net: add __pskb_copy_fclone and pskb_copy_for_clone
[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>
363ec392 35#include <linux/sched.h>
5203cd28 36#include <net/flow_keys.h>
1da177e4 37
78ea85f1
DB
38/* A. Checksumming of received packets by device.
39 *
40 * CHECKSUM_NONE:
41 *
42 * Device failed to checksum this packet e.g. due to lack of capabilities.
43 * The packet contains full (though not verified) checksum in packet but
44 * not in skb->csum. Thus, skb->csum is undefined in this case.
45 *
46 * CHECKSUM_UNNECESSARY:
47 *
48 * The hardware you're dealing with doesn't calculate the full checksum
49 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
50 * for specific protocols e.g. TCP/UDP/SCTP, then, for such packets it will
51 * set CHECKSUM_UNNECESSARY if their checksums are okay. skb->csum is still
52 * undefined in this case though. It is a bad option, but, unfortunately,
53 * nowadays most vendors do this. Apparently with the secret goal to sell
54 * you new devices, when you will add new protocol to your host, f.e. IPv6 8)
55 *
56 * CHECKSUM_COMPLETE:
57 *
58 * This is the most generic way. The device supplied checksum of the _whole_
59 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
60 * hardware doesn't need to parse L3/L4 headers to implement this.
61 *
62 * Note: Even if device supports only some protocols, but is able to produce
63 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
64 *
65 * CHECKSUM_PARTIAL:
66 *
67 * This is identical to the case for output below. This may occur on a packet
68 * received directly from another Linux OS, e.g., a virtualized Linux kernel
69 * on the same host. The packet can be treated in the same way as
70 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
71 * checksum must be filled in by the OS or the hardware.
72 *
73 * B. Checksumming on output.
74 *
75 * CHECKSUM_NONE:
76 *
77 * The skb was already checksummed by the protocol, or a checksum is not
78 * required.
79 *
80 * CHECKSUM_PARTIAL:
81 *
82 * The device is required to checksum the packet as seen by hard_start_xmit()
83 * from skb->csum_start up to the end, and to record/write the checksum at
84 * offset skb->csum_start + skb->csum_offset.
85 *
86 * The device must show its capabilities in dev->features, set up at device
87 * setup time, e.g. netdev_features.h:
88 *
89 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
90 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
91 * IPv4. Sigh. Vendors like this way for an unknown reason.
92 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
93 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
94 * NETIF_F_... - Well, you get the picture.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * Normally, the device will do per protocol specific checksumming. Protocol
99 * implementations that do not want the NIC to perform the checksum
100 * calculation should use this flag in their outgoing skbs.
101 *
102 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
103 * offload. Correspondingly, the FCoE protocol driver
104 * stack should use CHECKSUM_UNNECESSARY.
105 *
106 * Any questions? No questions, good. --ANK
107 */
108
60476372 109/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
110#define CHECKSUM_NONE 0
111#define CHECKSUM_UNNECESSARY 1
112#define CHECKSUM_COMPLETE 2
113#define CHECKSUM_PARTIAL 3
1da177e4
LT
114
115#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
116 ~(SMP_CACHE_BYTES - 1))
fc910a27 117#define SKB_WITH_OVERHEAD(X) \
deea84b0 118 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
119#define SKB_MAX_ORDER(X, ORDER) \
120 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
121#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
122#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
123
87fb4b7b
ED
124/* return minimum truesize of one skb containing X bytes of data */
125#define SKB_TRUESIZE(X) ((X) + \
126 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
127 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
128
1da177e4 129struct net_device;
716ea3a7 130struct scatterlist;
9c55e01c 131struct pipe_inode_info;
1da177e4 132
5f79e0f9 133#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
134struct nf_conntrack {
135 atomic_t use;
1da177e4 136};
5f79e0f9 137#endif
1da177e4
LT
138
139#ifdef CONFIG_BRIDGE_NETFILTER
140struct nf_bridge_info {
bf1ac5ca
ED
141 atomic_t use;
142 unsigned int mask;
143 struct net_device *physindev;
144 struct net_device *physoutdev;
145 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
146};
147#endif
148
1da177e4
LT
149struct sk_buff_head {
150 /* These two members must be first. */
151 struct sk_buff *next;
152 struct sk_buff *prev;
153
154 __u32 qlen;
155 spinlock_t lock;
156};
157
158struct sk_buff;
159
9d4dde52
IC
160/* To allow 64K frame to be packed as single skb without frag_list we
161 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
162 * buffers which do not start on a page boundary.
163 *
164 * Since GRO uses frags we allocate at least 16 regardless of page
165 * size.
a715dea3 166 */
9d4dde52 167#if (65536/PAGE_SIZE + 1) < 16
eec00954 168#define MAX_SKB_FRAGS 16UL
a715dea3 169#else
9d4dde52 170#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 171#endif
1da177e4
LT
172
173typedef struct skb_frag_struct skb_frag_t;
174
175struct skb_frag_struct {
a8605c60
IC
176 struct {
177 struct page *p;
178 } page;
cb4dfe56 179#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
180 __u32 page_offset;
181 __u32 size;
cb4dfe56
ED
182#else
183 __u16 page_offset;
184 __u16 size;
185#endif
1da177e4
LT
186};
187
9e903e08
ED
188static inline unsigned int skb_frag_size(const skb_frag_t *frag)
189{
190 return frag->size;
191}
192
193static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
194{
195 frag->size = size;
196}
197
198static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
199{
200 frag->size += delta;
201}
202
203static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
204{
205 frag->size -= delta;
206}
207
ac45f602
PO
208#define HAVE_HW_TIME_STAMP
209
210/**
d3a21be8 211 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
212 * @hwtstamp: hardware time stamp transformed into duration
213 * since arbitrary point in time
214 * @syststamp: hwtstamp transformed to system time base
215 *
216 * Software time stamps generated by ktime_get_real() are stored in
217 * skb->tstamp. The relation between the different kinds of time
218 * stamps is as follows:
219 *
220 * syststamp and tstamp can be compared against each other in
221 * arbitrary combinations. The accuracy of a
222 * syststamp/tstamp/"syststamp from other device" comparison is
223 * limited by the accuracy of the transformation into system time
224 * base. This depends on the device driver and its underlying
225 * hardware.
226 *
227 * hwtstamps can only be compared against other hwtstamps from
228 * the same device.
229 *
230 * This structure is attached to packets as part of the
231 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
232 */
233struct skb_shared_hwtstamps {
234 ktime_t hwtstamp;
235 ktime_t syststamp;
236};
237
2244d07b
OH
238/* Definitions for tx_flags in struct skb_shared_info */
239enum {
240 /* generate hardware time stamp */
241 SKBTX_HW_TSTAMP = 1 << 0,
242
243 /* generate software time stamp */
244 SKBTX_SW_TSTAMP = 1 << 1,
245
246 /* device driver is going to provide hardware time stamp */
247 SKBTX_IN_PROGRESS = 1 << 2,
248
a6686f2f 249 /* device driver supports TX zero-copy buffers */
62b1a8ab 250 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
251
252 /* generate wifi status information (where possible) */
62b1a8ab 253 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
254
255 /* This indicates at least one fragment might be overwritten
256 * (as in vmsplice(), sendfile() ...)
257 * If we need to compute a TX checksum, we'll need to copy
258 * all frags to avoid possible bad checksum
259 */
260 SKBTX_SHARED_FRAG = 1 << 5,
a6686f2f
SM
261};
262
263/*
264 * The callback notifies userspace to release buffers when skb DMA is done in
265 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
266 * The zerocopy_success argument is true if zero copy transmit occurred,
267 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
268 * The ctx field is used to track device context.
269 * The desc field is used to track userspace buffer index.
a6686f2f
SM
270 */
271struct ubuf_info {
e19d6763 272 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 273 void *ctx;
a6686f2f 274 unsigned long desc;
ac45f602
PO
275};
276
1da177e4
LT
277/* This data is invariant across clones and lives at
278 * the end of the header data, ie. at skb->end.
279 */
280struct skb_shared_info {
9f42f126
IC
281 unsigned char nr_frags;
282 __u8 tx_flags;
7967168c
HX
283 unsigned short gso_size;
284 /* Warning: this field is not always filled in (UFO)! */
285 unsigned short gso_segs;
286 unsigned short gso_type;
1da177e4 287 struct sk_buff *frag_list;
ac45f602 288 struct skb_shared_hwtstamps hwtstamps;
9f42f126 289 __be32 ip6_frag_id;
ec7d2f2c
ED
290
291 /*
292 * Warning : all fields before dataref are cleared in __alloc_skb()
293 */
294 atomic_t dataref;
295
69e3c75f
JB
296 /* Intermediate layers must ensure that destructor_arg
297 * remains valid until skb destructor */
298 void * destructor_arg;
a6686f2f 299
fed66381
ED
300 /* must be last field, see pskb_expand_head() */
301 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
302};
303
304/* We divide dataref into two halves. The higher 16 bits hold references
305 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
306 * the entire skb->data. A clone of a headerless skb holds the length of
307 * the header in skb->hdr_len.
1da177e4
LT
308 *
309 * All users must obey the rule that the skb->data reference count must be
310 * greater than or equal to the payload reference count.
311 *
312 * Holding a reference to the payload part means that the user does not
313 * care about modifications to the header part of skb->data.
314 */
315#define SKB_DATAREF_SHIFT 16
316#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
317
d179cd12
DM
318
319enum {
320 SKB_FCLONE_UNAVAILABLE,
321 SKB_FCLONE_ORIG,
322 SKB_FCLONE_CLONE,
323};
324
7967168c
HX
325enum {
326 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 327 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
328
329 /* This indicates the skb is from an untrusted source. */
330 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
331
332 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
333 SKB_GSO_TCP_ECN = 1 << 3,
334
335 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
336
337 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
338
339 SKB_GSO_GRE = 1 << 6,
73136267 340
cb32f511 341 SKB_GSO_IPIP = 1 << 7,
0d89d203 342
61c1db7f 343 SKB_GSO_SIT = 1 << 8,
cb32f511 344
61c1db7f
ED
345 SKB_GSO_UDP_TUNNEL = 1 << 9,
346
347 SKB_GSO_MPLS = 1 << 10,
0f4f4ffa
TH
348
349 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c
TH
350
351 SKB_GSO_GRE_CSUM = 1 << 12,
7967168c
HX
352};
353
2e07fa9c
ACM
354#if BITS_PER_LONG > 32
355#define NET_SKBUFF_DATA_USES_OFFSET 1
356#endif
357
358#ifdef NET_SKBUFF_DATA_USES_OFFSET
359typedef unsigned int sk_buff_data_t;
360#else
361typedef unsigned char *sk_buff_data_t;
362#endif
363
363ec392
ED
364/**
365 * struct skb_mstamp - multi resolution time stamps
366 * @stamp_us: timestamp in us resolution
367 * @stamp_jiffies: timestamp in jiffies
368 */
369struct skb_mstamp {
370 union {
371 u64 v64;
372 struct {
373 u32 stamp_us;
374 u32 stamp_jiffies;
375 };
376 };
377};
378
379/**
380 * skb_mstamp_get - get current timestamp
381 * @cl: place to store timestamps
382 */
383static inline void skb_mstamp_get(struct skb_mstamp *cl)
384{
385 u64 val = local_clock();
386
387 do_div(val, NSEC_PER_USEC);
388 cl->stamp_us = (u32)val;
389 cl->stamp_jiffies = (u32)jiffies;
390}
391
392/**
393 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
394 * @t1: pointer to newest sample
395 * @t0: pointer to oldest sample
396 */
397static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
398 const struct skb_mstamp *t0)
399{
400 s32 delta_us = t1->stamp_us - t0->stamp_us;
401 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
402
403 /* If delta_us is negative, this might be because interval is too big,
404 * or local_clock() drift is too big : fallback using jiffies.
