net: sctp: Rename SCTP_XMIT_NAGLE_DELAY to SCTP_XMIT_DELAY
[deliverable/linux.git] / include / linux / skbuff.h
CommitLineData
1da177e4
LT
1/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
1da177e4 17#include <linux/kernel.h>
fe55f6d5 18#include <linux/kmemcheck.h>
1da177e4
LT
19#include <linux/compiler.h>
20#include <linux/time.h>
187f1882 21#include <linux/bug.h>
1da177e4
LT
22#include <linux/cache.h>
23
60063497 24#include <linux/atomic.h>
1da177e4
LT
25#include <asm/types.h>
26#include <linux/spinlock.h>
1da177e4 27#include <linux/net.h>
3fc7e8a6 28#include <linux/textsearch.h>
1da177e4 29#include <net/checksum.h>
a80958f4 30#include <linux/rcupdate.h>
97fc2f08 31#include <linux/dmaengine.h>
b7aa0bf7 32#include <linux/hrtimer.h>
131ea667 33#include <linux/dma-mapping.h>
c8f44aff 34#include <linux/netdev_features.h>
363ec392 35#include <linux/sched.h>
5203cd28 36#include <net/flow_keys.h>
1da177e4 37
78ea85f1
DB
38/* A. Checksumming of received packets by device.
39 *
40 * CHECKSUM_NONE:
41 *
42 * Device failed to checksum this packet e.g. due to lack of capabilities.
43 * The packet contains full (though not verified) checksum in packet but
44 * not in skb->csum. Thus, skb->csum is undefined in this case.
45 *
46 * CHECKSUM_UNNECESSARY:
47 *
48 * The hardware you're dealing with doesn't calculate the full checksum
49 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
50 * for specific protocols e.g. TCP/UDP/SCTP, then, for such packets it will
51 * set CHECKSUM_UNNECESSARY if their checksums are okay. skb->csum is still
52 * undefined in this case though. It is a bad option, but, unfortunately,
53 * nowadays most vendors do this. Apparently with the secret goal to sell
54 * you new devices, when you will add new protocol to your host, f.e. IPv6 8)
55 *
56 * CHECKSUM_COMPLETE:
57 *
58 * This is the most generic way. The device supplied checksum of the _whole_
59 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
60 * hardware doesn't need to parse L3/L4 headers to implement this.
61 *
62 * Note: Even if device supports only some protocols, but is able to produce
63 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
64 *
65 * CHECKSUM_PARTIAL:
66 *
67 * This is identical to the case for output below. This may occur on a packet
68 * received directly from another Linux OS, e.g., a virtualized Linux kernel
69 * on the same host. The packet can be treated in the same way as
70 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
71 * checksum must be filled in by the OS or the hardware.
72 *
73 * B. Checksumming on output.
74 *
75 * CHECKSUM_NONE:
76 *
77 * The skb was already checksummed by the protocol, or a checksum is not
78 * required.
79 *
80 * CHECKSUM_PARTIAL:
81 *
82 * The device is required to checksum the packet as seen by hard_start_xmit()
83 * from skb->csum_start up to the end, and to record/write the checksum at
84 * offset skb->csum_start + skb->csum_offset.
85 *
86 * The device must show its capabilities in dev->features, set up at device
87 * setup time, e.g. netdev_features.h:
88 *
89 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
90 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
91 * IPv4. Sigh. Vendors like this way for an unknown reason.
92 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
93 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
94 * NETIF_F_... - Well, you get the picture.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * Normally, the device will do per protocol specific checksumming. Protocol
99 * implementations that do not want the NIC to perform the checksum
100 * calculation should use this flag in their outgoing skbs.
101 *
102 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
103 * offload. Correspondingly, the FCoE protocol driver
104 * stack should use CHECKSUM_UNNECESSARY.
105 *
106 * Any questions? No questions, good. --ANK
107 */
108
60476372 109/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
110#define CHECKSUM_NONE 0
111#define CHECKSUM_UNNECESSARY 1
112#define CHECKSUM_COMPLETE 2
113#define CHECKSUM_PARTIAL 3
1da177e4
LT
114
115#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
116 ~(SMP_CACHE_BYTES - 1))
fc910a27 117#define SKB_WITH_OVERHEAD(X) \
deea84b0 118 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
119#define SKB_MAX_ORDER(X, ORDER) \
120 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
121#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
122#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
123
87fb4b7b
ED
124/* return minimum truesize of one skb containing X bytes of data */
125#define SKB_TRUESIZE(X) ((X) + \
126 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
127 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
128
1da177e4 129struct net_device;
716ea3a7 130struct scatterlist;
9c55e01c 131struct pipe_inode_info;
1da177e4 132
5f79e0f9 133#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
134struct nf_conntrack {
135 atomic_t use;
1da177e4 136};
5f79e0f9 137#endif
1da177e4
LT
138
139#ifdef CONFIG_BRIDGE_NETFILTER
140struct nf_bridge_info {
bf1ac5ca
ED
141 atomic_t use;
142 unsigned int mask;
143 struct net_device *physindev;
144 struct net_device *physoutdev;
145 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
146};
147#endif
148
1da177e4
LT
149struct sk_buff_head {
150 /* These two members must be first. */
151 struct sk_buff *next;
152 struct sk_buff *prev;
153
154 __u32 qlen;
155 spinlock_t lock;
156};
157
158struct sk_buff;
159
9d4dde52
IC
160/* To allow 64K frame to be packed as single skb without frag_list we
161 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
162 * buffers which do not start on a page boundary.
163 *
164 * Since GRO uses frags we allocate at least 16 regardless of page
165 * size.
a715dea3 166 */
9d4dde52 167#if (65536/PAGE_SIZE + 1) < 16
eec00954 168#define MAX_SKB_FRAGS 16UL
a715dea3 169#else
9d4dde52 170#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 171#endif
1da177e4
LT
172
173typedef struct skb_frag_struct skb_frag_t;
174
175struct skb_frag_struct {
a8605c60
IC
176 struct {
177 struct page *p;
178 } page;
cb4dfe56 179#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
180 __u32 page_offset;
181 __u32 size;
cb4dfe56
ED
182#else
183 __u16 page_offset;
184 __u16 size;
185#endif
1da177e4
LT
186};
187
9e903e08
ED
188static inline unsigned int skb_frag_size(const skb_frag_t *frag)
189{
190 return frag->size;
191}
192
193static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
194{
195 frag->size = size;
196}
197
198static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
199{
200 frag->size += delta;
201}
202
203static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
204{
205 frag->size -= delta;
206}
207
ac45f602
PO
208#define HAVE_HW_TIME_STAMP
209
210/**
d3a21be8 211 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
212 * @hwtstamp: hardware time stamp transformed into duration
213 * since arbitrary point in time
26c4fdb0 214 * @syststamp: hwtstamp transformed to system time base (deprecated)
ac45f602
PO
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
26c4fdb0
WB
225 * hardware. The syststamp implementation is deprecated in favor
226 * of hwtstamps and hw PTP clock sources exposed directly to
227 * userspace.
ac45f602
PO
228 *
229 * hwtstamps can only be compared against other hwtstamps from
230 * the same device.
231 *
232 * This structure is attached to packets as part of the
233 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
234 */
235struct skb_shared_hwtstamps {
236 ktime_t hwtstamp;
237 ktime_t syststamp;
238};
239
2244d07b
OH
240/* Definitions for tx_flags in struct skb_shared_info */
241enum {
242 /* generate hardware time stamp */
243 SKBTX_HW_TSTAMP = 1 << 0,
244
245 /* generate software time stamp */
246 SKBTX_SW_TSTAMP = 1 << 1,
247
248 /* device driver is going to provide hardware time stamp */
249 SKBTX_IN_PROGRESS = 1 << 2,
250
a6686f2f 251 /* device driver supports TX zero-copy buffers */
62b1a8ab 252 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
253
254 /* generate wifi status information (where possible) */
62b1a8ab 255 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
256
257 /* This indicates at least one fragment might be overwritten
258 * (as in vmsplice(), sendfile() ...)
259 * If we need to compute a TX checksum, we'll need to copy
260 * all frags to avoid possible bad checksum
261 */
262 SKBTX_SHARED_FRAG = 1 << 5,
a6686f2f
SM
263};
264
265/*
266 * The callback notifies userspace to release buffers when skb DMA is done in
267 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
268 * The zerocopy_success argument is true if zero copy transmit occurred,
269 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
270 * The ctx field is used to track device context.
271 * The desc field is used to track userspace buffer index.
a6686f2f
SM
272 */
273struct ubuf_info {
e19d6763 274 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 275 void *ctx;
a6686f2f 276 unsigned long desc;
ac45f602
PO
277};
278
1da177e4
LT
279/* This data is invariant across clones and lives at
280 * the end of the header data, ie. at skb->end.
281 */
282struct skb_shared_info {
9f42f126
IC
283 unsigned char nr_frags;
284 __u8 tx_flags;
7967168c
HX
285 unsigned short gso_size;
286 /* Warning: this field is not always filled in (UFO)! */
287 unsigned short gso_segs;
288 unsigned short gso_type;
1da177e4 289 struct sk_buff *frag_list;
ac45f602 290 struct skb_shared_hwtstamps hwtstamps;
9f42f126 291 __be32 ip6_frag_id;
ec7d2f2c
ED
292
293 /*
294 * Warning : all fields before dataref are cleared in __alloc_skb()
295 */
296 atomic_t dataref;
297
69e3c75f
JB
298 /* Intermediate layers must ensure that destructor_arg
299 * remains valid until skb destructor */
300 void * destructor_arg;
a6686f2f 301
fed66381
ED
302 /* must be last field, see pskb_expand_head() */
303 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
304};
305
306/* We divide dataref into two halves. The higher 16 bits hold references
307 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
308 * the entire skb->data. A clone of a headerless skb holds the length of
309 * the header in skb->hdr_len.
1da177e4
LT
310 *
311 * All users must obey the rule that the skb->data reference count must be
312 * greater than or equal to the payload reference count.
313 *
314 * Holding a reference to the payload part means that the user does not
315 * care about modifications to the header part of skb->data.
316 */
317#define SKB_DATAREF_SHIFT 16
318#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
319
d179cd12
DM
320
321enum {
322 SKB_FCLONE_UNAVAILABLE,
323 SKB_FCLONE_ORIG,
324 SKB_FCLONE_CLONE,
325};
326
7967168c
HX
327enum {
328 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 329 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
330
331 /* This indicates the skb is from an untrusted source. */
332 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
333
334 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
335 SKB_GSO_TCP_ECN = 1 << 3,
336
337 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
338
339 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
340
341 SKB_GSO_GRE = 1 << 6,
73136267 342
4b28252c 343 SKB_GSO_GRE_CSUM = 1 << 7,
0d89d203 344
4b28252c 345 SKB_GSO_IPIP = 1 << 8,
cb32f511 346
4b28252c 347 SKB_GSO_SIT = 1 << 9,
61c1db7f 348
4b28252c 349 SKB_GSO_UDP_TUNNEL = 1 << 10,
0f4f4ffa
TH
350
351 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c 352
4b28252c
TH
353 SKB_GSO_MPLS = 1 << 12,
354
7967168c
HX
355};
356
2e07fa9c
ACM
357#if BITS_PER_LONG > 32
358#define NET_SKBUFF_DATA_USES_OFFSET 1
359#endif
360
361#ifdef NET_SKBUFF_DATA_USES_OFFSET
362typedef unsigned int sk_buff_data_t;
363#else
364typedef unsigned char *sk_buff_data_t;
365#endif
366
363ec392
ED
367/**
368 * struct skb_mstamp - multi resolution time stamps
369 * @stamp_us: timestamp in us resolution
370 * @stamp_jiffies: timestamp in jiffies
371 */
372struct skb_mstamp {
373 union {
374 u64 v64;
375 struct {
376 u32 stamp_us;
377 u32 stamp_jiffies;
378 };
379 };
380};
381
382/**
383 * skb_mstamp_get - get current timestamp
384 * @cl: place to store timestamps
385 */
386static inline void skb_mstamp_get(struct skb_mstamp *cl)
387{
388 u64 val = local_clock();
389
390 do_div(val, NSEC_PER_USEC);
391 cl->stamp_us = (u32)val;
392 cl->stamp_jiffies = (u32)jiffies;
393}
394
395/**
396 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
397 * @t1: pointer to newest sample
398 * @t0: pointer to oldest sample
399 */
400static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
401 const struct skb_mstamp *t0)
402{
403 s32 delta_us = t1->stamp_us - t0->stamp_us;
404 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
405
406 /* If delta_us is negative, this might be because interval is too big,
407 * or local_clock() drift is too big : fallback using jiffies.
