l2tp: call udp{6}_set_csum
[linux-2.6-block.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,
7967168c
HX
348};
349
2e07fa9c
ACM
350#if BITS_PER_LONG > 32
351#define NET_SKBUFF_DATA_USES_OFFSET 1
352#endif
353
354#ifdef NET_SKBUFF_DATA_USES_OFFSET
355typedef unsigned int sk_buff_data_t;
356#else
357typedef unsigned char *sk_buff_data_t;
358#endif
359
363ec392
ED
360/**
361 * struct skb_mstamp - multi resolution time stamps
362 * @stamp_us: timestamp in us resolution
363 * @stamp_jiffies: timestamp in jiffies
364 */
365struct skb_mstamp {
366 union {
367 u64 v64;
368 struct {
369 u32 stamp_us;
370 u32 stamp_jiffies;
371 };
372 };
373};
374
375/**
376 * skb_mstamp_get - get current timestamp
377 * @cl: place to store timestamps
378 */
379static inline void skb_mstamp_get(struct skb_mstamp *cl)
380{
381 u64 val = local_clock();
382
383 do_div(val, NSEC_PER_USEC);
384 cl->stamp_us = (u32)val;
385 cl->stamp_jiffies = (u32)jiffies;
386}
387
388/**
389 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
390 * @t1: pointer to newest sample
391 * @t0: pointer to oldest sample
392 */
393static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
394 const struct skb_mstamp *t0)
395{
396 s32 delta_us = t1->stamp_us - t0->stamp_us;
397 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
398
399 /* If delta_us is negative, this might be because interval is too big,
400 * or local_clock() drift is too big : fallback using jiffies.
401 */
402 if (delta_us <= 0 ||
403 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
404
405 delta_us = jiffies_to_usecs(delta_jiffies);
406
407 return delta_us;
408}
409
410
1da177e4
LT
411/**
412 * struct sk_buff - socket buffer
413 * @next: Next buffer in list
414 * @prev: Previous buffer in list
363ec392 415 * @tstamp: Time we arrived/left
d84e0bd7 416 * @sk: Socket we are owned by
1da177e4 417 * @dev: Device we arrived on/are leaving by
d84e0bd7 418 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 419 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 420 * @sp: the security path, used for xfrm
1da177e4
LT
421 * @len: Length of actual data
422 * @data_len: Data length
423 * @mac_len: Length of link layer header
334a8132 424 * @hdr_len: writable header length of cloned skb
663ead3b
HX
425 * @csum: Checksum (must include start/offset pair)
426 * @csum_start: Offset from skb->head where checksumming should start
427 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 428 * @priority: Packet queueing priority
60ff7467 429 * @ignore_df: allow local fragmentation
1da177e4 430 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 431 * @ip_summed: Driver fed us an IP checksum
1da177e4 432 * @nohdr: Payload reference only, must not modify header
d84e0bd7 433 * @nfctinfo: Relationship of this skb to the connection
1da177e4 434 * @pkt_type: Packet class
c83c2486 435 * @fclone: skbuff clone status
c83c2486 436 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
437 * @peeked: this packet has been seen already, so stats have been
438 * done for it, don't do them again
ba9dda3a 439 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
440 * @protocol: Packet protocol from driver
441 * @destructor: Destruct function
442 * @nfct: Associated connection, if any
1da177e4 443 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 444 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
445 * @tc_index: Traffic control index
446 * @tc_verd: traffic control verdict
61b905da 447 * @hash: the packet hash
d84e0bd7 448 * @queue_mapping: Queue mapping for multiqueue devices
553a5672 449 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 450 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 451 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 452 * ports.
6e3e939f
JB
453 * @wifi_acked_valid: wifi_acked was set
454 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 455 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
f4b8ea78
RD
456 * @dma_cookie: a cookie to one of several possible DMA operations
457 * done by skb DMA functions
06021292 458 * @napi_id: id of the NAPI struct this skb came from
984bc16c 459 * @secmark: security marking
d84e0bd7
DB
460 * @mark: Generic packet mark
461 * @dropcount: total number of sk_receive_queue overflows
86a9bad3 462 * @vlan_proto: vlan encapsulation protocol
6aa895b0 463 * @vlan_tci: vlan tag control information
0d89d203 464 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
465 * @inner_transport_header: Inner transport layer header (encapsulation)
466 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 467 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
468 * @transport_header: Transport layer header
469 * @network_header: Network layer header
470 * @mac_header: Link layer header
471 * @tail: Tail pointer
472 * @end: End pointer
473 * @head: Head of buffer
474 * @data: Data head pointer
475 * @truesize: Buffer size
476 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
477 */
478
479struct sk_buff {
480 /* These two members must be first. */
481 struct sk_buff *next;
482 struct sk_buff *prev;
483
363ec392
ED
484 union {
485 ktime_t tstamp;
486 struct skb_mstamp skb_mstamp;
487 };
da3f5cf1
FF
488
489 struct sock *sk;
1da177e4 490 struct net_device *dev;
1da177e4 491
1da177e4
LT
492 /*
493 * This is the control buffer. It is free to use for every
494 * layer. Please put your private variables there. If you
495 * want to keep them across layers you have to do a skb_clone()
496 * first. This is owned by whoever has the skb queued ATM.
497 */
da3f5cf1 498 char cb[48] __aligned(8);
1da177e4 499
7fee226a 500 unsigned long _skb_refdst;
da3f5cf1
FF
501#ifdef CONFIG_XFRM
502 struct sec_path *sp;
503#endif
1da177e4 504 unsigned int len,
334a8132
PM
505 data_len;
506 __u16 mac_len,
507 hdr_len;
ff1dcadb
AV
508 union {
509 __wsum csum;
663ead3b
HX
510 struct {
511 __u16 csum_start;
512 __u16 csum_offset;
513 };
ff1dcadb 514 };
1da177e4 515 __u32 priority;
fe55f6d5 516 kmemcheck_bitfield_begin(flags1);
60ff7467 517 __u8 ignore_df:1,
1cbb3380
TG
518 cloned:1,
519 ip_summed:2,
6869c4d8
HW
520 nohdr:1,
521 nfctinfo:3;
d179cd12 522 __u8 pkt_type:3,
b84f4cc9 523 fclone:2,
ba9dda3a 524 ipvs_property:1,
a59322be 525 peeked:1,
ba9dda3a 526 nf_trace:1;
fe55f6d5 527 kmemcheck_bitfield_end(flags1);
4ab408de 528 __be16 protocol;
1da177e4
LT
529
530 void (*destructor)(struct sk_buff *skb);
9fb9cbb1 531#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 532 struct nf_conntrack *nfct;
2fc72c7b 533#endif
1da177e4
LT
534#ifdef CONFIG_BRIDGE_NETFILTER
535 struct nf_bridge_info *nf_bridge;
536#endif
f25f4e44 537
8964be4a 538 int skb_iif;
4031ae6e 539
61b905da 540 __u32 hash;
4031ae6e 541
86a9bad3 542 __be16 vlan_proto;
4031ae6e
AD
543 __u16 vlan_tci;
544
1da177e4 545#ifdef CONFIG_NET_SCHED
b6b99eb5 546 __u16 tc_index; /* traffic control index */
1da177e4 547#ifdef CONFIG_NET_CLS_ACT
b6b99eb5 548 __u16 tc_verd; /* traffic control verdict */
1da177e4 549#endif
1da177e4 550#endif
fe55f6d5 551
0a14842f 552 __u16 queue_mapping;
fe55f6d5 553 kmemcheck_bitfield_begin(flags2);
de357cc0 554#ifdef CONFIG_IPV6_NDISC_NODETYPE
8a4eb573 555 __u8 ndisc_nodetype:2;
d0f09804 556#endif
c93bdd0e 557 __u8 pfmemalloc:1;
3853b584 558 __u8 ooo_okay:1;
61b905da 559 __u8 l4_hash:1;
6e3e939f
JB
560 __u8 wifi_acked_valid:1;
561 __u8 wifi_acked:1;
3bdc0eba 562 __u8 no_fcs:1;
d3836f21 563 __u8 head_frag:1;
6a674e9c
JG
564 /* Encapsulation protocol and NIC drivers should use
565 * this flag to indicate to each other if the skb contains
566 * encapsulated packet or not and maybe use the inner packet
567 * headers if needed
568 */
569 __u8 encapsulation:1;
45906723 570 /* 6/8 bit hole (depending on ndisc_nodetype presence) */
fe55f6d5
VN
571 kmemcheck_bitfield_end(flags2);
572
e0d1095a 573#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
06021292
ET
574 union {
575 unsigned int napi_id;
576 dma_cookie_t dma_cookie;
577 };
97fc2f08 578#endif
984bc16c
JM
579#ifdef CONFIG_NETWORK_SECMARK
580 __u32 secmark;
581#endif
3b885787
NH
582 union {
583 __u32 mark;
584 __u32 dropcount;
16fad69c 585 __u32 reserved_tailroom;
3b885787 586 };
1da177e4 587
0d89d203 588 __be16 inner_protocol;
1a37e412
SH
589 __u16 inner_transport_header;
590 __u16 inner_network_header;
591 __u16 inner_mac_header;
592 __u16 transport_header;
593 __u16 network_header;
594 __u16 mac_header;
1da177e4 595 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 596 sk_buff_data_t tail;
4305b541 597 sk_buff_data_t end;
1da177e4 598 unsigned char *head,
4305b541 599 *data;
27a884dc
ACM
600 unsigned int truesize;
601 atomic_t users;
1da177e4
LT
602};
603
604#ifdef __KERNEL__
605/*
606 * Handling routines are only of interest to the kernel
607 */
608#include <linux/slab.h>
609
1da177e4 610
c93bdd0e
MG
611#define SKB_ALLOC_FCLONE 0x01
612#define SKB_ALLOC_RX 0x02
613
614/* Returns true if the skb was allocated from PFMEMALLOC reserves */
615static inline bool skb_pfmemalloc(const struct sk_buff *skb)
616{
617 return unlikely(skb->pfmemalloc);
618}
619
7fee226a
ED
620/*
621 * skb might have a dst pointer attached, refcounted or not.
622 * _skb_refdst low order bit is set if refcount was _not_ taken
623 */
624#define SKB_DST_NOREF 1UL
625#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
626
627/**
628 * skb_dst - returns skb dst_entry
629 * @skb: buffer
630 *
631 * Returns skb dst_entry, regardless of reference taken or not.