405 */
406 if (delta_us <= 0 ||
407 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
408
409 delta_us = jiffies_to_usecs(delta_jiffies);
410
411 return delta_us;
412}
413
414
1da177e4
LT
415/**
416 * struct sk_buff - socket buffer
417 * @next: Next buffer in list
418 * @prev: Previous buffer in list
363ec392 419 * @tstamp: Time we arrived/left
d84e0bd7 420 * @sk: Socket we are owned by
1da177e4 421 * @dev: Device we arrived on/are leaving by
d84e0bd7 422 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 423 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 424 * @sp: the security path, used for xfrm
1da177e4
LT
425 * @len: Length of actual data
426 * @data_len: Data length
427 * @mac_len: Length of link layer header
334a8132 428 * @hdr_len: writable header length of cloned skb
663ead3b
HX
429 * @csum: Checksum (must include start/offset pair)
430 * @csum_start: Offset from skb->head where checksumming should start
431 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 432 * @priority: Packet queueing priority
60ff7467 433 * @ignore_df: allow local fragmentation
1da177e4 434 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 435 * @ip_summed: Driver fed us an IP checksum
1da177e4 436 * @nohdr: Payload reference only, must not modify header
d84e0bd7 437 * @nfctinfo: Relationship of this skb to the connection
1da177e4 438 * @pkt_type: Packet class
c83c2486 439 * @fclone: skbuff clone status
c83c2486 440 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
441 * @peeked: this packet has been seen already, so stats have been
442 * done for it, don't do them again
ba9dda3a 443 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
444 * @protocol: Packet protocol from driver
445 * @destructor: Destruct function
446 * @nfct: Associated connection, if any
1da177e4 447 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 448 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
449 * @tc_index: Traffic control index
450 * @tc_verd: traffic control verdict
61b905da 451 * @hash: the packet hash
d84e0bd7 452 * @queue_mapping: Queue mapping for multiqueue devices
553a5672 453 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 454 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 455 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 456 * ports.
6e3e939f
JB
457 * @wifi_acked_valid: wifi_acked was set
458 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 459 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
f4b8ea78
RD
460 * @dma_cookie: a cookie to one of several possible DMA operations
461 * done by skb DMA functions
06021292 462 * @napi_id: id of the NAPI struct this skb came from
984bc16c 463 * @secmark: security marking
d84e0bd7
DB
464 * @mark: Generic packet mark
465 * @dropcount: total number of sk_receive_queue overflows
86a9bad3 466 * @vlan_proto: vlan encapsulation protocol
6aa895b0 467 * @vlan_tci: vlan tag control information
0d89d203 468 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
469 * @inner_transport_header: Inner transport layer header (encapsulation)
470 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 471 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
472 * @transport_header: Transport layer header
473 * @network_header: Network layer header
474 * @mac_header: Link layer header
475 * @tail: Tail pointer
476 * @end: End pointer
477 * @head: Head of buffer
478 * @data: Data head pointer
479 * @truesize: Buffer size
480 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
481 */
482
483struct sk_buff {
484 /* These two members must be first. */
485 struct sk_buff *next;
486 struct sk_buff *prev;
487
363ec392
ED
488 union {
489 ktime_t tstamp;
490 struct skb_mstamp skb_mstamp;
491 };
da3f5cf1
FF
492
493 struct sock *sk;
1da177e4 494 struct net_device *dev;
1da177e4 495
1da177e4
LT
496 /*
497 * This is the control buffer. It is free to use for every
498 * layer. Please put your private variables there. If you
499 * want to keep them across layers you have to do a skb_clone()
500 * first. This is owned by whoever has the skb queued ATM.
501 */
da3f5cf1 502 char cb[48] __aligned(8);
1da177e4 503
7fee226a 504 unsigned long _skb_refdst;
da3f5cf1
FF
505#ifdef CONFIG_XFRM
506 struct sec_path *sp;
507#endif
1da177e4 508 unsigned int len,
334a8132
PM
509 data_len;
510 __u16 mac_len,
511 hdr_len;
ff1dcadb
AV
512 union {
513 __wsum csum;
663ead3b
HX
514 struct {
515 __u16 csum_start;
516 __u16 csum_offset;
517 };
ff1dcadb 518 };
1da177e4 519 __u32 priority;
fe55f6d5 520 kmemcheck_bitfield_begin(flags1);
60ff7467 521 __u8 ignore_df:1,
1cbb3380
TG
522 cloned:1,
523 ip_summed:2,
6869c4d8
HW
524 nohdr:1,
525 nfctinfo:3;
d179cd12 526 __u8 pkt_type:3,
b84f4cc9 527 fclone:2,
ba9dda3a 528 ipvs_property:1,
a59322be 529 peeked:1,
ba9dda3a 530 nf_trace:1;
fe55f6d5 531 kmemcheck_bitfield_end(flags1);
4ab408de 532 __be16 protocol;
1da177e4
LT
533
534 void (*destructor)(struct sk_buff *skb);
9fb9cbb1 535#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 536 struct nf_conntrack *nfct;
2fc72c7b 537#endif
1da177e4
LT
538#ifdef CONFIG_BRIDGE_NETFILTER
539 struct nf_bridge_info *nf_bridge;
540#endif
f25f4e44 541
8964be4a 542 int skb_iif;
4031ae6e 543
61b905da 544 __u32 hash;
4031ae6e 545
86a9bad3 546 __be16 vlan_proto;
4031ae6e
AD
547 __u16 vlan_tci;
548
1da177e4 549#ifdef CONFIG_NET_SCHED
b6b99eb5 550 __u16 tc_index; /* traffic control index */
1da177e4 551#ifdef CONFIG_NET_CLS_ACT
b6b99eb5 552 __u16 tc_verd; /* traffic control verdict */
1da177e4 553#endif
1da177e4 554#endif
fe55f6d5 555
0a14842f 556 __u16 queue_mapping;
fe55f6d5 557 kmemcheck_bitfield_begin(flags2);
de357cc0 558#ifdef CONFIG_IPV6_NDISC_NODETYPE
8a4eb573 559 __u8 ndisc_nodetype:2;
d0f09804 560#endif
c93bdd0e 561 __u8 pfmemalloc:1;
3853b584 562 __u8 ooo_okay:1;
61b905da 563 __u8 l4_hash:1;
6e3e939f
JB
564 __u8 wifi_acked_valid:1;
565 __u8 wifi_acked:1;
3bdc0eba 566 __u8 no_fcs:1;
d3836f21 567 __u8 head_frag:1;
6a674e9c
JG
568 /* Encapsulation protocol and NIC drivers should use
569 * this flag to indicate to each other if the skb contains
570 * encapsulated packet or not and maybe use the inner packet
571 * headers if needed
572 */
573 __u8 encapsulation:1;
7e2b10c1
TH
574 __u8 encap_hdr_csum:1;
575 /* 5/7 bit hole (depending on ndisc_nodetype presence) */
fe55f6d5
VN
576 kmemcheck_bitfield_end(flags2);
577
e0d1095a 578#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
06021292
ET
579 union {
580 unsigned int napi_id;
581 dma_cookie_t dma_cookie;
582 };
97fc2f08 583#endif
984bc16c
JM
584#ifdef CONFIG_NETWORK_SECMARK
585 __u32 secmark;
586#endif
3b885787
NH
587 union {
588 __u32 mark;
589 __u32 dropcount;
16fad69c 590 __u32 reserved_tailroom;
3b885787 591 };
1da177e4 592
0d89d203 593 __be16 inner_protocol;
1a37e412
SH
594 __u16 inner_transport_header;
595 __u16 inner_network_header;
596 __u16 inner_mac_header;
597 __u16 transport_header;
598 __u16 network_header;
599 __u16 mac_header;
1da177e4 600 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 601 sk_buff_data_t tail;
4305b541 602 sk_buff_data_t end;
1da177e4 603 unsigned char *head,
4305b541 604 *data;
27a884dc
ACM
605 unsigned int truesize;
606 atomic_t users;
1da177e4
LT
607};
608
609#ifdef __KERNEL__
610/*
611 * Handling routines are only of interest to the kernel
612 */
613#include <linux/slab.h>
614
1da177e4 615
c93bdd0e
MG
616#define SKB_ALLOC_FCLONE 0x01
617#define SKB_ALLOC_RX 0x02
618
619/* Returns true if the skb was allocated from PFMEMALLOC reserves */
620static inline bool skb_pfmemalloc(const struct sk_buff *skb)
621{
622 return unlikely(skb->pfmemalloc);
623}
624
7fee226a
ED
625/*
626 * skb might have a dst pointer attached, refcounted or not.
627 * _skb_refdst low order bit is set if refcount was _not_ taken
628 */
629#define SKB_DST_NOREF 1UL
630#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
631
632/**
633 * skb_dst - returns skb dst_entry
634 * @skb: buffer
635 *
636 * Returns skb dst_entry, regardless of reference taken or not.
637 */
adf30907
ED
638static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
639{
7fee226a
ED
640 /* If refdst was not refcounted, check we still are in a
641 * rcu_read_lock section
642 */
643 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
644 !rcu_read_lock_held() &&
645 !rcu_read_lock_bh_held());
646 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
647}
648
7fee226a
ED
649/**
650 * skb_dst_set - sets skb dst
651 * @skb: buffer
652 * @dst: dst entry
653 *
654 * Sets skb dst, assuming a reference was taken on dst and should
655 * be released by skb_dst_drop()
656 */
adf30907
ED
657static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
658{
7fee226a
ED
659 skb->_skb_refdst = (unsigned long)dst;
660}
661
7965bd4d
JP
662void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
663 bool force);
932bc4d7
JA
664
665/**
666 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
667 * @skb: buffer
668 * @dst: dst entry
669 *
670 * Sets skb dst, assuming a reference was not taken on dst.
671 * If dst entry is cached, we do not take reference and dst_release
672 * will be avoided by refdst_drop. If dst entry is not cached, we take
673 * reference, so that last dst_release can destroy the dst immediately.
674 */
675static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
676{
677 __skb_dst_set_noref(skb, dst, false);
678}
679
680/**
681 * skb_dst_set_noref_force - sets skb dst, without taking reference
682 * @skb: buffer
683 * @dst: dst entry
684 *
685 * Sets skb dst, assuming a reference was not taken on dst.
686 * No reference is taken and no dst_release will be called. While for
687 * cached dsts deferred reclaim is a basic feature, for entries that are
688 * not cached it is caller's job to guarantee that last dst_release for
689 * provided dst happens when nobody uses it, eg. after a RCU grace period.