408 */
409 if (delta_us <= 0 ||
410 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
411
412 delta_us = jiffies_to_usecs(delta_jiffies);
413
414 return delta_us;
415}
416
417
1da177e4
LT
418/**
419 * struct sk_buff - socket buffer
420 * @next: Next buffer in list
421 * @prev: Previous buffer in list
363ec392 422 * @tstamp: Time we arrived/left
d84e0bd7 423 * @sk: Socket we are owned by
1da177e4 424 * @dev: Device we arrived on/are leaving by
d84e0bd7 425 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 426 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 427 * @sp: the security path, used for xfrm
1da177e4
LT
428 * @len: Length of actual data
429 * @data_len: Data length
430 * @mac_len: Length of link layer header
334a8132 431 * @hdr_len: writable header length of cloned skb
663ead3b
HX
432 * @csum: Checksum (must include start/offset pair)
433 * @csum_start: Offset from skb->head where checksumming should start
434 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 435 * @priority: Packet queueing priority
60ff7467 436 * @ignore_df: allow local fragmentation
1da177e4 437 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 438 * @ip_summed: Driver fed us an IP checksum
1da177e4 439 * @nohdr: Payload reference only, must not modify header
d84e0bd7 440 * @nfctinfo: Relationship of this skb to the connection
1da177e4 441 * @pkt_type: Packet class
c83c2486 442 * @fclone: skbuff clone status
c83c2486 443 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
444 * @peeked: this packet has been seen already, so stats have been
445 * done for it, don't do them again
ba9dda3a 446 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
447 * @protocol: Packet protocol from driver
448 * @destructor: Destruct function
449 * @nfct: Associated connection, if any
1da177e4 450 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 451 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
452 * @tc_index: Traffic control index
453 * @tc_verd: traffic control verdict
61b905da 454 * @hash: the packet hash
d84e0bd7 455 * @queue_mapping: Queue mapping for multiqueue devices
553a5672 456 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 457 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 458 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 459 * ports.
a3b18ddb 460 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
461 * @wifi_acked_valid: wifi_acked was set
462 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 463 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
f4b8ea78
RD
464 * @dma_cookie: a cookie to one of several possible DMA operations
465 * done by skb DMA functions
06021292 466 * @napi_id: id of the NAPI struct this skb came from
984bc16c 467 * @secmark: security marking
d84e0bd7
DB
468 * @mark: Generic packet mark
469 * @dropcount: total number of sk_receive_queue overflows
86a9bad3 470 * @vlan_proto: vlan encapsulation protocol
6aa895b0 471 * @vlan_tci: vlan tag control information
0d89d203 472 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
473 * @inner_transport_header: Inner transport layer header (encapsulation)
474 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 475 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
476 * @transport_header: Transport layer header
477 * @network_header: Network layer header
478 * @mac_header: Link layer header
479 * @tail: Tail pointer
480 * @end: End pointer
481 * @head: Head of buffer
482 * @data: Data head pointer
483 * @truesize: Buffer size
484 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
485 */
486
487struct sk_buff {
488 /* These two members must be first. */
489 struct sk_buff *next;
490 struct sk_buff *prev;
491
363ec392
ED
492 union {
493 ktime_t tstamp;
494 struct skb_mstamp skb_mstamp;
495 };
da3f5cf1
FF
496
497 struct sock *sk;
1da177e4 498 struct net_device *dev;
1da177e4 499
1da177e4
LT
500 /*
501 * This is the control buffer. It is free to use for every
502 * layer. Please put your private variables there. If you
503 * want to keep them across layers you have to do a skb_clone()
504 * first. This is owned by whoever has the skb queued ATM.
505 */
da3f5cf1 506 char cb[48] __aligned(8);
1da177e4 507
7fee226a 508 unsigned long _skb_refdst;
da3f5cf1
FF
509#ifdef CONFIG_XFRM
510 struct sec_path *sp;
511#endif
1da177e4 512 unsigned int len,
334a8132
PM
513 data_len;
514 __u16 mac_len,
515 hdr_len;
ff1dcadb
AV
516 union {
517 __wsum csum;
663ead3b
HX
518 struct {
519 __u16 csum_start;
520 __u16 csum_offset;
521 };
ff1dcadb 522 };
1da177e4 523 __u32 priority;
fe55f6d5 524 kmemcheck_bitfield_begin(flags1);
60ff7467 525 __u8 ignore_df:1,
1cbb3380
TG
526 cloned:1,
527 ip_summed:2,
6869c4d8
HW
528 nohdr:1,
529 nfctinfo:3;
d179cd12 530 __u8 pkt_type:3,
b84f4cc9 531 fclone:2,
ba9dda3a 532 ipvs_property:1,
a59322be 533 peeked:1,
ba9dda3a 534 nf_trace:1;
fe55f6d5 535 kmemcheck_bitfield_end(flags1);
4ab408de 536 __be16 protocol;
1da177e4
LT
537
538 void (*destructor)(struct sk_buff *skb);
9fb9cbb1 539#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 540 struct nf_conntrack *nfct;
2fc72c7b 541#endif
1da177e4
LT
542#ifdef CONFIG_BRIDGE_NETFILTER
543 struct nf_bridge_info *nf_bridge;
544#endif
f25f4e44 545
8964be4a 546 int skb_iif;
4031ae6e 547
61b905da 548 __u32 hash;
4031ae6e 549
86a9bad3 550 __be16 vlan_proto;
4031ae6e
AD
551 __u16 vlan_tci;
552
1da177e4 553#ifdef CONFIG_NET_SCHED
b6b99eb5 554 __u16 tc_index; /* traffic control index */
1da177e4 555#ifdef CONFIG_NET_CLS_ACT
b6b99eb5 556 __u16 tc_verd; /* traffic control verdict */
1da177e4 557#endif
1da177e4 558#endif
fe55f6d5 559
0a14842f 560 __u16 queue_mapping;
fe55f6d5 561 kmemcheck_bitfield_begin(flags2);
de357cc0 562#ifdef CONFIG_IPV6_NDISC_NODETYPE
8a4eb573 563 __u8 ndisc_nodetype:2;
d0f09804 564#endif
c93bdd0e 565 __u8 pfmemalloc:1;
3853b584 566 __u8 ooo_okay:1;
61b905da 567 __u8 l4_hash:1;
a3b18ddb 568 __u8 sw_hash:1;
6e3e939f
JB
569 __u8 wifi_acked_valid:1;
570 __u8 wifi_acked:1;
3bdc0eba 571 __u8 no_fcs:1;
d3836f21 572 __u8 head_frag:1;
6a674e9c
JG
573 /* Encapsulation protocol and NIC drivers should use
574 * this flag to indicate to each other if the skb contains
575 * encapsulated packet or not and maybe use the inner packet
576 * headers if needed
577 */
578 __u8 encapsulation:1;
7e2b10c1 579 __u8 encap_hdr_csum:1;
5d0c2b95 580 __u8 csum_valid:1;
7e3cead5 581 __u8 csum_complete_sw:1;
a3b18ddb 582 /* 2/4 bit hole (depending on ndisc_nodetype presence) */
fe55f6d5
VN
583 kmemcheck_bitfield_end(flags2);
584
e0d1095a 585#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
06021292
ET
586 union {
587 unsigned int napi_id;
588 dma_cookie_t dma_cookie;
589 };
97fc2f08 590#endif
984bc16c
JM
591#ifdef CONFIG_NETWORK_SECMARK
592 __u32 secmark;
593#endif
3b885787
NH
594 union {
595 __u32 mark;
596 __u32 dropcount;
16fad69c 597 __u32 reserved_tailroom;
3b885787 598 };
1da177e4 599
0d89d203 600 __be16 inner_protocol;
1a37e412
SH
601 __u16 inner_transport_header;
602 __u16 inner_network_header;
603 __u16 inner_mac_header;
604 __u16 transport_header;
605 __u16 network_header;
606 __u16 mac_header;
1da177e4 607 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 608 sk_buff_data_t tail;
4305b541 609 sk_buff_data_t end;
1da177e4 610 unsigned char *head,
4305b541 611 *data;
27a884dc
ACM
612 unsigned int truesize;
613 atomic_t users;
1da177e4
LT
614};
615
616#ifdef __KERNEL__
617/*
618 * Handling routines are only of interest to the kernel
619 */
620#include <linux/slab.h>
621
1da177e4 622
c93bdd0e
MG
623#define SKB_ALLOC_FCLONE 0x01
624#define SKB_ALLOC_RX 0x02
625
626/* Returns true if the skb was allocated from PFMEMALLOC reserves */
627static inline bool skb_pfmemalloc(const struct sk_buff *skb)
628{
629 return unlikely(skb->pfmemalloc);
630}
631
7fee226a
ED
632/*
633 * skb might have a dst pointer attached, refcounted or not.
634 * _skb_refdst low order bit is set if refcount was _not_ taken
635 */
636#define SKB_DST_NOREF 1UL
637#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
638
639/**
640 * skb_dst - returns skb dst_entry
641 * @skb: buffer
642 *
643 * Returns skb dst_entry, regardless of reference taken or not.
644 */
adf30907
ED
645static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
646{
7fee226a
ED
647 /* If refdst was not refcounted, check we still are in a
648 * rcu_read_lock section
649 */
650 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
651 !rcu_read_lock_held() &&
652 !rcu_read_lock_bh_held());
653 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
654}
655
7fee226a
ED
656/**
657 * skb_dst_set - sets skb dst
658 * @skb: buffer
659 * @dst: dst entry
660 *
661 * Sets skb dst, assuming a reference was taken on dst and should
662 * be released by skb_dst_drop()
663 */
adf30907
ED
664static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
665{
7fee226a
ED
666 skb->_skb_refdst = (unsigned long)dst;
667}
668
7965bd4d
JP
669void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
670 bool force);
932bc4d7
JA
671
672/**
673 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
674 * @skb: buffer
675 * @dst: dst entry
676 *
677 * Sets skb dst, assuming a reference was not taken on dst.
678 * If dst entry is cached, we do not take reference and dst_release
679 * will be avoided by refdst_drop. If dst entry is not cached, we take
680 * reference, so that last dst_release can destroy the dst immediately.
681 */
682static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
683{
684 __skb_dst_set_noref(skb, dst, false);
685}
686
687/**
688 * skb_dst_set_noref_force - sets skb dst, without taking reference
689 * @skb: buffer
690 * @dst: dst entry
691 *
692 * Sets skb dst, assuming a reference was not taken on dst.
693 * No reference is taken and no dst_release will be called. While for
694 * cached dsts deferred reclaim is a basic feature, for entries that are
695 * not cached it is caller's job to guarantee that last dst_release for
696 * provided dst happens when nobody uses it, eg. after a RCU grace period.
