151 struct sk_buff *prev;
152
153 __u32 qlen;
154 spinlock_t lock;
155};
156
157struct sk_buff;
158
9d4dde52
IC
159/* To allow 64K frame to be packed as single skb without frag_list we
160 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
161 * buffers which do not start on a page boundary.
162 *
163 * Since GRO uses frags we allocate at least 16 regardless of page
164 * size.
a715dea3 165 */
9d4dde52 166#if (65536/PAGE_SIZE + 1) < 16
eec00954 167#define MAX_SKB_FRAGS 16UL
a715dea3 168#else
9d4dde52 169#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 170#endif
1da177e4
LT
171
172typedef struct skb_frag_struct skb_frag_t;
173
174struct skb_frag_struct {
a8605c60
IC
175 struct {
176 struct page *p;
177 } page;
cb4dfe56 178#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
179 __u32 page_offset;
180 __u32 size;
cb4dfe56
ED
181#else
182 __u16 page_offset;
183 __u16 size;
184#endif
1da177e4
LT
185};
186
9e903e08
ED
187static inline unsigned int skb_frag_size(const skb_frag_t *frag)
188{
189 return frag->size;
190}
191
192static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
193{
194 frag->size = size;
195}
196
197static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
198{
199 frag->size += delta;
200}
201
202static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
203{
204 frag->size -= delta;
205}
206
ac45f602
PO
207#define HAVE_HW_TIME_STAMP
208
209/**
d3a21be8 210 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
211 * @hwtstamp: hardware time stamp transformed into duration
212 * since arbitrary point in time
213 * @syststamp: hwtstamp transformed to system time base
214 *
215 * Software time stamps generated by ktime_get_real() are stored in
216 * skb->tstamp. The relation between the different kinds of time
217 * stamps is as follows:
218 *
219 * syststamp and tstamp can be compared against each other in
220 * arbitrary combinations. The accuracy of a
221 * syststamp/tstamp/"syststamp from other device" comparison is
222 * limited by the accuracy of the transformation into system time
223 * base. This depends on the device driver and its underlying
224 * hardware.
225 *
226 * hwtstamps can only be compared against other hwtstamps from
227 * the same device.
228 *
229 * This structure is attached to packets as part of the
230 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
231 */
232struct skb_shared_hwtstamps {
233 ktime_t hwtstamp;
234 ktime_t syststamp;
235};
236
2244d07b
OH
237/* Definitions for tx_flags in struct skb_shared_info */
238enum {
239 /* generate hardware time stamp */
240 SKBTX_HW_TSTAMP = 1 << 0,
241
242 /* generate software time stamp */
243 SKBTX_SW_TSTAMP = 1 << 1,
244
245 /* device driver is going to provide hardware time stamp */
246 SKBTX_IN_PROGRESS = 1 << 2,
247
a6686f2f 248 /* device driver supports TX zero-copy buffers */
62b1a8ab 249 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
250
251 /* generate wifi status information (where possible) */
62b1a8ab 252 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
253
254 /* This indicates at least one fragment might be overwritten
255 * (as in vmsplice(), sendfile() ...)
256 * If we need to compute a TX checksum, we'll need to copy
257 * all frags to avoid possible bad checksum
258 */
259 SKBTX_SHARED_FRAG = 1 << 5,
a6686f2f
SM
260};
261
262/*
263 * The callback notifies userspace to release buffers when skb DMA is done in
264 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
265 * The zerocopy_success argument is true if zero copy transmit occurred,
266 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
267 * The ctx field is used to track device context.
268 * The desc field is used to track userspace buffer index.
a6686f2f
SM
269 */
270struct ubuf_info {
e19d6763 271 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 272 void *ctx;
a6686f2f 273 unsigned long desc;
ac45f602
PO
274};
275
1da177e4
LT
276/* This data is invariant across clones and lives at
277 * the end of the header data, ie. at skb->end.
278 */
279struct skb_shared_info {
9f42f126
IC
280 unsigned char nr_frags;
281 __u8 tx_flags;
7967168c
HX
282 unsigned short gso_size;
283 /* Warning: this field is not always filled in (UFO)! */
284 unsigned short gso_segs;
285 unsigned short gso_type;
1da177e4 286 struct sk_buff *frag_list;
ac45f602 287 struct skb_shared_hwtstamps hwtstamps;
9f42f126 288 __be32 ip6_frag_id;
ec7d2f2c
ED
289
290 /*
291 * Warning : all fields before dataref are cleared in __alloc_skb()
292 */
293 atomic_t dataref;
294
69e3c75f
JB
295 /* Intermediate layers must ensure that destructor_arg
296 * remains valid until skb destructor */
297 void * destructor_arg;
a6686f2f 298
fed66381
ED
299 /* must be last field, see pskb_expand_head() */
300 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
301};
302
303/* We divide dataref into two halves. The higher 16 bits hold references
304 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
305 * the entire skb->data. A clone of a headerless skb holds the length of
306 * the header in skb->hdr_len.
1da177e4
LT
307 *
308 * All users must obey the rule that the skb->data reference count must be
309 * greater than or equal to the payload reference count.
310 *
311 * Holding a reference to the payload part means that the user does not
312 * care about modifications to the header part of skb->data.
313 */
314#define SKB_DATAREF_SHIFT 16
315#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
316
d179cd12
DM
317
318enum {
319 SKB_FCLONE_UNAVAILABLE,
320 SKB_FCLONE_ORIG,
321 SKB_FCLONE_CLONE,
322};
323
7967168c
HX
324enum {
325 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 326 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
327
328 /* This indicates the skb is from an untrusted source. */
329 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
330
331 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
332 SKB_GSO_TCP_ECN = 1 << 3,
333
334 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
335
336 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
337
338 SKB_GSO_GRE = 1 << 6,
73136267 339
cb32f511 340 SKB_GSO_IPIP = 1 << 7,
0d89d203 341
61c1db7f 342 SKB_GSO_SIT = 1 << 8,
cb32f511 343
61c1db7f
ED
344 SKB_GSO_UDP_TUNNEL = 1 << 9,
345
346 SKB_GSO_MPLS = 1 << 10,
7967168c
HX
347};
348
2e07fa9c
ACM
349#if BITS_PER_LONG > 32
350#define NET_SKBUFF_DATA_USES_OFFSET 1
351#endif
352
353#ifdef NET_SKBUFF_DATA_USES_OFFSET
354typedef unsigned int sk_buff_data_t;
355#else
356typedef unsigned char *sk_buff_data_t;
357#endif
358
1da177e4
LT
359/**
360 * struct sk_buff - socket buffer
361 * @next: Next buffer in list
362 * @prev: Previous buffer in list
325ed823 363 * @tstamp: Time we arrived
d84e0bd7 364 * @sk: Socket we are owned by
1da177e4 365 * @dev: Device we arrived on/are leaving by
d84e0bd7 366 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 367 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 368 * @sp: the security path, used for xfrm
1da177e4
LT
369 * @len: Length of actual data
370 * @data_len: Data length
371 * @mac_len: Length of link layer header
334a8132 372 * @hdr_len: writable header length of cloned skb
663ead3b
HX
373 * @csum: Checksum (must include start/offset pair)
374 * @csum_start: Offset from skb->head where checksumming should start
375 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 376 * @priority: Packet queueing priority
67be2dd1 377 * @local_df: allow local fragmentation
1da177e4 378 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 379 * @ip_summed: Driver fed us an IP checksum
1da177e4 380 * @nohdr: Payload reference only, must not modify header
d84e0bd7 381 * @nfctinfo: Relationship of this skb to the connection
1da177e4 382 * @pkt_type: Packet class
c83c2486 383 * @fclone: skbuff clone status
c83c2486 384 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
385 * @peeked: this packet has been seen already, so stats have been
386 * done for it, don't do them again
ba9dda3a 387 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
388 * @protocol: Packet protocol from driver
389 * @destructor: Destruct function
390 * @nfct: Associated connection, if any
1da177e4 391 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 392 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
393 * @tc_index: Traffic control index
394 * @tc_verd: traffic control verdict
d84e0bd7
DB
395 * @rxhash: the packet hash computed on receive
396 * @queue_mapping: Queue mapping for multiqueue devices
553a5672 397 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 398 * @ooo_okay: allow the mapping of a socket to a queue to be changed
4ca2462e
CG
399 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
400 * ports.
6e3e939f
JB
401 * @wifi_acked_valid: wifi_acked was set
402 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 403 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
f4b8ea78
RD
404 * @dma_cookie: a cookie to one of several possible DMA operations
405 * done by skb DMA functions
06021292 406 * @napi_id: id of the NAPI struct this skb came from
984bc16c 407 * @secmark: security marking
d84e0bd7
DB
408 * @mark: Generic packet mark
409 * @dropcount: total number of sk_receive_queue overflows
86a9bad3 410 * @vlan_proto: vlan encapsulation protocol
6aa895b0 411 * @vlan_tci: vlan tag control information
0d89d203 412 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
413 * @inner_transport_header: Inner transport layer header (encapsulation)
414 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 415 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
416 * @transport_header: Transport layer header
417 * @network_header: Network layer header
418 * @mac_header: Link layer header
419 * @tail: Tail pointer
420 * @end: End pointer
421 * @head: Head of buffer
422 * @data: Data head pointer
423 * @truesize: Buffer size
424 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
425 */
426
427struct sk_buff {
428 /* These two members must be first. */
429 struct sk_buff *next;
430 struct sk_buff *prev;
431
b7aa0bf7 432 ktime_t tstamp;
da3f5cf1
FF
433
434 struct sock *sk;
1da177e4 435 struct net_device *dev;
1da177e4 436
1da177e4
LT
437 /*
438 * This is the control buffer. It is free to use for every
439 * layer. Please put your private variables there. If you
440 * want to keep them across layers you have to do a skb_clone()
441 * first. This is owned by whoever has the skb queued ATM.