632 */
adf30907
ED
633static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
634{
7fee226a
ED
635 /* If refdst was not refcounted, check we still are in a
636 * rcu_read_lock section
637 */
638 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
639 !rcu_read_lock_held() &&
640 !rcu_read_lock_bh_held());
641 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
642}
643
7fee226a
ED
644/**
645 * skb_dst_set - sets skb dst
646 * @skb: buffer
647 * @dst: dst entry
648 *
649 * Sets skb dst, assuming a reference was taken on dst and should
650 * be released by skb_dst_drop()
651 */
adf30907
ED
652static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
653{
7fee226a
ED
654 skb->_skb_refdst = (unsigned long)dst;
655}
656
7965bd4d
JP
657void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
658 bool force);
932bc4d7
JA
659
660/**
661 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
662 * @skb: buffer
663 * @dst: dst entry
664 *
665 * Sets skb dst, assuming a reference was not taken on dst.
666 * If dst entry is cached, we do not take reference and dst_release
667 * will be avoided by refdst_drop. If dst entry is not cached, we take
668 * reference, so that last dst_release can destroy the dst immediately.
669 */
670static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
671{
672 __skb_dst_set_noref(skb, dst, false);
673}
674
675/**
676 * skb_dst_set_noref_force - sets skb dst, without taking reference
677 * @skb: buffer
678 * @dst: dst entry
679 *
680 * Sets skb dst, assuming a reference was not taken on dst.
681 * No reference is taken and no dst_release will be called. While for
682 * cached dsts deferred reclaim is a basic feature, for entries that are
683 * not cached it is caller's job to guarantee that last dst_release for
684 * provided dst happens when nobody uses it, eg. after a RCU grace period.
685 */
686static inline void skb_dst_set_noref_force(struct sk_buff *skb,
687 struct dst_entry *dst)
688{
689 __skb_dst_set_noref(skb, dst, true);
690}
7fee226a
ED
691
692/**
25985edc 693 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
694 * @skb: buffer
695 */
696static inline bool skb_dst_is_noref(const struct sk_buff *skb)
697{
698 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
699}
700
511c3f92
ED
701static inline struct rtable *skb_rtable(const struct sk_buff *skb)
702{
adf30907 703 return (struct rtable *)skb_dst(skb);
511c3f92
ED
704}
705
7965bd4d
JP
706void kfree_skb(struct sk_buff *skb);
707void kfree_skb_list(struct sk_buff *segs);
708void skb_tx_error(struct sk_buff *skb);
709void consume_skb(struct sk_buff *skb);
710void __kfree_skb(struct sk_buff *skb);
d7e8883c 711extern struct kmem_cache *skbuff_head_cache;
bad43ca8 712
7965bd4d
JP
713void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
714bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
715 bool *fragstolen, int *delta_truesize);
bad43ca8 716
7965bd4d
JP
717struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
718 int node);
719struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 720static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 721 gfp_t priority)
d179cd12 722{
564824b0 723 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
724}
725
726static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 727 gfp_t priority)
d179cd12 728{
c93bdd0e 729 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
730}
731
7965bd4d 732struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
733static inline struct sk_buff *alloc_skb_head(gfp_t priority)
734{
735 return __alloc_skb_head(priority, -1);
736}
737
7965bd4d
JP
738struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
739int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
740struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
741struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
742struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
743
744int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
745struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
746 unsigned int headroom);
747struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
748 int newtailroom, gfp_t priority);
25a91d8d
FD
749int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
750 int offset, int len);
7965bd4d
JP
751int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
752 int len);
753int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
754int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 755#define dev_kfree_skb(a) consume_skb(a)
1da177e4 756
7965bd4d
JP
757int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
758 int getfrag(void *from, char *to, int offset,
759 int len, int odd, struct sk_buff *skb),
760 void *from, int length);
e89e9cf5 761
d94d9fee 762struct skb_seq_state {
677e90ed
TG
763 __u32 lower_offset;
764 __u32 upper_offset;
765 __u32 frag_idx;
766 __u32 stepped_offset;
767 struct sk_buff *root_skb;
768 struct sk_buff *cur_skb;
769 __u8 *frag_data;
770};
771
7965bd4d
JP
772void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
773 unsigned int to, struct skb_seq_state *st);
774unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
775 struct skb_seq_state *st);
776void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 777
7965bd4d
JP
778unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
779 unsigned int to, struct ts_config *config,
780 struct ts_state *state);
3fc7e8a6 781
09323cc4
TH
782/*
783 * Packet hash types specify the type of hash in skb_set_hash.
784 *
785 * Hash types refer to the protocol layer addresses which are used to
786 * construct a packet's hash. The hashes are used to differentiate or identify
787 * flows of the protocol layer for the hash type. Hash types are either
788 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
789 *
790 * Properties of hashes:
791 *
792 * 1) Two packets in different flows have different hash values
793 * 2) Two packets in the same flow should have the same hash value
794 *
795 * A hash at a higher layer is considered to be more specific. A driver should
796 * set the most specific hash possible.
797 *
798 * A driver cannot indicate a more specific hash than the layer at which a hash
799 * was computed. For instance an L3 hash cannot be set as an L4 hash.
800 *
801 * A driver may indicate a hash level which is less specific than the
802 * actual layer the hash was computed on. For instance, a hash computed
803 * at L4 may be considered an L3 hash. This should only be done if the
804 * driver can't unambiguously determine that the HW computed the hash at
805 * the higher layer. Note that the "should" in the second property above
806 * permits this.
807 */
808enum pkt_hash_types {
809 PKT_HASH_TYPE_NONE, /* Undefined type */
810 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
811 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
812 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
813};
814
815static inline void
816skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
817{
61b905da
TH
818 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
819 skb->hash = hash;
09323cc4
TH
820}
821
3958afa1
TH
822void __skb_get_hash(struct sk_buff *skb);
823static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 824{
61b905da 825 if (!skb->l4_hash)
3958afa1 826 __skb_get_hash(skb);
bfb564e7 827
61b905da 828 return skb->hash;
bfb564e7
KK
829}
830
57bdf7f4
TH
831static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
832{
61b905da 833 return skb->hash;
57bdf7f4
TH
834}
835
7539fadc
TH
836static inline void skb_clear_hash(struct sk_buff *skb)
837{
61b905da
TH
838 skb->hash = 0;
839 skb->l4_hash = 0;
7539fadc
TH
840}
841
842static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
843{
61b905da 844 if (!skb->l4_hash)
7539fadc
TH
845 skb_clear_hash(skb);
846}
847
3df7a74e
TH
848static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
849{
61b905da
TH
850 to->hash = from->hash;
851 to->l4_hash = from->l4_hash;
3df7a74e
TH
852};
853
4305b541
ACM
854#ifdef NET_SKBUFF_DATA_USES_OFFSET
855static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
856{
857 return skb->head + skb->end;
858}
ec47ea82
AD
859
860static inline unsigned int skb_end_offset(const struct sk_buff *skb)
861{
862 return skb->end;
863}
4305b541
ACM
864#else
865static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
866{
867 return skb->end;
868}
ec47ea82
AD
869
870static inline unsigned int skb_end_offset(const struct sk_buff *skb)
871{
872 return skb->end - skb->head;
873}
4305b541
ACM
874#endif
875
1da177e4 876/* Internal */
4305b541 877#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 878
ac45f602
PO
879static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
880{
881 return &skb_shinfo(skb)->hwtstamps;
882}
883
1da177e4
LT
884/**
885 * skb_queue_empty - check if a queue is empty
886 * @list: queue head
887 *
888 * Returns true if the queue is empty, false otherwise.
889 */
890static inline int skb_queue_empty(const struct sk_buff_head *list)
891{
fd44b93c 892 return list->next == (const struct sk_buff *) list;
1da177e4
LT
893}
894
fc7ebb21
DM
895/**
896 * skb_queue_is_last - check if skb is the last entry in the queue
897 * @list: queue head
898 * @skb: buffer
899 *
900 * Returns true if @skb is the last buffer on the list.
901 */
902static inline bool skb_queue_is_last(const struct sk_buff_head *list,
903 const struct sk_buff *skb)
904{
fd44b93c 905 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
906}
907
832d11c5
IJ
908/**
909 * skb_queue_is_first - check if skb is the first entry in the queue
910 * @list: queue head
911 * @skb: buffer
912 *
913 * Returns true if @skb is the first buffer on the list.
914 */
915static inline bool skb_queue_is_first(const struct sk_buff_head *list,
916 const struct sk_buff *skb)
917{
fd44b93c 918 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
919}
920
249c8b42
DM
921/**
922 * skb_queue_next - return the next packet in the queue
923 * @list: queue head
924 * @skb: current buffer
925 *
926 * Return the next packet in @list after @skb. It is only valid to
927 * call this if skb_queue_is_last() evaluates to false.
928 */
929static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
930 const struct sk_buff *skb)
931{
932 /* This BUG_ON may seem severe, but if we just return then we
933 * are going to dereference garbage.
934 */
935 BUG_ON(skb_queue_is_last(list, skb));
936 return skb->next;
937}
938
832d11c5
IJ
939/**
940 * skb_queue_prev - return the prev packet in the queue
941 * @list: queue head
942 * @skb: current buffer
943 *
944 * Return the prev packet in @list before @skb. It is only valid to
945 * call this if skb_queue_is_first() evaluates to false.
946 */
947static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
948 const struct sk_buff *skb)
949{
950 /* This BUG_ON may seem severe, but if we just return then we
951 * are going to dereference garbage.
952 */
953 BUG_ON(skb_queue_is_first(list, skb));
954 return skb->prev;
955}
956
1da177e4
LT
957/**
958 * skb_get - reference buffer
959 * @skb: buffer to reference
960 *
961 * Makes another reference to a socket buffer and returns a pointer
962 * to the buffer.
963 */
964static inline struct sk_buff *skb_get(struct sk_buff *skb)
965{
966 atomic_inc(&skb->users);
967 return skb;
968}
969
970/*
971 * If users == 1, we are the only owner and are can avoid redundant
972 * atomic change.