697 */
698static inline void skb_dst_set_noref_force(struct sk_buff *skb,
699 struct dst_entry *dst)
700{
701 __skb_dst_set_noref(skb, dst, true);
702}
7fee226a
ED
703
704/**
25985edc 705 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
706 * @skb: buffer
707 */
708static inline bool skb_dst_is_noref(const struct sk_buff *skb)
709{
710 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
711}
712
511c3f92
ED
713static inline struct rtable *skb_rtable(const struct sk_buff *skb)
714{
adf30907 715 return (struct rtable *)skb_dst(skb);
511c3f92
ED
716}
717
7965bd4d
JP
718void kfree_skb(struct sk_buff *skb);
719void kfree_skb_list(struct sk_buff *segs);
720void skb_tx_error(struct sk_buff *skb);
721void consume_skb(struct sk_buff *skb);
722void __kfree_skb(struct sk_buff *skb);
d7e8883c 723extern struct kmem_cache *skbuff_head_cache;
bad43ca8 724
7965bd4d
JP
725void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
726bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
727 bool *fragstolen, int *delta_truesize);
bad43ca8 728
7965bd4d
JP
729struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
730 int node);
731struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 732static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 733 gfp_t priority)
d179cd12 734{
564824b0 735 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
736}
737
738static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 739 gfp_t priority)
d179cd12 740{
c93bdd0e 741 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
742}
743
7965bd4d 744struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
745static inline struct sk_buff *alloc_skb_head(gfp_t priority)
746{
747 return __alloc_skb_head(priority, -1);
748}
749
7965bd4d
JP
750struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
751int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
752struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
753struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
754struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
755 gfp_t gfp_mask, bool fclone);
756static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
757 gfp_t gfp_mask)
758{
759 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
760}
7965bd4d
JP
761
762int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
763struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
764 unsigned int headroom);
765struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
766 int newtailroom, gfp_t priority);
25a91d8d
FD
767int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
768 int offset, int len);
7965bd4d
JP
769int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
770 int len);
771int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
772int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 773#define dev_kfree_skb(a) consume_skb(a)
1da177e4 774
7965bd4d
JP
775int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
776 int getfrag(void *from, char *to, int offset,
777 int len, int odd, struct sk_buff *skb),
778 void *from, int length);
e89e9cf5 779
d94d9fee 780struct skb_seq_state {
677e90ed
TG
781 __u32 lower_offset;
782 __u32 upper_offset;
783 __u32 frag_idx;
784 __u32 stepped_offset;
785 struct sk_buff *root_skb;
786 struct sk_buff *cur_skb;
787 __u8 *frag_data;
788};
789
7965bd4d
JP
790void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
791 unsigned int to, struct skb_seq_state *st);
792unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
793 struct skb_seq_state *st);
794void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 795
7965bd4d
JP
796unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
797 unsigned int to, struct ts_config *config,
798 struct ts_state *state);
3fc7e8a6 799
09323cc4
TH
800/*
801 * Packet hash types specify the type of hash in skb_set_hash.
802 *
803 * Hash types refer to the protocol layer addresses which are used to
804 * construct a packet's hash. The hashes are used to differentiate or identify
805 * flows of the protocol layer for the hash type. Hash types are either
806 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
807 *
808 * Properties of hashes:
809 *
810 * 1) Two packets in different flows have different hash values
811 * 2) Two packets in the same flow should have the same hash value
812 *
813 * A hash at a higher layer is considered to be more specific. A driver should
814 * set the most specific hash possible.
815 *
816 * A driver cannot indicate a more specific hash than the layer at which a hash
817 * was computed. For instance an L3 hash cannot be set as an L4 hash.
818 *
819 * A driver may indicate a hash level which is less specific than the
820 * actual layer the hash was computed on. For instance, a hash computed
821 * at L4 may be considered an L3 hash. This should only be done if the
822 * driver can't unambiguously determine that the HW computed the hash at
823 * the higher layer. Note that the "should" in the second property above
824 * permits this.
825 */
826enum pkt_hash_types {
827 PKT_HASH_TYPE_NONE, /* Undefined type */
828 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
829 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
830 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
831};
832
833static inline void
834skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
835{
61b905da 836 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 837 skb->sw_hash = 0;
61b905da 838 skb->hash = hash;
09323cc4
TH
839}
840
3958afa1
TH
841void __skb_get_hash(struct sk_buff *skb);
842static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 843{
a3b18ddb 844 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 845 __skb_get_hash(skb);
bfb564e7 846
61b905da 847 return skb->hash;
bfb564e7
KK
848}
849
57bdf7f4
TH
850static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
851{
61b905da 852 return skb->hash;
57bdf7f4
TH
853}
854
7539fadc
TH
855static inline void skb_clear_hash(struct sk_buff *skb)
856{
61b905da 857 skb->hash = 0;
a3b18ddb 858 skb->sw_hash = 0;
61b905da 859 skb->l4_hash = 0;
7539fadc
TH
860}
861
862static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
863{
61b905da 864 if (!skb->l4_hash)
7539fadc
TH
865 skb_clear_hash(skb);
866}
867
3df7a74e
TH
868static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
869{
61b905da 870 to->hash = from->hash;
a3b18ddb 871 to->sw_hash = from->sw_hash;
61b905da 872 to->l4_hash = from->l4_hash;
3df7a74e
TH
873};
874
4305b541
ACM
875#ifdef NET_SKBUFF_DATA_USES_OFFSET
876static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
877{
878 return skb->head + skb->end;
879}
ec47ea82
AD
880
881static inline unsigned int skb_end_offset(const struct sk_buff *skb)
882{
883 return skb->end;
884}
4305b541
ACM
885#else
886static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
887{
888 return skb->end;
889}
ec47ea82
AD
890
891static inline unsigned int skb_end_offset(const struct sk_buff *skb)
892{
893 return skb->end - skb->head;
894}
4305b541
ACM
895#endif
896
1da177e4 897/* Internal */
4305b541 898#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 899
ac45f602
PO
900static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
901{
902 return &skb_shinfo(skb)->hwtstamps;
903}
904
1da177e4
LT
905/**
906 * skb_queue_empty - check if a queue is empty
907 * @list: queue head
908 *
909 * Returns true if the queue is empty, false otherwise.
910 */
911static inline int skb_queue_empty(const struct sk_buff_head *list)
912{
fd44b93c 913 return list->next == (const struct sk_buff *) list;
1da177e4
LT
914}
915
fc7ebb21
DM
916/**
917 * skb_queue_is_last - check if skb is the last entry in the queue
918 * @list: queue head
919 * @skb: buffer
920 *
921 * Returns true if @skb is the last buffer on the list.
922 */
923static inline bool skb_queue_is_last(const struct sk_buff_head *list,
924 const struct sk_buff *skb)
925{
fd44b93c 926 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
927}
928
832d11c5
IJ
929/**
930 * skb_queue_is_first - check if skb is the first entry in the queue
931 * @list: queue head
932 * @skb: buffer
933 *
934 * Returns true if @skb is the first buffer on the list.
935 */
936static inline bool skb_queue_is_first(const struct sk_buff_head *list,
937 const struct sk_buff *skb)
938{
fd44b93c 939 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
940}
941
249c8b42
DM
942/**
943 * skb_queue_next - return the next packet in the queue
944 * @list: queue head
945 * @skb: current buffer
946 *
947 * Return the next packet in @list after @skb. It is only valid to
948 * call this if skb_queue_is_last() evaluates to false.
949 */
950static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
951 const struct sk_buff *skb)
952{
953 /* This BUG_ON may seem severe, but if we just return then we
954 * are going to dereference garbage.
955 */
956 BUG_ON(skb_queue_is_last(list, skb));
957 return skb->next;
958}
959
832d11c5
IJ
960/**
961 * skb_queue_prev - return the prev packet in the queue
962 * @list: queue head
963 * @skb: current buffer
964 *
965 * Return the prev packet in @list before @skb. It is only valid to
966 * call this if skb_queue_is_first() evaluates to false.
967 */
968static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
969 const struct sk_buff *skb)
970{
971 /* This BUG_ON may seem severe, but if we just return then we
972 * are going to dereference garbage.
973 */
974 BUG_ON(skb_queue_is_first(list, skb));
975 return skb->prev;
976}
977
1da177e4
LT
978/**
979 * skb_get - reference buffer
980 * @skb: buffer to reference
981 *
982 * Makes another reference to a socket buffer and returns a pointer
983 * to the buffer.
984 */
985static inline struct sk_buff *skb_get(struct sk_buff *skb)
986{
987 atomic_inc(&skb->users);
988 return skb;
989}
990
991/*
992 * If users == 1, we are the only owner and are can avoid redundant
993 * atomic change.
994 */
995
1da177e4
LT
996/**
997 * skb_cloned - is the buffer a clone
998 * @skb: buffer to check
999 *
1000 * Returns true if the buffer was generated with skb_clone() and is
1001 * one of multiple shared copies of the buffer. Cloned buffers are
1002 * shared data so must not be written to under normal circumstances.
1003 */
1004static inline int skb_cloned(const struct sk_buff *skb)
1005{
1006 return skb->cloned &&
1007 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1008}
1009
14bbd6a5
PS
1010static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1011{
1012 might_sleep_if(pri & __GFP_WAIT);
1013
1014 if (skb_cloned(skb))
1015 return pskb_expand_head(skb, 0, 0, pri);
1016
1017 return 0;
1018}
1019
1da177e4
LT
1020/**
1021 * skb_header_cloned - is the header a clone
1022 * @skb: buffer to check
1023 *
1024 * Returns true if modifying the header part of the buffer requires
1025 * the data to be copied.
1026 */
1027static inline int skb_header_cloned(const struct sk_buff *skb)
1028{
1029 int dataref;
1030
1031 if (!skb->cloned)
1032 return 0;
1033
1034 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1035 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1036 return dataref != 1;
1037}
1038
1039/**
1040 * skb_header_release - release reference to header
1041 * @skb: buffer to operate on
1042 *
1043 * Drop a reference to the header part of the buffer. This is done
1044 * by acquiring a payload reference. You must not read from the header
1045 * part of skb->data after this.
1046 */
1047static inline void skb_header_release(struct sk_buff *skb)
1048{
1049 BUG_ON(skb->nohdr);
1050 skb->nohdr = 1;
1051 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1052}
1053
1054/**
1055 * skb_shared - is the buffer shared
1056 * @skb: buffer to check
1057 *
1058 * Returns true if more than one person has a reference to this
1059 * buffer.
1060 */
1061static inline int skb_shared(const struct sk_buff *skb)
1062{
1063 return atomic_read(&skb->users) != 1;
1064}
1065
1066/**
1067 * skb_share_check - check if buffer is shared and if so clone it
1068 * @skb: buffer to check
1069 * @pri: priority for memory allocation
1070 *
1071 * If the buffer is shared the buffer is cloned and the old copy
1072 * drops a reference. A new clone with a single reference is returned.
1073 * If the buffer is not shared the original buffer is returned. When
1074 * being called from interrupt status or with spinlocks held pri must
1075 * be GFP_ATOMIC.
1076 *
1077 * NULL is returned on a memory allocation failure.
1078 */
47061bc4 1079static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1080{
1081 might_sleep_if(pri & __GFP_WAIT);
1082 if (skb_shared(skb)) {
1083 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1084
1085 if (likely(nskb))
1086 consume_skb(skb);
1087 else
1088 kfree_skb(skb);
1da177e4
LT
1089 skb = nskb;
1090 }
1091 return skb;
1092}
1093
1094/*
1095 * Copy shared buffers into a new sk_buff. We effectively do COW on
1096 * packets to handle cases where we have a local reader and forward
1097 * and a couple of other messy ones. The normal one is tcpdumping
1098 * a packet thats being forwarded.
1099 */
1100
1101/**
1102 * skb_unshare - make a copy of a shared buffer
1103 * @skb: buffer to check
1104 * @pri: priority for memory allocation
1105 *
1106 * If the socket buffer is a clone then this function creates a new
1107 * copy of the data, drops a reference count on the old copy and returns
1108 * the new copy with the reference count at 1. If the buffer is not a clone
1109 * the original buffer is returned. When called with a spinlock held or
1110 * from interrupt state @pri must be %GFP_ATOMIC
1111 *
1112 * %NULL is returned on a memory allocation failure.