442 */
da3f5cf1 443 char cb[48] __aligned(8);
1da177e4 444
7fee226a 445 unsigned long _skb_refdst;
da3f5cf1
FF
446#ifdef CONFIG_XFRM
447 struct sec_path *sp;
448#endif
1da177e4 449 unsigned int len,
334a8132
PM
450 data_len;
451 __u16 mac_len,
452 hdr_len;
ff1dcadb
AV
453 union {
454 __wsum csum;
663ead3b
HX
455 struct {
456 __u16 csum_start;
457 __u16 csum_offset;
458 };
ff1dcadb 459 };
1da177e4 460 __u32 priority;
fe55f6d5 461 kmemcheck_bitfield_begin(flags1);
1cbb3380
TG
462 __u8 local_df:1,
463 cloned:1,
464 ip_summed:2,
6869c4d8
HW
465 nohdr:1,
466 nfctinfo:3;
d179cd12 467 __u8 pkt_type:3,
b84f4cc9 468 fclone:2,
ba9dda3a 469 ipvs_property:1,
a59322be 470 peeked:1,
ba9dda3a 471 nf_trace:1;
fe55f6d5 472 kmemcheck_bitfield_end(flags1);
4ab408de 473 __be16 protocol;
1da177e4
LT
474
475 void (*destructor)(struct sk_buff *skb);
9fb9cbb1 476#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 477 struct nf_conntrack *nfct;
2fc72c7b 478#endif
1da177e4
LT
479#ifdef CONFIG_BRIDGE_NETFILTER
480 struct nf_bridge_info *nf_bridge;
481#endif
f25f4e44 482
8964be4a 483 int skb_iif;
4031ae6e
AD
484
485 __u32 rxhash;
486
86a9bad3 487 __be16 vlan_proto;
4031ae6e
AD
488 __u16 vlan_tci;
489
1da177e4 490#ifdef CONFIG_NET_SCHED
b6b99eb5 491 __u16 tc_index; /* traffic control index */
1da177e4 492#ifdef CONFIG_NET_CLS_ACT
b6b99eb5 493 __u16 tc_verd; /* traffic control verdict */
1da177e4 494#endif
1da177e4 495#endif
fe55f6d5 496
0a14842f 497 __u16 queue_mapping;
fe55f6d5 498 kmemcheck_bitfield_begin(flags2);
de357cc0 499#ifdef CONFIG_IPV6_NDISC_NODETYPE
8a4eb573 500 __u8 ndisc_nodetype:2;
d0f09804 501#endif
c93bdd0e 502 __u8 pfmemalloc:1;
3853b584 503 __u8 ooo_okay:1;
bdeab991 504 __u8 l4_rxhash:1;
6e3e939f
JB
505 __u8 wifi_acked_valid:1;
506 __u8 wifi_acked:1;
3bdc0eba 507 __u8 no_fcs:1;
d3836f21 508 __u8 head_frag:1;
6a674e9c
JG
509 /* Encapsulation protocol and NIC drivers should use
510 * this flag to indicate to each other if the skb contains
511 * encapsulated packet or not and maybe use the inner packet
512 * headers if needed
513 */
514 __u8 encapsulation:1;
45906723 515 /* 6/8 bit hole (depending on ndisc_nodetype presence) */
fe55f6d5
VN
516 kmemcheck_bitfield_end(flags2);
517
e0d1095a 518#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
06021292
ET
519 union {
520 unsigned int napi_id;
521 dma_cookie_t dma_cookie;
522 };
97fc2f08 523#endif
984bc16c
JM
524#ifdef CONFIG_NETWORK_SECMARK
525 __u32 secmark;
526#endif
3b885787
NH
527 union {
528 __u32 mark;
529 __u32 dropcount;
16fad69c 530 __u32 reserved_tailroom;
3b885787 531 };
1da177e4 532
0d89d203 533 __be16 inner_protocol;
1a37e412
SH
534 __u16 inner_transport_header;
535 __u16 inner_network_header;
536 __u16 inner_mac_header;
537 __u16 transport_header;
538 __u16 network_header;
539 __u16 mac_header;
1da177e4 540 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 541 sk_buff_data_t tail;
4305b541 542 sk_buff_data_t end;
1da177e4 543 unsigned char *head,
4305b541 544 *data;
27a884dc
ACM
545 unsigned int truesize;
546 atomic_t users;
1da177e4
LT
547};
548
549#ifdef __KERNEL__
550/*
551 * Handling routines are only of interest to the kernel
552 */
553#include <linux/slab.h>
554
1da177e4 555
c93bdd0e
MG
556#define SKB_ALLOC_FCLONE 0x01
557#define SKB_ALLOC_RX 0x02
558
559/* Returns true if the skb was allocated from PFMEMALLOC reserves */
560static inline bool skb_pfmemalloc(const struct sk_buff *skb)
561{
562 return unlikely(skb->pfmemalloc);
563}
564
7fee226a
ED
565/*
566 * skb might have a dst pointer attached, refcounted or not.
567 * _skb_refdst low order bit is set if refcount was _not_ taken
568 */
569#define SKB_DST_NOREF 1UL
570#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
571
572/**
573 * skb_dst - returns skb dst_entry
574 * @skb: buffer
575 *
576 * Returns skb dst_entry, regardless of reference taken or not.
577 */
adf30907
ED
578static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
579{
7fee226a
ED
580 /* If refdst was not refcounted, check we still are in a
581 * rcu_read_lock section
582 */
583 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
584 !rcu_read_lock_held() &&
585 !rcu_read_lock_bh_held());
586 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
587}
588
7fee226a
ED
589/**
590 * skb_dst_set - sets skb dst
591 * @skb: buffer
592 * @dst: dst entry
593 *
594 * Sets skb dst, assuming a reference was taken on dst and should
595 * be released by skb_dst_drop()
596 */
adf30907
ED
597static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
598{
7fee226a
ED
599 skb->_skb_refdst = (unsigned long)dst;
600}
601
7965bd4d
JP
602void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
603 bool force);
932bc4d7
JA
604
605/**
606 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
607 * @skb: buffer
608 * @dst: dst entry
609 *
610 * Sets skb dst, assuming a reference was not taken on dst.
611 * If dst entry is cached, we do not take reference and dst_release
612 * will be avoided by refdst_drop. If dst entry is not cached, we take
613 * reference, so that last dst_release can destroy the dst immediately.
614 */
615static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
616{
617 __skb_dst_set_noref(skb, dst, false);
618}
619
620/**
621 * skb_dst_set_noref_force - sets skb dst, without taking reference
622 * @skb: buffer
623 * @dst: dst entry
624 *
625 * Sets skb dst, assuming a reference was not taken on dst.
626 * No reference is taken and no dst_release will be called. While for
627 * cached dsts deferred reclaim is a basic feature, for entries that are
628 * not cached it is caller's job to guarantee that last dst_release for
629 * provided dst happens when nobody uses it, eg. after a RCU grace period.
630 */
631static inline void skb_dst_set_noref_force(struct sk_buff *skb,
632 struct dst_entry *dst)
633{
634 __skb_dst_set_noref(skb, dst, true);
635}
7fee226a
ED
636
637/**
25985edc 638 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
639 * @skb: buffer
640 */
641static inline bool skb_dst_is_noref(const struct sk_buff *skb)
642{
643 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
644}
645
511c3f92
ED
646static inline struct rtable *skb_rtable(const struct sk_buff *skb)
647{
adf30907 648 return (struct rtable *)skb_dst(skb);
511c3f92
ED
649}
650
7965bd4d
JP
651void kfree_skb(struct sk_buff *skb);
652void kfree_skb_list(struct sk_buff *segs);
653void skb_tx_error(struct sk_buff *skb);
654void consume_skb(struct sk_buff *skb);
655void __kfree_skb(struct sk_buff *skb);
d7e8883c 656extern struct kmem_cache *skbuff_head_cache;
bad43ca8 657
7965bd4d
JP
658void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
659bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
660 bool *fragstolen, int *delta_truesize);
bad43ca8 661
7965bd4d
JP
662struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
663 int node);
664struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 665static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 666 gfp_t priority)
d179cd12 667{
564824b0 668 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
669}
670
671static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 672 gfp_t priority)
d179cd12 673{
c93bdd0e 674 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
675}
676
7965bd4d 677struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
678static inline struct sk_buff *alloc_skb_head(gfp_t priority)
679{
680 return __alloc_skb_head(priority, -1);
681}
682
7965bd4d
JP
683struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
684int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
685struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
686struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
687struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
688
689int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
690struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
691 unsigned int headroom);
692struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
693 int newtailroom, gfp_t priority);
694int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
695 int len);
696int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
697int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 698#define dev_kfree_skb(a) consume_skb(a)
1da177e4 699
7965bd4d
JP
700int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
701 int getfrag(void *from, char *to, int offset,
702 int len, int odd, struct sk_buff *skb),
703 void *from, int length);
e89e9cf5 704
d94d9fee 705struct skb_seq_state {
677e90ed
TG
706 __u32 lower_offset;
707 __u32 upper_offset;
708 __u32 frag_idx;
709 __u32 stepped_offset;
710 struct sk_buff *root_skb;
711 struct sk_buff *cur_skb;
712 __u8 *frag_data;
713};
714
7965bd4d
JP
715void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
716 unsigned int to, struct skb_seq_state *st);
717unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
718 struct skb_seq_state *st);
719void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 720
7965bd4d
JP
721unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
722 unsigned int to, struct ts_config *config,
723 struct ts_state *state);
3fc7e8a6 724
09323cc4
TH
725/*
726 * Packet hash types specify the type of hash in skb_set_hash.
727 *
728 * Hash types refer to the protocol layer addresses which are used to
729 * construct a packet's hash. The hashes are used to differentiate or identify
730 * flows of the protocol layer for the hash type. Hash types are either
731 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
732 *
733 * Properties of hashes:
734 *
735 * 1) Two packets in different flows have different hash values
736 * 2) Two packets in the same flow should have the same hash value
737 *
738 * A hash at a higher layer is considered to be more specific. A driver should
739 * set the most specific hash possible.
740 *
741 * A driver cannot indicate a more specific hash than the layer at which a hash
742 * was computed. For instance an L3 hash cannot be set as an L4 hash.
743 *
744 * A driver may indicate a hash level which is less specific than the
745 * actual layer the hash was computed on. For instance, a hash computed
746 * at L4 may be considered an L3 hash. This should only be done if the
747 * driver can't unambiguously determine that the HW computed the hash at
748 * the higher layer. Note that the "should" in the second property above
749 * permits this.