973 */
974
1da177e4
LT
975/**
976 * skb_cloned - is the buffer a clone
977 * @skb: buffer to check
978 *
979 * Returns true if the buffer was generated with skb_clone() and is
980 * one of multiple shared copies of the buffer. Cloned buffers are
981 * shared data so must not be written to under normal circumstances.
982 */
983static inline int skb_cloned(const struct sk_buff *skb)
984{
985 return skb->cloned &&
986 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
987}
988
14bbd6a5
PS
989static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
990{
991 might_sleep_if(pri & __GFP_WAIT);
992
993 if (skb_cloned(skb))
994 return pskb_expand_head(skb, 0, 0, pri);
995
996 return 0;
997}
998
1da177e4
LT
999/**
1000 * skb_header_cloned - is the header a clone
1001 * @skb: buffer to check
1002 *
1003 * Returns true if modifying the header part of the buffer requires
1004 * the data to be copied.
1005 */
1006static inline int skb_header_cloned(const struct sk_buff *skb)
1007{
1008 int dataref;
1009
1010 if (!skb->cloned)
1011 return 0;
1012
1013 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1014 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1015 return dataref != 1;
1016}
1017
1018/**
1019 * skb_header_release - release reference to header
1020 * @skb: buffer to operate on
1021 *
1022 * Drop a reference to the header part of the buffer. This is done
1023 * by acquiring a payload reference. You must not read from the header
1024 * part of skb->data after this.
1025 */
1026static inline void skb_header_release(struct sk_buff *skb)
1027{
1028 BUG_ON(skb->nohdr);
1029 skb->nohdr = 1;
1030 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1031}
1032
1033/**
1034 * skb_shared - is the buffer shared
1035 * @skb: buffer to check
1036 *
1037 * Returns true if more than one person has a reference to this
1038 * buffer.
1039 */
1040static inline int skb_shared(const struct sk_buff *skb)
1041{
1042 return atomic_read(&skb->users) != 1;
1043}
1044
1045/**
1046 * skb_share_check - check if buffer is shared and if so clone it
1047 * @skb: buffer to check
1048 * @pri: priority for memory allocation
1049 *
1050 * If the buffer is shared the buffer is cloned and the old copy
1051 * drops a reference. A new clone with a single reference is returned.
1052 * If the buffer is not shared the original buffer is returned. When
1053 * being called from interrupt status or with spinlocks held pri must
1054 * be GFP_ATOMIC.
1055 *
1056 * NULL is returned on a memory allocation failure.
1057 */
47061bc4 1058static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1059{
1060 might_sleep_if(pri & __GFP_WAIT);
1061 if (skb_shared(skb)) {
1062 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1063
1064 if (likely(nskb))
1065 consume_skb(skb);
1066 else
1067 kfree_skb(skb);
1da177e4
LT
1068 skb = nskb;
1069 }
1070 return skb;
1071}
1072
1073/*
1074 * Copy shared buffers into a new sk_buff. We effectively do COW on
1075 * packets to handle cases where we have a local reader and forward
1076 * and a couple of other messy ones. The normal one is tcpdumping
1077 * a packet thats being forwarded.
1078 */
1079
1080/**
1081 * skb_unshare - make a copy of a shared buffer
1082 * @skb: buffer to check
1083 * @pri: priority for memory allocation
1084 *
1085 * If the socket buffer is a clone then this function creates a new
1086 * copy of the data, drops a reference count on the old copy and returns
1087 * the new copy with the reference count at 1. If the buffer is not a clone
1088 * the original buffer is returned. When called with a spinlock held or
1089 * from interrupt state @pri must be %GFP_ATOMIC
1090 *
1091 * %NULL is returned on a memory allocation failure.
1092 */
e2bf521d 1093static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1094 gfp_t pri)
1da177e4
LT
1095{
1096 might_sleep_if(pri & __GFP_WAIT);
1097 if (skb_cloned(skb)) {
1098 struct sk_buff *nskb = skb_copy(skb, pri);
1099 kfree_skb(skb); /* Free our shared copy */
1100 skb = nskb;
1101 }
1102 return skb;
1103}
1104
1105/**
1a5778aa 1106 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1107 * @list_: list to peek at
1108 *
1109 * Peek an &sk_buff. Unlike most other operations you _MUST_
1110 * be careful with this one. A peek leaves the buffer on the
1111 * list and someone else may run off with it. You must hold
1112 * the appropriate locks or have a private queue to do this.
1113 *
1114 * Returns %NULL for an empty list or a pointer to the head element.
1115 * The reference count is not incremented and the reference is therefore
1116 * volatile. Use with caution.
1117 */
05bdd2f1 1118static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1119{
18d07000
ED
1120 struct sk_buff *skb = list_->next;
1121
1122 if (skb == (struct sk_buff *)list_)
1123 skb = NULL;
1124 return skb;
1da177e4
LT
1125}
1126
da5ef6e5
PE
1127/**
1128 * skb_peek_next - peek skb following the given one from a queue
1129 * @skb: skb to start from
1130 * @list_: list to peek at
1131 *
1132 * Returns %NULL when the end of the list is met or a pointer to the
1133 * next element. The reference count is not incremented and the
1134 * reference is therefore volatile. Use with caution.
1135 */
1136static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1137 const struct sk_buff_head *list_)
1138{
1139 struct sk_buff *next = skb->next;
18d07000 1140
da5ef6e5
PE
1141 if (next == (struct sk_buff *)list_)
1142 next = NULL;
1143 return next;
1144}
1145
1da177e4 1146/**
1a5778aa 1147 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1148 * @list_: list to peek at
1149 *
1150 * Peek an &sk_buff. Unlike most other operations you _MUST_
1151 * be careful with this one. A peek leaves the buffer on the
1152 * list and someone else may run off with it. You must hold
1153 * the appropriate locks or have a private queue to do this.
1154 *
1155 * Returns %NULL for an empty list or a pointer to the tail element.
1156 * The reference count is not incremented and the reference is therefore
1157 * volatile. Use with caution.
1158 */
05bdd2f1 1159static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1160{
18d07000
ED
1161 struct sk_buff *skb = list_->prev;
1162
1163 if (skb == (struct sk_buff *)list_)
1164 skb = NULL;
1165 return skb;
1166
1da177e4
LT
1167}
1168
1169/**
1170 * skb_queue_len - get queue length
1171 * @list_: list to measure
1172 *
1173 * Return the length of an &sk_buff queue.
1174 */
1175static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1176{
1177 return list_->qlen;
1178}
1179
67fed459
DM
1180/**
1181 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1182 * @list: queue to initialize
1183 *
1184 * This initializes only the list and queue length aspects of
1185 * an sk_buff_head object. This allows to initialize the list
1186 * aspects of an sk_buff_head without reinitializing things like
1187 * the spinlock. It can also be used for on-stack sk_buff_head
1188 * objects where the spinlock is known to not be used.
1189 */
1190static inline void __skb_queue_head_init(struct sk_buff_head *list)
1191{
1192 list->prev = list->next = (struct sk_buff *)list;
1193 list->qlen = 0;
1194}
1195
76f10ad0
AV
1196/*
1197 * This function creates a split out lock class for each invocation;
1198 * this is needed for now since a whole lot of users of the skb-queue
1199 * infrastructure in drivers have different locking usage (in hardirq)
1200 * than the networking core (in softirq only). In the long run either the
1201 * network layer or drivers should need annotation to consolidate the
1202 * main types of usage into 3 classes.
1203 */
1da177e4
LT
1204static inline void skb_queue_head_init(struct sk_buff_head *list)
1205{
1206 spin_lock_init(&list->lock);
67fed459 1207 __skb_queue_head_init(list);
1da177e4
LT
1208}
1209
c2ecba71
PE
1210static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1211 struct lock_class_key *class)
1212{
1213 skb_queue_head_init(list);
1214 lockdep_set_class(&list->lock, class);
1215}
1216
1da177e4 1217/*
bf299275 1218 * Insert an sk_buff on a list.
1da177e4
LT
1219 *
1220 * The "__skb_xxxx()" functions are the non-atomic ones that
1221 * can only be called with interrupts disabled.
1222 */
7965bd4d
JP
1223void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1224 struct sk_buff_head *list);
bf299275
GR
1225static inline void __skb_insert(struct sk_buff *newsk,
1226 struct sk_buff *prev, struct sk_buff *next,
1227 struct sk_buff_head *list)
1228{
1229 newsk->next = next;
1230 newsk->prev = prev;
1231 next->prev = prev->next = newsk;
1232 list->qlen++;
1233}
1da177e4 1234
67fed459
DM
1235static inline void __skb_queue_splice(const struct sk_buff_head *list,
1236 struct sk_buff *prev,
1237 struct sk_buff *next)
1238{
1239 struct sk_buff *first = list->next;
1240 struct sk_buff *last = list->prev;
1241
1242 first->prev = prev;
1243 prev->next = first;
1244
1245 last->next = next;
1246 next->prev = last;
1247}
1248
1249/**
1250 * skb_queue_splice - join two skb lists, this is designed for stacks
1251 * @list: the new list to add
1252 * @head: the place to add it in the first list
1253 */
1254static inline void skb_queue_splice(const struct sk_buff_head *list,
1255 struct sk_buff_head *head)
1256{
1257 if (!skb_queue_empty(list)) {
1258 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1259 head->qlen += list->qlen;
67fed459
DM
1260 }
1261}
1262
1263/**
d9619496 1264 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1265 * @list: the new list to add
1266 * @head: the place to add it in the first list
1267 *
1268 * The list at @list is reinitialised
1269 */
1270static inline void skb_queue_splice_init(struct sk_buff_head *list,
1271 struct sk_buff_head *head)
1272{
1273 if (!skb_queue_empty(list)) {
1274 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1275 head->qlen += list->qlen;
67fed459
DM
1276 __skb_queue_head_init(list);
1277 }
1278}
1279
1280/**
1281 * skb_queue_splice_tail - join two skb lists, each list being a queue
1282 * @list: the new list to add
1283 * @head: the place to add it in the first list
1284 */
1285static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1286 struct sk_buff_head *head)
1287{
1288 if (!skb_queue_empty(list)) {
1289 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1290 head->qlen += list->qlen;
67fed459
DM
1291 }
1292}
1293
1294/**
d9619496 1295 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1296 * @list: the new list to add
1297 * @head: the place to add it in the first list
1298 *
1299 * Each of the lists is a queue.