1113 */
e2bf521d 1114static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1115 gfp_t pri)
1da177e4
LT
1116{
1117 might_sleep_if(pri & __GFP_WAIT);
1118 if (skb_cloned(skb)) {
1119 struct sk_buff *nskb = skb_copy(skb, pri);
1120 kfree_skb(skb); /* Free our shared copy */
1121 skb = nskb;
1122 }
1123 return skb;
1124}
1125
1126/**
1a5778aa 1127 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1128 * @list_: list to peek at
1129 *
1130 * Peek an &sk_buff. Unlike most other operations you _MUST_
1131 * be careful with this one. A peek leaves the buffer on the
1132 * list and someone else may run off with it. You must hold
1133 * the appropriate locks or have a private queue to do this.
1134 *
1135 * Returns %NULL for an empty list or a pointer to the head element.
1136 * The reference count is not incremented and the reference is therefore
1137 * volatile. Use with caution.
1138 */
05bdd2f1 1139static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1140{
18d07000
ED
1141 struct sk_buff *skb = list_->next;
1142
1143 if (skb == (struct sk_buff *)list_)
1144 skb = NULL;
1145 return skb;
1da177e4
LT
1146}
1147
da5ef6e5
PE
1148/**
1149 * skb_peek_next - peek skb following the given one from a queue
1150 * @skb: skb to start from
1151 * @list_: list to peek at
1152 *
1153 * Returns %NULL when the end of the list is met or a pointer to the
1154 * next element. The reference count is not incremented and the
1155 * reference is therefore volatile. Use with caution.
1156 */
1157static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1158 const struct sk_buff_head *list_)
1159{
1160 struct sk_buff *next = skb->next;
18d07000 1161
da5ef6e5
PE
1162 if (next == (struct sk_buff *)list_)
1163 next = NULL;
1164 return next;
1165}
1166
1da177e4 1167/**
1a5778aa 1168 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1169 * @list_: list to peek at
1170 *
1171 * Peek an &sk_buff. Unlike most other operations you _MUST_
1172 * be careful with this one. A peek leaves the buffer on the
1173 * list and someone else may run off with it. You must hold
1174 * the appropriate locks or have a private queue to do this.
1175 *
1176 * Returns %NULL for an empty list or a pointer to the tail element.
1177 * The reference count is not incremented and the reference is therefore
1178 * volatile. Use with caution.
1179 */
05bdd2f1 1180static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1181{
18d07000
ED
1182 struct sk_buff *skb = list_->prev;
1183
1184 if (skb == (struct sk_buff *)list_)
1185 skb = NULL;
1186 return skb;
1187
1da177e4
LT
1188}
1189
1190/**
1191 * skb_queue_len - get queue length
1192 * @list_: list to measure
1193 *
1194 * Return the length of an &sk_buff queue.
1195 */
1196static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1197{
1198 return list_->qlen;
1199}
1200
67fed459
DM
1201/**
1202 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1203 * @list: queue to initialize
1204 *
1205 * This initializes only the list and queue length aspects of
1206 * an sk_buff_head object. This allows to initialize the list
1207 * aspects of an sk_buff_head without reinitializing things like
1208 * the spinlock. It can also be used for on-stack sk_buff_head
1209 * objects where the spinlock is known to not be used.
1210 */
1211static inline void __skb_queue_head_init(struct sk_buff_head *list)
1212{
1213 list->prev = list->next = (struct sk_buff *)list;
1214 list->qlen = 0;
1215}
1216
76f10ad0
AV
1217/*
1218 * This function creates a split out lock class for each invocation;
1219 * this is needed for now since a whole lot of users of the skb-queue
1220 * infrastructure in drivers have different locking usage (in hardirq)
1221 * than the networking core (in softirq only). In the long run either the
1222 * network layer or drivers should need annotation to consolidate the
1223 * main types of usage into 3 classes.
1224 */
1da177e4
LT
1225static inline void skb_queue_head_init(struct sk_buff_head *list)
1226{
1227 spin_lock_init(&list->lock);
67fed459 1228 __skb_queue_head_init(list);
1da177e4
LT
1229}
1230
c2ecba71
PE
1231static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1232 struct lock_class_key *class)
1233{
1234 skb_queue_head_init(list);
1235 lockdep_set_class(&list->lock, class);
1236}
1237
1da177e4 1238/*
bf299275 1239 * Insert an sk_buff on a list.
1da177e4
LT
1240 *
1241 * The "__skb_xxxx()" functions are the non-atomic ones that
1242 * can only be called with interrupts disabled.
1243 */
7965bd4d
JP
1244void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1245 struct sk_buff_head *list);
bf299275
GR
1246static inline void __skb_insert(struct sk_buff *newsk,
1247 struct sk_buff *prev, struct sk_buff *next,
1248 struct sk_buff_head *list)
1249{
1250 newsk->next = next;
1251 newsk->prev = prev;
1252 next->prev = prev->next = newsk;
1253 list->qlen++;
1254}
1da177e4 1255
67fed459
DM
1256static inline void __skb_queue_splice(const struct sk_buff_head *list,
1257 struct sk_buff *prev,
1258 struct sk_buff *next)
1259{
1260 struct sk_buff *first = list->next;
1261 struct sk_buff *last = list->prev;
1262
1263 first->prev = prev;
1264 prev->next = first;
1265
1266 last->next = next;
1267 next->prev = last;
1268}
1269
1270/**
1271 * skb_queue_splice - join two skb lists, this is designed for stacks
1272 * @list: the new list to add
1273 * @head: the place to add it in the first list
1274 */
1275static inline void skb_queue_splice(const struct sk_buff_head *list,
1276 struct sk_buff_head *head)
1277{
1278 if (!skb_queue_empty(list)) {
1279 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1280 head->qlen += list->qlen;
67fed459
DM
1281 }
1282}
1283
1284/**
d9619496 1285 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1286 * @list: the new list to add
1287 * @head: the place to add it in the first list
1288 *
1289 * The list at @list is reinitialised
1290 */
1291static inline void skb_queue_splice_init(struct sk_buff_head *list,
1292 struct sk_buff_head *head)
1293{
1294 if (!skb_queue_empty(list)) {
1295 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1296 head->qlen += list->qlen;
67fed459
DM
1297 __skb_queue_head_init(list);
1298 }
1299}
1300
1301/**
1302 * skb_queue_splice_tail - join two skb lists, each list being a queue
1303 * @list: the new list to add
1304 * @head: the place to add it in the first list
1305 */
1306static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1307 struct sk_buff_head *head)
1308{
1309 if (!skb_queue_empty(list)) {
1310 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1311 head->qlen += list->qlen;
67fed459
DM
1312 }
1313}
1314
1315/**
d9619496 1316 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1317 * @list: the new list to add
1318 * @head: the place to add it in the first list
1319 *
1320 * Each of the lists is a queue.
1321 * The list at @list is reinitialised
1322 */
1323static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1324 struct sk_buff_head *head)
1325{
1326 if (!skb_queue_empty(list)) {
1327 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1328 head->qlen += list->qlen;
67fed459
DM
1329 __skb_queue_head_init(list);
1330 }
1331}
1332
1da177e4 1333/**
300ce174 1334 * __skb_queue_after - queue a buffer at the list head
1da177e4 1335 * @list: list to use
300ce174 1336 * @prev: place after this buffer
1da177e4
LT
1337 * @newsk: buffer to queue
1338 *
300ce174 1339 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1340 * and you must therefore hold required locks before calling it.
1341 *
1342 * A buffer cannot be placed on two lists at the same time.
1343 */
300ce174
SH
1344static inline void __skb_queue_after(struct sk_buff_head *list,
1345 struct sk_buff *prev,
1346 struct sk_buff *newsk)
1da177e4 1347{
bf299275 1348 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1349}
1350
7965bd4d
JP
1351void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1352 struct sk_buff_head *list);
7de6c033 1353
f5572855
GR
1354static inline void __skb_queue_before(struct sk_buff_head *list,
1355 struct sk_buff *next,
1356 struct sk_buff *newsk)
1357{
1358 __skb_insert(newsk, next->prev, next, list);
1359}
1360
300ce174
SH
1361/**
1362 * __skb_queue_head - queue a buffer at the list head
1363 * @list: list to use
1364 * @newsk: buffer to queue
1365 *
1366 * Queue a buffer at the start of a list. This function takes no locks
1367 * and you must therefore hold required locks before calling it.
1368 *
1369 * A buffer cannot be placed on two lists at the same time.
1370 */
7965bd4d 1371void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1372static inline void __skb_queue_head(struct sk_buff_head *list,
1373 struct sk_buff *newsk)
1374{
1375 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1376}
1377
1da177e4
LT
1378/**
1379 * __skb_queue_tail - queue a buffer at the list tail
1380 * @list: list to use
1381 * @newsk: buffer to queue
1382 *
1383 * Queue a buffer at the end of a list. This function takes no locks
1384 * and you must therefore hold required locks before calling it.
1385 *
1386 * A buffer cannot be placed on two lists at the same time.
1387 */
7965bd4d 1388void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1389static inline void __skb_queue_tail(struct sk_buff_head *list,
1390 struct sk_buff *newsk)
1391{
f5572855 1392 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1393}
1394
1da177e4
LT
1395/*
1396 * remove sk_buff from list. _Must_ be called atomically, and with
1397 * the list known..
1398 */
7965bd4d 1399void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1400static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1401{
1402 struct sk_buff *next, *prev;
1403
1404 list->qlen--;
1405 next = skb->next;
1406 prev = skb->prev;
1407 skb->next = skb->prev = NULL;
1da177e4
LT
1408 next->prev = prev;
1409 prev->next = next;
1410}
1411
f525c06d
GR
1412/**
1413 * __skb_dequeue - remove from the head of the queue
1414 * @list: list to dequeue from
1415 *
1416 * Remove the head of the list. This function does not take any locks
1417 * so must be used with appropriate locks held only. The head item is
1418 * returned or %NULL if the list is empty.
1419 */
7965bd4d 1420struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1421static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1422{
1423 struct sk_buff *skb = skb_peek(list);
1424 if (skb)
1425 __skb_unlink(skb, list);
1426 return skb;
1427}
1da177e4
LT
1428
1429/**
1430 * __skb_dequeue_tail - remove from the tail of the queue
1431 * @list: list to dequeue from
1432 *
1433 * Remove the tail of the list. This function does not take any locks
1434 * so must be used with appropriate locks held only. The tail item is
1435 * returned or %NULL if the list is empty.
1436 */
7965bd4d 1437struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1438static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1439{
1440 struct sk_buff *skb = skb_peek_tail(list);
1441 if (skb)
1442 __skb_unlink(skb, list);
1443 return skb;
1444}
1445
1446
bdcc0924 1447static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1448{
1449 return skb->data_len;
1450}
1451
1452static inline unsigned int skb_headlen(const struct sk_buff *skb)
1453{
1454 return skb->len - skb->data_len;
1455}
1456
1457static inline int skb_pagelen(const struct sk_buff *skb)
1458{
1459 int i, len = 0;
1460
1461 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1462 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1463 return len + skb_headlen(skb);
1464}
1465
131ea667
IC
1466/**
1467 * __skb_fill_page_desc - initialise a paged fragment in an skb
1468 * @skb: buffer containing fragment to be initialised
1469 * @i: paged fragment index to initialise
1470 * @page: the page to use for this fragment
1471 * @off: the offset to the data with @page
1472 * @size: the length of the data
1473 *
1474 * Initialises the @i'th fragment of @skb to point to &size bytes at
1475 * offset @off within @page.
1476 *
1477 * Does not take any additional reference on the fragment.
1478 */
1479static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1480 struct page *page, int off, int size)
1da177e4
LT
1481{
1482 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1483
c48a11c7
MG
1484 /*
1485 * Propagate page->pfmemalloc to the skb if we can. The problem is
1486 * that not all callers have unique ownership of the page. If
1487 * pfmemalloc is set, we check the mapping as a mapping implies
1488 * page->index is set (index and pfmemalloc share space).
1489 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1490 * do not lose pfmemalloc information as the pages would not be
1491 * allocated using __GFP_MEMALLOC.