690 */
691static inline void skb_dst_set_noref_force(struct sk_buff *skb,
692 struct dst_entry *dst)
693{
694 __skb_dst_set_noref(skb, dst, true);
695}
7fee226a
ED
696
697/**
25985edc 698 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
699 * @skb: buffer
700 */
701static inline bool skb_dst_is_noref(const struct sk_buff *skb)
702{
703 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
704}
705
511c3f92
ED
706static inline struct rtable *skb_rtable(const struct sk_buff *skb)
707{
adf30907 708 return (struct rtable *)skb_dst(skb);
511c3f92
ED
709}
710
7965bd4d
JP
711void kfree_skb(struct sk_buff *skb);
712void kfree_skb_list(struct sk_buff *segs);
713void skb_tx_error(struct sk_buff *skb);
714void consume_skb(struct sk_buff *skb);
715void __kfree_skb(struct sk_buff *skb);
d7e8883c 716extern struct kmem_cache *skbuff_head_cache;
bad43ca8 717
7965bd4d
JP
718void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
719bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
720 bool *fragstolen, int *delta_truesize);
bad43ca8 721
7965bd4d
JP
722struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
723 int node);
724struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 725static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 726 gfp_t priority)
d179cd12 727{
564824b0 728 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
729}
730
731static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 732 gfp_t priority)
d179cd12 733{
c93bdd0e 734 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
735}
736
7965bd4d 737struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
738static inline struct sk_buff *alloc_skb_head(gfp_t priority)
739{
740 return __alloc_skb_head(priority, -1);
741}
742
7965bd4d
JP
743struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
744int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
745struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
746struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
747struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
748 gfp_t gfp_mask, bool fclone);
749static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
750 gfp_t gfp_mask)
751{
752 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
753}
7965bd4d
JP
754
755int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
756struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
757 unsigned int headroom);
758struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
759 int newtailroom, gfp_t priority);
25a91d8d
FD
760int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
761 int offset, int len);
7965bd4d
JP
762int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
763 int len);
764int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
765int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 766#define dev_kfree_skb(a) consume_skb(a)
1da177e4 767
7965bd4d
JP
768int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
769 int getfrag(void *from, char *to, int offset,
770 int len, int odd, struct sk_buff *skb),
771 void *from, int length);
e89e9cf5 772
d94d9fee 773struct skb_seq_state {
677e90ed
TG
774 __u32 lower_offset;
775 __u32 upper_offset;
776 __u32 frag_idx;
777 __u32 stepped_offset;
778 struct sk_buff *root_skb;
779 struct sk_buff *cur_skb;
780 __u8 *frag_data;
781};
782
7965bd4d
JP
783void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
784 unsigned int to, struct skb_seq_state *st);
785unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
786 struct skb_seq_state *st);
787void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 788
7965bd4d
JP
789unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
790 unsigned int to, struct ts_config *config,
791 struct ts_state *state);
3fc7e8a6 792
09323cc4
TH
793/*
794 * Packet hash types specify the type of hash in skb_set_hash.
795 *
796 * Hash types refer to the protocol layer addresses which are used to
797 * construct a packet's hash. The hashes are used to differentiate or identify
798 * flows of the protocol layer for the hash type. Hash types are either
799 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
800 *
801 * Properties of hashes:
802 *
803 * 1) Two packets in different flows have different hash values
804 * 2) Two packets in the same flow should have the same hash value
805 *
806 * A hash at a higher layer is considered to be more specific. A driver should
807 * set the most specific hash possible.
808 *
809 * A driver cannot indicate a more specific hash than the layer at which a hash
810 * was computed. For instance an L3 hash cannot be set as an L4 hash.
811 *
812 * A driver may indicate a hash level which is less specific than the
813 * actual layer the hash was computed on. For instance, a hash computed
814 * at L4 may be considered an L3 hash. This should only be done if the
815 * driver can't unambiguously determine that the HW computed the hash at
816 * the higher layer. Note that the "should" in the second property above
817 * permits this.
818 */
819enum pkt_hash_types {
820 PKT_HASH_TYPE_NONE, /* Undefined type */
821 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
822 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
823 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
824};
825
826static inline void
827skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
828{
61b905da
TH
829 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
830 skb->hash = hash;
09323cc4
TH
831}
832
3958afa1
TH
833void __skb_get_hash(struct sk_buff *skb);
834static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 835{
61b905da 836 if (!skb->l4_hash)
3958afa1 837 __skb_get_hash(skb);
bfb564e7 838
61b905da 839 return skb->hash;
bfb564e7
KK
840}
841
57bdf7f4
TH
842static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
843{
61b905da 844 return skb->hash;
57bdf7f4
TH
845}
846
7539fadc
TH
847static inline void skb_clear_hash(struct sk_buff *skb)
848{
61b905da
TH
849 skb->hash = 0;
850 skb->l4_hash = 0;
7539fadc
TH
851}
852
853static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
854{
61b905da 855 if (!skb->l4_hash)
7539fadc
TH
856 skb_clear_hash(skb);
857}
858
3df7a74e
TH
859static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
860{
61b905da
TH
861 to->hash = from->hash;
862 to->l4_hash = from->l4_hash;
3df7a74e
TH
863};
864
4305b541
ACM
865#ifdef NET_SKBUFF_DATA_USES_OFFSET
866static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
867{
868 return skb->head + skb->end;
869}
ec47ea82
AD
870
871static inline unsigned int skb_end_offset(const struct sk_buff *skb)
872{
873 return skb->end;
874}
4305b541
ACM
875#else
876static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
877{
878 return skb->end;
879}
ec47ea82
AD
880
881static inline unsigned int skb_end_offset(const struct sk_buff *skb)
882{
883 return skb->end - skb->head;
884}
4305b541
ACM
885#endif
886
1da177e4 887/* Internal */
4305b541 888#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 889
ac45f602
PO
890static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
891{
892 return &skb_shinfo(skb)->hwtstamps;
893}
894
1da177e4
LT
895/**
896 * skb_queue_empty - check if a queue is empty
897 * @list: queue head
898 *
899 * Returns true if the queue is empty, false otherwise.
900 */
901static inline int skb_queue_empty(const struct sk_buff_head *list)
902{
fd44b93c 903 return list->next == (const struct sk_buff *) list;
1da177e4
LT
904}
905
fc7ebb21
DM
906/**
907 * skb_queue_is_last - check if skb is the last entry in the queue
908 * @list: queue head
909 * @skb: buffer
910 *
911 * Returns true if @skb is the last buffer on the list.
912 */
913static inline bool skb_queue_is_last(const struct sk_buff_head *list,
914 const struct sk_buff *skb)
915{
fd44b93c 916 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
917}
918
832d11c5
IJ
919/**
920 * skb_queue_is_first - check if skb is the first entry in the queue
921 * @list: queue head
922 * @skb: buffer
923 *
924 * Returns true if @skb is the first buffer on the list.
925 */
926static inline bool skb_queue_is_first(const struct sk_buff_head *list,
927 const struct sk_buff *skb)
928{
fd44b93c 929 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
930}
931
249c8b42
DM
932/**
933 * skb_queue_next - return the next packet in the queue
934 * @list: queue head
935 * @skb: current buffer
936 *
937 * Return the next packet in @list after @skb. It is only valid to
938 * call this if skb_queue_is_last() evaluates to false.
939 */
940static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
941 const struct sk_buff *skb)
942{
943 /* This BUG_ON may seem severe, but if we just return then we
944 * are going to dereference garbage.
945 */
946 BUG_ON(skb_queue_is_last(list, skb));
947 return skb->next;
948}
949
832d11c5
IJ
950/**
951 * skb_queue_prev - return the prev packet in the queue
952 * @list: queue head
953 * @skb: current buffer
954 *
955 * Return the prev packet in @list before @skb. It is only valid to
956 * call this if skb_queue_is_first() evaluates to false.
957 */
958static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
959 const struct sk_buff *skb)
960{
961 /* This BUG_ON may seem severe, but if we just return then we
962 * are going to dereference garbage.
963 */
964 BUG_ON(skb_queue_is_first(list, skb));
965 return skb->prev;
966}
967
1da177e4
LT
968/**
969 * skb_get - reference buffer
970 * @skb: buffer to reference
971 *
972 * Makes another reference to a socket buffer and returns a pointer
973 * to the buffer.
974 */
975static inline struct sk_buff *skb_get(struct sk_buff *skb)
976{
977 atomic_inc(&skb->users);
978 return skb;
979}
980
981/*
982 * If users == 1, we are the only owner and are can avoid redundant
983 * atomic change.
984 */
985
1da177e4
LT
986/**
987 * skb_cloned - is the buffer a clone
988 * @skb: buffer to check
989 *
990 * Returns true if the buffer was generated with skb_clone() and is
991 * one of multiple shared copies of the buffer. Cloned buffers are
992 * shared data so must not be written to under normal circumstances.
993 */
994static inline int skb_cloned(const struct sk_buff *skb)
995{
996 return skb->cloned &&
997 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
998}
999
14bbd6a5
PS
1000static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1001{
1002 might_sleep_if(pri & __GFP_WAIT);
1003
1004 if (skb_cloned(skb))
1005 return pskb_expand_head(skb, 0, 0, pri);
1006
1007 return 0;
1008}
1009
1da177e4
LT
1010/**
1011 * skb_header_cloned - is the header a clone
1012 * @skb: buffer to check
1013 *
1014 * Returns true if modifying the header part of the buffer requires
1015 * the data to be copied.
1016 */
1017static inline int skb_header_cloned(const struct sk_buff *skb)
1018{
1019 int dataref;
1020
1021 if (!skb->cloned)
1022 return 0;
1023
1024 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1025 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1026 return dataref != 1;
1027}
1028
1029/**
1030 * skb_header_release - release reference to header
1031 * @skb: buffer to operate on
1032 *
1033 * Drop a reference to the header part of the buffer. This is done
1034 * by acquiring a payload reference. You must not read from the header
1035 * part of skb->data after this.
1036 */
1037static inline void skb_header_release(struct sk_buff *skb)
1038{
1039 BUG_ON(skb->nohdr);
1040 skb->nohdr = 1;
1041 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1042}
1043
1044/**
1045 * skb_shared - is the buffer shared
1046 * @skb: buffer to check
1047 *
1048 * Returns true if more than one person has a reference to this
1049 * buffer.
1050 */
1051static inline int skb_shared(const struct sk_buff *skb)
1052{
1053 return atomic_read(&skb->users) != 1;
1054}
1055
1056/**
1057 * skb_share_check - check if buffer is shared and if so clone it
1058 * @skb: buffer to check
1059 * @pri: priority for memory allocation
1060 *
1061 * If the buffer is shared the buffer is cloned and the old copy
1062 * drops a reference. A new clone with a single reference is returned.
1063 * If the buffer is not shared the original buffer is returned. When
1064 * being called from interrupt status or with spinlocks held pri must
1065 * be GFP_ATOMIC.
1066 *
1067 * NULL is returned on a memory allocation failure.
1068 */
47061bc4 1069static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1070{
1071 might_sleep_if(pri & __GFP_WAIT);
1072 if (skb_shared(skb)) {
1073 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1074
1075 if (likely(nskb))
1076 consume_skb(skb);
1077 else
1078 kfree_skb(skb);
1da177e4
LT
1079 skb = nskb;
1080 }
1081 return skb;
1082}
1083
1084/*
1085 * Copy shared buffers into a new sk_buff. We effectively do COW on
1086 * packets to handle cases where we have a local reader and forward
1087 * and a couple of other messy ones. The normal one is tcpdumping
1088 * a packet thats being forwarded.
1089 */
1090
1091/**
1092 * skb_unshare - make a copy of a shared buffer
1093 * @skb: buffer to check
1094 * @pri: priority for memory allocation
1095 *
1096 * If the socket buffer is a clone then this function creates a new
1097 * copy of the data, drops a reference count on the old copy and returns
1098 * the new copy with the reference count at 1. If the buffer is not a clone
1099 * the original buffer is returned. When called with a spinlock held or
1100 * from interrupt state @pri must be %GFP_ATOMIC
1101 *
1102 * %NULL is returned on a memory allocation failure.
1103 */
e2bf521d 1104static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1105 gfp_t pri)
1da177e4
LT
1106{
1107 might_sleep_if(pri & __GFP_WAIT);
1108 if (skb_cloned(skb)) {
1109 struct sk_buff *nskb = skb_copy(skb, pri);
1110 kfree_skb(skb); /* Free our shared copy */
1111 skb = nskb;
1112 }
1113 return skb;
1114}
1115
1116/**
1a5778aa 1117 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1118 * @list_: list to peek at
1119 *
1120 * Peek an &sk_buff. Unlike most other operations you _MUST_
1121 * be careful with this one. A peek leaves the buffer on the
1122 * list and someone else may run off with it. You must hold
1123 * the appropriate locks or have a private queue to do this.