1300 * The list at @list is reinitialised
1301 */
1302static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1303 struct sk_buff_head *head)
1304{
1305 if (!skb_queue_empty(list)) {
1306 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1307 head->qlen += list->qlen;
67fed459
DM
1308 __skb_queue_head_init(list);
1309 }
1310}
1311
1da177e4 1312/**
300ce174 1313 * __skb_queue_after - queue a buffer at the list head
1da177e4 1314 * @list: list to use
300ce174 1315 * @prev: place after this buffer
1da177e4
LT
1316 * @newsk: buffer to queue
1317 *
300ce174 1318 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1319 * and you must therefore hold required locks before calling it.
1320 *
1321 * A buffer cannot be placed on two lists at the same time.
1322 */
300ce174
SH
1323static inline void __skb_queue_after(struct sk_buff_head *list,
1324 struct sk_buff *prev,
1325 struct sk_buff *newsk)
1da177e4 1326{
bf299275 1327 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1328}
1329
7965bd4d
JP
1330void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1331 struct sk_buff_head *list);
7de6c033 1332
f5572855
GR
1333static inline void __skb_queue_before(struct sk_buff_head *list,
1334 struct sk_buff *next,
1335 struct sk_buff *newsk)
1336{
1337 __skb_insert(newsk, next->prev, next, list);
1338}
1339
300ce174
SH
1340/**
1341 * __skb_queue_head - queue a buffer at the list head
1342 * @list: list to use
1343 * @newsk: buffer to queue
1344 *
1345 * Queue a buffer at the start of a list. This function takes no locks
1346 * and you must therefore hold required locks before calling it.
1347 *
1348 * A buffer cannot be placed on two lists at the same time.
1349 */
7965bd4d 1350void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1351static inline void __skb_queue_head(struct sk_buff_head *list,
1352 struct sk_buff *newsk)
1353{
1354 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1355}
1356
1da177e4
LT
1357/**
1358 * __skb_queue_tail - queue a buffer at the list tail
1359 * @list: list to use
1360 * @newsk: buffer to queue
1361 *
1362 * Queue a buffer at the end of a list. This function takes no locks
1363 * and you must therefore hold required locks before calling it.
1364 *
1365 * A buffer cannot be placed on two lists at the same time.
1366 */
7965bd4d 1367void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1368static inline void __skb_queue_tail(struct sk_buff_head *list,
1369 struct sk_buff *newsk)
1370{
f5572855 1371 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1372}
1373
1da177e4
LT
1374/*
1375 * remove sk_buff from list. _Must_ be called atomically, and with
1376 * the list known..
1377 */
7965bd4d 1378void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1379static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1380{
1381 struct sk_buff *next, *prev;
1382
1383 list->qlen--;
1384 next = skb->next;
1385 prev = skb->prev;
1386 skb->next = skb->prev = NULL;
1da177e4
LT
1387 next->prev = prev;
1388 prev->next = next;
1389}
1390
f525c06d
GR
1391/**
1392 * __skb_dequeue - remove from the head of the queue
1393 * @list: list to dequeue from
1394 *
1395 * Remove the head of the list. This function does not take any locks
1396 * so must be used with appropriate locks held only. The head item is
1397 * returned or %NULL if the list is empty.
1398 */
7965bd4d 1399struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1400static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1401{
1402 struct sk_buff *skb = skb_peek(list);
1403 if (skb)
1404 __skb_unlink(skb, list);
1405 return skb;
1406}
1da177e4
LT
1407
1408/**
1409 * __skb_dequeue_tail - remove from the tail of the queue
1410 * @list: list to dequeue from
1411 *
1412 * Remove the tail of the list. This function does not take any locks
1413 * so must be used with appropriate locks held only. The tail item is
1414 * returned or %NULL if the list is empty.
1415 */
7965bd4d 1416struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1417static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1418{
1419 struct sk_buff *skb = skb_peek_tail(list);
1420 if (skb)
1421 __skb_unlink(skb, list);
1422 return skb;
1423}
1424
1425
bdcc0924 1426static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1427{
1428 return skb->data_len;
1429}
1430
1431static inline unsigned int skb_headlen(const struct sk_buff *skb)
1432{
1433 return skb->len - skb->data_len;
1434}
1435
1436static inline int skb_pagelen(const struct sk_buff *skb)
1437{
1438 int i, len = 0;
1439
1440 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1441 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1442 return len + skb_headlen(skb);
1443}
1444
131ea667
IC
1445/**
1446 * __skb_fill_page_desc - initialise a paged fragment in an skb
1447 * @skb: buffer containing fragment to be initialised
1448 * @i: paged fragment index to initialise
1449 * @page: the page to use for this fragment
1450 * @off: the offset to the data with @page
1451 * @size: the length of the data
1452 *
1453 * Initialises the @i'th fragment of @skb to point to &size bytes at
1454 * offset @off within @page.
1455 *
1456 * Does not take any additional reference on the fragment.
1457 */
1458static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1459 struct page *page, int off, int size)
1da177e4
LT
1460{
1461 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1462
c48a11c7
MG
1463 /*
1464 * Propagate page->pfmemalloc to the skb if we can. The problem is
1465 * that not all callers have unique ownership of the page. If
1466 * pfmemalloc is set, we check the mapping as a mapping implies
1467 * page->index is set (index and pfmemalloc share space).
1468 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1469 * do not lose pfmemalloc information as the pages would not be
1470 * allocated using __GFP_MEMALLOC.
1471 */
a8605c60 1472 frag->page.p = page;
1da177e4 1473 frag->page_offset = off;
9e903e08 1474 skb_frag_size_set(frag, size);
cca7af38
PE
1475
1476 page = compound_head(page);
1477 if (page->pfmemalloc && !page->mapping)
1478 skb->pfmemalloc = true;
131ea667
IC
1479}
1480
1481/**
1482 * skb_fill_page_desc - initialise a paged fragment in an skb
1483 * @skb: buffer containing fragment to be initialised
1484 * @i: paged fragment index to initialise
1485 * @page: the page to use for this fragment
1486 * @off: the offset to the data with @page
1487 * @size: the length of the data
1488 *
1489 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1490 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1491 * addition updates @skb such that @i is the last fragment.
1492 *
1493 * Does not take any additional reference on the fragment.
1494 */
1495static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1496 struct page *page, int off, int size)
1497{
1498 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1499 skb_shinfo(skb)->nr_frags = i + 1;
1500}
1501
7965bd4d
JP
1502void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1503 int size, unsigned int truesize);
654bed16 1504
f8e617e1
JW
1505void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1506 unsigned int truesize);
1507
1da177e4 1508#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1509#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1510#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1511
27a884dc
ACM
1512#ifdef NET_SKBUFF_DATA_USES_OFFSET
1513static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1514{
1515 return skb->head + skb->tail;
1516}
1517
1518static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1519{
1520 skb->tail = skb->data - skb->head;
1521}
1522
1523static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1524{
1525 skb_reset_tail_pointer(skb);
1526 skb->tail += offset;
1527}
7cc46190 1528
27a884dc
ACM
1529#else /* NET_SKBUFF_DATA_USES_OFFSET */
1530static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1531{
1532 return skb->tail;
1533}
1534
1535static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1536{
1537 skb->tail = skb->data;
1538}
1539
1540static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1541{
1542 skb->tail = skb->data + offset;
1543}
4305b541 1544
27a884dc
ACM
1545#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1546
1da177e4
LT
1547/*
1548 * Add data to an sk_buff
1549 */
0c7ddf36 1550unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1551unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1552static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1553{
27a884dc 1554 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1555 SKB_LINEAR_ASSERT(skb);
1556 skb->tail += len;
1557 skb->len += len;
1558 return tmp;
1559}
1560
7965bd4d 1561unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1562static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1563{
1564 skb->data -= len;
1565 skb->len += len;
1566 return skb->data;
1567}
1568
7965bd4d 1569unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1570static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1571{
1572 skb->len -= len;
1573 BUG_ON(skb->len < skb->data_len);
1574 return skb->data += len;
1575}
1576
47d29646
DM
1577static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1578{
1579 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1580}
1581
7965bd4d 1582unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1583
1584static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1585{
1586 if (len > skb_headlen(skb) &&
987c402a 1587 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1588 return NULL;
1589 skb->len -= len;
1590 return skb->data += len;
1591}
1592
1593static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1594{
1595 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1596}
1597
1598static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1599{
1600 if (likely(len <= skb_headlen(skb)))
1601 return 1;
1602 if (unlikely(len > skb->len))
1603 return 0;
987c402a 1604 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1605}
1606
1607/**
1608 * skb_headroom - bytes at buffer head
1609 * @skb: buffer to check
1610 *
1611 * Return the number of bytes of free space at the head of an &sk_buff.
1612 */
c2636b4d 1613static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1614{
1615 return skb->data - skb->head;
1616}
1617
1618/**
1619 * skb_tailroom - bytes at buffer end
1620 * @skb: buffer to check
1621 *
1622 * Return the number of bytes of free space at the tail of an sk_buff
1623 */
1624static inline int skb_tailroom(const struct sk_buff *skb)
1625{
4305b541 1626 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1627}
1628
a21d4572
ED
1629/**
1630 * skb_availroom - 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 * allocated by sk_stream_alloc()
1635 */
1636static inline int skb_availroom(const struct sk_buff *skb)
1637{
16fad69c
ED
1638 if (skb_is_nonlinear(skb))
1639 return 0;
1640
1641 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1642}
1643
1da177e4
LT
1644/**
1645 * skb_reserve - adjust headroom
1646 * @skb: buffer to alter
1647 * @len: bytes to move
1648 *
1649 * Increase the headroom of an empty &sk_buff by reducing the tail
1650 * room. This is only allowed for an empty buffer.