1492 */
a8605c60 1493 frag->page.p = page;
1da177e4 1494 frag->page_offset = off;
9e903e08 1495 skb_frag_size_set(frag, size);
cca7af38
PE
1496
1497 page = compound_head(page);
1498 if (page->pfmemalloc && !page->mapping)
1499 skb->pfmemalloc = true;
131ea667
IC
1500}
1501
1502/**
1503 * skb_fill_page_desc - initialise a paged fragment in an skb
1504 * @skb: buffer containing fragment to be initialised
1505 * @i: paged fragment index to initialise
1506 * @page: the page to use for this fragment
1507 * @off: the offset to the data with @page
1508 * @size: the length of the data
1509 *
1510 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1511 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1512 * addition updates @skb such that @i is the last fragment.
1513 *
1514 * Does not take any additional reference on the fragment.
1515 */
1516static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1517 struct page *page, int off, int size)
1518{
1519 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1520 skb_shinfo(skb)->nr_frags = i + 1;
1521}
1522
7965bd4d
JP
1523void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1524 int size, unsigned int truesize);
654bed16 1525
f8e617e1
JW
1526void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1527 unsigned int truesize);
1528
1da177e4 1529#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1530#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1531#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1532
27a884dc
ACM
1533#ifdef NET_SKBUFF_DATA_USES_OFFSET
1534static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1535{
1536 return skb->head + skb->tail;
1537}
1538
1539static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1540{
1541 skb->tail = skb->data - skb->head;
1542}
1543
1544static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1545{
1546 skb_reset_tail_pointer(skb);
1547 skb->tail += offset;
1548}
7cc46190 1549
27a884dc
ACM
1550#else /* NET_SKBUFF_DATA_USES_OFFSET */
1551static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1552{
1553 return skb->tail;
1554}
1555
1556static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1557{
1558 skb->tail = skb->data;
1559}
1560
1561static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1562{
1563 skb->tail = skb->data + offset;
1564}
4305b541 1565
27a884dc
ACM
1566#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1567
1da177e4
LT
1568/*
1569 * Add data to an sk_buff
1570 */
0c7ddf36 1571unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1572unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1573static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1574{
27a884dc 1575 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1576 SKB_LINEAR_ASSERT(skb);
1577 skb->tail += len;
1578 skb->len += len;
1579 return tmp;
1580}
1581
7965bd4d 1582unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1583static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1584{
1585 skb->data -= len;
1586 skb->len += len;
1587 return skb->data;
1588}
1589
7965bd4d 1590unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1591static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1592{
1593 skb->len -= len;
1594 BUG_ON(skb->len < skb->data_len);
1595 return skb->data += len;
1596}
1597
47d29646
DM
1598static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1599{
1600 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1601}
1602
7965bd4d 1603unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1604
1605static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1606{
1607 if (len > skb_headlen(skb) &&
987c402a 1608 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1609 return NULL;
1610 skb->len -= len;
1611 return skb->data += len;
1612}
1613
1614static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1615{
1616 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1617}
1618
1619static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1620{
1621 if (likely(len <= skb_headlen(skb)))
1622 return 1;
1623 if (unlikely(len > skb->len))
1624 return 0;
987c402a 1625 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1626}
1627
1628/**
1629 * skb_headroom - bytes at buffer head
1630 * @skb: buffer to check
1631 *
1632 * Return the number of bytes of free space at the head of an &sk_buff.
1633 */
c2636b4d 1634static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1635{
1636 return skb->data - skb->head;
1637}
1638
1639/**
1640 * skb_tailroom - bytes at buffer end
1641 * @skb: buffer to check
1642 *
1643 * Return the number of bytes of free space at the tail of an sk_buff
1644 */
1645static inline int skb_tailroom(const struct sk_buff *skb)
1646{
4305b541 1647 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1648}
1649
a21d4572
ED
1650/**
1651 * skb_availroom - bytes at buffer end
1652 * @skb: buffer to check
1653 *
1654 * Return the number of bytes of free space at the tail of an sk_buff
1655 * allocated by sk_stream_alloc()
1656 */
1657static inline int skb_availroom(const struct sk_buff *skb)
1658{
16fad69c
ED
1659 if (skb_is_nonlinear(skb))
1660 return 0;
1661
1662 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1663}
1664
1da177e4
LT
1665/**
1666 * skb_reserve - adjust headroom
1667 * @skb: buffer to alter
1668 * @len: bytes to move
1669 *
1670 * Increase the headroom of an empty &sk_buff by reducing the tail
1671 * room. This is only allowed for an empty buffer.
1672 */
8243126c 1673static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1674{
1675 skb->data += len;
1676 skb->tail += len;
1677}
1678
6a674e9c
JG
1679static inline void skb_reset_inner_headers(struct sk_buff *skb)
1680{
aefbd2b3 1681 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1682 skb->inner_network_header = skb->network_header;
1683 skb->inner_transport_header = skb->transport_header;
1684}
1685
0b5c9db1
JP
1686static inline void skb_reset_mac_len(struct sk_buff *skb)
1687{
1688 skb->mac_len = skb->network_header - skb->mac_header;
1689}
1690
6a674e9c
JG
1691static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1692 *skb)
1693{
1694 return skb->head + skb->inner_transport_header;
1695}
1696
1697static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1698{
1699 skb->inner_transport_header = skb->data - skb->head;
1700}
1701
1702static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1703 const int offset)
1704{
1705 skb_reset_inner_transport_header(skb);
1706 skb->inner_transport_header += offset;
1707}
1708
1709static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1710{
1711 return skb->head + skb->inner_network_header;
1712}
1713
1714static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1715{
1716 skb->inner_network_header = skb->data - skb->head;
1717}
1718
1719static inline void skb_set_inner_network_header(struct sk_buff *skb,
1720 const int offset)
1721{
1722 skb_reset_inner_network_header(skb);
1723 skb->inner_network_header += offset;
1724}
1725
aefbd2b3
PS
1726static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1727{
1728 return skb->head + skb->inner_mac_header;
1729}
1730
1731static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1732{
1733 skb->inner_mac_header = skb->data - skb->head;
1734}
1735
1736static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1737 const int offset)
1738{
1739 skb_reset_inner_mac_header(skb);
1740 skb->inner_mac_header += offset;
1741}
fda55eca
ED
1742static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1743{
35d04610 1744 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1745}
1746
9c70220b
ACM
1747static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1748{
2e07fa9c 1749 return skb->head + skb->transport_header;
9c70220b
ACM
1750}
1751
badff6d0
ACM
1752static inline void skb_reset_transport_header(struct sk_buff *skb)
1753{
2e07fa9c 1754 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1755}
1756
967b05f6
ACM
1757static inline void skb_set_transport_header(struct sk_buff *skb,
1758 const int offset)
1759{
2e07fa9c
ACM
1760 skb_reset_transport_header(skb);
1761 skb->transport_header += offset;
ea2ae17d
ACM
1762}
1763
d56f90a7
ACM
1764static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1765{
2e07fa9c 1766 return skb->head + skb->network_header;
d56f90a7
ACM
1767}
1768
c1d2bbe1
ACM
1769static inline void skb_reset_network_header(struct sk_buff *skb)
1770{
2e07fa9c 1771 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1772}
1773
c14d2450
ACM
1774static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1775{
2e07fa9c
ACM
1776 skb_reset_network_header(skb);
1777 skb->network_header += offset;
c14d2450
ACM
1778}
1779
2e07fa9c 1780static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1781{
2e07fa9c 1782 return skb->head + skb->mac_header;
bbe735e4
ACM
1783}
1784
2e07fa9c 1785static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1786{
35d04610 1787 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1788}
1789
1790static inline void skb_reset_mac_header(struct sk_buff *skb)
1791{
1792 skb->mac_header = skb->data - skb->head;
1793}
1794
1795static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1796{
1797 skb_reset_mac_header(skb);
1798 skb->mac_header += offset;
1799}
1800
0e3da5bb
TT
1801static inline void skb_pop_mac_header(struct sk_buff *skb)
1802{
1803 skb->mac_header = skb->network_header;
1804}
1805
fbbdb8f0
YX
1806static inline void skb_probe_transport_header(struct sk_buff *skb,
1807 const int offset_hint)
1808{
1809 struct flow_keys keys;
1810
1811 if (skb_transport_header_was_set(skb))
1812 return;
1813 else if (skb_flow_dissect(skb, &keys))
1814 skb_set_transport_header(skb, keys.thoff);
1815 else
1816 skb_set_transport_header(skb, offset_hint);
1817}
1818
03606895
ED
1819static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1820{
1821 if (skb_mac_header_was_set(skb)) {
1822 const unsigned char *old_mac = skb_mac_header(skb);
1823
1824 skb_set_mac_header(skb, -skb->mac_len);
1825 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1826 }
1827}
1828
04fb451e
MM
1829static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1830{
1831 return skb->csum_start - skb_headroom(skb);
1832}
1833
2e07fa9c
ACM
1834static inline int skb_transport_offset(const struct sk_buff *skb)
1835{
1836 return skb_transport_header(skb) - skb->data;
1837}
1838
1839static inline u32 skb_network_header_len(const struct sk_buff *skb)
1840{
1841 return skb->transport_header - skb->network_header;
1842}
1843
6a674e9c
JG
1844static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1845{
1846 return skb->inner_transport_header - skb->inner_network_header;
1847}
1848
2e07fa9c
ACM
1849static inline int skb_network_offset(const struct sk_buff *skb)
1850{
1851 return skb_network_header(skb) - skb->data;
1852}
48d49d0c 1853
6a674e9c
JG
1854static inline int skb_inner_network_offset(const struct sk_buff *skb)
1855{
1856 return skb_inner_network_header(skb) - skb->data;
1857}
1858
f9599ce1
CG
1859static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1860{
1861 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1862}
1863
e5eb4e30
TH
1864static inline void skb_pop_rcv_encapsulation(struct sk_buff *skb)
1865{
1866 /* Only continue with checksum unnecessary if device indicated
1867 * it is valid across encapsulation (skb->encapsulation was set).
1868 */
1869 if (skb->ip_summed == CHECKSUM_UNNECESSARY && !skb->encapsulation)
1870 skb->ip_summed = CHECKSUM_NONE;
1871
1872 skb->encapsulation = 0;
1873 skb->csum_valid = 0;
1874}
1875
1da177e4
LT
1876/*
1877 * CPUs often take a performance hit when accessing unaligned memory
1878 * locations. The actual performance hit varies, it can be small if the
1879 * hardware handles it or large if we have to take an exception and fix it
1880 * in software.
1881 *
1882 * Since an ethernet header is 14 bytes network drivers often end up with
1883 * the IP header at an unaligned offset. The IP header can be aligned by
1884 * shifting the start of the packet by 2 bytes. Drivers should do this
1885 * with:
1886 *
8660c124 1887 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1888 *
1889 * The downside to this alignment of the IP header is that the DMA is now
1890 * unaligned. On some architectures the cost of an unaligned DMA is high
1891 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1892 *
1da177e4
LT
1893 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1894 * to be overridden.
1895 */
1896#ifndef NET_IP_ALIGN
1897#define NET_IP_ALIGN 2
1898#endif
1899
025be81e
AB
1900/*
1901 * The networking layer reserves some headroom in skb data (via
1902 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1903 * the header has to grow. In the default case, if the header has to grow
d6301d3d 1904 * 32 bytes or less we avoid the reallocation.
025be81e
AB
1905 *
1906 * Unfortunately this headroom changes the DMA alignment of the resulting
1907 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1908 * on some architectures. An architecture can override this value,
1909 * perhaps setting it to a cacheline in size (since that will maintain
1910 * cacheline alignment of the DMA). It must be a power of 2.
1911 *
d6301d3d 1912 * Various parts of the networking layer expect at least 32 bytes of
025be81e 1913 * headroom, you should not reduce this.
5933dd2f
ED
1914 *
1915 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1916 * to reduce average number of cache lines per packet.