750 */
751enum pkt_hash_types {
752 PKT_HASH_TYPE_NONE, /* Undefined type */
753 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
754 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
755 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
756};
757
758static inline void
759skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
760{
761 skb->l4_rxhash = (type == PKT_HASH_TYPE_L4);
762 skb->rxhash = hash;
763}
764
3958afa1
TH
765void __skb_get_hash(struct sk_buff *skb);
766static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 767{
ecd5cf5d 768 if (!skb->l4_rxhash)
3958afa1 769 __skb_get_hash(skb);
bfb564e7
KK
770
771 return skb->rxhash;
772}
773
57bdf7f4
TH
774static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
775{
776 return skb->rxhash;
777}
778
7539fadc
TH
779static inline void skb_clear_hash(struct sk_buff *skb)
780{
781 skb->rxhash = 0;
782 skb->l4_rxhash = 0;
783}
784
785static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
786{
787 if (!skb->l4_rxhash)
788 skb_clear_hash(skb);
789}
790
3df7a74e
TH
791static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
792{
793 to->rxhash = from->rxhash;
794 to->l4_rxhash = from->l4_rxhash;
795};
796
4305b541
ACM
797#ifdef NET_SKBUFF_DATA_USES_OFFSET
798static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
799{
800 return skb->head + skb->end;
801}
ec47ea82
AD
802
803static inline unsigned int skb_end_offset(const struct sk_buff *skb)
804{
805 return skb->end;
806}
4305b541
ACM
807#else
808static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
809{
810 return skb->end;
811}
ec47ea82
AD
812
813static inline unsigned int skb_end_offset(const struct sk_buff *skb)
814{
815 return skb->end - skb->head;
816}
4305b541
ACM
817#endif
818
1da177e4 819/* Internal */
4305b541 820#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 821
ac45f602
PO
822static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
823{
824 return &skb_shinfo(skb)->hwtstamps;
825}
826
1da177e4
LT
827/**
828 * skb_queue_empty - check if a queue is empty
829 * @list: queue head
830 *
831 * Returns true if the queue is empty, false otherwise.
832 */
833static inline int skb_queue_empty(const struct sk_buff_head *list)
834{
fd44b93c 835 return list->next == (const struct sk_buff *) list;
1da177e4
LT
836}
837
fc7ebb21
DM
838/**
839 * skb_queue_is_last - check if skb is the last entry in the queue
840 * @list: queue head
841 * @skb: buffer
842 *
843 * Returns true if @skb is the last buffer on the list.
844 */
845static inline bool skb_queue_is_last(const struct sk_buff_head *list,
846 const struct sk_buff *skb)
847{
fd44b93c 848 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
849}
850
832d11c5
IJ
851/**
852 * skb_queue_is_first - check if skb is the first entry in the queue
853 * @list: queue head
854 * @skb: buffer
855 *
856 * Returns true if @skb is the first buffer on the list.
857 */
858static inline bool skb_queue_is_first(const struct sk_buff_head *list,
859 const struct sk_buff *skb)
860{
fd44b93c 861 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
862}
863
249c8b42
DM
864/**
865 * skb_queue_next - return the next packet in the queue
866 * @list: queue head
867 * @skb: current buffer
868 *
869 * Return the next packet in @list after @skb. It is only valid to
870 * call this if skb_queue_is_last() evaluates to false.
871 */
872static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
873 const struct sk_buff *skb)
874{
875 /* This BUG_ON may seem severe, but if we just return then we
876 * are going to dereference garbage.
877 */
878 BUG_ON(skb_queue_is_last(list, skb));
879 return skb->next;
880}
881
832d11c5
IJ
882/**
883 * skb_queue_prev - return the prev packet in the queue
884 * @list: queue head
885 * @skb: current buffer
886 *
887 * Return the prev packet in @list before @skb. It is only valid to
888 * call this if skb_queue_is_first() evaluates to false.
889 */
890static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
891 const struct sk_buff *skb)
892{
893 /* This BUG_ON may seem severe, but if we just return then we
894 * are going to dereference garbage.
895 */
896 BUG_ON(skb_queue_is_first(list, skb));
897 return skb->prev;
898}
899
1da177e4
LT
900/**
901 * skb_get - reference buffer
902 * @skb: buffer to reference
903 *
904 * Makes another reference to a socket buffer and returns a pointer
905 * to the buffer.
906 */
907static inline struct sk_buff *skb_get(struct sk_buff *skb)
908{
909 atomic_inc(&skb->users);
910 return skb;
911}
912
913/*
914 * If users == 1, we are the only owner and are can avoid redundant
915 * atomic change.
916 */
917
1da177e4
LT
918/**
919 * skb_cloned - is the buffer a clone
920 * @skb: buffer to check
921 *
922 * Returns true if the buffer was generated with skb_clone() and is
923 * one of multiple shared copies of the buffer. Cloned buffers are
924 * shared data so must not be written to under normal circumstances.
925 */
926static inline int skb_cloned(const struct sk_buff *skb)
927{
928 return skb->cloned &&
929 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
930}
931
14bbd6a5
PS
932static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
933{
934 might_sleep_if(pri & __GFP_WAIT);
935
936 if (skb_cloned(skb))
937 return pskb_expand_head(skb, 0, 0, pri);
938
939 return 0;
940}
941
1da177e4
LT
942/**
943 * skb_header_cloned - is the header a clone
944 * @skb: buffer to check
945 *
946 * Returns true if modifying the header part of the buffer requires
947 * the data to be copied.
948 */
949static inline int skb_header_cloned(const struct sk_buff *skb)
950{
951 int dataref;
952
953 if (!skb->cloned)
954 return 0;
955
956 dataref = atomic_read(&skb_shinfo(skb)->dataref);
957 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
958 return dataref != 1;
959}
960
961/**
962 * skb_header_release - release reference to header
963 * @skb: buffer to operate on
964 *
965 * Drop a reference to the header part of the buffer. This is done
966 * by acquiring a payload reference. You must not read from the header
967 * part of skb->data after this.
968 */
969static inline void skb_header_release(struct sk_buff *skb)
970{
971 BUG_ON(skb->nohdr);
972 skb->nohdr = 1;
973 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
974}
975
976/**
977 * skb_shared - is the buffer shared
978 * @skb: buffer to check
979 *
980 * Returns true if more than one person has a reference to this
981 * buffer.
982 */
983static inline int skb_shared(const struct sk_buff *skb)
984{
985 return atomic_read(&skb->users) != 1;
986}
987
988/**
989 * skb_share_check - check if buffer is shared and if so clone it
990 * @skb: buffer to check
991 * @pri: priority for memory allocation
992 *
993 * If the buffer is shared the buffer is cloned and the old copy
994 * drops a reference. A new clone with a single reference is returned.
995 * If the buffer is not shared the original buffer is returned. When
996 * being called from interrupt status or with spinlocks held pri must
997 * be GFP_ATOMIC.
998 *
999 * NULL is returned on a memory allocation failure.
1000 */
47061bc4 1001static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1002{
1003 might_sleep_if(pri & __GFP_WAIT);
1004 if (skb_shared(skb)) {
1005 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1006
1007 if (likely(nskb))
1008 consume_skb(skb);
1009 else
1010 kfree_skb(skb);
1da177e4
LT
1011 skb = nskb;
1012 }
1013 return skb;
1014}
1015
1016/*
1017 * Copy shared buffers into a new sk_buff. We effectively do COW on
1018 * packets to handle cases where we have a local reader and forward
1019 * and a couple of other messy ones. The normal one is tcpdumping
1020 * a packet thats being forwarded.
1021 */
1022
1023/**
1024 * skb_unshare - make a copy of a shared buffer
1025 * @skb: buffer to check
1026 * @pri: priority for memory allocation
1027 *
1028 * If the socket buffer is a clone then this function creates a new
1029 * copy of the data, drops a reference count on the old copy and returns
1030 * the new copy with the reference count at 1. If the buffer is not a clone
1031 * the original buffer is returned. When called with a spinlock held or
1032 * from interrupt state @pri must be %GFP_ATOMIC
1033 *
1034 * %NULL is returned on a memory allocation failure.
1035 */
e2bf521d 1036static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1037 gfp_t pri)
1da177e4
LT
1038{
1039 might_sleep_if(pri & __GFP_WAIT);
1040 if (skb_cloned(skb)) {
1041 struct sk_buff *nskb = skb_copy(skb, pri);
1042 kfree_skb(skb); /* Free our shared copy */
1043 skb = nskb;
1044 }
1045 return skb;
1046}
1047
1048/**
1a5778aa 1049 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1050 * @list_: list to peek at
1051 *
1052 * Peek an &sk_buff. Unlike most other operations you _MUST_
1053 * be careful with this one. A peek leaves the buffer on the
1054 * list and someone else may run off with it. You must hold
1055 * the appropriate locks or have a private queue to do this.
1056 *
1057 * Returns %NULL for an empty list or a pointer to the head element.
1058 * The reference count is not incremented and the reference is therefore
1059 * volatile. Use with caution.
1060 */
05bdd2f1 1061static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1062{
18d07000
ED
1063 struct sk_buff *skb = list_->next;
1064
1065 if (skb == (struct sk_buff *)list_)
1066 skb = NULL;
1067 return skb;
1da177e4
LT
1068}
1069
da5ef6e5
PE
1070/**
1071 * skb_peek_next - peek skb following the given one from a queue
1072 * @skb: skb to start from
1073 * @list_: list to peek at
1074 *
1075 * Returns %NULL when the end of the list is met or a pointer to the
1076 * next element. The reference count is not incremented and the
1077 * reference is therefore volatile. Use with caution.
1078 */
1079static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1080 const struct sk_buff_head *list_)
1081{
1082 struct sk_buff *next = skb->next;
18d07000 1083
da5ef6e5
PE
1084 if (next == (struct sk_buff *)list_)
1085 next = NULL;
1086 return next;
1087}
1088
1da177e4 1089/**
1a5778aa 1090 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1091 * @list_: list to peek at
1092 *
1093 * Peek an &sk_buff. Unlike most other operations you _MUST_
1094 * be careful with this one. A peek leaves the buffer on the
1095 * list and someone else may run off with it. You must hold
1096 * the appropriate locks or have a private queue to do this.
1097 *
1098 * Returns %NULL for an empty list or a pointer to the tail element.
1099 * The reference count is not incremented and the reference is therefore
1100 * volatile. Use with caution.
1101 */
05bdd2f1 1102static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1103{
18d07000
ED
1104 struct sk_buff *skb = list_->prev;
1105
1106 if (skb == (struct sk_buff *)list_)
1107 skb = NULL;
1108 return skb;
1109
1da177e4
LT
1110}
1111
1112/**
1113 * skb_queue_len - get queue length
1114 * @list_: list to measure
1115 *
1116 * Return the length of an &sk_buff queue.