1124 *
1125 * Returns %NULL for an empty list or a pointer to the head element.
1126 * The reference count is not incremented and the reference is therefore
1127 * volatile. Use with caution.
1128 */
05bdd2f1 1129static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1130{
18d07000
ED
1131 struct sk_buff *skb = list_->next;
1132
1133 if (skb == (struct sk_buff *)list_)
1134 skb = NULL;
1135 return skb;
1da177e4
LT
1136}
1137
da5ef6e5
PE
1138/**
1139 * skb_peek_next - peek skb following the given one from a queue
1140 * @skb: skb to start from
1141 * @list_: list to peek at
1142 *
1143 * Returns %NULL when the end of the list is met or a pointer to the
1144 * next element. The reference count is not incremented and the
1145 * reference is therefore volatile. Use with caution.
1146 */
1147static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1148 const struct sk_buff_head *list_)
1149{
1150 struct sk_buff *next = skb->next;
18d07000 1151
da5ef6e5
PE
1152 if (next == (struct sk_buff *)list_)
1153 next = NULL;
1154 return next;
1155}
1156
1da177e4 1157/**
1a5778aa 1158 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1159 * @list_: list to peek at
1160 *
1161 * Peek an &sk_buff. Unlike most other operations you _MUST_
1162 * be careful with this one. A peek leaves the buffer on the
1163 * list and someone else may run off with it. You must hold
1164 * the appropriate locks or have a private queue to do this.
1165 *
1166 * Returns %NULL for an empty list or a pointer to the tail element.
1167 * The reference count is not incremented and the reference is therefore
1168 * volatile. Use with caution.
1169 */
05bdd2f1 1170static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1171{
18d07000
ED
1172 struct sk_buff *skb = list_->prev;
1173
1174 if (skb == (struct sk_buff *)list_)
1175 skb = NULL;
1176 return skb;
1177
1da177e4
LT
1178}
1179
1180/**
1181 * skb_queue_len - get queue length
1182 * @list_: list to measure
1183 *
1184 * Return the length of an &sk_buff queue.
1185 */
1186static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1187{
1188 return list_->qlen;
1189}
1190
67fed459
DM
1191/**
1192 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1193 * @list: queue to initialize
1194 *
1195 * This initializes only the list and queue length aspects of
1196 * an sk_buff_head object. This allows to initialize the list
1197 * aspects of an sk_buff_head without reinitializing things like
1198 * the spinlock. It can also be used for on-stack sk_buff_head
1199 * objects where the spinlock is known to not be used.
1200 */
1201static inline void __skb_queue_head_init(struct sk_buff_head *list)
1202{
1203 list->prev = list->next = (struct sk_buff *)list;
1204 list->qlen = 0;
1205}
1206
76f10ad0
AV
1207/*
1208 * This function creates a split out lock class for each invocation;
1209 * this is needed for now since a whole lot of users of the skb-queue
1210 * infrastructure in drivers have different locking usage (in hardirq)
1211 * than the networking core (in softirq only). In the long run either the
1212 * network layer or drivers should need annotation to consolidate the
1213 * main types of usage into 3 classes.
1214 */
1da177e4
LT
1215static inline void skb_queue_head_init(struct sk_buff_head *list)
1216{
1217 spin_lock_init(&list->lock);
67fed459 1218 __skb_queue_head_init(list);
1da177e4
LT
1219}
1220
c2ecba71
PE
1221static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1222 struct lock_class_key *class)
1223{
1224 skb_queue_head_init(list);
1225 lockdep_set_class(&list->lock, class);
1226}
1227
1da177e4 1228/*
bf299275 1229 * Insert an sk_buff on a list.
1da177e4
LT
1230 *
1231 * The "__skb_xxxx()" functions are the non-atomic ones that
1232 * can only be called with interrupts disabled.
1233 */
7965bd4d
JP
1234void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1235 struct sk_buff_head *list);
bf299275
GR
1236static inline void __skb_insert(struct sk_buff *newsk,
1237 struct sk_buff *prev, struct sk_buff *next,
1238 struct sk_buff_head *list)
1239{
1240 newsk->next = next;
1241 newsk->prev = prev;
1242 next->prev = prev->next = newsk;
1243 list->qlen++;
1244}
1da177e4 1245
67fed459
DM
1246static inline void __skb_queue_splice(const struct sk_buff_head *list,
1247 struct sk_buff *prev,
1248 struct sk_buff *next)
1249{
1250 struct sk_buff *first = list->next;
1251 struct sk_buff *last = list->prev;
1252
1253 first->prev = prev;
1254 prev->next = first;
1255
1256 last->next = next;
1257 next->prev = last;
1258}
1259
1260/**
1261 * skb_queue_splice - join two skb lists, this is designed for stacks
1262 * @list: the new list to add
1263 * @head: the place to add it in the first list
1264 */
1265static inline void skb_queue_splice(const struct sk_buff_head *list,
1266 struct sk_buff_head *head)
1267{
1268 if (!skb_queue_empty(list)) {
1269 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1270 head->qlen += list->qlen;
67fed459
DM
1271 }
1272}
1273
1274/**
d9619496 1275 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1276 * @list: the new list to add
1277 * @head: the place to add it in the first list
1278 *
1279 * The list at @list is reinitialised
1280 */
1281static inline void skb_queue_splice_init(struct sk_buff_head *list,
1282 struct sk_buff_head *head)
1283{
1284 if (!skb_queue_empty(list)) {
1285 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1286 head->qlen += list->qlen;
67fed459
DM
1287 __skb_queue_head_init(list);
1288 }
1289}
1290
1291/**
1292 * skb_queue_splice_tail - join two skb lists, each list being a queue
1293 * @list: the new list to add
1294 * @head: the place to add it in the first list
1295 */
1296static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1297 struct sk_buff_head *head)
1298{
1299 if (!skb_queue_empty(list)) {
1300 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1301 head->qlen += list->qlen;
67fed459
DM
1302 }
1303}
1304
1305/**
d9619496 1306 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1307 * @list: the new list to add
1308 * @head: the place to add it in the first list
1309 *
1310 * Each of the lists is a queue.
1311 * The list at @list is reinitialised
1312 */
1313static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1314 struct sk_buff_head *head)
1315{
1316 if (!skb_queue_empty(list)) {
1317 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1318 head->qlen += list->qlen;
67fed459
DM
1319 __skb_queue_head_init(list);
1320 }
1321}
1322
1da177e4 1323/**
300ce174 1324 * __skb_queue_after - queue a buffer at the list head
1da177e4 1325 * @list: list to use
300ce174 1326 * @prev: place after this buffer
1da177e4
LT
1327 * @newsk: buffer to queue
1328 *
300ce174 1329 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1330 * and you must therefore hold required locks before calling it.
1331 *
1332 * A buffer cannot be placed on two lists at the same time.
1333 */
300ce174
SH
1334static inline void __skb_queue_after(struct sk_buff_head *list,
1335 struct sk_buff *prev,
1336 struct sk_buff *newsk)
1da177e4 1337{
bf299275 1338 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1339}
1340
7965bd4d
JP
1341void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1342 struct sk_buff_head *list);
7de6c033 1343
f5572855
GR
1344static inline void __skb_queue_before(struct sk_buff_head *list,
1345 struct sk_buff *next,
1346 struct sk_buff *newsk)
1347{
1348 __skb_insert(newsk, next->prev, next, list);
1349}
1350
300ce174
SH
1351/**
1352 * __skb_queue_head - queue a buffer at the list head
1353 * @list: list to use
1354 * @newsk: buffer to queue
1355 *
1356 * Queue a buffer at the start of a list. This function takes no locks
1357 * and you must therefore hold required locks before calling it.
1358 *
1359 * A buffer cannot be placed on two lists at the same time.
1360 */
7965bd4d 1361void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1362static inline void __skb_queue_head(struct sk_buff_head *list,
1363 struct sk_buff *newsk)
1364{
1365 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1366}
1367
1da177e4
LT
1368/**
1369 * __skb_queue_tail - queue a buffer at the list tail
1370 * @list: list to use
1371 * @newsk: buffer to queue
1372 *
1373 * Queue a buffer at the end of a list. This function takes no locks
1374 * and you must therefore hold required locks before calling it.
1375 *
1376 * A buffer cannot be placed on two lists at the same time.
1377 */
7965bd4d 1378void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1379static inline void __skb_queue_tail(struct sk_buff_head *list,
1380 struct sk_buff *newsk)
1381{
f5572855 1382 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1383}
1384
1da177e4
LT
1385/*
1386 * remove sk_buff from list. _Must_ be called atomically, and with
1387 * the list known..
1388 */
7965bd4d 1389void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1390static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1391{
1392 struct sk_buff *next, *prev;
1393
1394 list->qlen--;
1395 next = skb->next;
1396 prev = skb->prev;
1397 skb->next = skb->prev = NULL;
1da177e4
LT
1398 next->prev = prev;
1399 prev->next = next;
1400}
1401
f525c06d
GR
1402/**
1403 * __skb_dequeue - remove from the head of the queue
1404 * @list: list to dequeue from
1405 *
1406 * Remove the head of the list. This function does not take any locks
1407 * so must be used with appropriate locks held only. The head item is
1408 * returned or %NULL if the list is empty.
1409 */
7965bd4d 1410struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1411static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1412{
1413 struct sk_buff *skb = skb_peek(list);
1414 if (skb)
1415 __skb_unlink(skb, list);
1416 return skb;
1417}
1da177e4
LT
1418
1419/**
1420 * __skb_dequeue_tail - remove from the tail of the queue
1421 * @list: list to dequeue from
1422 *
1423 * Remove the tail of the list. This function does not take any locks
1424 * so must be used with appropriate locks held only. The tail item is
1425 * returned or %NULL if the list is empty.
1426 */
7965bd4d 1427struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1428static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1429{
1430 struct sk_buff *skb = skb_peek_tail(list);
1431 if (skb)
1432 __skb_unlink(skb, list);
1433 return skb;
1434}
1435
1436
bdcc0924 1437static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1438{
1439 return skb->data_len;
1440}
1441
1442static inline unsigned int skb_headlen(const struct sk_buff *skb)
1443{
1444 return skb->len - skb->data_len;
1445}
1446
1447static inline int skb_pagelen(const struct sk_buff *skb)
1448{
1449 int i, len = 0;
1450
1451 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1452 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1453 return len + skb_headlen(skb);
1454}
1455
131ea667
IC
1456/**
1457 * __skb_fill_page_desc - initialise a paged fragment in an skb
1458 * @skb: buffer containing fragment to be initialised
1459 * @i: paged fragment index to initialise
1460 * @page: the page to use for this fragment
1461 * @off: the offset to the data with @page
1462 * @size: the length of the data
1463 *
1464 * Initialises the @i'th fragment of @skb to point to &size bytes at
1465 * offset @off within @page.
1466 *
1467 * Does not take any additional reference on the fragment.
1468 */
1469static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1470 struct page *page, int off, int size)
1da177e4
LT
1471{
1472 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1473
c48a11c7
MG
1474 /*
1475 * Propagate page->pfmemalloc to the skb if we can. The problem is
1476 * that not all callers have unique ownership of the page. If
1477 * pfmemalloc is set, we check the mapping as a mapping implies
1478 * page->index is set (index and pfmemalloc share space).
1479 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1480 * do not lose pfmemalloc information as the pages would not be
1481 * allocated using __GFP_MEMALLOC.