1917 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 1918 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
1919 */
1920#ifndef NET_SKB_PAD
5933dd2f 1921#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
1922#endif
1923
7965bd4d 1924int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1925
1926static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1927{
c4264f27 1928 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
1929 WARN_ON(1);
1930 return;
1931 }
27a884dc
ACM
1932 skb->len = len;
1933 skb_set_tail_pointer(skb, len);
1da177e4
LT
1934}
1935
7965bd4d 1936void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1937
1938static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1939{
3cc0e873
HX
1940 if (skb->data_len)
1941 return ___pskb_trim(skb, len);
1942 __skb_trim(skb, len);
1943 return 0;
1da177e4
LT
1944}
1945
1946static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1947{
1948 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1949}
1950
e9fa4f7b
HX
1951/**
1952 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1953 * @skb: buffer to alter
1954 * @len: new length
1955 *
1956 * This is identical to pskb_trim except that the caller knows that
1957 * the skb is not cloned so we should never get an error due to out-
1958 * of-memory.
1959 */
1960static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1961{
1962 int err = pskb_trim(skb, len);
1963 BUG_ON(err);
1964}
1965
1da177e4
LT
1966/**
1967 * skb_orphan - orphan a buffer
1968 * @skb: buffer to orphan
1969 *
1970 * If a buffer currently has an owner then we call the owner's
1971 * destructor function and make the @skb unowned. The buffer continues
1972 * to exist but is no longer charged to its former owner.
1973 */
1974static inline void skb_orphan(struct sk_buff *skb)
1975{
c34a7612 1976 if (skb->destructor) {
1da177e4 1977 skb->destructor(skb);
c34a7612
ED
1978 skb->destructor = NULL;
1979 skb->sk = NULL;
376c7311
ED
1980 } else {
1981 BUG_ON(skb->sk);
c34a7612 1982 }
1da177e4
LT
1983}
1984
a353e0ce
MT
1985/**
1986 * skb_orphan_frags - orphan the frags contained in a buffer
1987 * @skb: buffer to orphan frags from
1988 * @gfp_mask: allocation mask for replacement pages
1989 *
1990 * For each frag in the SKB which needs a destructor (i.e. has an
1991 * owner) create a copy of that frag and release the original
1992 * page by calling the destructor.
1993 */
1994static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1995{
1996 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1997 return 0;
1998 return skb_copy_ubufs(skb, gfp_mask);
1999}
2000
1da177e4
LT
2001/**
2002 * __skb_queue_purge - empty a list
2003 * @list: list to empty
2004 *
2005 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2006 * the list and one reference dropped. This function does not take the
2007 * list lock and the caller must hold the relevant locks to use it.
2008 */
7965bd4d 2009void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2010static inline void __skb_queue_purge(struct sk_buff_head *list)
2011{
2012 struct sk_buff *skb;
2013 while ((skb = __skb_dequeue(list)) != NULL)
2014 kfree_skb(skb);
2015}
2016
e5e67305
AD
2017#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2018#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2019#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2020
7965bd4d 2021void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2022
7965bd4d
JP
2023struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2024 gfp_t gfp_mask);
8af27456
CH
2025
2026/**
2027 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2028 * @dev: network device to receive on
2029 * @length: length to allocate
2030 *
2031 * Allocate a new &sk_buff and assign it a usage count of one. The
2032 * buffer has unspecified headroom built in. Users should allocate
2033 * the headroom they think they need without accounting for the
2034 * built in space. The built in space is used for optimisations.
2035 *
2036 * %NULL is returned if there is no free memory. Although this function
2037 * allocates memory it can be called from an interrupt.
2038 */
2039static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2040 unsigned int length)
8af27456
CH
2041{
2042 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2043}
2044
6f532612
ED
2045/* legacy helper around __netdev_alloc_skb() */
2046static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2047 gfp_t gfp_mask)
2048{
2049 return __netdev_alloc_skb(NULL, length, gfp_mask);
2050}
2051
2052/* legacy helper around netdev_alloc_skb() */
2053static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2054{
2055 return netdev_alloc_skb(NULL, length);
2056}
2057
2058
4915a0de
ED
2059static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2060 unsigned int length, gfp_t gfp)
61321bbd 2061{
4915a0de 2062 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2063
2064 if (NET_IP_ALIGN && skb)
2065 skb_reserve(skb, NET_IP_ALIGN);
2066 return skb;
2067}
2068
4915a0de
ED
2069static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2070 unsigned int length)
2071{
2072 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2073}
2074
bc6fc9fa
FF
2075/**
2076 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
2077 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2078 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2079 * @order: size of the allocation
2080 *
2081 * Allocate a new page.
2082 *
2083 * %NULL is returned if there is no free memory.
2084*/
2085static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2086 struct sk_buff *skb,
2087 unsigned int order)
2088{
2089 struct page *page;
2090
2091 gfp_mask |= __GFP_COLD;
2092
2093 if (!(gfp_mask & __GFP_NOMEMALLOC))
2094 gfp_mask |= __GFP_MEMALLOC;
2095
2096 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2097 if (skb && page && page->pfmemalloc)
2098 skb->pfmemalloc = true;
2099
2100 return page;
2101}
2102
2103/**
2104 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2105 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2106 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2107 *
2108 * Allocate a new page.
2109 *
2110 * %NULL is returned if there is no free memory.
2111 */
2112static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2113 struct sk_buff *skb)
2114{
2115 return __skb_alloc_pages(gfp_mask, skb, 0);
2116}
2117
2118/**
2119 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2120 * @page: The page that was allocated from skb_alloc_page
2121 * @skb: The skb that may need pfmemalloc set
2122 */
2123static inline void skb_propagate_pfmemalloc(struct page *page,
2124 struct sk_buff *skb)
2125{
2126 if (page && page->pfmemalloc)
2127 skb->pfmemalloc = true;
2128}
2129
131ea667 2130/**
e227867f 2131 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2132 * @frag: the paged fragment
2133 *
2134 * Returns the &struct page associated with @frag.
2135 */
2136static inline struct page *skb_frag_page(const skb_frag_t *frag)
2137{
a8605c60 2138 return frag->page.p;
131ea667
IC
2139}
2140
2141/**
2142 * __skb_frag_ref - take an addition reference on a paged fragment.
2143 * @frag: the paged fragment
2144 *
2145 * Takes an additional reference on the paged fragment @frag.
2146 */
2147static inline void __skb_frag_ref(skb_frag_t *frag)
2148{
2149 get_page(skb_frag_page(frag));
2150}
2151
2152/**
2153 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2154 * @skb: the buffer
2155 * @f: the fragment offset.
2156 *
2157 * Takes an additional reference on the @f'th paged fragment of @skb.
2158 */
2159static inline void skb_frag_ref(struct sk_buff *skb, int f)
2160{
2161 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2162}
2163
2164/**
2165 * __skb_frag_unref - release a reference on a paged fragment.
2166 * @frag: the paged fragment
2167 *
2168 * Releases a reference on the paged fragment @frag.
2169 */
2170static inline void __skb_frag_unref(skb_frag_t *frag)
2171{
2172 put_page(skb_frag_page(frag));
2173}
2174
2175/**
2176 * skb_frag_unref - release a reference on a paged fragment of an skb.
2177 * @skb: the buffer
2178 * @f: the fragment offset
2179 *
2180 * Releases a reference on the @f'th paged fragment of @skb.
2181 */
2182static inline void skb_frag_unref(struct sk_buff *skb, int f)
2183{
2184 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2185}
2186
2187/**
2188 * skb_frag_address - gets the address of the data contained in a paged fragment
2189 * @frag: the paged fragment buffer
2190 *
2191 * Returns the address of the data within @frag. The page must already
2192 * be mapped.
2193 */
2194static inline void *skb_frag_address(const skb_frag_t *frag)
2195{
2196 return page_address(skb_frag_page(frag)) + frag->page_offset;
2197}
2198
2199/**
2200 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2201 * @frag: the paged fragment buffer
2202 *
2203 * Returns the address of the data within @frag. Checks that the page
2204 * is mapped and returns %NULL otherwise.
2205 */
2206static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2207{
2208 void *ptr = page_address(skb_frag_page(frag));
2209 if (unlikely(!ptr))
2210 return NULL;
2211
2212 return ptr + frag->page_offset;
2213}
2214
2215/**
2216 * __skb_frag_set_page - sets the page contained in a paged fragment
2217 * @frag: the paged fragment
2218 * @page: the page to set
2219 *
2220 * Sets the fragment @frag to contain @page.
2221 */
2222static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2223{
a8605c60 2224 frag->page.p = page;
131ea667
IC
2225}
2226
2227/**
2228 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2229 * @skb: the buffer
2230 * @f: the fragment offset
2231 * @page: the page to set
2232 *
2233 * Sets the @f'th fragment of @skb to contain @page.
2234 */
2235static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2236 struct page *page)
2237{
2238 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2239}
2240
400dfd3a
ED
2241bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2242
131ea667
IC
2243/**
2244 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2245 * @dev: the device to map the fragment to
131ea667
IC
2246 * @frag: the paged fragment to map
2247 * @offset: the offset within the fragment (starting at the
2248 * fragment's own offset)
2249 * @size: the number of bytes to map
f83347df 2250 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2251 *
2252 * Maps the page associated with @frag to @device.
2253 */
2254static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2255 const skb_frag_t *frag,
2256 size_t offset, size_t size,
2257 enum dma_data_direction dir)
2258{
2259 return dma_map_page(dev, skb_frag_page(frag),
2260 frag->page_offset + offset, size, dir);
2261}
2262
117632e6
ED
2263static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2264 gfp_t gfp_mask)
2265{
2266 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2267}
2268
bad93e9d
OP
2269
2270static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2271 gfp_t gfp_mask)
2272{
2273 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2274}
2275
2276
334a8132
PM
2277/**
2278 * skb_clone_writable - is the header of a clone writable
2279 * @skb: buffer to check
2280 * @len: length up to which to write
2281 *
2282 * Returns true if modifying the header part of the cloned buffer
2283 * does not requires the data to be copied.
2284 */
05bdd2f1 2285static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2286{
2287 return !skb_header_cloned(skb) &&
2288 skb_headroom(skb) + len <= skb->hdr_len;
2289}
2290
d9cc2048
HX
2291static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2292 int cloned)
2293{
2294 int delta = 0;
2295
d9cc2048
HX
2296 if (headroom > skb_headroom(skb))
2297 delta = headroom - skb_headroom(skb);
2298
2299 if (delta || cloned)
2300 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2301 GFP_ATOMIC);
2302 return 0;
2303}
2304
1da177e4
LT
2305/**
2306 * skb_cow - copy header of skb when it is required
2307 * @skb: buffer to cow
2308 * @headroom: needed headroom
2309 *
2310 * If the skb passed lacks sufficient headroom or its data part
2311 * is shared, data is reallocated. If reallocation fails, an error
2312 * is returned and original skb is not changed.
2313 *
2314 * The result is skb with writable area skb->head...skb->tail
2315 * and at least @headroom of space at head.
2316 */
2317static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2318{
d9cc2048
HX
2319 return __skb_cow(skb, headroom, skb_cloned(skb));
2320}
1da177e4 2321
d9cc2048
HX
2322/**
2323 * skb_cow_head - skb_cow but only making the head writable
2324 * @skb: buffer to cow
2325 * @headroom: needed headroom
2326 *
2327 * This function is identical to skb_cow except that we replace the
2328 * skb_cloned check by skb_header_cloned. It should be used when
2329 * you only need to push on some header and do not need to modify
2330 * the data.
2331 */
2332static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2333{
2334 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2335}
2336
2337/**
2338 * skb_padto - pad an skbuff up to a minimal size
2339 * @skb: buffer to pad
2340 * @len: minimal length
2341 *
2342 * Pads up a buffer to ensure the trailing bytes exist and are
2343 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2344 * is untouched. Otherwise it is extended. Returns zero on
2345 * success. The skb is freed on error.