1651 */
8243126c 1652static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1653{
1654 skb->data += len;
1655 skb->tail += len;
1656}
1657
6a674e9c
JG
1658static inline void skb_reset_inner_headers(struct sk_buff *skb)
1659{
aefbd2b3 1660 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1661 skb->inner_network_header = skb->network_header;
1662 skb->inner_transport_header = skb->transport_header;
1663}
1664
0b5c9db1
JP
1665static inline void skb_reset_mac_len(struct sk_buff *skb)
1666{
1667 skb->mac_len = skb->network_header - skb->mac_header;
1668}
1669
6a674e9c
JG
1670static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1671 *skb)
1672{
1673 return skb->head + skb->inner_transport_header;
1674}
1675
1676static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1677{
1678 skb->inner_transport_header = skb->data - skb->head;
1679}
1680
1681static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1682 const int offset)
1683{
1684 skb_reset_inner_transport_header(skb);
1685 skb->inner_transport_header += offset;
1686}
1687
1688static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1689{
1690 return skb->head + skb->inner_network_header;
1691}
1692
1693static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1694{
1695 skb->inner_network_header = skb->data - skb->head;
1696}
1697
1698static inline void skb_set_inner_network_header(struct sk_buff *skb,
1699 const int offset)
1700{
1701 skb_reset_inner_network_header(skb);
1702 skb->inner_network_header += offset;
1703}
1704
aefbd2b3
PS
1705static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1706{
1707 return skb->head + skb->inner_mac_header;
1708}
1709
1710static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1711{
1712 skb->inner_mac_header = skb->data - skb->head;
1713}
1714
1715static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1716 const int offset)
1717{
1718 skb_reset_inner_mac_header(skb);
1719 skb->inner_mac_header += offset;
1720}
fda55eca
ED
1721static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1722{
35d04610 1723 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1724}
1725
9c70220b
ACM
1726static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1727{
2e07fa9c 1728 return skb->head + skb->transport_header;
9c70220b
ACM
1729}
1730
badff6d0
ACM
1731static inline void skb_reset_transport_header(struct sk_buff *skb)
1732{
2e07fa9c 1733 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1734}
1735
967b05f6
ACM
1736static inline void skb_set_transport_header(struct sk_buff *skb,
1737 const int offset)
1738{
2e07fa9c
ACM
1739 skb_reset_transport_header(skb);
1740 skb->transport_header += offset;
ea2ae17d
ACM
1741}
1742
d56f90a7
ACM
1743static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1744{
2e07fa9c 1745 return skb->head + skb->network_header;
d56f90a7
ACM
1746}
1747
c1d2bbe1
ACM
1748static inline void skb_reset_network_header(struct sk_buff *skb)
1749{
2e07fa9c 1750 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1751}
1752
c14d2450
ACM
1753static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1754{
2e07fa9c
ACM
1755 skb_reset_network_header(skb);
1756 skb->network_header += offset;
c14d2450
ACM
1757}
1758
2e07fa9c 1759static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1760{
2e07fa9c 1761 return skb->head + skb->mac_header;
bbe735e4
ACM
1762}
1763
2e07fa9c 1764static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1765{
35d04610 1766 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1767}
1768
1769static inline void skb_reset_mac_header(struct sk_buff *skb)
1770{
1771 skb->mac_header = skb->data - skb->head;
1772}
1773
1774static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1775{
1776 skb_reset_mac_header(skb);
1777 skb->mac_header += offset;
1778}
1779
0e3da5bb
TT
1780static inline void skb_pop_mac_header(struct sk_buff *skb)
1781{
1782 skb->mac_header = skb->network_header;
1783}
1784
fbbdb8f0
YX
1785static inline void skb_probe_transport_header(struct sk_buff *skb,
1786 const int offset_hint)
1787{
1788 struct flow_keys keys;
1789
1790 if (skb_transport_header_was_set(skb))
1791 return;
1792 else if (skb_flow_dissect(skb, &keys))
1793 skb_set_transport_header(skb, keys.thoff);
1794 else
1795 skb_set_transport_header(skb, offset_hint);
1796}
1797
03606895
ED
1798static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1799{
1800 if (skb_mac_header_was_set(skb)) {
1801 const unsigned char *old_mac = skb_mac_header(skb);
1802
1803 skb_set_mac_header(skb, -skb->mac_len);
1804 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1805 }
1806}
1807
04fb451e
MM
1808static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1809{
1810 return skb->csum_start - skb_headroom(skb);
1811}
1812
2e07fa9c
ACM
1813static inline int skb_transport_offset(const struct sk_buff *skb)
1814{
1815 return skb_transport_header(skb) - skb->data;
1816}
1817
1818static inline u32 skb_network_header_len(const struct sk_buff *skb)
1819{
1820 return skb->transport_header - skb->network_header;
1821}
1822
6a674e9c
JG
1823static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1824{
1825 return skb->inner_transport_header - skb->inner_network_header;
1826}
1827
2e07fa9c
ACM
1828static inline int skb_network_offset(const struct sk_buff *skb)
1829{
1830 return skb_network_header(skb) - skb->data;
1831}
48d49d0c 1832
6a674e9c
JG
1833static inline int skb_inner_network_offset(const struct sk_buff *skb)
1834{
1835 return skb_inner_network_header(skb) - skb->data;
1836}
1837
f9599ce1
CG
1838static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1839{
1840 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1841}
1842
1da177e4
LT
1843/*
1844 * CPUs often take a performance hit when accessing unaligned memory
1845 * locations. The actual performance hit varies, it can be small if the
1846 * hardware handles it or large if we have to take an exception and fix it
1847 * in software.
1848 *
1849 * Since an ethernet header is 14 bytes network drivers often end up with
1850 * the IP header at an unaligned offset. The IP header can be aligned by
1851 * shifting the start of the packet by 2 bytes. Drivers should do this
1852 * with:
1853 *
8660c124 1854 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1855 *
1856 * The downside to this alignment of the IP header is that the DMA is now
1857 * unaligned. On some architectures the cost of an unaligned DMA is high
1858 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1859 *
1da177e4
LT
1860 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1861 * to be overridden.
1862 */
1863#ifndef NET_IP_ALIGN
1864#define NET_IP_ALIGN 2
1865#endif
1866
025be81e
AB
1867/*
1868 * The networking layer reserves some headroom in skb data (via
1869 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1870 * the header has to grow. In the default case, if the header has to grow
d6301d3d 1871 * 32 bytes or less we avoid the reallocation.
025be81e
AB
1872 *
1873 * Unfortunately this headroom changes the DMA alignment of the resulting
1874 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1875 * on some architectures. An architecture can override this value,
1876 * perhaps setting it to a cacheline in size (since that will maintain
1877 * cacheline alignment of the DMA). It must be a power of 2.
1878 *
d6301d3d 1879 * Various parts of the networking layer expect at least 32 bytes of
025be81e 1880 * headroom, you should not reduce this.
5933dd2f
ED
1881 *
1882 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1883 * to reduce average number of cache lines per packet.
1884 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 1885 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
1886 */
1887#ifndef NET_SKB_PAD
5933dd2f 1888#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
1889#endif
1890
7965bd4d 1891int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1892
1893static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1894{
c4264f27 1895 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
1896 WARN_ON(1);
1897 return;
1898 }
27a884dc
ACM
1899 skb->len = len;
1900 skb_set_tail_pointer(skb, len);
1da177e4
LT
1901}
1902
7965bd4d 1903void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1904
1905static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1906{
3cc0e873
HX
1907 if (skb->data_len)
1908 return ___pskb_trim(skb, len);
1909 __skb_trim(skb, len);
1910 return 0;
1da177e4
LT
1911}
1912
1913static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1914{
1915 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1916}
1917
e9fa4f7b
HX
1918/**
1919 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1920 * @skb: buffer to alter
1921 * @len: new length
1922 *
1923 * This is identical to pskb_trim except that the caller knows that
1924 * the skb is not cloned so we should never get an error due to out-
1925 * of-memory.
1926 */
1927static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1928{
1929 int err = pskb_trim(skb, len);
1930 BUG_ON(err);
1931}
1932
1da177e4
LT
1933/**
1934 * skb_orphan - orphan a buffer
1935 * @skb: buffer to orphan
1936 *
1937 * If a buffer currently has an owner then we call the owner's
1938 * destructor function and make the @skb unowned. The buffer continues
1939 * to exist but is no longer charged to its former owner.
1940 */
1941static inline void skb_orphan(struct sk_buff *skb)
1942{
c34a7612 1943 if (skb->destructor) {
1da177e4 1944 skb->destructor(skb);
c34a7612
ED
1945 skb->destructor = NULL;
1946 skb->sk = NULL;
376c7311
ED
1947 } else {
1948 BUG_ON(skb->sk);
c34a7612 1949 }
1da177e4
LT
1950}
1951
a353e0ce
MT
1952/**
1953 * skb_orphan_frags - orphan the frags contained in a buffer
1954 * @skb: buffer to orphan frags from
1955 * @gfp_mask: allocation mask for replacement pages
1956 *
1957 * For each frag in the SKB which needs a destructor (i.e. has an
1958 * owner) create a copy of that frag and release the original
1959 * page by calling the destructor.
1960 */
1961static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1962{
1963 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1964 return 0;
1965 return skb_copy_ubufs(skb, gfp_mask);
1966}
1967
1da177e4
LT
1968/**
1969 * __skb_queue_purge - empty a list
1970 * @list: list to empty
1971 *
1972 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1973 * the list and one reference dropped. This function does not take the
1974 * list lock and the caller must hold the relevant locks to use it.
1975 */
7965bd4d 1976void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
1977static inline void __skb_queue_purge(struct sk_buff_head *list)
1978{
1979 struct sk_buff *skb;
1980 while ((skb = __skb_dequeue(list)) != NULL)
1981 kfree_skb(skb);
1982}
1983
e5e67305
AD
1984#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1985#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1986#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1987
7965bd4d 1988void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 1989
7965bd4d
JP
1990struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1991 gfp_t gfp_mask);
8af27456
CH
1992
1993/**
1994 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1995 * @dev: network device to receive on
1996 * @length: length to allocate
1997 *
1998 * Allocate a new &sk_buff and assign it a usage count of one. The
1999 * buffer has unspecified headroom built in. Users should allocate
2000 * the headroom they think they need without accounting for the
2001 * built in space. The built in space is used for optimisations.
2002 *
2003 * %NULL is returned if there is no free memory. Although this function
2004 * allocates memory it can be called from an interrupt.
