Merge tag 'iwlwifi-for-kalle-2015-06-12' of https://git.kernel.org/pub/scm/linux...
[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 22#include <linux/cache.h>
56b17425 23#include <linux/rbtree.h>
51f3d02b 24#include <linux/socket.h>
1da177e4 25
60063497 26#include <linux/atomic.h>
1da177e4
LT
27#include <asm/types.h>
28#include <linux/spinlock.h>
1da177e4 29#include <linux/net.h>
3fc7e8a6 30#include <linux/textsearch.h>
1da177e4 31#include <net/checksum.h>
a80958f4 32#include <linux/rcupdate.h>
b7aa0bf7 33#include <linux/hrtimer.h>
131ea667 34#include <linux/dma-mapping.h>
c8f44aff 35#include <linux/netdev_features.h>
363ec392 36#include <linux/sched.h>
1bd758eb 37#include <net/flow_dissector.h>
a60e3cc7 38#include <linux/splice.h>
72b31f72 39#include <linux/in6.h>
1da177e4 40
78ea85f1
DB
41/* A. Checksumming of received packets by device.
42 *
43 * CHECKSUM_NONE:
44 *
45 * Device failed to checksum this packet e.g. due to lack of capabilities.
46 * The packet contains full (though not verified) checksum in packet but
47 * not in skb->csum. Thus, skb->csum is undefined in this case.
48 *
49 * CHECKSUM_UNNECESSARY:
50 *
51 * The hardware you're dealing with doesn't calculate the full checksum
52 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
53 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
54 * if their checksums are okay. skb->csum is still undefined in this case
55 * though. It is a bad option, but, unfortunately, nowadays most vendors do
56 * this. Apparently with the secret goal to sell you new devices, when you
57 * will add new protocol to your host, f.e. IPv6 8)
58 *
59 * CHECKSUM_UNNECESSARY is applicable to following protocols:
60 * TCP: IPv6 and IPv4.
61 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
62 * zero UDP checksum for either IPv4 or IPv6, the networking stack
63 * may perform further validation in this case.
64 * GRE: only if the checksum is present in the header.
65 * SCTP: indicates the CRC in SCTP header has been validated.
66 *
67 * skb->csum_level indicates the number of consecutive checksums found in
68 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
69 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
70 * and a device is able to verify the checksums for UDP (possibly zero),
71 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
72 * two. If the device were only able to verify the UDP checksum and not
73 * GRE, either because it doesn't support GRE checksum of because GRE
74 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
75 * not considered in this case).
78ea85f1
DB
76 *
77 * CHECKSUM_COMPLETE:
78 *
79 * This is the most generic way. The device supplied checksum of the _whole_
80 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
81 * hardware doesn't need to parse L3/L4 headers to implement this.
82 *
83 * Note: Even if device supports only some protocols, but is able to produce
84 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
85 *
86 * CHECKSUM_PARTIAL:
87 *
6edec0e6
TH
88 * A checksum is set up to be offloaded to a device as described in the
89 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 90 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
91 * on the same host, or it may be set in the input path in GRO or remote
92 * checksum offload. For the purposes of checksum verification, the checksum
93 * referred to by skb->csum_start + skb->csum_offset and any preceding
94 * checksums in the packet are considered verified. Any checksums in the
95 * packet that are after the checksum being offloaded are not considered to
96 * be verified.
78ea85f1
DB
97 *
98 * B. Checksumming on output.
99 *
100 * CHECKSUM_NONE:
101 *
102 * The skb was already checksummed by the protocol, or a checksum is not
103 * required.
104 *
105 * CHECKSUM_PARTIAL:
106 *
107 * The device is required to checksum the packet as seen by hard_start_xmit()
108 * from skb->csum_start up to the end, and to record/write the checksum at
109 * offset skb->csum_start + skb->csum_offset.
110 *
111 * The device must show its capabilities in dev->features, set up at device
112 * setup time, e.g. netdev_features.h:
113 *
114 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
115 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
116 * IPv4. Sigh. Vendors like this way for an unknown reason.
117 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
118 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
119 * NETIF_F_... - Well, you get the picture.
120 *
121 * CHECKSUM_UNNECESSARY:
122 *
123 * Normally, the device will do per protocol specific checksumming. Protocol
124 * implementations that do not want the NIC to perform the checksum
125 * calculation should use this flag in their outgoing skbs.
126 *
127 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
128 * offload. Correspondingly, the FCoE protocol driver
129 * stack should use CHECKSUM_UNNECESSARY.