1da177e4
LT
2346 */
2347
5b057c6b 2348static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2349{
2350 unsigned int size = skb->len;
2351 if (likely(size >= len))
5b057c6b 2352 return 0;
987c402a 2353 return skb_pad(skb, len - size);
1da177e4
LT
2354}
2355
2356static inline int skb_add_data(struct sk_buff *skb,
2357 char __user *from, int copy)
2358{
2359 const int off = skb->len;
2360
2361 if (skb->ip_summed == CHECKSUM_NONE) {
2362 int err = 0;
5084205f 2363 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2364 copy, 0, &err);
2365 if (!err) {
2366 skb->csum = csum_block_add(skb->csum, csum, off);
2367 return 0;
2368 }
2369 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2370 return 0;
2371
2372 __skb_trim(skb, off);
2373 return -EFAULT;
2374}
2375
38ba0a65
ED
2376static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2377 const struct page *page, int off)
1da177e4
LT
2378{
2379 if (i) {
9e903e08 2380 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2381
ea2ab693 2382 return page == skb_frag_page(frag) &&
9e903e08 2383 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2384 }
38ba0a65 2385 return false;
1da177e4
LT
2386}
2387
364c6bad
HX
2388static inline int __skb_linearize(struct sk_buff *skb)
2389{
2390 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2391}
2392
1da177e4
LT
2393/**
2394 * skb_linearize - convert paged skb to linear one
2395 * @skb: buffer to linarize
1da177e4
LT
2396 *
2397 * If there is no free memory -ENOMEM is returned, otherwise zero
2398 * is returned and the old skb data released.
2399 */
364c6bad
HX
2400static inline int skb_linearize(struct sk_buff *skb)
2401{
2402 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2403}
2404
cef401de
ED
2405/**
2406 * skb_has_shared_frag - can any frag be overwritten
2407 * @skb: buffer to test
2408 *
2409 * Return true if the skb has at least one frag that might be modified
2410 * by an external entity (as in vmsplice()/sendfile())
2411 */
2412static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2413{
c9af6db4
PS
2414 return skb_is_nonlinear(skb) &&
2415 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2416}
2417
364c6bad
HX
2418/**
2419 * skb_linearize_cow - make sure skb is linear and writable
2420 * @skb: buffer to process
2421 *
2422 * If there is no free memory -ENOMEM is returned, otherwise zero
2423 * is returned and the old skb data released.
2424 */
2425static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2426{
364c6bad
HX
2427 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2428 __skb_linearize(skb) : 0;
1da177e4
LT
2429}
2430
2431/**
2432 * skb_postpull_rcsum - update checksum for received skb after pull
2433 * @skb: buffer to update
2434 * @start: start of data before pull
2435 * @len: length of data pulled
2436 *
2437 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2438 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2439 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2440 */
2441
2442static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2443 const void *start, unsigned int len)
1da177e4 2444{
84fa7933 2445 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2446 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2447}
2448
cbb042f9
HX
2449unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2450
7ce5a27f
DM
2451/**
2452 * pskb_trim_rcsum - trim received skb and update checksum
2453 * @skb: buffer to trim
2454 * @len: new length
2455 *
2456 * This is exactly the same as pskb_trim except that it ensures the
2457 * checksum of received packets are still valid after the operation.
2458 */
2459
2460static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2461{
2462 if (likely(len >= skb->len))
2463 return 0;
2464 if (skb->ip_summed == CHECKSUM_COMPLETE)
2465 skb->ip_summed = CHECKSUM_NONE;
2466 return __pskb_trim(skb, len);
2467}
2468
1da177e4
LT
2469#define skb_queue_walk(queue, skb) \
2470 for (skb = (queue)->next; \
a1e4891f 2471 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2472 skb = skb->next)
2473
46f8914e
JC
2474#define skb_queue_walk_safe(queue, skb, tmp) \
2475 for (skb = (queue)->next, tmp = skb->next; \
2476 skb != (struct sk_buff *)(queue); \
2477 skb = tmp, tmp = skb->next)
2478
1164f52a 2479#define skb_queue_walk_from(queue, skb) \
a1e4891f 2480 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2481 skb = skb->next)
2482
2483#define skb_queue_walk_from_safe(queue, skb, tmp) \
2484 for (tmp = skb->next; \
2485 skb != (struct sk_buff *)(queue); \
2486 skb = tmp, tmp = skb->next)
2487
300ce174
SH
2488#define skb_queue_reverse_walk(queue, skb) \
2489 for (skb = (queue)->prev; \
a1e4891f 2490 skb != (struct sk_buff *)(queue); \
300ce174
SH
2491 skb = skb->prev)
2492
686a2955
DM
2493#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2494 for (skb = (queue)->prev, tmp = skb->prev; \
2495 skb != (struct sk_buff *)(queue); \
2496 skb = tmp, tmp = skb->prev)
2497
2498#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2499 for (tmp = skb->prev; \
2500 skb != (struct sk_buff *)(queue); \
2501 skb = tmp, tmp = skb->prev)
1da177e4 2502
21dc3301 2503static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2504{
2505 return skb_shinfo(skb)->frag_list != NULL;
2506}
2507
2508static inline void skb_frag_list_init(struct sk_buff *skb)
2509{
2510 skb_shinfo(skb)->frag_list = NULL;
2511}
2512
2513static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2514{
2515 frag->next = skb_shinfo(skb)->frag_list;
2516 skb_shinfo(skb)->frag_list = frag;
2517}
2518
2519#define skb_walk_frags(skb, iter) \
2520 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2521
7965bd4d
JP
2522struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2523 int *peeked, int *off, int *err);
2524struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2525 int *err);
2526unsigned int datagram_poll(struct file *file, struct socket *sock,
2527 struct poll_table_struct *wait);
2528int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2529 struct iovec *to, int size);
2530int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2531 struct iovec *iov);
2532int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2533 const struct iovec *from, int from_offset,
2534 int len);
2535int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2536 int offset, size_t count);
2537int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2538 const struct iovec *to, int to_offset,
2539 int size);
2540void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2541void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2542int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2543int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2544int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2545__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2546 int len, __wsum csum);
2547int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2548 struct pipe_inode_info *pipe, unsigned int len,
2549 unsigned int flags);
2550void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2551unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2552int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2553 int len, int hlen);
7965bd4d
JP
2554void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2555int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2556void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2557unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2558struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
20380731 2559
2817a336
DB
2560struct skb_checksum_ops {
2561 __wsum (*update)(const void *mem, int len, __wsum wsum);
2562 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2563};
2564
2565__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2566 __wsum csum, const struct skb_checksum_ops *ops);
2567__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2568 __wsum csum);
2569
1da177e4
LT
2570static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2571 int len, void *buffer)
2572{
2573 int hlen = skb_headlen(skb);
2574
55820ee2 2575 if (hlen - offset >= len)
1da177e4
LT
2576 return skb->data + offset;
2577
2578 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2579 return NULL;
2580
2581 return buffer;
2582}
2583
4262e5cc
DB
2584/**
2585 * skb_needs_linearize - check if we need to linearize a given skb
2586 * depending on the given device features.
2587 * @skb: socket buffer to check
2588 * @features: net device features
2589 *
2590 * Returns true if either:
2591 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2592 * 2. skb is fragmented and the device does not support SG.
2593 */
2594static inline bool skb_needs_linearize(struct sk_buff *skb,
2595 netdev_features_t features)
2596{
2597 return skb_is_nonlinear(skb) &&
2598 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2599 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2600}
2601
d626f62b
ACM
2602static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2603 void *to,
2604 const unsigned int len)
2605{
2606 memcpy(to, skb->data, len);
2607}
2608
2609static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2610 const int offset, void *to,
2611 const unsigned int len)
2612{
2613 memcpy(to, skb->data + offset, len);
2614}
2615
27d7ff46
ACM
2616static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2617 const void *from,
2618 const unsigned int len)
2619{
2620 memcpy(skb->data, from, len);
2621}
2622
2623static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2624 const int offset,
2625 const void *from,
2626 const unsigned int len)
2627{
2628 memcpy(skb->data + offset, from, len);
2629}
2630
7965bd4d 2631void skb_init(void);
1da177e4 2632
ac45f602
PO
2633static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2634{
2635 return skb->tstamp;
2636}
2637
a61bbcf2
PM
2638/**
2639 * skb_get_timestamp - get timestamp from a skb
2640 * @skb: skb to get stamp from
2641 * @stamp: pointer to struct timeval to store stamp in
2642 *
2643 * Timestamps are stored in the skb as offsets to a base timestamp.
2644 * This function converts the offset back to a struct timeval and stores
2645 * it in stamp.
2646 */
ac45f602
PO
2647static inline void skb_get_timestamp(const struct sk_buff *skb,
2648 struct timeval *stamp)
a61bbcf2 2649{
b7aa0bf7 2650 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2651}
2652
ac45f602
PO
2653static inline void skb_get_timestampns(const struct sk_buff *skb,
2654 struct timespec *stamp)
2655{
2656 *stamp = ktime_to_timespec(skb->tstamp);
2657}
2658
b7aa0bf7 2659static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2660{
b7aa0bf7 2661 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2662}
2663
164891aa
SH
2664static inline ktime_t net_timedelta(ktime_t t)
2665{
2666 return ktime_sub(ktime_get_real(), t);
2667}
2668
b9ce204f
IJ
2669static inline ktime_t net_invalid_timestamp(void)
2670{
2671 return ktime_set(0, 0);
2672}
a61bbcf2 2673
c1f19b51
RC
2674#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2675
7965bd4d
JP
2676void skb_clone_tx_timestamp(struct sk_buff *skb);
2677bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2678
2679#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2680
2681static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2682{
2683}
2684
2685static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2686{
2687 return false;
2688}
2689
2690#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2691
2692/**
2693 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2694 *
da92b194
RC
2695 * PHY drivers may accept clones of transmitted packets for
2696 * timestamping via their phy_driver.txtstamp method. These drivers
2697 * must call this function to return the skb back to the stack, with
2698 * or without a timestamp.
2699 *
c1f19b51 2700 * @skb: clone of the the original outgoing packet
da92b194 2701 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2702 *
2703 */
2704void skb_complete_tx_timestamp(struct sk_buff *skb,
2705 struct skb_shared_hwtstamps *hwtstamps);
2706
ac45f602
PO
2707/**
2708 * skb_tstamp_tx - queue clone of skb with send time stamps
2709 * @orig_skb: the original outgoing packet
2710 * @hwtstamps: hardware time stamps, may be NULL if not available
2711 *
2712 * If the skb has a socket associated, then this function clones the
2713 * skb (thus sharing the actual data and optional structures), stores
2714 * the optional hardware time stamping information (if non NULL) or
2715 * generates a software time stamp (otherwise), then queues the clone
2716 * to the error queue of the socket. Errors are silently ignored.
2717 */
7965bd4d
JP
2718void skb_tstamp_tx(struct sk_buff *orig_skb,
2719 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2720
4507a715
RC
2721static inline void sw_tx_timestamp(struct sk_buff *skb)
2722{
2244d07b
OH
2723 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2724 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2725 skb_tstamp_tx(skb, NULL);
2726}
2727
2728/**
2729 * skb_tx_timestamp() - Driver hook for transmit timestamping
2730 *
2731 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2732 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2733 *
73409f3b
DM
2734 * Specifically, one should make absolutely sure that this function is
2735 * called before TX completion of this packet can trigger. Otherwise
2736 * the packet could potentially already be freed.
2737 *
4507a715
RC
2738 * @skb: A socket buffer.