1117 */
1118static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1119{
1120 return list_->qlen;
1121}
1122
67fed459
DM
1123/**
1124 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1125 * @list: queue to initialize
1126 *
1127 * This initializes only the list and queue length aspects of
1128 * an sk_buff_head object. This allows to initialize the list
1129 * aspects of an sk_buff_head without reinitializing things like
1130 * the spinlock. It can also be used for on-stack sk_buff_head
1131 * objects where the spinlock is known to not be used.
1132 */
1133static inline void __skb_queue_head_init(struct sk_buff_head *list)
1134{
1135 list->prev = list->next = (struct sk_buff *)list;
1136 list->qlen = 0;
1137}
1138
76f10ad0
AV
1139/*
1140 * This function creates a split out lock class for each invocation;
1141 * this is needed for now since a whole lot of users of the skb-queue
1142 * infrastructure in drivers have different locking usage (in hardirq)
1143 * than the networking core (in softirq only). In the long run either the
1144 * network layer or drivers should need annotation to consolidate the
1145 * main types of usage into 3 classes.
1146 */
1da177e4
LT
1147static inline void skb_queue_head_init(struct sk_buff_head *list)
1148{
1149 spin_lock_init(&list->lock);
67fed459 1150 __skb_queue_head_init(list);
1da177e4
LT
1151}
1152
c2ecba71
PE
1153static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1154 struct lock_class_key *class)
1155{
1156 skb_queue_head_init(list);
1157 lockdep_set_class(&list->lock, class);
1158}
1159
1da177e4 1160/*
bf299275 1161 * Insert an sk_buff on a list.
1da177e4
LT
1162 *
1163 * The "__skb_xxxx()" functions are the non-atomic ones that
1164 * can only be called with interrupts disabled.
1165 */
7965bd4d
JP
1166void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1167 struct sk_buff_head *list);
bf299275
GR
1168static inline void __skb_insert(struct sk_buff *newsk,
1169 struct sk_buff *prev, struct sk_buff *next,
1170 struct sk_buff_head *list)
1171{
1172 newsk->next = next;
1173 newsk->prev = prev;
1174 next->prev = prev->next = newsk;
1175 list->qlen++;
1176}
1da177e4 1177
67fed459
DM
1178static inline void __skb_queue_splice(const struct sk_buff_head *list,
1179 struct sk_buff *prev,
1180 struct sk_buff *next)
1181{
1182 struct sk_buff *first = list->next;
1183 struct sk_buff *last = list->prev;
1184
1185 first->prev = prev;
1186 prev->next = first;
1187
1188 last->next = next;
1189 next->prev = last;
1190}
1191
1192/**
1193 * skb_queue_splice - join two skb lists, this is designed for stacks
1194 * @list: the new list to add
1195 * @head: the place to add it in the first list
1196 */
1197static inline void skb_queue_splice(const struct sk_buff_head *list,
1198 struct sk_buff_head *head)
1199{
1200 if (!skb_queue_empty(list)) {
1201 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1202 head->qlen += list->qlen;
67fed459
DM
1203 }
1204}
1205
1206/**
d9619496 1207 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1208 * @list: the new list to add
1209 * @head: the place to add it in the first list
1210 *
1211 * The list at @list is reinitialised
1212 */
1213static inline void skb_queue_splice_init(struct sk_buff_head *list,
1214 struct sk_buff_head *head)
1215{
1216 if (!skb_queue_empty(list)) {
1217 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1218 head->qlen += list->qlen;
67fed459
DM
1219 __skb_queue_head_init(list);
1220 }
1221}
1222
1223/**
1224 * skb_queue_splice_tail - join two skb lists, each list being a queue
1225 * @list: the new list to add
1226 * @head: the place to add it in the first list
1227 */
1228static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1229 struct sk_buff_head *head)
1230{
1231 if (!skb_queue_empty(list)) {
1232 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1233 head->qlen += list->qlen;
67fed459
DM
1234 }
1235}
1236
1237/**
d9619496 1238 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1239 * @list: the new list to add
1240 * @head: the place to add it in the first list
1241 *
1242 * Each of the lists is a queue.
1243 * The list at @list is reinitialised
1244 */
1245static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1246 struct sk_buff_head *head)
1247{
1248 if (!skb_queue_empty(list)) {
1249 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1250 head->qlen += list->qlen;
67fed459
DM
1251 __skb_queue_head_init(list);
1252 }
1253}
1254
1da177e4 1255/**
300ce174 1256 * __skb_queue_after - queue a buffer at the list head
1da177e4 1257 * @list: list to use
300ce174 1258 * @prev: place after this buffer
1da177e4
LT
1259 * @newsk: buffer to queue
1260 *
300ce174 1261 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1262 * and you must therefore hold required locks before calling it.
1263 *
1264 * A buffer cannot be placed on two lists at the same time.
1265 */
300ce174
SH
1266static inline void __skb_queue_after(struct sk_buff_head *list,
1267 struct sk_buff *prev,
1268 struct sk_buff *newsk)
1da177e4 1269{
bf299275 1270 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1271}
1272
7965bd4d
JP
1273void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1274 struct sk_buff_head *list);
7de6c033 1275
f5572855
GR
1276static inline void __skb_queue_before(struct sk_buff_head *list,
1277 struct sk_buff *next,
1278 struct sk_buff *newsk)
1279{
1280 __skb_insert(newsk, next->prev, next, list);
1281}
1282
300ce174
SH
1283/**
1284 * __skb_queue_head - queue a buffer at the list head
1285 * @list: list to use
1286 * @newsk: buffer to queue
1287 *
1288 * Queue a buffer at the start of a list. This function takes no locks
1289 * and you must therefore hold required locks before calling it.
1290 *
1291 * A buffer cannot be placed on two lists at the same time.
1292 */
7965bd4d 1293void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1294static inline void __skb_queue_head(struct sk_buff_head *list,
1295 struct sk_buff *newsk)
1296{
1297 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1298}
1299
1da177e4
LT
1300/**
1301 * __skb_queue_tail - queue a buffer at the list tail
1302 * @list: list to use
1303 * @newsk: buffer to queue
1304 *
1305 * Queue a buffer at the end of a list. This function takes no locks
1306 * and you must therefore hold required locks before calling it.
1307 *
1308 * A buffer cannot be placed on two lists at the same time.
1309 */
7965bd4d 1310void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1311static inline void __skb_queue_tail(struct sk_buff_head *list,
1312 struct sk_buff *newsk)
1313{
f5572855 1314 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1315}
1316
1da177e4
LT
1317/*
1318 * remove sk_buff from list. _Must_ be called atomically, and with
1319 * the list known..
1320 */
7965bd4d 1321void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1322static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1323{
1324 struct sk_buff *next, *prev;
1325
1326 list->qlen--;
1327 next = skb->next;
1328 prev = skb->prev;
1329 skb->next = skb->prev = NULL;
1da177e4
LT
1330 next->prev = prev;
1331 prev->next = next;
1332}
1333
f525c06d
GR
1334/**
1335 * __skb_dequeue - remove from the head of the queue
1336 * @list: list to dequeue from
1337 *
1338 * Remove the head of the list. This function does not take any locks
1339 * so must be used with appropriate locks held only. The head item is
1340 * returned or %NULL if the list is empty.
1341 */
7965bd4d 1342struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1343static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1344{
1345 struct sk_buff *skb = skb_peek(list);
1346 if (skb)
1347 __skb_unlink(skb, list);
1348 return skb;
1349}
1da177e4
LT
1350
1351/**
1352 * __skb_dequeue_tail - remove from the tail of the queue
1353 * @list: list to dequeue from
1354 *
1355 * Remove the tail of the list. This function does not take any locks
1356 * so must be used with appropriate locks held only. The tail item is
1357 * returned or %NULL if the list is empty.
1358 */
7965bd4d 1359struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1360static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1361{
1362 struct sk_buff *skb = skb_peek_tail(list);
1363 if (skb)
1364 __skb_unlink(skb, list);
1365 return skb;
1366}
1367
1368
bdcc0924 1369static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1370{
1371 return skb->data_len;
1372}
1373
1374static inline unsigned int skb_headlen(const struct sk_buff *skb)
1375{
1376 return skb->len - skb->data_len;
1377}
1378
1379static inline int skb_pagelen(const struct sk_buff *skb)
1380{
1381 int i, len = 0;
1382
1383 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1384 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1385 return len + skb_headlen(skb);
1386}
1387
131ea667
IC
1388/**
1389 * __skb_fill_page_desc - initialise a paged fragment in an skb
1390 * @skb: buffer containing fragment to be initialised
1391 * @i: paged fragment index to initialise
1392 * @page: the page to use for this fragment
1393 * @off: the offset to the data with @page
1394 * @size: the length of the data
1395 *
1396 * Initialises the @i'th fragment of @skb to point to &size bytes at
1397 * offset @off within @page.
1398 *
1399 * Does not take any additional reference on the fragment.
1400 */
1401static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1402 struct page *page, int off, int size)
1da177e4
LT
1403{
1404 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1405
c48a11c7
MG
1406 /*
1407 * Propagate page->pfmemalloc to the skb if we can. The problem is
1408 * that not all callers have unique ownership of the page. If
1409 * pfmemalloc is set, we check the mapping as a mapping implies
1410 * page->index is set (index and pfmemalloc share space).
1411 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1412 * do not lose pfmemalloc information as the pages would not be
1413 * allocated using __GFP_MEMALLOC.
1414 */
a8605c60 1415 frag->page.p = page;
1da177e4 1416 frag->page_offset = off;
9e903e08 1417 skb_frag_size_set(frag, size);
cca7af38
PE
1418
1419 page = compound_head(page);
1420 if (page->pfmemalloc && !page->mapping)
1421 skb->pfmemalloc = true;
131ea667
IC
1422}
1423
1424/**
1425 * skb_fill_page_desc - initialise a paged fragment in an skb
1426 * @skb: buffer containing fragment to be initialised
1427 * @i: paged fragment index to initialise
1428 * @page: the page to use for this fragment
1429 * @off: the offset to the data with @page
1430 * @size: the length of the data
1431 *
1432 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1433 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1434 * addition updates @skb such that @i is the last fragment.
1435 *
1436 * Does not take any additional reference on the fragment.