2005 */
2006static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2007 unsigned int length)
8af27456
CH
2008{
2009 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2010}
2011
6f532612
ED
2012/* legacy helper around __netdev_alloc_skb() */
2013static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2014 gfp_t gfp_mask)
2015{
2016 return __netdev_alloc_skb(NULL, length, gfp_mask);
2017}
2018
2019/* legacy helper around netdev_alloc_skb() */
2020static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2021{
2022 return netdev_alloc_skb(NULL, length);
2023}
2024
2025
4915a0de
ED
2026static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2027 unsigned int length, gfp_t gfp)
61321bbd 2028{
4915a0de 2029 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2030
2031 if (NET_IP_ALIGN && skb)
2032 skb_reserve(skb, NET_IP_ALIGN);
2033 return skb;
2034}
2035
4915a0de
ED
2036static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2037 unsigned int length)
2038{
2039 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2040}
2041
bc6fc9fa
FF
2042/**
2043 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
2044 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2045 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2046 * @order: size of the allocation
2047 *
2048 * Allocate a new page.
2049 *
2050 * %NULL is returned if there is no free memory.
2051*/
2052static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2053 struct sk_buff *skb,
2054 unsigned int order)
2055{
2056 struct page *page;
2057
2058 gfp_mask |= __GFP_COLD;
2059
2060 if (!(gfp_mask & __GFP_NOMEMALLOC))
2061 gfp_mask |= __GFP_MEMALLOC;
2062
2063 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2064 if (skb && page && page->pfmemalloc)
2065 skb->pfmemalloc = true;
2066
2067 return page;
2068}
2069
2070/**
2071 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2072 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2073 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2074 *
2075 * Allocate a new page.
2076 *
2077 * %NULL is returned if there is no free memory.
2078 */
2079static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2080 struct sk_buff *skb)
2081{
2082 return __skb_alloc_pages(gfp_mask, skb, 0);
2083}
2084
2085/**
2086 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2087 * @page: The page that was allocated from skb_alloc_page
2088 * @skb: The skb that may need pfmemalloc set
2089 */
2090static inline void skb_propagate_pfmemalloc(struct page *page,
2091 struct sk_buff *skb)
2092{
2093 if (page && page->pfmemalloc)
2094 skb->pfmemalloc = true;
2095}
2096
131ea667 2097/**
e227867f 2098 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2099 * @frag: the paged fragment
2100 *
2101 * Returns the &struct page associated with @frag.
2102 */
2103static inline struct page *skb_frag_page(const skb_frag_t *frag)
2104{
a8605c60 2105 return frag->page.p;
131ea667
IC
2106}
2107
2108/**
2109 * __skb_frag_ref - take an addition reference on a paged fragment.
2110 * @frag: the paged fragment
2111 *
2112 * Takes an additional reference on the paged fragment @frag.
2113 */
2114static inline void __skb_frag_ref(skb_frag_t *frag)
2115{
2116 get_page(skb_frag_page(frag));
2117}
2118
2119/**
2120 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2121 * @skb: the buffer
2122 * @f: the fragment offset.
2123 *
2124 * Takes an additional reference on the @f'th paged fragment of @skb.
2125 */
2126static inline void skb_frag_ref(struct sk_buff *skb, int f)
2127{
2128 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2129}
2130
2131/**
2132 * __skb_frag_unref - release a reference on a paged fragment.
2133 * @frag: the paged fragment
2134 *
2135 * Releases a reference on the paged fragment @frag.
2136 */
2137static inline void __skb_frag_unref(skb_frag_t *frag)
2138{
2139 put_page(skb_frag_page(frag));
2140}
2141
2142/**
2143 * skb_frag_unref - release a reference on a paged fragment of an skb.
2144 * @skb: the buffer
2145 * @f: the fragment offset
2146 *
2147 * Releases a reference on the @f'th paged fragment of @skb.
2148 */
2149static inline void skb_frag_unref(struct sk_buff *skb, int f)
2150{
2151 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2152}
2153
2154/**
2155 * skb_frag_address - gets the address of the data contained in a paged fragment
2156 * @frag: the paged fragment buffer
2157 *
2158 * Returns the address of the data within @frag. The page must already
2159 * be mapped.
2160 */
2161static inline void *skb_frag_address(const skb_frag_t *frag)
2162{
2163 return page_address(skb_frag_page(frag)) + frag->page_offset;
2164}
2165
2166/**
2167 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2168 * @frag: the paged fragment buffer
2169 *
2170 * Returns the address of the data within @frag. Checks that the page
2171 * is mapped and returns %NULL otherwise.
2172 */
2173static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2174{
2175 void *ptr = page_address(skb_frag_page(frag));
2176 if (unlikely(!ptr))
2177 return NULL;
2178
2179 return ptr + frag->page_offset;
2180}
2181
2182/**
2183 * __skb_frag_set_page - sets the page contained in a paged fragment
2184 * @frag: the paged fragment
2185 * @page: the page to set
2186 *
2187 * Sets the fragment @frag to contain @page.
2188 */
2189static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2190{
a8605c60 2191 frag->page.p = page;
131ea667
IC
2192}
2193
2194/**
2195 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2196 * @skb: the buffer
2197 * @f: the fragment offset
2198 * @page: the page to set
2199 *
2200 * Sets the @f'th fragment of @skb to contain @page.
2201 */
2202static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2203 struct page *page)
2204{
2205 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2206}
2207
400dfd3a
ED
2208bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2209
131ea667
IC
2210/**
2211 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2212 * @dev: the device to map the fragment to
131ea667
IC
2213 * @frag: the paged fragment to map
2214 * @offset: the offset within the fragment (starting at the
2215 * fragment's own offset)
2216 * @size: the number of bytes to map
f83347df 2217 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2218 *
2219 * Maps the page associated with @frag to @device.
2220 */
2221static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2222 const skb_frag_t *frag,
2223 size_t offset, size_t size,
2224 enum dma_data_direction dir)
2225{
2226 return dma_map_page(dev, skb_frag_page(frag),
2227 frag->page_offset + offset, size, dir);
2228}
2229
117632e6
ED
2230static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2231 gfp_t gfp_mask)
2232{
2233 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2234}
2235
334a8132
PM
2236/**
2237 * skb_clone_writable - is the header of a clone writable
2238 * @skb: buffer to check
2239 * @len: length up to which to write
2240 *
2241 * Returns true if modifying the header part of the cloned buffer
2242 * does not requires the data to be copied.
2243 */
05bdd2f1 2244static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2245{
2246 return !skb_header_cloned(skb) &&
2247 skb_headroom(skb) + len <= skb->hdr_len;
2248}
2249
d9cc2048
HX
2250static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2251 int cloned)
2252{
2253 int delta = 0;
2254
d9cc2048
HX
2255 if (headroom > skb_headroom(skb))
2256 delta = headroom - skb_headroom(skb);
2257
2258 if (delta || cloned)
2259 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2260 GFP_ATOMIC);
2261 return 0;
2262}
2263
1da177e4
LT
2264/**
2265 * skb_cow - copy header of skb when it is required
2266 * @skb: buffer to cow
2267 * @headroom: needed headroom
2268 *
2269 * If the skb passed lacks sufficient headroom or its data part
2270 * is shared, data is reallocated. If reallocation fails, an error
2271 * is returned and original skb is not changed.
2272 *
2273 * The result is skb with writable area skb->head...skb->tail
2274 * and at least @headroom of space at head.
2275 */
2276static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2277{
d9cc2048
HX
2278 return __skb_cow(skb, headroom, skb_cloned(skb));
2279}
1da177e4 2280
d9cc2048
HX
2281/**
2282 * skb_cow_head - skb_cow but only making the head writable
2283 * @skb: buffer to cow
2284 * @headroom: needed headroom
2285 *
2286 * This function is identical to skb_cow except that we replace the
2287 * skb_cloned check by skb_header_cloned. It should be used when
2288 * you only need to push on some header and do not need to modify
2289 * the data.
2290 */
2291static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2292{
2293 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2294}
2295
2296/**
2297 * skb_padto - pad an skbuff up to a minimal size
2298 * @skb: buffer to pad
2299 * @len: minimal length
2300 *
2301 * Pads up a buffer to ensure the trailing bytes exist and are
2302 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2303 * is untouched. Otherwise it is extended. Returns zero on
2304 * success. The skb is freed on error.
1da177e4
LT
2305 */
2306
5b057c6b 2307static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2308{
2309 unsigned int size = skb->len;
2310 if (likely(size >= len))
5b057c6b 2311 return 0;
987c402a 2312 return skb_pad(skb, len - size);
1da177e4
LT
2313}
2314
2315static inline int skb_add_data(struct sk_buff *skb,
2316 char __user *from, int copy)
2317{
2318 const int off = skb->len;
2319
2320 if (skb->ip_summed == CHECKSUM_NONE) {
2321 int err = 0;
5084205f 2322 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2323 copy, 0, &err);
2324 if (!err) {
2325 skb->csum = csum_block_add(skb->csum, csum, off);
2326 return 0;
2327 }
2328 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2329 return 0;
2330
2331 __skb_trim(skb, off);
2332 return -EFAULT;
2333}
2334
38ba0a65
ED
2335static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2336 const struct page *page, int off)
1da177e4
LT
2337{
2338 if (i) {
9e903e08 2339 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2340
ea2ab693 2341 return page == skb_frag_page(frag) &&
9e903e08 2342 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2343 }
38ba0a65 2344 return false;
1da177e4
LT
2345}
2346
364c6bad
HX
2347static inline int __skb_linearize(struct sk_buff *skb)
2348{
2349 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2350}
2351
1da177e4
LT
2352/**
2353 * skb_linearize - convert paged skb to linear one
2354 * @skb: buffer to linarize
1da177e4
LT
2355 *
2356 * If there is no free memory -ENOMEM is returned, otherwise zero
2357 * is returned and the old skb data released.
2358 */
364c6bad
HX
2359static inline int skb_linearize(struct sk_buff *skb)
2360{
2361 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2362}
2363
cef401de
ED
2364/**
2365 * skb_has_shared_frag - can any frag be overwritten
2366 * @skb: buffer to test
2367 *
2368 * Return true if the skb has at least one frag that might be modified
2369 * by an external entity (as in vmsplice()/sendfile())
2370 */
2371static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2372{
c9af6db4
PS
2373 return skb_is_nonlinear(skb) &&
2374 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2375}
2376
364c6bad
HX
2377/**
2378 * skb_linearize_cow - make sure skb is linear and writable
2379 * @skb: buffer to process
2380 *
2381 * If there is no free memory -ENOMEM is returned, otherwise zero
2382 * is returned and the old skb data released.