2739 */
2740static inline void skb_tx_timestamp(struct sk_buff *skb)
2741{
c1f19b51 2742 skb_clone_tx_timestamp(skb);
4507a715
RC
2743 sw_tx_timestamp(skb);
2744}
2745
6e3e939f
JB
2746/**
2747 * skb_complete_wifi_ack - deliver skb with wifi status
2748 *
2749 * @skb: the original outgoing packet
2750 * @acked: ack status
2751 *
2752 */
2753void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2754
7965bd4d
JP
2755__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2756__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2757
60476372
HX
2758static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2759{
5d0c2b95 2760 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
60476372
HX
2761}
2762
fb286bb2
HX
2763/**
2764 * skb_checksum_complete - Calculate checksum of an entire packet
2765 * @skb: packet to process
2766 *
2767 * This function calculates the checksum over the entire packet plus
2768 * the value of skb->csum. The latter can be used to supply the
2769 * checksum of a pseudo header as used by TCP/UDP. It returns the
2770 * checksum.
2771 *
2772 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2773 * this function can be used to verify that checksum on received
2774 * packets. In that case the function should return zero if the
2775 * checksum is correct. In particular, this function will return zero
2776 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2777 * hardware has already verified the correctness of the checksum.
2778 */
4381ca3c 2779static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2780{
60476372
HX
2781 return skb_csum_unnecessary(skb) ?
2782 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2783}
2784
76ba0aae
TH
2785/* Check if we need to perform checksum complete validation.
2786 *
2787 * Returns true if checksum complete is needed, false otherwise
2788 * (either checksum is unnecessary or zero checksum is allowed).
2789 */
2790static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2791 bool zero_okay,
2792 __sum16 check)
2793{
5d0c2b95
TH
2794 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2795 skb->csum_valid = 1;
76ba0aae
TH
2796 return false;
2797 }
2798
2799 return true;
2800}
2801
2802/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2803 * in checksum_init.
2804 */
2805#define CHECKSUM_BREAK 76
2806
2807/* Validate (init) checksum based on checksum complete.
2808 *
2809 * Return values:
2810 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2811 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2812 * checksum is stored in skb->csum for use in __skb_checksum_complete
2813 * non-zero: value of invalid checksum
2814 *
2815 */
2816static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2817 bool complete,
2818 __wsum psum)
2819{
2820 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2821 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 2822 skb->csum_valid = 1;
76ba0aae
TH
2823 return 0;
2824 }
2825 }
2826
2827 skb->csum = psum;
2828
5d0c2b95
TH
2829 if (complete || skb->len <= CHECKSUM_BREAK) {
2830 __sum16 csum;
2831
2832 csum = __skb_checksum_complete(skb);
2833 skb->csum_valid = !csum;
2834 return csum;
2835 }
76ba0aae
TH
2836
2837 return 0;
2838}
2839
2840static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2841{
2842 return 0;
2843}
2844
2845/* Perform checksum validate (init). Note that this is a macro since we only
2846 * want to calculate the pseudo header which is an input function if necessary.
2847 * First we try to validate without any computation (checksum unnecessary) and
2848 * then calculate based on checksum complete calling the function to compute
2849 * pseudo header.
2850 *
2851 * Return values:
2852 * 0: checksum is validated or try to in skb_checksum_complete
2853 * non-zero: value of invalid checksum
2854 */
2855#define __skb_checksum_validate(skb, proto, complete, \
2856 zero_okay, check, compute_pseudo) \
2857({ \
2858 __sum16 __ret = 0; \
5d0c2b95 2859 skb->csum_valid = 0; \
76ba0aae
TH
2860 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2861 __ret = __skb_checksum_validate_complete(skb, \
2862 complete, compute_pseudo(skb, proto)); \
2863 __ret; \
2864})
2865
2866#define skb_checksum_init(skb, proto, compute_pseudo) \
2867 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2868
2869#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2870 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2871
2872#define skb_checksum_validate(skb, proto, compute_pseudo) \
2873 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2874
2875#define skb_checksum_validate_zero_check(skb, proto, check, \
2876 compute_pseudo) \
2877 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2878
2879#define skb_checksum_simple_validate(skb) \
2880 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2881
5f79e0f9 2882#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 2883void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
2884static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2885{
2886 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 2887 nf_conntrack_destroy(nfct);
1da177e4
LT
2888}
2889static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2890{
2891 if (nfct)
2892 atomic_inc(&nfct->use);
2893}
2fc72c7b 2894#endif
1da177e4
LT
2895#ifdef CONFIG_BRIDGE_NETFILTER
2896static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2897{
2898 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2899 kfree(nf_bridge);
2900}
2901static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2902{
2903 if (nf_bridge)
2904 atomic_inc(&nf_bridge->use);
2905}
2906#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
2907static inline void nf_reset(struct sk_buff *skb)
2908{
5f79e0f9 2909#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
2910 nf_conntrack_put(skb->nfct);
2911 skb->nfct = NULL;
2fc72c7b 2912#endif
a193a4ab
PM
2913#ifdef CONFIG_BRIDGE_NETFILTER
2914 nf_bridge_put(skb->nf_bridge);
2915 skb->nf_bridge = NULL;
2916#endif
2917}
2918
124dff01
PM
2919static inline void nf_reset_trace(struct sk_buff *skb)
2920{
478b360a 2921#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
2922 skb->nf_trace = 0;
2923#endif
a193a4ab
PM
2924}
2925
edda553c
YK
2926/* Note: This doesn't put any conntrack and bridge info in dst. */
2927static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2928{
5f79e0f9 2929#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
2930 dst->nfct = src->nfct;
2931 nf_conntrack_get(src->nfct);
2932 dst->nfctinfo = src->nfctinfo;
2fc72c7b 2933#endif
edda553c
YK
2934#ifdef CONFIG_BRIDGE_NETFILTER
2935 dst->nf_bridge = src->nf_bridge;
2936 nf_bridge_get(src->nf_bridge);
2937#endif
478b360a
FW
2938#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2939 dst->nf_trace = src->nf_trace;
2940#endif
edda553c
YK
2941}
2942
e7ac05f3
YK
2943static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2944{
e7ac05f3 2945#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 2946 nf_conntrack_put(dst->nfct);
2fc72c7b 2947#endif
e7ac05f3
YK
2948#ifdef CONFIG_BRIDGE_NETFILTER
2949 nf_bridge_put(dst->nf_bridge);
2950#endif
2951 __nf_copy(dst, src);
2952}
2953
984bc16c
JM
2954#ifdef CONFIG_NETWORK_SECMARK
2955static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2956{
2957 to->secmark = from->secmark;
2958}
2959
2960static inline void skb_init_secmark(struct sk_buff *skb)
2961{
2962 skb->secmark = 0;
2963}
2964#else
2965static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2966{ }
2967
2968static inline void skb_init_secmark(struct sk_buff *skb)
2969{ }
2970#endif
2971
574f7194
EB
2972static inline bool skb_irq_freeable(const struct sk_buff *skb)
2973{
2974 return !skb->destructor &&
2975#if IS_ENABLED(CONFIG_XFRM)
2976 !skb->sp &&
2977#endif
2978#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2979 !skb->nfct &&
2980#endif
2981 !skb->_skb_refdst &&
2982 !skb_has_frag_list(skb);
2983}
2984
f25f4e44
PWJ
2985static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2986{
f25f4e44 2987 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
2988}
2989
9247744e 2990static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 2991{
4e3ab47a 2992 return skb->queue_mapping;
4e3ab47a
PE
2993}
2994
f25f4e44
PWJ
2995static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2996{
f25f4e44 2997 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
2998}
2999
d5a9e24a
DM
3000static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3001{
3002 skb->queue_mapping = rx_queue + 1;
3003}
3004
9247744e 3005static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3006{
3007 return skb->queue_mapping - 1;
3008}
3009
9247744e 3010static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3011{
a02cec21 3012 return skb->queue_mapping != 0;
d5a9e24a
DM
3013}
3014
0e001614 3015u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
7965bd4d 3016 unsigned int num_tx_queues);
9247744e 3017
def8b4fa
AD
3018static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3019{
0b3d8e08 3020#ifdef CONFIG_XFRM
def8b4fa 3021 return skb->sp;
def8b4fa 3022#else
def8b4fa 3023 return NULL;
def8b4fa 3024#endif
0b3d8e08 3025}
def8b4fa 3026
68c33163
PS
3027/* Keeps track of mac header offset relative to skb->head.
3028 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3029 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3030 * tunnel skb it points to outer mac header.
3031 * Keeps track of level of encapsulation of network headers.
3032 */
68c33163 3033struct skb_gso_cb {
3347c960
ED
3034 int mac_offset;
3035 int encap_level;
7e2b10c1 3036 __u16 csum_start;
68c33163
PS
3037};
3038#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3039
3040static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3041{
3042 return (skb_mac_header(inner_skb) - inner_skb->head) -
3043 SKB_GSO_CB(inner_skb)->mac_offset;
3044}
3045
1e2bd517
PS
3046static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3047{
3048 int new_headroom, headroom;
3049 int ret;
3050
3051 headroom = skb_headroom(skb);
3052 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3053 if (ret)
3054 return ret;
3055
3056 new_headroom = skb_headroom(skb);
3057 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3058 return 0;
3059}
3060
7e2b10c1
TH
3061/* Compute the checksum for a gso segment. First compute the checksum value
3062 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3063 * then add in skb->csum (checksum from csum_start to end of packet).
3064 * skb->csum and csum_start are then updated to reflect the checksum of the
3065 * resultant packet starting from the transport header-- the resultant checksum
3066 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3067 * header.
3068 */
3069static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3070{
3071 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3072 skb_transport_offset(skb);
3073 __u16 csum;
3074
3075 csum = csum_fold(csum_partial(skb_transport_header(skb),
3076 plen, skb->csum));
3077 skb->csum = res;
3078 SKB_GSO_CB(skb)->csum_start -= plen;
3079
3080 return csum;
3081}
3082
bdcc0924 3083static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3084{
3085 return skb_shinfo(skb)->gso_size;
3086}
3087
36a8f39e 3088/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3089static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3090{
3091 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3092}
3093
7965bd4d 3094void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3095
3096static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3097{
3098 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3099 * wanted then gso_type will be set. */
05bdd2f1
ED
3100 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3101
b78462eb
AD
3102 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3103 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3104 __skb_warn_lro_forwarding(skb);
3105 return true;
3106 }
3107 return false;
3108}
3109
35fc92a9
HX
3110static inline void skb_forward_csum(struct sk_buff *skb)
3111{
3112 /* Unfortunately we don't support this one. Any brave souls? */
3113 if (skb->ip_summed == CHECKSUM_COMPLETE)
3114 skb->ip_summed = CHECKSUM_NONE;
3115}
3116
bc8acf2c
ED
3117/**
3118 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3119 * @skb: skb to check
3120 *
3121 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3122 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3123 * use this helper, to document places where we make this assertion.
3124 */
05bdd2f1 3125static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3126{
3127#ifdef DEBUG
3128 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3129#endif
3130}
3131
f35d9d8a 3132bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3133
ed1f50c3
PD
3134int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3135
f77668dc
DB
3136u32 __skb_get_poff(const struct sk_buff *skb);
3137
3a7c1ee4
AD
3138/**
3139 * skb_head_is_locked - Determine if the skb->head is locked down
3140 * @skb: skb to check
3141 *
3142 * The head on skbs build around a head frag can be removed if they are
3143 * not cloned. This function returns true if the skb head is locked down
3144 * due to either being allocated via kmalloc, or by being a clone with
3145 * multiple references to the head.
3146 */
3147static inline bool skb_head_is_locked(const struct sk_buff *skb)
3148{
3149 return !skb->head_frag || skb_cloned(skb);
3150}
fe6cc55f
FW
3151
3152/**
3153 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3154 *
3155 * @skb: GSO skb
3156 *
3157 * skb_gso_network_seglen is used to determine the real size of the
3158 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3159 *
3160 * The MAC/L2 header is not accounted for.
3161 */
3162static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3163{
3164 unsigned int hdr_len = skb_transport_header(skb) -
3165 skb_network_header(skb);
3166 return hdr_len + skb_gso_transport_seglen(skb);
3167}
1da177e4
LT
3168#endif /* __KERNEL__ */
3169#endif /* _LINUX_SKBUFF_H */
This page took 1.69006 seconds and 5 git commands to generate.