1437 */
1438static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1439 struct page *page, int off, int size)
1440{
1441 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1442 skb_shinfo(skb)->nr_frags = i + 1;
1443}
1444
7965bd4d
JP
1445void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1446 int size, unsigned int truesize);
654bed16 1447
f8e617e1
JW
1448void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1449 unsigned int truesize);
1450
1da177e4 1451#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1452#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1453#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1454
27a884dc
ACM
1455#ifdef NET_SKBUFF_DATA_USES_OFFSET
1456static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1457{
1458 return skb->head + skb->tail;
1459}
1460
1461static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1462{
1463 skb->tail = skb->data - skb->head;
1464}
1465
1466static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1467{
1468 skb_reset_tail_pointer(skb);
1469 skb->tail += offset;
1470}
7cc46190 1471
27a884dc
ACM
1472#else /* NET_SKBUFF_DATA_USES_OFFSET */
1473static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1474{
1475 return skb->tail;
1476}
1477
1478static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1479{
1480 skb->tail = skb->data;
1481}
1482
1483static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1484{
1485 skb->tail = skb->data + offset;
1486}
4305b541 1487
27a884dc
ACM
1488#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1489
1da177e4
LT
1490/*
1491 * Add data to an sk_buff
1492 */
0c7ddf36 1493unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1494unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1495static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1496{
27a884dc 1497 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1498 SKB_LINEAR_ASSERT(skb);
1499 skb->tail += len;
1500 skb->len += len;
1501 return tmp;
1502}
1503
7965bd4d 1504unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1505static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1506{
1507 skb->data -= len;
1508 skb->len += len;
1509 return skb->data;
1510}
1511
7965bd4d 1512unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1513static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1514{
1515 skb->len -= len;
1516 BUG_ON(skb->len < skb->data_len);
1517 return skb->data += len;
1518}
1519
47d29646
DM
1520static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1521{
1522 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1523}
1524
7965bd4d 1525unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1526
1527static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1528{
1529 if (len > skb_headlen(skb) &&
987c402a 1530 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1531 return NULL;
1532 skb->len -= len;
1533 return skb->data += len;
1534}
1535
1536static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1537{
1538 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1539}
1540
1541static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1542{
1543 if (likely(len <= skb_headlen(skb)))
1544 return 1;
1545 if (unlikely(len > skb->len))
1546 return 0;
987c402a 1547 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1548}
1549
1550/**
1551 * skb_headroom - bytes at buffer head
1552 * @skb: buffer to check
1553 *
1554 * Return the number of bytes of free space at the head of an &sk_buff.
1555 */
c2636b4d 1556static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1557{
1558 return skb->data - skb->head;
1559}
1560
1561/**
1562 * skb_tailroom - bytes at buffer end
1563 * @skb: buffer to check
1564 *
1565 * Return the number of bytes of free space at the tail of an sk_buff
1566 */
1567static inline int skb_tailroom(const struct sk_buff *skb)
1568{
4305b541 1569 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1570}
1571
a21d4572
ED
1572/**
1573 * skb_availroom - bytes at buffer end
1574 * @skb: buffer to check
1575 *
1576 * Return the number of bytes of free space at the tail of an sk_buff
1577 * allocated by sk_stream_alloc()
1578 */
1579static inline int skb_availroom(const struct sk_buff *skb)
1580{
16fad69c
ED
1581 if (skb_is_nonlinear(skb))
1582 return 0;
1583
1584 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1585}
1586
1da177e4
LT
1587/**
1588 * skb_reserve - adjust headroom
1589 * @skb: buffer to alter
1590 * @len: bytes to move
1591 *
1592 * Increase the headroom of an empty &sk_buff by reducing the tail
1593 * room. This is only allowed for an empty buffer.
1594 */
8243126c 1595static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1596{
1597 skb->data += len;
1598 skb->tail += len;
1599}
1600
6a674e9c
JG
1601static inline void skb_reset_inner_headers(struct sk_buff *skb)
1602{
aefbd2b3 1603 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1604 skb->inner_network_header = skb->network_header;
1605 skb->inner_transport_header = skb->transport_header;
1606}
1607
0b5c9db1
JP
1608static inline void skb_reset_mac_len(struct sk_buff *skb)
1609{
1610 skb->mac_len = skb->network_header - skb->mac_header;
1611}
1612
6a674e9c
JG
1613static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1614 *skb)
1615{
1616 return skb->head + skb->inner_transport_header;
1617}
1618
1619static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1620{
1621 skb->inner_transport_header = skb->data - skb->head;
1622}
1623
1624static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1625 const int offset)
1626{
1627 skb_reset_inner_transport_header(skb);
1628 skb->inner_transport_header += offset;
1629}
1630
1631static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1632{
1633 return skb->head + skb->inner_network_header;
1634}
1635
1636static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1637{
1638 skb->inner_network_header = skb->data - skb->head;
1639}
1640
1641static inline void skb_set_inner_network_header(struct sk_buff *skb,
1642 const int offset)
1643{
1644 skb_reset_inner_network_header(skb);
1645 skb->inner_network_header += offset;
1646}
1647
aefbd2b3
PS
1648static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1649{
1650 return skb->head + skb->inner_mac_header;
1651}
1652
1653static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1654{
1655 skb->inner_mac_header = skb->data - skb->head;
1656}
1657
1658static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1659 const int offset)
1660{
1661 skb_reset_inner_mac_header(skb);
1662 skb->inner_mac_header += offset;
1663}
fda55eca
ED
1664static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1665{
35d04610 1666 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1667}
1668
9c70220b
ACM
1669static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1670{
2e07fa9c 1671 return skb->head + skb->transport_header;
9c70220b
ACM
1672}
1673
badff6d0
ACM
1674static inline void skb_reset_transport_header(struct sk_buff *skb)
1675{
2e07fa9c 1676 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1677}
1678
967b05f6
ACM
1679static inline void skb_set_transport_header(struct sk_buff *skb,
1680 const int offset)
1681{
2e07fa9c
ACM
1682 skb_reset_transport_header(skb);
1683 skb->transport_header += offset;
ea2ae17d
ACM
1684}
1685
d56f90a7
ACM
1686static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1687{
2e07fa9c 1688 return skb->head + skb->network_header;
d56f90a7
ACM
1689}
1690
c1d2bbe1
ACM
1691static inline void skb_reset_network_header(struct sk_buff *skb)
1692{
2e07fa9c 1693 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1694}
1695
c14d2450
ACM
1696static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1697{
2e07fa9c
ACM
1698 skb_reset_network_header(skb);
1699 skb->network_header += offset;
c14d2450
ACM
1700}
1701
2e07fa9c 1702static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1703{
2e07fa9c 1704 return skb->head + skb->mac_header;
bbe735e4
ACM
1705}
1706
2e07fa9c 1707static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1708{
35d04610 1709 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1710}
1711
1712static inline void skb_reset_mac_header(struct sk_buff *skb)
1713{
1714 skb->mac_header = skb->data - skb->head;
1715}
1716
1717static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1718{
1719 skb_reset_mac_header(skb);
1720 skb->mac_header += offset;
1721}
1722
0e3da5bb
TT
1723static inline void skb_pop_mac_header(struct sk_buff *skb)
1724{
1725 skb->mac_header = skb->network_header;
1726}
1727
fbbdb8f0
YX
1728static inline void skb_probe_transport_header(struct sk_buff *skb,
1729 const int offset_hint)
1730{
1731 struct flow_keys keys;
1732
1733 if (skb_transport_header_was_set(skb))
1734 return;
1735 else if (skb_flow_dissect(skb, &keys))
1736 skb_set_transport_header(skb, keys.thoff);
1737 else
1738 skb_set_transport_header(skb, offset_hint);
1739}
1740
03606895
ED
1741static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1742{
1743 if (skb_mac_header_was_set(skb)) {
1744 const unsigned char *old_mac = skb_mac_header(skb);
1745
1746 skb_set_mac_header(skb, -skb->mac_len);
1747 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1748 }
1749}
1750
04fb451e
MM
1751static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1752{
1753 return skb->csum_start - skb_headroom(skb);
1754}
1755
2e07fa9c
ACM
1756static inline int skb_transport_offset(const struct sk_buff *skb)
1757{
1758 return skb_transport_header(skb) - skb->data;
1759}
1760
1761static inline u32 skb_network_header_len(const struct sk_buff *skb)
1762{
1763 return skb->transport_header - skb->network_header;
1764}
1765
6a674e9c
JG
1766static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1767{
1768 return skb->inner_transport_header - skb->inner_network_header;
1769}
1770
2e07fa9c
ACM
1771static inline int skb_network_offset(const struct sk_buff *skb)
1772{
1773 return skb_network_header(skb) - skb->data;
1774}
48d49d0c 1775
6a674e9c
JG
1776static inline int skb_inner_network_offset(const struct sk_buff *skb)
1777{
1778 return skb_inner_network_header(skb) - skb->data;
1779}
1780
f9599ce1
CG
1781static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1782{
1783 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1784}
1785
1da177e4
LT
1786/*
1787 * CPUs often take a performance hit when accessing unaligned memory
1788 * locations. The actual performance hit varies, it can be small if the
1789 * hardware handles it or large if we have to take an exception and fix it
1790 * in software.
1791 *
1792 * Since an ethernet header is 14 bytes network drivers often end up with
1793 * the IP header at an unaligned offset. The IP header can be aligned by
1794 * shifting the start of the packet by 2 bytes. Drivers should do this
1795 * with:
1796 *
8660c124 1797 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1798 *
1799 * The downside to this alignment of the IP header is that the DMA is now
1800 * unaligned. On some architectures the cost of an unaligned DMA is high
1801 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1802 *
1da177e4
LT
1803 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1804 * to be overridden.
1805 */
1806#ifndef NET_IP_ALIGN
1807#define NET_IP_ALIGN 2
1808#endif
1809
025be81e
AB
1810/*
1811 * The networking layer reserves some headroom in skb data (via
1812 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1813 * the header has to grow. In the default case, if the header has to grow
d6301d3d 1814 * 32 bytes or less we avoid the reallocation.
025be81e
AB
1815 *
1816 * Unfortunately this headroom changes the DMA alignment of the resulting
1817 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1818 * on some architectures. An architecture can override this value,
1819 * perhaps setting it to a cacheline in size (since that will maintain
1820 * cacheline alignment of the DMA). It must be a power of 2.
1821 *
d6301d3d 1822 * Various parts of the networking layer expect at least 32 bytes of
025be81e 1823 * headroom, you should not reduce this.
5933dd2f
ED
1824 *
1825 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1826 * to reduce average number of cache lines per packet.