2383 */
2384static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2385{
364c6bad
HX
2386 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2387 __skb_linearize(skb) : 0;
1da177e4
LT
2388}
2389
2390/**
2391 * skb_postpull_rcsum - update checksum for received skb after pull
2392 * @skb: buffer to update
2393 * @start: start of data before pull
2394 * @len: length of data pulled
2395 *
2396 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2397 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2398 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2399 */
2400
2401static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2402 const void *start, unsigned int len)
1da177e4 2403{
84fa7933 2404 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2405 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2406}
2407
cbb042f9
HX
2408unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2409
7ce5a27f
DM
2410/**
2411 * pskb_trim_rcsum - trim received skb and update checksum
2412 * @skb: buffer to trim
2413 * @len: new length
2414 *
2415 * This is exactly the same as pskb_trim except that it ensures the
2416 * checksum of received packets are still valid after the operation.
2417 */
2418
2419static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2420{
2421 if (likely(len >= skb->len))
2422 return 0;
2423 if (skb->ip_summed == CHECKSUM_COMPLETE)
2424 skb->ip_summed = CHECKSUM_NONE;
2425 return __pskb_trim(skb, len);
2426}
2427
1da177e4
LT
2428#define skb_queue_walk(queue, skb) \
2429 for (skb = (queue)->next; \
a1e4891f 2430 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2431 skb = skb->next)
2432
46f8914e
JC
2433#define skb_queue_walk_safe(queue, skb, tmp) \
2434 for (skb = (queue)->next, tmp = skb->next; \
2435 skb != (struct sk_buff *)(queue); \
2436 skb = tmp, tmp = skb->next)
2437
1164f52a 2438#define skb_queue_walk_from(queue, skb) \
a1e4891f 2439 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2440 skb = skb->next)
2441
2442#define skb_queue_walk_from_safe(queue, skb, tmp) \
2443 for (tmp = skb->next; \
2444 skb != (struct sk_buff *)(queue); \
2445 skb = tmp, tmp = skb->next)
2446
300ce174
SH
2447#define skb_queue_reverse_walk(queue, skb) \
2448 for (skb = (queue)->prev; \
a1e4891f 2449 skb != (struct sk_buff *)(queue); \
300ce174
SH
2450 skb = skb->prev)
2451
686a2955
DM
2452#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2453 for (skb = (queue)->prev, tmp = skb->prev; \
2454 skb != (struct sk_buff *)(queue); \
2455 skb = tmp, tmp = skb->prev)
2456
2457#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2458 for (tmp = skb->prev; \
2459 skb != (struct sk_buff *)(queue); \
2460 skb = tmp, tmp = skb->prev)
1da177e4 2461
21dc3301 2462static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2463{
2464 return skb_shinfo(skb)->frag_list != NULL;
2465}
2466
2467static inline void skb_frag_list_init(struct sk_buff *skb)
2468{
2469 skb_shinfo(skb)->frag_list = NULL;
2470}
2471
2472static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2473{
2474 frag->next = skb_shinfo(skb)->frag_list;
2475 skb_shinfo(skb)->frag_list = frag;
2476}
2477
2478#define skb_walk_frags(skb, iter) \
2479 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2480
7965bd4d
JP
2481struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2482 int *peeked, int *off, int *err);
2483struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2484 int *err);
2485unsigned int datagram_poll(struct file *file, struct socket *sock,
2486 struct poll_table_struct *wait);
2487int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2488 struct iovec *to, int size);
2489int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2490 struct iovec *iov);
2491int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2492 const struct iovec *from, int from_offset,
2493 int len);
2494int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2495 int offset, size_t count);
2496int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2497 const struct iovec *to, int to_offset,
2498 int size);
2499void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2500void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2501int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2502int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2503int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2504__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2505 int len, __wsum csum);
2506int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2507 struct pipe_inode_info *pipe, unsigned int len,
2508 unsigned int flags);
2509void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2510unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2511int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2512 int len, int hlen);
7965bd4d
JP
2513void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2514int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2515void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2516unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2517struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
20380731 2518
2817a336
DB
2519struct skb_checksum_ops {
2520 __wsum (*update)(const void *mem, int len, __wsum wsum);
2521 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2522};
2523
2524__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2525 __wsum csum, const struct skb_checksum_ops *ops);
2526__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2527 __wsum csum);
2528
1da177e4
LT
2529static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2530 int len, void *buffer)
2531{
2532 int hlen = skb_headlen(skb);
2533
55820ee2 2534 if (hlen - offset >= len)
1da177e4
LT
2535 return skb->data + offset;
2536
2537 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2538 return NULL;
2539
2540 return buffer;
2541}
2542
4262e5cc
DB
2543/**
2544 * skb_needs_linearize - check if we need to linearize a given skb
2545 * depending on the given device features.
2546 * @skb: socket buffer to check
2547 * @features: net device features
2548 *
2549 * Returns true if either:
2550 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2551 * 2. skb is fragmented and the device does not support SG.
2552 */
2553static inline bool skb_needs_linearize(struct sk_buff *skb,
2554 netdev_features_t features)
2555{
2556 return skb_is_nonlinear(skb) &&
2557 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2558 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2559}
2560
d626f62b
ACM
2561static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2562 void *to,
2563 const unsigned int len)
2564{
2565 memcpy(to, skb->data, len);
2566}
2567
2568static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2569 const int offset, void *to,
2570 const unsigned int len)
2571{
2572 memcpy(to, skb->data + offset, len);
2573}
2574
27d7ff46
ACM
2575static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2576 const void *from,
2577 const unsigned int len)
2578{
2579 memcpy(skb->data, from, len);
2580}
2581
2582static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2583 const int offset,
2584 const void *from,
2585 const unsigned int len)
2586{
2587 memcpy(skb->data + offset, from, len);
2588}
2589
7965bd4d 2590void skb_init(void);
1da177e4 2591
ac45f602
PO
2592static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2593{
2594 return skb->tstamp;
2595}
2596
a61bbcf2
PM
2597/**
2598 * skb_get_timestamp - get timestamp from a skb
2599 * @skb: skb to get stamp from
2600 * @stamp: pointer to struct timeval to store stamp in
2601 *
2602 * Timestamps are stored in the skb as offsets to a base timestamp.
2603 * This function converts the offset back to a struct timeval and stores
2604 * it in stamp.
2605 */
ac45f602
PO
2606static inline void skb_get_timestamp(const struct sk_buff *skb,
2607 struct timeval *stamp)
a61bbcf2 2608{
b7aa0bf7 2609 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2610}
2611
ac45f602
PO
2612static inline void skb_get_timestampns(const struct sk_buff *skb,
2613 struct timespec *stamp)
2614{
2615 *stamp = ktime_to_timespec(skb->tstamp);
2616}
2617
b7aa0bf7 2618static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2619{
b7aa0bf7 2620 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2621}
2622
164891aa
SH
2623static inline ktime_t net_timedelta(ktime_t t)
2624{
2625 return ktime_sub(ktime_get_real(), t);
2626}
2627
b9ce204f
IJ
2628static inline ktime_t net_invalid_timestamp(void)
2629{
2630 return ktime_set(0, 0);
2631}
a61bbcf2 2632
c1f19b51
RC
2633#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2634
7965bd4d
JP
2635void skb_clone_tx_timestamp(struct sk_buff *skb);
2636bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2637
2638#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2639
2640static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2641{
2642}
2643
2644static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2645{
2646 return false;
2647}
2648
2649#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2650
2651/**
2652 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2653 *
da92b194
RC
2654 * PHY drivers may accept clones of transmitted packets for
2655 * timestamping via their phy_driver.txtstamp method. These drivers
2656 * must call this function to return the skb back to the stack, with
2657 * or without a timestamp.
2658 *
c1f19b51 2659 * @skb: clone of the the original outgoing packet
da92b194 2660 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2661 *
2662 */
2663void skb_complete_tx_timestamp(struct sk_buff *skb,
2664 struct skb_shared_hwtstamps *hwtstamps);
2665
ac45f602
PO
2666/**
2667 * skb_tstamp_tx - queue clone of skb with send time stamps
2668 * @orig_skb: the original outgoing packet
2669 * @hwtstamps: hardware time stamps, may be NULL if not available
2670 *
2671 * If the skb has a socket associated, then this function clones the
2672 * skb (thus sharing the actual data and optional structures), stores
2673 * the optional hardware time stamping information (if non NULL) or
2674 * generates a software time stamp (otherwise), then queues the clone
2675 * to the error queue of the socket. Errors are silently ignored.
2676 */
7965bd4d
JP
2677void skb_tstamp_tx(struct sk_buff *orig_skb,
2678 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2679
4507a715
RC
2680static inline void sw_tx_timestamp(struct sk_buff *skb)
2681{
2244d07b
OH
2682 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2683 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2684 skb_tstamp_tx(skb, NULL);
2685}
2686
2687/**
2688 * skb_tx_timestamp() - Driver hook for transmit timestamping
2689 *
2690 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2691 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2692 *
73409f3b
DM
2693 * Specifically, one should make absolutely sure that this function is
2694 * called before TX completion of this packet can trigger. Otherwise
2695 * the packet could potentially already be freed.
2696 *
4507a715
RC
2697 * @skb: A socket buffer.
2698 */
2699static inline void skb_tx_timestamp(struct sk_buff *skb)
2700{
c1f19b51 2701 skb_clone_tx_timestamp(skb);
4507a715
RC
2702 sw_tx_timestamp(skb);
2703}
2704
6e3e939f
JB
2705/**
2706 * skb_complete_wifi_ack - deliver skb with wifi status
2707 *
2708 * @skb: the original outgoing packet
2709 * @acked: ack status
2710 *
2711 */
2712void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2713
7965bd4d
JP
2714__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2715__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2716
60476372
HX
2717static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2718{
2719 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2720}
2721
fb286bb2
HX
2722/**
2723 * skb_checksum_complete - Calculate checksum of an entire packet
2724 * @skb: packet to process
2725 *
2726 * This function calculates the checksum over the entire packet plus
2727 * the value of skb->csum. The latter can be used to supply the
2728 * checksum of a pseudo header as used by TCP/UDP. It returns the
2729 * checksum.
2730 *
2731 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2732 * this function can be used to verify that checksum on received
2733 * packets. In that case the function should return zero if the
2734 * checksum is correct. In particular, this function will return zero
2735 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2736 * hardware has already verified the correctness of the checksum.