130 *
131 * Any questions? No questions, good. --ANK
132 */
133
60476372 134/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
135#define CHECKSUM_NONE 0
136#define CHECKSUM_UNNECESSARY 1
137#define CHECKSUM_COMPLETE 2
138#define CHECKSUM_PARTIAL 3
1da177e4 139
77cffe23
TH
140/* Maximum value in skb->csum_level */
141#define SKB_MAX_CSUM_LEVEL 3
142
0bec8c88 143#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 144#define SKB_WITH_OVERHEAD(X) \
deea84b0 145 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
146#define SKB_MAX_ORDER(X, ORDER) \
147 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
148#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
149#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
150
87fb4b7b
ED
151/* return minimum truesize of one skb containing X bytes of data */
152#define SKB_TRUESIZE(X) ((X) + \
153 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
154 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
155
1da177e4 156struct net_device;
716ea3a7 157struct scatterlist;
9c55e01c 158struct pipe_inode_info;
a8f820aa 159struct iov_iter;
fd11a83d 160struct napi_struct;
1da177e4 161
5f79e0f9 162#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
163struct nf_conntrack {
164 atomic_t use;
1da177e4 165};
5f79e0f9 166#endif
1da177e4 167
34666d46 168#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 169struct nf_bridge_info {
bf1ac5ca 170 atomic_t use;
3eaf4025
FW
171 enum {
172 BRNF_PROTO_UNCHANGED,
173 BRNF_PROTO_8021Q,
174 BRNF_PROTO_PPPOE
7fb48c5b 175 } orig_proto:8;
a1e67951 176 bool pkt_otherhost;
411ffb4f 177 __u16 frag_max_size;
bf1ac5ca
ED
178 unsigned int mask;
179 struct net_device *physindev;
7fb48c5b
FW
180 union {
181 struct net_device *physoutdev;
182 char neigh_header[8];
183 };
72b31f72
BT
184 union {
185 __be32 ipv4_daddr;
186 struct in6_addr ipv6_daddr;
187 };
1da177e4
LT
188};
189#endif
190
1da177e4
LT
191struct sk_buff_head {
192 /* These two members must be first. */
193 struct sk_buff *next;
194 struct sk_buff *prev;
195
196 __u32 qlen;
197 spinlock_t lock;
198};
199
200struct sk_buff;
201
9d4dde52
IC
202/* To allow 64K frame to be packed as single skb without frag_list we
203 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
204 * buffers which do not start on a page boundary.
205 *
206 * Since GRO uses frags we allocate at least 16 regardless of page
207 * size.
a715dea3 208 */
9d4dde52 209#if (65536/PAGE_SIZE + 1) < 16
eec00954 210#define MAX_SKB_FRAGS 16UL
a715dea3 211#else
9d4dde52 212#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 213#endif
1da177e4
LT
214
215typedef struct skb_frag_struct skb_frag_t;
216
217struct skb_frag_struct {
a8605c60
IC
218 struct {
219 struct page *p;
220 } page;
cb4dfe56 221#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
222 __u32 page_offset;
223 __u32 size;
cb4dfe56
ED
224#else
225 __u16 page_offset;
226 __u16 size;
227#endif
1da177e4
LT
228};
229
9e903e08
ED
230static inline unsigned int skb_frag_size(const skb_frag_t *frag)
231{
232 return frag->size;
233}
234
235static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
236{
237 frag->size = size;
238}
239
240static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
241{
242 frag->size += delta;
243}
244
245static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
246{
247 frag->size -= delta;
248}
249
ac45f602
PO
250#define HAVE_HW_TIME_STAMP
251
252/**
d3a21be8 253 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
254 * @hwtstamp: hardware time stamp transformed into duration
255 * since arbitrary point in time
ac45f602
PO
256 *
257 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 258 * skb->tstamp.
ac45f602
PO
259 *
260 * hwtstamps can only be compared against other hwtstamps from
261 * the same device.
262 *
263 * This structure is attached to packets as part of the
264 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
265 */
266struct skb_shared_hwtstamps {
267 ktime_t hwtstamp;
ac45f602
PO
268};
269
2244d07b
OH
270/* Definitions for tx_flags in struct skb_shared_info */
271enum {
272 /* generate hardware time stamp */
273 SKBTX_HW_TSTAMP = 1 << 0,
274
e7fd2885 275 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
276 SKBTX_SW_TSTAMP = 1 << 1,
277
278 /* device driver is going to provide hardware time stamp */
279 SKBTX_IN_PROGRESS = 1 << 2,
280
a6686f2f 281 /* device driver supports TX zero-copy buffers */
62b1a8ab 282 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
283
284 /* generate wifi status information (where possible) */
62b1a8ab 285 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
286
287 /* This indicates at least one fragment might be overwritten
288 * (as in vmsplice(), sendfile() ...)
289 * If we need to compute a TX checksum, we'll need to copy
290 * all frags to avoid possible bad checksum
291 */
292 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
293
294 /* generate software time stamp when entering packet scheduling */
295 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
296
297 /* generate software timestamp on peer data acknowledgment */
298 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
299};
300
e1c8a607
WB
301#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
302 SKBTX_SCHED_TSTAMP | \
303 SKBTX_ACK_TSTAMP)
f24b9be5
WB
304#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
305
a6686f2f
SM
306/*
307 * The callback notifies userspace to release buffers when skb DMA is done in
308 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
309 * The zerocopy_success argument is true if zero copy transmit occurred,
310 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
311 * The ctx field is used to track device context.