1827 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 1828 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
1829 */
1830#ifndef NET_SKB_PAD
5933dd2f 1831#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
1832#endif
1833
7965bd4d 1834int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1835
1836static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1837{
c4264f27 1838 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
1839 WARN_ON(1);
1840 return;
1841 }
27a884dc
ACM
1842 skb->len = len;
1843 skb_set_tail_pointer(skb, len);
1da177e4
LT
1844}
1845
7965bd4d 1846void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1847
1848static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1849{
3cc0e873
HX
1850 if (skb->data_len)
1851 return ___pskb_trim(skb, len);
1852 __skb_trim(skb, len);
1853 return 0;
1da177e4
LT
1854}
1855
1856static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1857{
1858 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1859}
1860
e9fa4f7b
HX
1861/**
1862 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1863 * @skb: buffer to alter
1864 * @len: new length
1865 *
1866 * This is identical to pskb_trim except that the caller knows that
1867 * the skb is not cloned so we should never get an error due to out-
1868 * of-memory.
1869 */
1870static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1871{
1872 int err = pskb_trim(skb, len);
1873 BUG_ON(err);
1874}
1875
1da177e4
LT
1876/**
1877 * skb_orphan - orphan a buffer
1878 * @skb: buffer to orphan
1879 *
1880 * If a buffer currently has an owner then we call the owner's
1881 * destructor function and make the @skb unowned. The buffer continues
1882 * to exist but is no longer charged to its former owner.
1883 */
1884static inline void skb_orphan(struct sk_buff *skb)
1885{
c34a7612 1886 if (skb->destructor) {
1da177e4 1887 skb->destructor(skb);
c34a7612
ED
1888 skb->destructor = NULL;
1889 skb->sk = NULL;
376c7311
ED
1890 } else {
1891 BUG_ON(skb->sk);
c34a7612 1892 }
1da177e4
LT
1893}
1894
a353e0ce
MT
1895/**
1896 * skb_orphan_frags - orphan the frags contained in a buffer
1897 * @skb: buffer to orphan frags from
1898 * @gfp_mask: allocation mask for replacement pages
1899 *
1900 * For each frag in the SKB which needs a destructor (i.e. has an
1901 * owner) create a copy of that frag and release the original
1902 * page by calling the destructor.
1903 */
1904static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1905{
1906 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1907 return 0;
1908 return skb_copy_ubufs(skb, gfp_mask);
1909}
1910
1da177e4
LT
1911/**
1912 * __skb_queue_purge - empty a list
1913 * @list: list to empty
1914 *
1915 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1916 * the list and one reference dropped. This function does not take the
1917 * list lock and the caller must hold the relevant locks to use it.
1918 */
7965bd4d 1919void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
1920static inline void __skb_queue_purge(struct sk_buff_head *list)
1921{
1922 struct sk_buff *skb;
1923 while ((skb = __skb_dequeue(list)) != NULL)
1924 kfree_skb(skb);
1925}
1926
e5e67305
AD
1927#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1928#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1929#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1930
7965bd4d 1931void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 1932
7965bd4d
JP
1933struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1934 gfp_t gfp_mask);
8af27456
CH
1935
1936/**
1937 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1938 * @dev: network device to receive on
1939 * @length: length to allocate
1940 *
1941 * Allocate a new &sk_buff and assign it a usage count of one. The
1942 * buffer has unspecified headroom built in. Users should allocate
1943 * the headroom they think they need without accounting for the
1944 * built in space. The built in space is used for optimisations.
1945 *
1946 * %NULL is returned if there is no free memory. Although this function
1947 * allocates memory it can be called from an interrupt.
1948 */
1949static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 1950 unsigned int length)
8af27456
CH
1951{
1952 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1953}
1954
6f532612
ED
1955/* legacy helper around __netdev_alloc_skb() */
1956static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1957 gfp_t gfp_mask)
1958{
1959 return __netdev_alloc_skb(NULL, length, gfp_mask);
1960}
1961
1962/* legacy helper around netdev_alloc_skb() */
1963static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1964{
1965 return netdev_alloc_skb(NULL, length);
1966}
1967
1968
4915a0de
ED
1969static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1970 unsigned int length, gfp_t gfp)
61321bbd 1971{
4915a0de 1972 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
1973
1974 if (NET_IP_ALIGN && skb)
1975 skb_reserve(skb, NET_IP_ALIGN);
1976 return skb;
1977}
1978
4915a0de
ED
1979static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1980 unsigned int length)
1981{
1982 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1983}
1984
bc6fc9fa
FF
1985/**
1986 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
1987 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1988 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1989 * @order: size of the allocation
1990 *
1991 * Allocate a new page.
1992 *
1993 * %NULL is returned if there is no free memory.
1994*/
1995static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1996 struct sk_buff *skb,
1997 unsigned int order)
1998{
1999 struct page *page;
2000
2001 gfp_mask |= __GFP_COLD;
2002
2003 if (!(gfp_mask & __GFP_NOMEMALLOC))
2004 gfp_mask |= __GFP_MEMALLOC;
2005
2006 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2007 if (skb && page && page->pfmemalloc)
2008 skb->pfmemalloc = true;
2009
2010 return page;
2011}
2012
2013/**
2014 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2015 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2016 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2017 *
2018 * Allocate a new page.
2019 *
2020 * %NULL is returned if there is no free memory.
2021 */
2022static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2023 struct sk_buff *skb)
2024{
2025 return __skb_alloc_pages(gfp_mask, skb, 0);
2026}
2027
2028/**
2029 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2030 * @page: The page that was allocated from skb_alloc_page
2031 * @skb: The skb that may need pfmemalloc set
2032 */
2033static inline void skb_propagate_pfmemalloc(struct page *page,
2034 struct sk_buff *skb)
2035{
2036 if (page && page->pfmemalloc)
2037 skb->pfmemalloc = true;
2038}
2039
131ea667
IC
2040/**
2041 * skb_frag_page - retrieve the page refered to by a paged fragment
2042 * @frag: the paged fragment
2043 *
2044 * Returns the &struct page associated with @frag.
2045 */
2046static inline struct page *skb_frag_page(const skb_frag_t *frag)
2047{
a8605c60 2048 return frag->page.p;
131ea667
IC
2049}
2050
2051/**
2052 * __skb_frag_ref - take an addition reference on a paged fragment.
2053 * @frag: the paged fragment
2054 *
2055 * Takes an additional reference on the paged fragment @frag.
2056 */
2057static inline void __skb_frag_ref(skb_frag_t *frag)
2058{
2059 get_page(skb_frag_page(frag));
2060}
2061
2062/**
2063 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2064 * @skb: the buffer
2065 * @f: the fragment offset.
2066 *
2067 * Takes an additional reference on the @f'th paged fragment of @skb.
2068 */
2069static inline void skb_frag_ref(struct sk_buff *skb, int f)
2070{
2071 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2072}
2073
2074/**
2075 * __skb_frag_unref - release a reference on a paged fragment.
2076 * @frag: the paged fragment
2077 *
2078 * Releases a reference on the paged fragment @frag.
2079 */
2080static inline void __skb_frag_unref(skb_frag_t *frag)
2081{
2082 put_page(skb_frag_page(frag));
2083}
2084
2085/**
2086 * skb_frag_unref - release a reference on a paged fragment of an skb.
2087 * @skb: the buffer
2088 * @f: the fragment offset
2089 *
2090 * Releases a reference on the @f'th paged fragment of @skb.
2091 */
2092static inline void skb_frag_unref(struct sk_buff *skb, int f)
2093{
2094 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2095}
2096
2097/**
2098 * skb_frag_address - gets the address of the data contained in a paged fragment
2099 * @frag: the paged fragment buffer
2100 *
2101 * Returns the address of the data within @frag. The page must already
2102 * be mapped.
2103 */
2104static inline void *skb_frag_address(const skb_frag_t *frag)
2105{
2106 return page_address(skb_frag_page(frag)) + frag->page_offset;
2107}
2108
2109/**
2110 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2111 * @frag: the paged fragment buffer
2112 *
2113 * Returns the address of the data within @frag. Checks that the page
2114 * is mapped and returns %NULL otherwise.
2115 */
2116static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2117{
2118 void *ptr = page_address(skb_frag_page(frag));
2119 if (unlikely(!ptr))
2120 return NULL;
2121
2122 return ptr + frag->page_offset;
2123}
2124
2125/**
2126 * __skb_frag_set_page - sets the page contained in a paged fragment
2127 * @frag: the paged fragment
2128 * @page: the page to set
2129 *
2130 * Sets the fragment @frag to contain @page.
2131 */
2132static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2133{
a8605c60 2134 frag->page.p = page;
131ea667
IC
2135}
2136
2137/**
2138 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2139 * @skb: the buffer
2140 * @f: the fragment offset
2141 * @page: the page to set
2142 *
2143 * Sets the @f'th fragment of @skb to contain @page.
2144 */
2145static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2146 struct page *page)
2147{
2148 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2149}
2150
400dfd3a
ED
2151bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2152
131ea667
IC
2153/**
2154 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2155 * @dev: the device to map the fragment to
131ea667
IC
2156 * @frag: the paged fragment to map
2157 * @offset: the offset within the fragment (starting at the
2158 * fragment's own offset)
2159 * @size: the number of bytes to map
f83347df 2160 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2161 *
2162 * Maps the page associated with @frag to @device.
2163 */
2164static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2165 const skb_frag_t *frag,
2166 size_t offset, size_t size,
2167 enum dma_data_direction dir)
2168{
2169 return dma_map_page(dev, skb_frag_page(frag),
2170 frag->page_offset + offset, size, dir);
2171}
2172
117632e6
ED
2173static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2174 gfp_t gfp_mask)
2175{
2176 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2177}
2178
334a8132
PM
2179/**
2180 * skb_clone_writable - is the header of a clone writable
2181 * @skb: buffer to check
2182 * @len: length up to which to write
2183 *
2184 * Returns true if modifying the header part of the cloned buffer
2185 * does not requires the data to be copied.
2186 */
05bdd2f1 2187static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2188{
2189 return !skb_header_cloned(skb) &&
2190 skb_headroom(skb) + len <= skb->hdr_len;
2191}
2192
d9cc2048
HX
2193static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2194 int cloned)
2195{
2196 int delta = 0;
2197
d9cc2048
HX
2198 if (headroom > skb_headroom(skb))
2199 delta = headroom - skb_headroom(skb);
2200
2201 if (delta || cloned)
2202 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2203 GFP_ATOMIC);
2204 return 0;
2205}
2206
1da177e4
LT
2207/**
2208 * skb_cow - copy header of skb when it is required
2209 * @skb: buffer to cow
2210 * @headroom: needed headroom
2211 *
2212 * If the skb passed lacks sufficient headroom or its data part
2213 * is shared, data is reallocated. If reallocation fails, an error
2214 * is returned and original skb is not changed.