1482 */
a8605c60 1483 frag->page.p = page;
1da177e4 1484 frag->page_offset = off;
9e903e08 1485 skb_frag_size_set(frag, size);
cca7af38
PE
1486
1487 page = compound_head(page);
1488 if (page->pfmemalloc && !page->mapping)
1489 skb->pfmemalloc = true;
131ea667
IC
1490}
1491
1492/**
1493 * skb_fill_page_desc - initialise a paged fragment in an skb
1494 * @skb: buffer containing fragment to be initialised
1495 * @i: paged fragment index to initialise
1496 * @page: the page to use for this fragment
1497 * @off: the offset to the data with @page
1498 * @size: the length of the data
1499 *
1500 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1501 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1502 * addition updates @skb such that @i is the last fragment.
1503 *
1504 * Does not take any additional reference on the fragment.
1505 */
1506static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1507 struct page *page, int off, int size)
1508{
1509 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1510 skb_shinfo(skb)->nr_frags = i + 1;
1511}
1512
7965bd4d
JP
1513void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1514 int size, unsigned int truesize);
654bed16 1515
f8e617e1
JW
1516void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1517 unsigned int truesize);
1518
1da177e4 1519#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1520#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1521#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1522
27a884dc
ACM
1523#ifdef NET_SKBUFF_DATA_USES_OFFSET
1524static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1525{
1526 return skb->head + skb->tail;
1527}
1528
1529static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1530{
1531 skb->tail = skb->data - skb->head;
1532}
1533
1534static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1535{
1536 skb_reset_tail_pointer(skb);
1537 skb->tail += offset;
1538}
7cc46190 1539
27a884dc
ACM
1540#else /* NET_SKBUFF_DATA_USES_OFFSET */
1541static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1542{
1543 return skb->tail;
1544}
1545
1546static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1547{
1548 skb->tail = skb->data;
1549}
1550
1551static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1552{
1553 skb->tail = skb->data + offset;
1554}
4305b541 1555
27a884dc
ACM
1556#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1557
1da177e4
LT
1558/*
1559 * Add data to an sk_buff
1560 */
0c7ddf36 1561unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1562unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1563static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1564{
27a884dc 1565 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1566 SKB_LINEAR_ASSERT(skb);
1567 skb->tail += len;
1568 skb->len += len;
1569 return tmp;
1570}
1571
7965bd4d 1572unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1573static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1574{
1575 skb->data -= len;
1576 skb->len += len;
1577 return skb->data;
1578}
1579
7965bd4d 1580unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1581static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1582{
1583 skb->len -= len;
1584 BUG_ON(skb->len < skb->data_len);
1585 return skb->data += len;
1586}
1587
47d29646
DM
1588static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1589{
1590 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1591}
1592
7965bd4d 1593unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1594
1595static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1596{
1597 if (len > skb_headlen(skb) &&
987c402a 1598 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1599 return NULL;
1600 skb->len -= len;
1601 return skb->data += len;
1602}
1603
1604static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1605{
1606 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1607}
1608
1609static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1610{
1611 if (likely(len <= skb_headlen(skb)))
1612 return 1;
1613 if (unlikely(len > skb->len))
1614 return 0;
987c402a 1615 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1616}
1617
1618/**
1619 * skb_headroom - bytes at buffer head
1620 * @skb: buffer to check
1621 *
1622 * Return the number of bytes of free space at the head of an &sk_buff.
1623 */
c2636b4d 1624static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1625{
1626 return skb->data - skb->head;
1627}
1628
1629/**
1630 * skb_tailroom - bytes at buffer end
1631 * @skb: buffer to check
1632 *
1633 * Return the number of bytes of free space at the tail of an sk_buff
1634 */
1635static inline int skb_tailroom(const struct sk_buff *skb)
1636{
4305b541 1637 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1638}
1639
a21d4572
ED
1640/**
1641 * skb_availroom - bytes at buffer end
1642 * @skb: buffer to check
1643 *
1644 * Return the number of bytes of free space at the tail of an sk_buff
1645 * allocated by sk_stream_alloc()
1646 */
1647static inline int skb_availroom(const struct sk_buff *skb)
1648{
16fad69c
ED
1649 if (skb_is_nonlinear(skb))
1650 return 0;
1651
1652 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1653}
1654
1da177e4
LT
1655/**
1656 * skb_reserve - adjust headroom
1657 * @skb: buffer to alter
1658 * @len: bytes to move
1659 *
1660 * Increase the headroom of an empty &sk_buff by reducing the tail
1661 * room. This is only allowed for an empty buffer.
1662 */
8243126c 1663static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1664{
1665 skb->data += len;
1666 skb->tail += len;
1667}
1668
6a674e9c
JG
1669static inline void skb_reset_inner_headers(struct sk_buff *skb)
1670{
aefbd2b3 1671 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1672 skb->inner_network_header = skb->network_header;
1673 skb->inner_transport_header = skb->transport_header;
1674}
1675
0b5c9db1
JP
1676static inline void skb_reset_mac_len(struct sk_buff *skb)
1677{
1678 skb->mac_len = skb->network_header - skb->mac_header;
1679}
1680
6a674e9c
JG
1681static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1682 *skb)
1683{
1684 return skb->head + skb->inner_transport_header;
1685}
1686
1687static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1688{
1689 skb->inner_transport_header = skb->data - skb->head;
1690}
1691
1692static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1693 const int offset)
1694{
1695 skb_reset_inner_transport_header(skb);
1696 skb->inner_transport_header += offset;
1697}
1698
1699static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1700{
1701 return skb->head + skb->inner_network_header;
1702}
1703
1704static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1705{
1706 skb->inner_network_header = skb->data - skb->head;
1707}
1708
1709static inline void skb_set_inner_network_header(struct sk_buff *skb,
1710 const int offset)
1711{
1712 skb_reset_inner_network_header(skb);
1713 skb->inner_network_header += offset;
1714}
1715
aefbd2b3
PS
1716static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1717{
1718 return skb->head + skb->inner_mac_header;
1719}
1720
1721static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1722{
1723 skb->inner_mac_header = skb->data - skb->head;
1724}
1725
1726static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1727 const int offset)
1728{
1729 skb_reset_inner_mac_header(skb);
1730 skb->inner_mac_header += offset;
1731}
fda55eca
ED
1732static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1733{
35d04610 1734 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1735}
1736
9c70220b
ACM
1737static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1738{
2e07fa9c 1739 return skb->head + skb->transport_header;
9c70220b
ACM
1740}
1741
badff6d0
ACM
1742static inline void skb_reset_transport_header(struct sk_buff *skb)
1743{
2e07fa9c 1744 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1745}
1746
967b05f6
ACM
1747static inline void skb_set_transport_header(struct sk_buff *skb,
1748 const int offset)
1749{
2e07fa9c
ACM
1750 skb_reset_transport_header(skb);
1751 skb->transport_header += offset;
ea2ae17d
ACM
1752}
1753
d56f90a7
ACM
1754static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1755{
2e07fa9c 1756 return skb->head + skb->network_header;
d56f90a7
ACM
1757}
1758
c1d2bbe1
ACM
1759static inline void skb_reset_network_header(struct sk_buff *skb)
1760{
2e07fa9c 1761 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1762}
1763
c14d2450
ACM
1764static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1765{
2e07fa9c
ACM
1766 skb_reset_network_header(skb);
1767 skb->network_header += offset;
c14d2450
ACM
1768}
1769
2e07fa9c 1770static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1771{
2e07fa9c 1772 return skb->head + skb->mac_header;
bbe735e4
ACM
1773}
1774
2e07fa9c 1775static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1776{
35d04610 1777 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1778}
1779
1780static inline void skb_reset_mac_header(struct sk_buff *skb)
1781{
1782 skb->mac_header = skb->data - skb->head;
1783}
1784
1785static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1786{
1787 skb_reset_mac_header(skb);
1788 skb->mac_header += offset;
1789}
1790
0e3da5bb
TT
1791static inline void skb_pop_mac_header(struct sk_buff *skb)
1792{
1793 skb->mac_header = skb->network_header;
1794}
1795
fbbdb8f0
YX
1796static inline void skb_probe_transport_header(struct sk_buff *skb,
1797 const int offset_hint)
1798{
1799 struct flow_keys keys;
1800
1801 if (skb_transport_header_was_set(skb))
1802 return;
1803 else if (skb_flow_dissect(skb, &keys))
1804 skb_set_transport_header(skb, keys.thoff);
1805 else
1806 skb_set_transport_header(skb, offset_hint);
1807}
1808
03606895
ED
1809static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1810{
1811 if (skb_mac_header_was_set(skb)) {
1812 const unsigned char *old_mac = skb_mac_header(skb);
1813
1814 skb_set_mac_header(skb, -skb->mac_len);
1815 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1816 }
1817}
1818
04fb451e
MM
1819static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1820{
1821 return skb->csum_start - skb_headroom(skb);
1822}
1823
2e07fa9c
ACM
1824static inline int skb_transport_offset(const struct sk_buff *skb)
1825{
1826 return skb_transport_header(skb) - skb->data;
1827}
1828
1829static inline u32 skb_network_header_len(const struct sk_buff *skb)
1830{
1831 return skb->transport_header - skb->network_header;
1832}
1833
6a674e9c
JG
1834static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1835{
1836 return skb->inner_transport_header - skb->inner_network_header;
1837}
1838
2e07fa9c
ACM
1839static inline int skb_network_offset(const struct sk_buff *skb)
1840{
1841 return skb_network_header(skb) - skb->data;
1842}
48d49d0c 1843
6a674e9c
JG
1844static inline int skb_inner_network_offset(const struct sk_buff *skb)
1845{
1846 return skb_inner_network_header(skb) - skb->data;
1847}
1848
f9599ce1
CG
1849static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1850{
1851 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1852}
1853
1da177e4
LT
1854/*
1855 * CPUs often take a performance hit when accessing unaligned memory
1856 * locations. The actual performance hit varies, it can be small if the
1857 * hardware handles it or large if we have to take an exception and fix it
1858 * in software.
1859 *
1860 * Since an ethernet header is 14 bytes network drivers often end up with
1861 * the IP header at an unaligned offset. The IP header can be aligned by
1862 * shifting the start of the packet by 2 bytes. Drivers should do this
1863 * with:
1864 *
8660c124 1865 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1866 *
1867 * The downside to this alignment of the IP header is that the DMA is now
1868 * unaligned. On some architectures the cost of an unaligned DMA is high
1869 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1870 *
1da177e4
LT
1871 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1872 * to be overridden.
1873 */
1874#ifndef NET_IP_ALIGN
1875#define NET_IP_ALIGN 2
1876#endif
1877
025be81e
AB
1878/*
1879 * The networking layer reserves some headroom in skb data (via
1880 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1881 * the header has to grow. In the default case, if the header has to grow
d6301d3d 1882 * 32 bytes or less we avoid the reallocation.
025be81e
AB
1883 *
1884 * Unfortunately this headroom changes the DMA alignment of the resulting
1885 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1886 * on some architectures. An architecture can override this value,
1887 * perhaps setting it to a cacheline in size (since that will maintain
1888 * cacheline alignment of the DMA). It must be a power of 2.
1889 *
d6301d3d 1890 * Various parts of the networking layer expect at least 32 bytes of
025be81e 1891 * headroom, you should not reduce this.
5933dd2f
ED
1892 *
1893 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1894 * to reduce average number of cache lines per packet.
1895 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 1896 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
1897 */
1898#ifndef NET_SKB_PAD
5933dd2f 1899#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
1900#endif
1901
7965bd4d 1902int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1903
1904static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1905{
c4264f27 1906 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
1907 WARN_ON(1);
1908 return;
1909 }
27a884dc
ACM
1910 skb->len = len;
1911 skb_set_tail_pointer(skb, len);
1da177e4
LT
1912}
1913
7965bd4d 1914void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1915
1916static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1917{
3cc0e873
HX
1918 if (skb->data_len)
1919 return ___pskb_trim(skb, len);
1920 __skb_trim(skb, len);
1921 return 0;
1da177e4
LT
1922}
1923
1924static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1925{
1926 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1927}
1928
e9fa4f7b
HX
1929/**
1930 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1931 * @skb: buffer to alter
1932 * @len: new length
1933 *
1934 * This is identical to pskb_trim except that the caller knows that
1935 * the skb is not cloned so we should never get an error due to out-
1936 * of-memory.