2737 */
4381ca3c 2738static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2739{
60476372
HX
2740 return skb_csum_unnecessary(skb) ?
2741 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2742}
2743
76ba0aae
TH
2744/* Check if we need to perform checksum complete validation.
2745 *
2746 * Returns true if checksum complete is needed, false otherwise
2747 * (either checksum is unnecessary or zero checksum is allowed).
2748 */
2749static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2750 bool zero_okay,
2751 __sum16 check)
2752{
2753 if (skb_csum_unnecessary(skb)) {
2754 return false;
2755 } else if (zero_okay && !check) {
2756 skb->ip_summed = CHECKSUM_UNNECESSARY;
2757 return false;
2758 }
2759
2760 return true;
2761}
2762
2763/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2764 * in checksum_init.
2765 */
2766#define CHECKSUM_BREAK 76
2767
2768/* Validate (init) checksum based on checksum complete.
2769 *
2770 * Return values:
2771 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2772 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2773 * checksum is stored in skb->csum for use in __skb_checksum_complete
2774 * non-zero: value of invalid checksum
2775 *
2776 */
2777static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2778 bool complete,
2779 __wsum psum)
2780{
2781 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2782 if (!csum_fold(csum_add(psum, skb->csum))) {
2783 skb->ip_summed = CHECKSUM_UNNECESSARY;
2784 return 0;
2785 }
2786 }
2787
2788 skb->csum = psum;
2789
2790 if (complete || skb->len <= CHECKSUM_BREAK)
2791 return __skb_checksum_complete(skb);
2792
2793 return 0;
2794}
2795
2796static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2797{
2798 return 0;
2799}
2800
2801/* Perform checksum validate (init). Note that this is a macro since we only
2802 * want to calculate the pseudo header which is an input function if necessary.
2803 * First we try to validate without any computation (checksum unnecessary) and
2804 * then calculate based on checksum complete calling the function to compute
2805 * pseudo header.
2806 *
2807 * Return values:
2808 * 0: checksum is validated or try to in skb_checksum_complete
2809 * non-zero: value of invalid checksum
2810 */
2811#define __skb_checksum_validate(skb, proto, complete, \
2812 zero_okay, check, compute_pseudo) \
2813({ \
2814 __sum16 __ret = 0; \
2815 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2816 __ret = __skb_checksum_validate_complete(skb, \
2817 complete, compute_pseudo(skb, proto)); \
2818 __ret; \
2819})
2820
2821#define skb_checksum_init(skb, proto, compute_pseudo) \
2822 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2823
2824#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2825 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2826
2827#define skb_checksum_validate(skb, proto, compute_pseudo) \
2828 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2829
2830#define skb_checksum_validate_zero_check(skb, proto, check, \
2831 compute_pseudo) \
2832 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2833
2834#define skb_checksum_simple_validate(skb) \
2835 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2836
5f79e0f9 2837#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 2838void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
2839static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2840{
2841 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 2842 nf_conntrack_destroy(nfct);
1da177e4
LT
2843}
2844static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2845{
2846 if (nfct)
2847 atomic_inc(&nfct->use);
2848}
2fc72c7b 2849#endif
1da177e4
LT
2850#ifdef CONFIG_BRIDGE_NETFILTER
2851static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2852{
2853 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2854 kfree(nf_bridge);
2855}
2856static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2857{
2858 if (nf_bridge)
2859 atomic_inc(&nf_bridge->use);
2860}
2861#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
2862static inline void nf_reset(struct sk_buff *skb)
2863{
5f79e0f9 2864#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
2865 nf_conntrack_put(skb->nfct);
2866 skb->nfct = NULL;
2fc72c7b 2867#endif
a193a4ab
PM
2868#ifdef CONFIG_BRIDGE_NETFILTER
2869 nf_bridge_put(skb->nf_bridge);
2870 skb->nf_bridge = NULL;
2871#endif
2872}
2873
124dff01
PM
2874static inline void nf_reset_trace(struct sk_buff *skb)
2875{
478b360a 2876#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
2877 skb->nf_trace = 0;
2878#endif
a193a4ab
PM
2879}
2880
edda553c
YK
2881/* Note: This doesn't put any conntrack and bridge info in dst. */
2882static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2883{
5f79e0f9 2884#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
2885 dst->nfct = src->nfct;
2886 nf_conntrack_get(src->nfct);
2887 dst->nfctinfo = src->nfctinfo;
2fc72c7b 2888#endif
edda553c
YK
2889#ifdef CONFIG_BRIDGE_NETFILTER
2890 dst->nf_bridge = src->nf_bridge;
2891 nf_bridge_get(src->nf_bridge);
2892#endif
478b360a
FW
2893#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2894 dst->nf_trace = src->nf_trace;
2895#endif
edda553c
YK
2896}
2897
e7ac05f3
YK
2898static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2899{
e7ac05f3 2900#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 2901 nf_conntrack_put(dst->nfct);
2fc72c7b 2902#endif
e7ac05f3
YK
2903#ifdef CONFIG_BRIDGE_NETFILTER
2904 nf_bridge_put(dst->nf_bridge);
2905#endif
2906 __nf_copy(dst, src);
2907}
2908
984bc16c
JM
2909#ifdef CONFIG_NETWORK_SECMARK
2910static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2911{
2912 to->secmark = from->secmark;
2913}
2914
2915static inline void skb_init_secmark(struct sk_buff *skb)
2916{
2917 skb->secmark = 0;
2918}
2919#else
2920static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2921{ }
2922
2923static inline void skb_init_secmark(struct sk_buff *skb)
2924{ }
2925#endif
2926
574f7194
EB
2927static inline bool skb_irq_freeable(const struct sk_buff *skb)
2928{
2929 return !skb->destructor &&
2930#if IS_ENABLED(CONFIG_XFRM)
2931 !skb->sp &&
2932#endif
2933#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2934 !skb->nfct &&
2935#endif
2936 !skb->_skb_refdst &&
2937 !skb_has_frag_list(skb);
2938}
2939
f25f4e44
PWJ
2940static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2941{
f25f4e44 2942 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
2943}
2944
9247744e 2945static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 2946{
4e3ab47a 2947 return skb->queue_mapping;
4e3ab47a
PE
2948}
2949
f25f4e44
PWJ
2950static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2951{
f25f4e44 2952 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
2953}
2954
d5a9e24a
DM
2955static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2956{
2957 skb->queue_mapping = rx_queue + 1;
2958}
2959
9247744e 2960static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
2961{
2962 return skb->queue_mapping - 1;
2963}
2964
9247744e 2965static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 2966{
a02cec21 2967 return skb->queue_mapping != 0;
d5a9e24a
DM
2968}
2969
7965bd4d
JP
2970u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2971 unsigned int num_tx_queues);
9247744e 2972
def8b4fa
AD
2973static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2974{
0b3d8e08 2975#ifdef CONFIG_XFRM
def8b4fa 2976 return skb->sp;
def8b4fa 2977#else
def8b4fa 2978 return NULL;
def8b4fa 2979#endif
0b3d8e08 2980}
def8b4fa 2981
68c33163
PS
2982/* Keeps track of mac header offset relative to skb->head.
2983 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2984 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
2985 * tunnel skb it points to outer mac header.
2986 * Keeps track of level of encapsulation of network headers.
2987 */
68c33163 2988struct skb_gso_cb {
3347c960
ED
2989 int mac_offset;
2990 int encap_level;
68c33163
PS
2991};
2992#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2993
2994static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2995{
2996 return (skb_mac_header(inner_skb) - inner_skb->head) -
2997 SKB_GSO_CB(inner_skb)->mac_offset;
2998}
2999
1e2bd517
PS
3000static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3001{
3002 int new_headroom, headroom;
3003 int ret;
3004
3005 headroom = skb_headroom(skb);
3006 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3007 if (ret)
3008 return ret;
3009
3010 new_headroom = skb_headroom(skb);
3011 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3012 return 0;
3013}
3014
bdcc0924 3015static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3016{
3017 return skb_shinfo(skb)->gso_size;
3018}
3019
36a8f39e 3020/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3021static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3022{
3023 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3024}
3025
7965bd4d 3026void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3027
3028static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3029{
3030 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3031 * wanted then gso_type will be set. */
05bdd2f1
ED
3032 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3033
b78462eb
AD
3034 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3035 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3036 __skb_warn_lro_forwarding(skb);
3037 return true;
3038 }
3039 return false;
3040}
3041
35fc92a9
HX
3042static inline void skb_forward_csum(struct sk_buff *skb)
3043{
3044 /* Unfortunately we don't support this one. Any brave souls? */
3045 if (skb->ip_summed == CHECKSUM_COMPLETE)
3046 skb->ip_summed = CHECKSUM_NONE;
3047}
3048
bc8acf2c
ED
3049/**
3050 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3051 * @skb: skb to check
3052 *
3053 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3054 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3055 * use this helper, to document places where we make this assertion.
3056 */
05bdd2f1 3057static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3058{
3059#ifdef DEBUG
3060 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3061#endif
3062}
3063
f35d9d8a 3064bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3065
ed1f50c3
PD
3066int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3067
f77668dc
DB
3068u32 __skb_get_poff(const struct sk_buff *skb);
3069
3a7c1ee4
AD
3070/**
3071 * skb_head_is_locked - Determine if the skb->head is locked down
3072 * @skb: skb to check
3073 *
3074 * The head on skbs build around a head frag can be removed if they are
3075 * not cloned. This function returns true if the skb head is locked down
3076 * due to either being allocated via kmalloc, or by being a clone with
3077 * multiple references to the head.
3078 */
3079static inline bool skb_head_is_locked(const struct sk_buff *skb)
3080{
3081 return !skb->head_frag || skb_cloned(skb);
3082}
fe6cc55f
FW
3083
3084/**
3085 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3086 *
3087 * @skb: GSO skb
3088 *
3089 * skb_gso_network_seglen is used to determine the real size of the
3090 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3091 *
3092 * The MAC/L2 header is not accounted for.
3093 */
3094static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3095{
3096 unsigned int hdr_len = skb_transport_header(skb) -
3097 skb_network_header(skb);
3098 return hdr_len + skb_gso_transport_seglen(skb);
3099}
1da177e4
LT
3100#endif /* __KERNEL__ */
3101#endif /* _LINUX_SKBUFF_H */