2215 *
2216 * The result is skb with writable area skb->head...skb->tail
2217 * and at least @headroom of space at head.
2218 */
2219static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2220{
d9cc2048
HX
2221 return __skb_cow(skb, headroom, skb_cloned(skb));
2222}
1da177e4 2223
d9cc2048
HX
2224/**
2225 * skb_cow_head - skb_cow but only making the head writable
2226 * @skb: buffer to cow
2227 * @headroom: needed headroom
2228 *
2229 * This function is identical to skb_cow except that we replace the
2230 * skb_cloned check by skb_header_cloned. It should be used when
2231 * you only need to push on some header and do not need to modify
2232 * the data.
2233 */
2234static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2235{
2236 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2237}
2238
2239/**
2240 * skb_padto - pad an skbuff up to a minimal size
2241 * @skb: buffer to pad
2242 * @len: minimal length
2243 *
2244 * Pads up a buffer to ensure the trailing bytes exist and are
2245 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2246 * is untouched. Otherwise it is extended. Returns zero on
2247 * success. The skb is freed on error.
1da177e4
LT
2248 */
2249
5b057c6b 2250static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2251{
2252 unsigned int size = skb->len;
2253 if (likely(size >= len))
5b057c6b 2254 return 0;
987c402a 2255 return skb_pad(skb, len - size);
1da177e4
LT
2256}
2257
2258static inline int skb_add_data(struct sk_buff *skb,
2259 char __user *from, int copy)
2260{
2261 const int off = skb->len;
2262
2263 if (skb->ip_summed == CHECKSUM_NONE) {
2264 int err = 0;
5084205f 2265 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2266 copy, 0, &err);
2267 if (!err) {
2268 skb->csum = csum_block_add(skb->csum, csum, off);
2269 return 0;
2270 }
2271 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2272 return 0;
2273
2274 __skb_trim(skb, off);
2275 return -EFAULT;
2276}
2277
38ba0a65
ED
2278static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2279 const struct page *page, int off)
1da177e4
LT
2280{
2281 if (i) {
9e903e08 2282 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2283
ea2ab693 2284 return page == skb_frag_page(frag) &&
9e903e08 2285 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2286 }
38ba0a65 2287 return false;
1da177e4
LT
2288}
2289
364c6bad
HX
2290static inline int __skb_linearize(struct sk_buff *skb)
2291{
2292 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2293}
2294
1da177e4
LT
2295/**
2296 * skb_linearize - convert paged skb to linear one
2297 * @skb: buffer to linarize
1da177e4
LT
2298 *
2299 * If there is no free memory -ENOMEM is returned, otherwise zero
2300 * is returned and the old skb data released.
2301 */
364c6bad
HX
2302static inline int skb_linearize(struct sk_buff *skb)
2303{
2304 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2305}
2306
cef401de
ED
2307/**
2308 * skb_has_shared_frag - can any frag be overwritten
2309 * @skb: buffer to test
2310 *
2311 * Return true if the skb has at least one frag that might be modified
2312 * by an external entity (as in vmsplice()/sendfile())
2313 */
2314static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2315{
c9af6db4
PS
2316 return skb_is_nonlinear(skb) &&
2317 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2318}
2319
364c6bad
HX
2320/**
2321 * skb_linearize_cow - make sure skb is linear and writable
2322 * @skb: buffer to process
2323 *
2324 * If there is no free memory -ENOMEM is returned, otherwise zero
2325 * is returned and the old skb data released.
2326 */
2327static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2328{
364c6bad
HX
2329 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2330 __skb_linearize(skb) : 0;
1da177e4
LT
2331}
2332
2333/**
2334 * skb_postpull_rcsum - update checksum for received skb after pull
2335 * @skb: buffer to update
2336 * @start: start of data before pull
2337 * @len: length of data pulled
2338 *
2339 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2340 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2341 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2342 */
2343
2344static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2345 const void *start, unsigned int len)
1da177e4 2346{
84fa7933 2347 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2348 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2349}
2350
cbb042f9
HX
2351unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2352
7ce5a27f
DM
2353/**
2354 * pskb_trim_rcsum - trim received skb and update checksum
2355 * @skb: buffer to trim
2356 * @len: new length
2357 *
2358 * This is exactly the same as pskb_trim except that it ensures the
2359 * checksum of received packets are still valid after the operation.
2360 */
2361
2362static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2363{
2364 if (likely(len >= skb->len))
2365 return 0;
2366 if (skb->ip_summed == CHECKSUM_COMPLETE)
2367 skb->ip_summed = CHECKSUM_NONE;
2368 return __pskb_trim(skb, len);
2369}
2370
1da177e4
LT
2371#define skb_queue_walk(queue, skb) \
2372 for (skb = (queue)->next; \
a1e4891f 2373 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2374 skb = skb->next)
2375
46f8914e
JC
2376#define skb_queue_walk_safe(queue, skb, tmp) \
2377 for (skb = (queue)->next, tmp = skb->next; \
2378 skb != (struct sk_buff *)(queue); \
2379 skb = tmp, tmp = skb->next)
2380
1164f52a 2381#define skb_queue_walk_from(queue, skb) \
a1e4891f 2382 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2383 skb = skb->next)
2384
2385#define skb_queue_walk_from_safe(queue, skb, tmp) \
2386 for (tmp = skb->next; \
2387 skb != (struct sk_buff *)(queue); \
2388 skb = tmp, tmp = skb->next)
2389
300ce174
SH
2390#define skb_queue_reverse_walk(queue, skb) \
2391 for (skb = (queue)->prev; \
a1e4891f 2392 skb != (struct sk_buff *)(queue); \
300ce174
SH
2393 skb = skb->prev)
2394
686a2955
DM
2395#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2396 for (skb = (queue)->prev, tmp = skb->prev; \
2397 skb != (struct sk_buff *)(queue); \
2398 skb = tmp, tmp = skb->prev)
2399
2400#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2401 for (tmp = skb->prev; \
2402 skb != (struct sk_buff *)(queue); \
2403 skb = tmp, tmp = skb->prev)
1da177e4 2404
21dc3301 2405static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2406{
2407 return skb_shinfo(skb)->frag_list != NULL;
2408}
2409
2410static inline void skb_frag_list_init(struct sk_buff *skb)
2411{
2412 skb_shinfo(skb)->frag_list = NULL;
2413}
2414
2415static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2416{
2417 frag->next = skb_shinfo(skb)->frag_list;
2418 skb_shinfo(skb)->frag_list = frag;
2419}
2420
2421#define skb_walk_frags(skb, iter) \
2422 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2423
7965bd4d
JP
2424struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2425 int *peeked, int *off, int *err);
2426struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2427 int *err);
2428unsigned int datagram_poll(struct file *file, struct socket *sock,
2429 struct poll_table_struct *wait);
2430int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2431 struct iovec *to, int size);
2432int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2433 struct iovec *iov);
2434int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2435 const struct iovec *from, int from_offset,
2436 int len);
2437int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2438 int offset, size_t count);
2439int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2440 const struct iovec *to, int to_offset,
2441 int size);
2442void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2443void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2444int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2445int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2446int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2447__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2448 int len, __wsum csum);
2449int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2450 struct pipe_inode_info *pipe, unsigned int len,
2451 unsigned int flags);
2452void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8
TG
2453unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2454void skb_zerocopy(struct sk_buff *to, const struct sk_buff *from,
2455 int len, int hlen);
7965bd4d
JP
2456void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2457int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2458void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2459unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2460struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
20380731 2461
2817a336
DB
2462struct skb_checksum_ops {
2463 __wsum (*update)(const void *mem, int len, __wsum wsum);
2464 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2465};
2466
2467__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2468 __wsum csum, const struct skb_checksum_ops *ops);
2469__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2470 __wsum csum);
2471
1da177e4
LT
2472static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2473 int len, void *buffer)
2474{
2475 int hlen = skb_headlen(skb);
2476
55820ee2 2477 if (hlen - offset >= len)
1da177e4
LT
2478 return skb->data + offset;
2479
2480 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2481 return NULL;
2482
2483 return buffer;
2484}
2485
4262e5cc
DB
2486/**
2487 * skb_needs_linearize - check if we need to linearize a given skb
2488 * depending on the given device features.
2489 * @skb: socket buffer to check
2490 * @features: net device features
2491 *
2492 * Returns true if either:
2493 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2494 * 2. skb is fragmented and the device does not support SG.
2495 */
2496static inline bool skb_needs_linearize(struct sk_buff *skb,
2497 netdev_features_t features)
2498{
2499 return skb_is_nonlinear(skb) &&
2500 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2501 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2502}
2503
d626f62b
ACM
2504static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2505 void *to,
2506 const unsigned int len)
2507{
2508 memcpy(to, skb->data, len);
2509}
2510
2511static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2512 const int offset, void *to,
2513 const unsigned int len)
2514{
2515 memcpy(to, skb->data + offset, len);
2516}
2517
27d7ff46
ACM
2518static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2519 const void *from,
2520 const unsigned int len)
2521{
2522 memcpy(skb->data, from, len);
2523}
2524
2525static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2526 const int offset,
2527 const void *from,
2528 const unsigned int len)
2529{
2530 memcpy(skb->data + offset, from, len);
2531}
2532
7965bd4d 2533void skb_init(void);
1da177e4 2534
ac45f602
PO
2535static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2536{
2537 return skb->tstamp;
2538}
2539
a61bbcf2
PM
2540/**
2541 * skb_get_timestamp - get timestamp from a skb
2542 * @skb: skb to get stamp from
2543 * @stamp: pointer to struct timeval to store stamp in
2544 *
2545 * Timestamps are stored in the skb as offsets to a base timestamp.
2546 * This function converts the offset back to a struct timeval and stores
2547 * it in stamp.