1937 */
1938static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1939{
1940 int err = pskb_trim(skb, len);
1941 BUG_ON(err);
1942}
1943
1da177e4
LT
1944/**
1945 * skb_orphan - orphan a buffer
1946 * @skb: buffer to orphan
1947 *
1948 * If a buffer currently has an owner then we call the owner's
1949 * destructor function and make the @skb unowned. The buffer continues
1950 * to exist but is no longer charged to its former owner.
1951 */
1952static inline void skb_orphan(struct sk_buff *skb)
1953{
c34a7612 1954 if (skb->destructor) {
1da177e4 1955 skb->destructor(skb);
c34a7612
ED
1956 skb->destructor = NULL;
1957 skb->sk = NULL;
376c7311
ED
1958 } else {
1959 BUG_ON(skb->sk);
c34a7612 1960 }
1da177e4
LT
1961}
1962
a353e0ce
MT
1963/**
1964 * skb_orphan_frags - orphan the frags contained in a buffer
1965 * @skb: buffer to orphan frags from
1966 * @gfp_mask: allocation mask for replacement pages
1967 *
1968 * For each frag in the SKB which needs a destructor (i.e. has an
1969 * owner) create a copy of that frag and release the original
1970 * page by calling the destructor.
1971 */
1972static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1973{
1974 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1975 return 0;
1976 return skb_copy_ubufs(skb, gfp_mask);
1977}
1978
1da177e4
LT
1979/**
1980 * __skb_queue_purge - empty a list
1981 * @list: list to empty
1982 *
1983 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1984 * the list and one reference dropped. This function does not take the
1985 * list lock and the caller must hold the relevant locks to use it.
1986 */
7965bd4d 1987void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
1988static inline void __skb_queue_purge(struct sk_buff_head *list)
1989{
1990 struct sk_buff *skb;
1991 while ((skb = __skb_dequeue(list)) != NULL)
1992 kfree_skb(skb);
1993}
1994
e5e67305
AD
1995#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1996#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1997#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1998
7965bd4d 1999void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2000
7965bd4d
JP
2001struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2002 gfp_t gfp_mask);
8af27456
CH
2003
2004/**
2005 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2006 * @dev: network device to receive on
2007 * @length: length to allocate
2008 *
2009 * Allocate a new &sk_buff and assign it a usage count of one. The
2010 * buffer has unspecified headroom built in. Users should allocate
2011 * the headroom they think they need without accounting for the
2012 * built in space. The built in space is used for optimisations.
2013 *
2014 * %NULL is returned if there is no free memory. Although this function
2015 * allocates memory it can be called from an interrupt.
2016 */
2017static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2018 unsigned int length)
8af27456
CH
2019{
2020 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2021}
2022
6f532612
ED
2023/* legacy helper around __netdev_alloc_skb() */
2024static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2025 gfp_t gfp_mask)
2026{
2027 return __netdev_alloc_skb(NULL, length, gfp_mask);
2028}
2029
2030/* legacy helper around netdev_alloc_skb() */
2031static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2032{
2033 return netdev_alloc_skb(NULL, length);
2034}
2035
2036
4915a0de
ED
2037static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2038 unsigned int length, gfp_t gfp)
61321bbd 2039{
4915a0de 2040 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2041
2042 if (NET_IP_ALIGN && skb)
2043 skb_reserve(skb, NET_IP_ALIGN);
2044 return skb;
2045}
2046
4915a0de
ED
2047static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2048 unsigned int length)
2049{
2050 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2051}
2052
bc6fc9fa
FF
2053/**
2054 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
2055 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2056 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2057 * @order: size of the allocation
2058 *
2059 * Allocate a new page.
2060 *
2061 * %NULL is returned if there is no free memory.
2062*/
2063static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2064 struct sk_buff *skb,
2065 unsigned int order)
2066{
2067 struct page *page;
2068
2069 gfp_mask |= __GFP_COLD;
2070
2071 if (!(gfp_mask & __GFP_NOMEMALLOC))
2072 gfp_mask |= __GFP_MEMALLOC;
2073
2074 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2075 if (skb && page && page->pfmemalloc)
2076 skb->pfmemalloc = true;
2077
2078 return page;
2079}
2080
2081/**
2082 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2083 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2084 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2085 *
2086 * Allocate a new page.
2087 *
2088 * %NULL is returned if there is no free memory.
2089 */
2090static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2091 struct sk_buff *skb)
2092{
2093 return __skb_alloc_pages(gfp_mask, skb, 0);
2094}
2095
2096/**
2097 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2098 * @page: The page that was allocated from skb_alloc_page
2099 * @skb: The skb that may need pfmemalloc set
2100 */
2101static inline void skb_propagate_pfmemalloc(struct page *page,
2102 struct sk_buff *skb)
2103{
2104 if (page && page->pfmemalloc)
2105 skb->pfmemalloc = true;
2106}
2107
131ea667 2108/**
e227867f 2109 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2110 * @frag: the paged fragment
2111 *
2112 * Returns the &struct page associated with @frag.
2113 */
2114static inline struct page *skb_frag_page(const skb_frag_t *frag)
2115{
a8605c60 2116 return frag->page.p;
131ea667
IC
2117}
2118
2119/**
2120 * __skb_frag_ref - take an addition reference on a paged fragment.
2121 * @frag: the paged fragment
2122 *
2123 * Takes an additional reference on the paged fragment @frag.
2124 */
2125static inline void __skb_frag_ref(skb_frag_t *frag)
2126{
2127 get_page(skb_frag_page(frag));
2128}
2129
2130/**
2131 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2132 * @skb: the buffer
2133 * @f: the fragment offset.
2134 *
2135 * Takes an additional reference on the @f'th paged fragment of @skb.
2136 */
2137static inline void skb_frag_ref(struct sk_buff *skb, int f)
2138{
2139 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2140}
2141
2142/**
2143 * __skb_frag_unref - release a reference on a paged fragment.
2144 * @frag: the paged fragment
2145 *
2146 * Releases a reference on the paged fragment @frag.
2147 */
2148static inline void __skb_frag_unref(skb_frag_t *frag)
2149{
2150 put_page(skb_frag_page(frag));
2151}
2152
2153/**
2154 * skb_frag_unref - release a reference on a paged fragment of an skb.
2155 * @skb: the buffer
2156 * @f: the fragment offset
2157 *
2158 * Releases a reference on the @f'th paged fragment of @skb.
2159 */
2160static inline void skb_frag_unref(struct sk_buff *skb, int f)
2161{
2162 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2163}
2164
2165/**
2166 * skb_frag_address - gets the address of the data contained in a paged fragment
2167 * @frag: the paged fragment buffer
2168 *
2169 * Returns the address of the data within @frag. The page must already
2170 * be mapped.
2171 */
2172static inline void *skb_frag_address(const skb_frag_t *frag)
2173{
2174 return page_address(skb_frag_page(frag)) + frag->page_offset;
2175}
2176
2177/**
2178 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2179 * @frag: the paged fragment buffer
2180 *
2181 * Returns the address of the data within @frag. Checks that the page
2182 * is mapped and returns %NULL otherwise.
2183 */
2184static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2185{
2186 void *ptr = page_address(skb_frag_page(frag));
2187 if (unlikely(!ptr))
2188 return NULL;
2189
2190 return ptr + frag->page_offset;
2191}
2192
2193/**
2194 * __skb_frag_set_page - sets the page contained in a paged fragment
2195 * @frag: the paged fragment
2196 * @page: the page to set
2197 *
2198 * Sets the fragment @frag to contain @page.
2199 */
2200static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2201{
a8605c60 2202 frag->page.p = page;
131ea667
IC
2203}
2204
2205/**
2206 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2207 * @skb: the buffer
2208 * @f: the fragment offset
2209 * @page: the page to set
2210 *
2211 * Sets the @f'th fragment of @skb to contain @page.
2212 */
2213static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2214 struct page *page)
2215{
2216 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2217}
2218
400dfd3a
ED
2219bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2220
131ea667
IC
2221/**
2222 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2223 * @dev: the device to map the fragment to
131ea667
IC
2224 * @frag: the paged fragment to map
2225 * @offset: the offset within the fragment (starting at the
2226 * fragment's own offset)
2227 * @size: the number of bytes to map
f83347df 2228 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2229 *
2230 * Maps the page associated with @frag to @device.
2231 */
2232static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2233 const skb_frag_t *frag,
2234 size_t offset, size_t size,
2235 enum dma_data_direction dir)
2236{
2237 return dma_map_page(dev, skb_frag_page(frag),
2238 frag->page_offset + offset, size, dir);
2239}
2240
117632e6
ED
2241static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2242 gfp_t gfp_mask)
2243{
2244 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2245}
2246
bad93e9d
OP
2247
2248static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2249 gfp_t gfp_mask)
2250{
2251 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2252}
2253
2254
334a8132
PM
2255/**
2256 * skb_clone_writable - is the header of a clone writable
2257 * @skb: buffer to check
2258 * @len: length up to which to write
2259 *
2260 * Returns true if modifying the header part of the cloned buffer
2261 * does not requires the data to be copied.
2262 */
05bdd2f1 2263static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2264{
2265 return !skb_header_cloned(skb) &&
2266 skb_headroom(skb) + len <= skb->hdr_len;
2267}
2268
d9cc2048
HX
2269static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2270 int cloned)
2271{
2272 int delta = 0;
2273
d9cc2048
HX
2274 if (headroom > skb_headroom(skb))
2275 delta = headroom - skb_headroom(skb);
2276
2277 if (delta || cloned)
2278 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2279 GFP_ATOMIC);
2280 return 0;
2281}
2282
1da177e4
LT
2283/**
2284 * skb_cow - copy header of skb when it is required
2285 * @skb: buffer to cow
2286 * @headroom: needed headroom
2287 *
2288 * If the skb passed lacks sufficient headroom or its data part
2289 * is shared, data is reallocated. If reallocation fails, an error
2290 * is returned and original skb is not changed.
2291 *
2292 * The result is skb with writable area skb->head...skb->tail
2293 * and at least @headroom of space at head.
2294 */
2295static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2296{
d9cc2048
HX
2297 return __skb_cow(skb, headroom, skb_cloned(skb));
2298}
1da177e4 2299
d9cc2048
HX
2300/**
2301 * skb_cow_head - skb_cow but only making the head writable
2302 * @skb: buffer to cow
2303 * @headroom: needed headroom
2304 *
2305 * This function is identical to skb_cow except that we replace the
2306 * skb_cloned check by skb_header_cloned. It should be used when
2307 * you only need to push on some header and do not need to modify
2308 * the data.
2309 */
2310static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2311{
2312 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2313}
2314
2315/**
2316 * skb_padto - pad an skbuff up to a minimal size
2317 * @skb: buffer to pad
2318 * @len: minimal length
2319 *
2320 * Pads up a buffer to ensure the trailing bytes exist and are
2321 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2322 * is untouched. Otherwise it is extended. Returns zero on
2323 * success. The skb is freed on error.
1da177e4
LT
2324 */
2325
5b057c6b 2326static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2327{
2328 unsigned int size = skb->len;
2329 if (likely(size >= len))
5b057c6b 2330 return 0;
987c402a 2331 return skb_pad(skb, len - size);
1da177e4
LT
2332}
2333
2334static inline int skb_add_data(struct sk_buff *skb,
2335 char __user *from, int copy)
2336{
2337 const int off = skb->len;
2338
2339 if (skb->ip_summed == CHECKSUM_NONE) {
2340 int err = 0;
5084205f 2341 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2342 copy, 0, &err);
2343 if (!err) {
2344 skb->csum = csum_block_add(skb->csum, csum, off);
2345 return 0;
2346 }
2347 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2348 return 0;
2349
2350 __skb_trim(skb, off);
2351 return -EFAULT;
2352}
2353
38ba0a65
ED
2354static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2355 const struct page *page, int off)
1da177e4
LT
2356{
2357 if (i) {
9e903e08 2358 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2359
ea2ab693 2360 return page == skb_frag_page(frag) &&
9e903e08 2361 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2362 }
38ba0a65 2363 return false;
1da177e4
LT
2364}
2365
364c6bad
HX
2366static inline int __skb_linearize(struct sk_buff *skb)
2367{
2368 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2369}
2370
1da177e4
LT
2371/**
2372 * skb_linearize - convert paged skb to linear one
2373 * @skb: buffer to linarize
1da177e4
LT
2374 *
2375 * If there is no free memory -ENOMEM is returned, otherwise zero
2376 * is returned and the old skb data released.
2377 */
364c6bad
HX
2378static inline int skb_linearize(struct sk_buff *skb)
2379{
2380 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2381}
2382
cef401de
ED
2383/**
2384 * skb_has_shared_frag - can any frag be overwritten
2385 * @skb: buffer to test
2386 *
2387 * Return true if the skb has at least one frag that might be modified
2388 * by an external entity (as in vmsplice()/sendfile())
2389 */
2390static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2391{
c9af6db4
PS
2392 return skb_is_nonlinear(skb) &&
2393 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2394}
2395
364c6bad
HX
2396/**
2397 * skb_linearize_cow - make sure skb is linear and writable
2398 * @skb: buffer to process
2399 *
2400 * If there is no free memory -ENOMEM is returned, otherwise zero
2401 * is returned and the old skb data released.
2402 */
2403static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2404{
364c6bad
HX
2405 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2406 __skb_linearize(skb) : 0;
1da177e4
LT
2407}
2408
2409/**
2410 * skb_postpull_rcsum - update checksum for received skb after pull
2411 * @skb: buffer to update
2412 * @start: start of data before pull
2413 * @len: length of data pulled
2414 *
2415 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2416 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2417 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2418 */
2419
2420static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2421 const void *start, unsigned int len)
1da177e4 2422{
84fa7933 2423 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2424 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2425}
2426
cbb042f9
HX
2427unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2428
7ce5a27f
DM
2429/**
2430 * pskb_trim_rcsum - trim received skb and update checksum
2431 * @skb: buffer to trim
2432 * @len: new length
2433 *
2434 * This is exactly the same as pskb_trim except that it ensures the
2435 * checksum of received packets are still valid after the operation.
2436 */
2437
2438static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2439{
2440 if (likely(len >= skb->len))
2441 return 0;
2442 if (skb->ip_summed == CHECKSUM_COMPLETE)
2443 skb->ip_summed = CHECKSUM_NONE;
2444 return __pskb_trim(skb, len);
2445}
2446
1da177e4
LT
2447#define skb_queue_walk(queue, skb) \
2448 for (skb = (queue)->next; \
a1e4891f 2449 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2450 skb = skb->next)
2451
46f8914e
JC
2452#define skb_queue_walk_safe(queue, skb, tmp) \
2453 for (skb = (queue)->next, tmp = skb->next; \
2454 skb != (struct sk_buff *)(queue); \
2455 skb = tmp, tmp = skb->next)
2456
1164f52a 2457#define skb_queue_walk_from(queue, skb) \
a1e4891f 2458 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2459 skb = skb->next)
2460
2461#define skb_queue_walk_from_safe(queue, skb, tmp) \
2462 for (tmp = skb->next; \
2463 skb != (struct sk_buff *)(queue); \
2464 skb = tmp, tmp = skb->next)
2465
300ce174
SH
2466#define skb_queue_reverse_walk(queue, skb) \
2467 for (skb = (queue)->prev; \
a1e4891f 2468 skb != (struct sk_buff *)(queue); \
300ce174
SH
2469 skb = skb->prev)
2470
686a2955
DM
2471#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2472 for (skb = (queue)->prev, tmp = skb->prev; \
2473 skb != (struct sk_buff *)(queue); \
2474 skb = tmp, tmp = skb->prev)
2475
2476#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2477 for (tmp = skb->prev; \
2478 skb != (struct sk_buff *)(queue); \
2479 skb = tmp, tmp = skb->prev)
1da177e4 2480
21dc3301 2481static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2482{
2483 return skb_shinfo(skb)->frag_list != NULL;
2484}
2485
2486static inline void skb_frag_list_init(struct sk_buff *skb)
2487{
2488 skb_shinfo(skb)->frag_list = NULL;
2489}
2490
2491static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2492{
2493 frag->next = skb_shinfo(skb)->frag_list;
2494 skb_shinfo(skb)->frag_list = frag;
2495}
2496
2497#define skb_walk_frags(skb, iter) \
2498 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2499
7965bd4d
JP
2500struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2501 int *peeked, int *off, int *err);
2502struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2503 int *err);
2504unsigned int datagram_poll(struct file *file, struct socket *sock,
2505 struct poll_table_struct *wait);
2506int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2507 struct iovec *to, int size);
2508int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2509 struct iovec *iov);
2510int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2511 const struct iovec *from, int from_offset,
2512 int len);
2513int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2514 int offset, size_t count);
2515int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2516 const struct iovec *to, int to_offset,
2517 int size);
2518void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2519void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2520int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2521int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2522int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2523__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2524 int len, __wsum csum);
2525int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2526 struct pipe_inode_info *pipe, unsigned int len,
2527 unsigned int flags);
2528void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2529unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2530int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2531 int len, int hlen);
7965bd4d
JP
2532void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2533int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2534void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2535unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2536struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
20380731 2537
2817a336
DB
2538struct skb_checksum_ops {
2539 __wsum (*update)(const void *mem, int len, __wsum wsum);
2540 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2541};
2542
2543__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2544 __wsum csum, const struct skb_checksum_ops *ops);
2545__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2546 __wsum csum);
2547
1da177e4
LT
2548static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2549 int len, void *buffer)
2550{
2551 int hlen = skb_headlen(skb);
2552
55820ee2 2553 if (hlen - offset >= len)
1da177e4
LT
2554 return skb->data + offset;
2555
2556 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2557 return NULL;
2558
2559 return buffer;
2560}
2561
4262e5cc
DB
2562/**
2563 * skb_needs_linearize - check if we need to linearize a given skb
2564 * depending on the given device features.
2565 * @skb: socket buffer to check
2566 * @features: net device features
2567 *
2568 * Returns true if either:
2569 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2570 * 2. skb is fragmented and the device does not support SG.
2571 */
2572static inline bool skb_needs_linearize(struct sk_buff *skb,
2573 netdev_features_t features)
2574{
2575 return skb_is_nonlinear(skb) &&
2576 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2577 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2578}
2579
d626f62b
ACM
2580static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2581 void *to,
2582 const unsigned int len)
2583{
2584 memcpy(to, skb->data, len);
2585}
2586
2587static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2588 const int offset, void *to,
2589 const unsigned int len)
2590{
2591 memcpy(to, skb->data + offset, len);
2592}
2593
27d7ff46
ACM
2594static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2595 const void *from,
2596 const unsigned int len)
2597{
2598 memcpy(skb->data, from, len);
2599}
2600
2601static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2602 const int offset,
2603 const void *from,
2604 const unsigned int len)
2605{
2606 memcpy(skb->data + offset, from, len);
2607}
2608
7965bd4d 2609void skb_init(void);
1da177e4 2610
ac45f602
PO
2611static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2612{
2613 return skb->tstamp;
2614}
2615
a61bbcf2
PM
2616/**
2617 * skb_get_timestamp - get timestamp from a skb
2618 * @skb: skb to get stamp from
2619 * @stamp: pointer to struct timeval to store stamp in
2620 *
2621 * Timestamps are stored in the skb as offsets to a base timestamp.
2622 * This function converts the offset back to a struct timeval and stores
2623 * it in stamp.
2624 */
ac45f602
PO
2625static inline void skb_get_timestamp(const struct sk_buff *skb,
2626 struct timeval *stamp)
a61bbcf2 2627{
b7aa0bf7 2628 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2629}
2630
ac45f602
PO
2631static inline void skb_get_timestampns(const struct sk_buff *skb,
2632 struct timespec *stamp)
2633{
2634 *stamp = ktime_to_timespec(skb->tstamp);
2635}
2636
b7aa0bf7 2637static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2638{
b7aa0bf7 2639 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2640}
2641
164891aa
SH
2642static inline ktime_t net_timedelta(ktime_t t)
2643{
2644 return ktime_sub(ktime_get_real(), t);
2645}
2646
b9ce204f
IJ
2647static inline ktime_t net_invalid_timestamp(void)
2648{
2649 return ktime_set(0, 0);
2650}
a61bbcf2 2651
c1f19b51
RC
2652#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2653
7965bd4d
JP
2654void skb_clone_tx_timestamp(struct sk_buff *skb);
2655bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2656
2657#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2658
2659static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2660{
2661}
2662
2663static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2664{
2665 return false;
2666}
2667
2668#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2669
2670/**
2671 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2672 *
da92b194
RC
2673 * PHY drivers may accept clones of transmitted packets for
2674 * timestamping via their phy_driver.txtstamp method. These drivers
2675 * must call this function to return the skb back to the stack, with
2676 * or without a timestamp.
2677 *
c1f19b51 2678 * @skb: clone of the the original outgoing packet
da92b194 2679 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2680 *
2681 */
2682void skb_complete_tx_timestamp(struct sk_buff *skb,
2683 struct skb_shared_hwtstamps *hwtstamps);
2684
ac45f602
PO
2685/**
2686 * skb_tstamp_tx - queue clone of skb with send time stamps
2687 * @orig_skb: the original outgoing packet
2688 * @hwtstamps: hardware time stamps, may be NULL if not available
2689 *
2690 * If the skb has a socket associated, then this function clones the
2691 * skb (thus sharing the actual data and optional structures), stores
2692 * the optional hardware time stamping information (if non NULL) or
2693 * generates a software time stamp (otherwise), then queues the clone
2694 * to the error queue of the socket. Errors are silently ignored.
2695 */
7965bd4d
JP
2696void skb_tstamp_tx(struct sk_buff *orig_skb,
2697 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2698
4507a715
RC
2699static inline void sw_tx_timestamp(struct sk_buff *skb)
2700{
2244d07b
OH
2701 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2702 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2703 skb_tstamp_tx(skb, NULL);
2704}
2705
2706/**
2707 * skb_tx_timestamp() - Driver hook for transmit timestamping
2708 *
2709 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2710 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2711 *
73409f3b
DM
2712 * Specifically, one should make absolutely sure that this function is
2713 * called before TX completion of this packet can trigger. Otherwise
2714 * the packet could potentially already be freed.
2715 *
4507a715
RC
2716 * @skb: A socket buffer.
2717 */
2718static inline void skb_tx_timestamp(struct sk_buff *skb)
2719{
c1f19b51 2720 skb_clone_tx_timestamp(skb);
4507a715
RC
2721 sw_tx_timestamp(skb);
2722}
2723
6e3e939f
JB
2724/**
2725 * skb_complete_wifi_ack - deliver skb with wifi status
2726 *
2727 * @skb: the original outgoing packet
2728 * @acked: ack status
2729 *
2730 */
2731void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2732
7965bd4d
JP
2733__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2734__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2735
60476372
HX
2736static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2737{
2738 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2739}
2740
fb286bb2
HX
2741/**
2742 * skb_checksum_complete - Calculate checksum of an entire packet
2743 * @skb: packet to process
2744 *
2745 * This function calculates the checksum over the entire packet plus
2746 * the value of skb->csum. The latter can be used to supply the
2747 * checksum of a pseudo header as used by TCP/UDP. It returns the
2748 * checksum.