2548 */
ac45f602
PO
2549static inline void skb_get_timestamp(const struct sk_buff *skb,
2550 struct timeval *stamp)
a61bbcf2 2551{
b7aa0bf7 2552 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2553}
2554
ac45f602
PO
2555static inline void skb_get_timestampns(const struct sk_buff *skb,
2556 struct timespec *stamp)
2557{
2558 *stamp = ktime_to_timespec(skb->tstamp);
2559}
2560
b7aa0bf7 2561static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2562{
b7aa0bf7 2563 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2564}
2565
164891aa
SH
2566static inline ktime_t net_timedelta(ktime_t t)
2567{
2568 return ktime_sub(ktime_get_real(), t);
2569}
2570
b9ce204f
IJ
2571static inline ktime_t net_invalid_timestamp(void)
2572{
2573 return ktime_set(0, 0);
2574}
a61bbcf2 2575
7965bd4d 2576void skb_timestamping_init(void);
c1f19b51
RC
2577
2578#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2579
7965bd4d
JP
2580void skb_clone_tx_timestamp(struct sk_buff *skb);
2581bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2582
2583#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2584
2585static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2586{
2587}
2588
2589static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2590{
2591 return false;
2592}
2593
2594#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2595
2596/**
2597 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2598 *
da92b194
RC
2599 * PHY drivers may accept clones of transmitted packets for
2600 * timestamping via their phy_driver.txtstamp method. These drivers
2601 * must call this function to return the skb back to the stack, with
2602 * or without a timestamp.
2603 *
c1f19b51 2604 * @skb: clone of the the original outgoing packet
da92b194 2605 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2606 *
2607 */
2608void skb_complete_tx_timestamp(struct sk_buff *skb,
2609 struct skb_shared_hwtstamps *hwtstamps);
2610
ac45f602
PO
2611/**
2612 * skb_tstamp_tx - queue clone of skb with send time stamps
2613 * @orig_skb: the original outgoing packet
2614 * @hwtstamps: hardware time stamps, may be NULL if not available
2615 *
2616 * If the skb has a socket associated, then this function clones the
2617 * skb (thus sharing the actual data and optional structures), stores
2618 * the optional hardware time stamping information (if non NULL) or
2619 * generates a software time stamp (otherwise), then queues the clone
2620 * to the error queue of the socket. Errors are silently ignored.
2621 */
7965bd4d
JP
2622void skb_tstamp_tx(struct sk_buff *orig_skb,
2623 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2624
4507a715
RC
2625static inline void sw_tx_timestamp(struct sk_buff *skb)
2626{
2244d07b
OH
2627 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2628 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2629 skb_tstamp_tx(skb, NULL);
2630}
2631
2632/**
2633 * skb_tx_timestamp() - Driver hook for transmit timestamping
2634 *
2635 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2636 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2637 *
73409f3b
DM
2638 * Specifically, one should make absolutely sure that this function is
2639 * called before TX completion of this packet can trigger. Otherwise
2640 * the packet could potentially already be freed.
2641 *
4507a715
RC
2642 * @skb: A socket buffer.
2643 */
2644static inline void skb_tx_timestamp(struct sk_buff *skb)
2645{
c1f19b51 2646 skb_clone_tx_timestamp(skb);
4507a715
RC
2647 sw_tx_timestamp(skb);
2648}
2649
6e3e939f
JB
2650/**
2651 * skb_complete_wifi_ack - deliver skb with wifi status
2652 *
2653 * @skb: the original outgoing packet
2654 * @acked: ack status
2655 *
2656 */
2657void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2658
7965bd4d
JP
2659__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2660__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2661
60476372
HX
2662static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2663{
2664 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2665}
2666
fb286bb2
HX
2667/**
2668 * skb_checksum_complete - Calculate checksum of an entire packet
2669 * @skb: packet to process
2670 *
2671 * This function calculates the checksum over the entire packet plus
2672 * the value of skb->csum. The latter can be used to supply the
2673 * checksum of a pseudo header as used by TCP/UDP. It returns the
2674 * checksum.
2675 *
2676 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2677 * this function can be used to verify that checksum on received
2678 * packets. In that case the function should return zero if the
2679 * checksum is correct. In particular, this function will return zero
2680 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2681 * hardware has already verified the correctness of the checksum.
2682 */
4381ca3c 2683static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2684{
60476372
HX
2685 return skb_csum_unnecessary(skb) ?
2686 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2687}
2688
5f79e0f9 2689#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 2690void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
2691static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2692{
2693 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 2694 nf_conntrack_destroy(nfct);
1da177e4
LT
2695}
2696static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2697{
2698 if (nfct)
2699 atomic_inc(&nfct->use);
2700}
2fc72c7b 2701#endif
1da177e4
LT
2702#ifdef CONFIG_BRIDGE_NETFILTER
2703static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2704{
2705 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2706 kfree(nf_bridge);
2707}
2708static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2709{
2710 if (nf_bridge)
2711 atomic_inc(&nf_bridge->use);
2712}
2713#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
2714static inline void nf_reset(struct sk_buff *skb)
2715{
5f79e0f9 2716#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
2717 nf_conntrack_put(skb->nfct);
2718 skb->nfct = NULL;
2fc72c7b 2719#endif
a193a4ab
PM
2720#ifdef CONFIG_BRIDGE_NETFILTER
2721 nf_bridge_put(skb->nf_bridge);
2722 skb->nf_bridge = NULL;
2723#endif
2724}
2725
124dff01
PM
2726static inline void nf_reset_trace(struct sk_buff *skb)
2727{
478b360a 2728#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
2729 skb->nf_trace = 0;
2730#endif
a193a4ab
PM
2731}
2732
edda553c
YK
2733/* Note: This doesn't put any conntrack and bridge info in dst. */
2734static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2735{
5f79e0f9 2736#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
2737 dst->nfct = src->nfct;
2738 nf_conntrack_get(src->nfct);
2739 dst->nfctinfo = src->nfctinfo;
2fc72c7b 2740#endif
edda553c
YK
2741#ifdef CONFIG_BRIDGE_NETFILTER
2742 dst->nf_bridge = src->nf_bridge;
2743 nf_bridge_get(src->nf_bridge);
2744#endif
478b360a
FW
2745#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2746 dst->nf_trace = src->nf_trace;
2747#endif
edda553c
YK
2748}
2749
e7ac05f3
YK
2750static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2751{
e7ac05f3 2752#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 2753 nf_conntrack_put(dst->nfct);
2fc72c7b 2754#endif
e7ac05f3
YK
2755#ifdef CONFIG_BRIDGE_NETFILTER
2756 nf_bridge_put(dst->nf_bridge);
2757#endif
2758 __nf_copy(dst, src);
2759}
2760
984bc16c
JM
2761#ifdef CONFIG_NETWORK_SECMARK
2762static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2763{
2764 to->secmark = from->secmark;
2765}
2766
2767static inline void skb_init_secmark(struct sk_buff *skb)
2768{
2769 skb->secmark = 0;
2770}
2771#else
2772static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2773{ }
2774
2775static inline void skb_init_secmark(struct sk_buff *skb)
2776{ }
2777#endif
2778
f25f4e44
PWJ
2779static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2780{
f25f4e44 2781 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
2782}
2783
9247744e 2784static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 2785{
4e3ab47a 2786 return skb->queue_mapping;
4e3ab47a
PE
2787}
2788
f25f4e44
PWJ
2789static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2790{
f25f4e44 2791 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
2792}
2793
d5a9e24a
DM
2794static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2795{
2796 skb->queue_mapping = rx_queue + 1;
2797}
2798
9247744e 2799static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
2800{
2801 return skb->queue_mapping - 1;
2802}
2803
9247744e 2804static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 2805{
a02cec21 2806 return skb->queue_mapping != 0;
d5a9e24a
DM
2807}
2808
7965bd4d
JP
2809u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2810 unsigned int num_tx_queues);
9247744e 2811
def8b4fa
AD
2812static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2813{
0b3d8e08 2814#ifdef CONFIG_XFRM
def8b4fa 2815 return skb->sp;
def8b4fa 2816#else
def8b4fa 2817 return NULL;
def8b4fa 2818#endif
0b3d8e08 2819}
def8b4fa 2820
68c33163
PS
2821/* Keeps track of mac header offset relative to skb->head.
2822 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2823 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
2824 * tunnel skb it points to outer mac header.
2825 * Keeps track of level of encapsulation of network headers.
2826 */
68c33163 2827struct skb_gso_cb {
3347c960
ED
2828 int mac_offset;
2829 int encap_level;
68c33163
PS
2830};
2831#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2832
2833static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2834{
2835 return (skb_mac_header(inner_skb) - inner_skb->head) -
2836 SKB_GSO_CB(inner_skb)->mac_offset;
2837}
2838
1e2bd517
PS
2839static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
2840{
2841 int new_headroom, headroom;
2842 int ret;
2843
2844 headroom = skb_headroom(skb);
2845 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
2846 if (ret)
2847 return ret;
2848
2849 new_headroom = skb_headroom(skb);
2850 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
2851 return 0;
2852}
2853
bdcc0924 2854static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
2855{
2856 return skb_shinfo(skb)->gso_size;
2857}
2858
36a8f39e 2859/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 2860static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
2861{
2862 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2863}
2864
7965bd4d 2865void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
2866
2867static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2868{
2869 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2870 * wanted then gso_type will be set. */
05bdd2f1
ED
2871 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2872
b78462eb
AD
2873 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2874 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
2875 __skb_warn_lro_forwarding(skb);
2876 return true;
2877 }
2878 return false;
2879}
2880
35fc92a9
HX
2881static inline void skb_forward_csum(struct sk_buff *skb)
2882{
2883 /* Unfortunately we don't support this one. Any brave souls? */
2884 if (skb->ip_summed == CHECKSUM_COMPLETE)
2885 skb->ip_summed = CHECKSUM_NONE;
2886}
2887
bc8acf2c
ED
2888/**
2889 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2890 * @skb: skb to check
2891 *
2892 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2893 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2894 * use this helper, to document places where we make this assertion.
2895 */
05bdd2f1 2896static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
2897{
2898#ifdef DEBUG
2899 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2900#endif
2901}
2902
f35d9d8a 2903bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 2904
ed1f50c3
PD
2905int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
2906
f77668dc
DB
2907u32 __skb_get_poff(const struct sk_buff *skb);
2908
3a7c1ee4
AD
2909/**
2910 * skb_head_is_locked - Determine if the skb->head is locked down
2911 * @skb: skb to check
2912 *
2913 * The head on skbs build around a head frag can be removed if they are
2914 * not cloned. This function returns true if the skb head is locked down
2915 * due to either being allocated via kmalloc, or by being a clone with
2916 * multiple references to the head.
2917 */
2918static inline bool skb_head_is_locked(const struct sk_buff *skb)
2919{
2920 return !skb->head_frag || skb_cloned(skb);
2921}
1da177e4
LT
2922#endif /* __KERNEL__ */
2923#endif /* _LINUX_SKBUFF_H */
This page took 2.729946 seconds and 5 git commands to generate.