2749 *
2750 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2751 * this function can be used to verify that checksum on received
2752 * packets. In that case the function should return zero if the
2753 * checksum is correct. In particular, this function will return zero
2754 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2755 * hardware has already verified the correctness of the checksum.
2756 */
4381ca3c 2757static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2758{
60476372
HX
2759 return skb_csum_unnecessary(skb) ?
2760 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2761}
2762
76ba0aae
TH
2763/* Check if we need to perform checksum complete validation.
2764 *
2765 * Returns true if checksum complete is needed, false otherwise
2766 * (either checksum is unnecessary or zero checksum is allowed).
2767 */
2768static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2769 bool zero_okay,
2770 __sum16 check)
2771{
2772 if (skb_csum_unnecessary(skb)) {
2773 return false;
2774 } else if (zero_okay && !check) {
2775 skb->ip_summed = CHECKSUM_UNNECESSARY;
2776 return false;
2777 }
2778
2779 return true;
2780}
2781
2782/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2783 * in checksum_init.
2784 */
2785#define CHECKSUM_BREAK 76
2786
2787/* Validate (init) checksum based on checksum complete.
2788 *
2789 * Return values:
2790 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2791 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2792 * checksum is stored in skb->csum for use in __skb_checksum_complete
2793 * non-zero: value of invalid checksum
2794 *
2795 */
2796static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2797 bool complete,
2798 __wsum psum)
2799{
2800 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2801 if (!csum_fold(csum_add(psum, skb->csum))) {
2802 skb->ip_summed = CHECKSUM_UNNECESSARY;
2803 return 0;
2804 }
2805 }
2806
2807 skb->csum = psum;
2808
2809 if (complete || skb->len <= CHECKSUM_BREAK)
2810 return __skb_checksum_complete(skb);
2811
2812 return 0;
2813}
2814
2815static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2816{
2817 return 0;
2818}
2819
2820/* Perform checksum validate (init). Note that this is a macro since we only
2821 * want to calculate the pseudo header which is an input function if necessary.
2822 * First we try to validate without any computation (checksum unnecessary) and
2823 * then calculate based on checksum complete calling the function to compute
2824 * pseudo header.
2825 *
2826 * Return values:
2827 * 0: checksum is validated or try to in skb_checksum_complete
2828 * non-zero: value of invalid checksum
2829 */
2830#define __skb_checksum_validate(skb, proto, complete, \
2831 zero_okay, check, compute_pseudo) \
2832({ \
2833 __sum16 __ret = 0; \
2834 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2835 __ret = __skb_checksum_validate_complete(skb, \
2836 complete, compute_pseudo(skb, proto)); \
2837 __ret; \
2838})
2839
2840#define skb_checksum_init(skb, proto, compute_pseudo) \
2841 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2842
2843#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2844 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2845
2846#define skb_checksum_validate(skb, proto, compute_pseudo) \
2847 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2848
2849#define skb_checksum_validate_zero_check(skb, proto, check, \
2850 compute_pseudo) \
2851 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2852
2853#define skb_checksum_simple_validate(skb) \
2854 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2855
5f79e0f9 2856#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 2857void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
2858static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2859{
2860 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 2861 nf_conntrack_destroy(nfct);
1da177e4
LT
2862}
2863static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2864{
2865 if (nfct)
2866 atomic_inc(&nfct->use);
2867}
2fc72c7b 2868#endif
1da177e4
LT
2869#ifdef CONFIG_BRIDGE_NETFILTER
2870static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2871{
2872 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2873 kfree(nf_bridge);
2874}
2875static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2876{
2877 if (nf_bridge)
2878 atomic_inc(&nf_bridge->use);
2879}
2880#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
2881static inline void nf_reset(struct sk_buff *skb)
2882{
5f79e0f9 2883#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
2884 nf_conntrack_put(skb->nfct);
2885 skb->nfct = NULL;
2fc72c7b 2886#endif
a193a4ab
PM
2887#ifdef CONFIG_BRIDGE_NETFILTER
2888 nf_bridge_put(skb->nf_bridge);
2889 skb->nf_bridge = NULL;
2890#endif
2891}
2892
124dff01
PM
2893static inline void nf_reset_trace(struct sk_buff *skb)
2894{
478b360a 2895#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
2896 skb->nf_trace = 0;
2897#endif
a193a4ab
PM
2898}
2899
edda553c
YK
2900/* Note: This doesn't put any conntrack and bridge info in dst. */
2901static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2902{
5f79e0f9 2903#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
2904 dst->nfct = src->nfct;
2905 nf_conntrack_get(src->nfct);
2906 dst->nfctinfo = src->nfctinfo;
2fc72c7b 2907#endif
edda553c
YK
2908#ifdef CONFIG_BRIDGE_NETFILTER
2909 dst->nf_bridge = src->nf_bridge;
2910 nf_bridge_get(src->nf_bridge);
2911#endif
478b360a
FW
2912#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2913 dst->nf_trace = src->nf_trace;
2914#endif
edda553c
YK
2915}
2916
e7ac05f3
YK
2917static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2918{
e7ac05f3 2919#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 2920 nf_conntrack_put(dst->nfct);
2fc72c7b 2921#endif
e7ac05f3
YK
2922#ifdef CONFIG_BRIDGE_NETFILTER
2923 nf_bridge_put(dst->nf_bridge);
2924#endif
2925 __nf_copy(dst, src);
2926}
2927
984bc16c
JM
2928#ifdef CONFIG_NETWORK_SECMARK
2929static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2930{
2931 to->secmark = from->secmark;
2932}
2933
2934static inline void skb_init_secmark(struct sk_buff *skb)
2935{
2936 skb->secmark = 0;
2937}
2938#else
2939static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2940{ }
2941
2942static inline void skb_init_secmark(struct sk_buff *skb)
2943{ }
2944#endif
2945
574f7194
EB
2946static inline bool skb_irq_freeable(const struct sk_buff *skb)
2947{
2948 return !skb->destructor &&
2949#if IS_ENABLED(CONFIG_XFRM)
2950 !skb->sp &&
2951#endif
2952#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2953 !skb->nfct &&
2954#endif
2955 !skb->_skb_refdst &&
2956 !skb_has_frag_list(skb);
2957}
2958
f25f4e44
PWJ
2959static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2960{
f25f4e44 2961 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
2962}
2963
9247744e 2964static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 2965{
4e3ab47a 2966 return skb->queue_mapping;
4e3ab47a
PE
2967}
2968
f25f4e44
PWJ
2969static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2970{
f25f4e44 2971 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
2972}
2973
d5a9e24a
DM
2974static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2975{
2976 skb->queue_mapping = rx_queue + 1;
2977}
2978
9247744e 2979static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
2980{
2981 return skb->queue_mapping - 1;
2982}
2983
9247744e 2984static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 2985{
a02cec21 2986 return skb->queue_mapping != 0;
d5a9e24a
DM
2987}
2988
7965bd4d
JP
2989u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2990 unsigned int num_tx_queues);
9247744e 2991
def8b4fa
AD
2992static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2993{
0b3d8e08 2994#ifdef CONFIG_XFRM
def8b4fa 2995 return skb->sp;
def8b4fa 2996#else
def8b4fa 2997 return NULL;
def8b4fa 2998#endif
0b3d8e08 2999}
def8b4fa 3000
68c33163
PS
3001/* Keeps track of mac header offset relative to skb->head.
3002 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3003 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3004 * tunnel skb it points to outer mac header.
3005 * Keeps track of level of encapsulation of network headers.
3006 */
68c33163 3007struct skb_gso_cb {
3347c960
ED
3008 int mac_offset;
3009 int encap_level;
7e2b10c1 3010 __u16 csum_start;
68c33163
PS
3011};
3012#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3013
3014static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3015{
3016 return (skb_mac_header(inner_skb) - inner_skb->head) -
3017 SKB_GSO_CB(inner_skb)->mac_offset;
3018}
3019
1e2bd517
PS
3020static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3021{
3022 int new_headroom, headroom;
3023 int ret;
3024
3025 headroom = skb_headroom(skb);
3026 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3027 if (ret)
3028 return ret;
3029
3030 new_headroom = skb_headroom(skb);
3031 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3032 return 0;
3033}
3034
7e2b10c1
TH
3035/* Compute the checksum for a gso segment. First compute the checksum value
3036 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3037 * then add in skb->csum (checksum from csum_start to end of packet).
3038 * skb->csum and csum_start are then updated to reflect the checksum of the
3039 * resultant packet starting from the transport header-- the resultant checksum
3040 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3041 * header.
3042 */
3043static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3044{
3045 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3046 skb_transport_offset(skb);
3047 __u16 csum;
3048
3049 csum = csum_fold(csum_partial(skb_transport_header(skb),
3050 plen, skb->csum));
3051 skb->csum = res;
3052 SKB_GSO_CB(skb)->csum_start -= plen;
3053
3054 return csum;
3055}
3056
bdcc0924 3057static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3058{
3059 return skb_shinfo(skb)->gso_size;
3060}
3061
36a8f39e 3062/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3063static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3064{
3065 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3066}
3067
7965bd4d 3068void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3069
3070static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3071{
3072 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3073 * wanted then gso_type will be set. */
05bdd2f1
ED
3074 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3075
b78462eb
AD
3076 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3077 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3078 __skb_warn_lro_forwarding(skb);
3079 return true;
3080 }
3081 return false;
3082}
3083
35fc92a9
HX
3084static inline void skb_forward_csum(struct sk_buff *skb)
3085{
3086 /* Unfortunately we don't support this one. Any brave souls? */
3087 if (skb->ip_summed == CHECKSUM_COMPLETE)
3088 skb->ip_summed = CHECKSUM_NONE;
3089}
3090
bc8acf2c
ED
3091/**
3092 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3093 * @skb: skb to check
3094 *
3095 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3096 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3097 * use this helper, to document places where we make this assertion.
3098 */
05bdd2f1 3099static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3100{
3101#ifdef DEBUG
3102 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3103#endif
3104}
3105
f35d9d8a 3106bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3107
ed1f50c3
PD
3108int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3109
f77668dc
DB
3110u32 __skb_get_poff(const struct sk_buff *skb);
3111
3a7c1ee4
AD
3112/**
3113 * skb_head_is_locked - Determine if the skb->head is locked down
3114 * @skb: skb to check
3115 *
3116 * The head on skbs build around a head frag can be removed if they are
3117 * not cloned. This function returns true if the skb head is locked down
3118 * due to either being allocated via kmalloc, or by being a clone with
3119 * multiple references to the head.
3120 */
3121static inline bool skb_head_is_locked(const struct sk_buff *skb)
3122{
3123 return !skb->head_frag || skb_cloned(skb);
3124}
fe6cc55f
FW
3125
3126/**
3127 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3128 *
3129 * @skb: GSO skb
3130 *
3131 * skb_gso_network_seglen is used to determine the real size of the
3132 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3133 *
3134 * The MAC/L2 header is not accounted for.
3135 */
3136static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3137{
3138 unsigned int hdr_len = skb_transport_header(skb) -
3139 skb_network_header(skb);
3140 return hdr_len + skb_gso_transport_seglen(skb);
3141}
1da177e4
LT
3142#endif /* __KERNEL__ */
3143#endif /* _LINUX_SKBUFF_H */
This page took 1.891817 seconds and 5 git commands to generate.