312 * The desc field is used to track userspace buffer index.
a6686f2f
SM
313 */
314struct ubuf_info {
e19d6763 315 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 316 void *ctx;
a6686f2f 317 unsigned long desc;
ac45f602
PO
318};
319
1da177e4
LT
320/* This data is invariant across clones and lives at
321 * the end of the header data, ie. at skb->end.
322 */
323struct skb_shared_info {
9f42f126
IC
324 unsigned char nr_frags;
325 __u8 tx_flags;
7967168c
HX
326 unsigned short gso_size;
327 /* Warning: this field is not always filled in (UFO)! */
328 unsigned short gso_segs;
329 unsigned short gso_type;
1da177e4 330 struct sk_buff *frag_list;
ac45f602 331 struct skb_shared_hwtstamps hwtstamps;
09c2d251 332 u32 tskey;
9f42f126 333 __be32 ip6_frag_id;
ec7d2f2c
ED
334
335 /*
336 * Warning : all fields before dataref are cleared in __alloc_skb()
337 */
338 atomic_t dataref;
339
69e3c75f
JB
340 /* Intermediate layers must ensure that destructor_arg
341 * remains valid until skb destructor */
342 void * destructor_arg;
a6686f2f 343
fed66381
ED
344 /* must be last field, see pskb_expand_head() */
345 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
346};
347
348/* We divide dataref into two halves. The higher 16 bits hold references
349 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
350 * the entire skb->data. A clone of a headerless skb holds the length of
351 * the header in skb->hdr_len.
1da177e4
LT
352 *
353 * All users must obey the rule that the skb->data reference count must be
354 * greater than or equal to the payload reference count.
355 *
356 * Holding a reference to the payload part means that the user does not
357 * care about modifications to the header part of skb->data.
358 */
359#define SKB_DATAREF_SHIFT 16
360#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
361
d179cd12
DM
362
363enum {
c8753d55
VS
364 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
365 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
366 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
367};
368
7967168c
HX
369enum {
370 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 371 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
372
373 /* This indicates the skb is from an untrusted source. */
374 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
375
376 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
377 SKB_GSO_TCP_ECN = 1 << 3,
378
379 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
380
381 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
382
383 SKB_GSO_GRE = 1 << 6,
73136267 384
4b28252c 385 SKB_GSO_GRE_CSUM = 1 << 7,
0d89d203 386
4b28252c 387 SKB_GSO_IPIP = 1 << 8,
cb32f511 388
4b28252c 389 SKB_GSO_SIT = 1 << 9,
61c1db7f 390
4b28252c 391 SKB_GSO_UDP_TUNNEL = 1 << 10,
0f4f4ffa
TH
392
393 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c 394
59b93b41 395 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
7967168c
HX
396};
397
2e07fa9c
ACM
398#if BITS_PER_LONG > 32
399#define NET_SKBUFF_DATA_USES_OFFSET 1
400#endif
401
402#ifdef NET_SKBUFF_DATA_USES_OFFSET
403typedef unsigned int sk_buff_data_t;
404#else
405typedef unsigned char *sk_buff_data_t;
406#endif
407
363ec392
ED
408/**
409 * struct skb_mstamp - multi resolution time stamps
410 * @stamp_us: timestamp in us resolution
411 * @stamp_jiffies: timestamp in jiffies
412 */
413struct skb_mstamp {
414 union {
415 u64 v64;
416 struct {
417 u32 stamp_us;
418 u32 stamp_jiffies;
419 };
420 };
421};
422
423/**
424 * skb_mstamp_get - get current timestamp
425 * @cl: place to store timestamps
426 */
427static inline void skb_mstamp_get(struct skb_mstamp *cl)
428{
429 u64 val = local_clock();
430
431 do_div(val, NSEC_PER_USEC);
432 cl->stamp_us = (u32)val;
433 cl->stamp_jiffies = (u32)jiffies;
434}
435
436/**
437 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
438 * @t1: pointer to newest sample
439 * @t0: pointer to oldest sample
440 */
441static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
442 const struct skb_mstamp *t0)
443{
444 s32 delta_us = t1->stamp_us - t0->stamp_us;
445 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
446
447 /* If delta_us is negative, this might be because interval is too big,
448 * or local_clock() drift is too big : fallback using jiffies.
449 */
450 if (delta_us <= 0 ||
451 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
452
453 delta_us = jiffies_to_usecs(delta_jiffies);
454
455 return delta_us;
456}
457
458
1da177e4
LT
459/**
460 * struct sk_buff - socket buffer
461 * @next: Next buffer in list
462 * @prev: Previous buffer in list
363ec392 463 * @tstamp: Time we arrived/left
56b17425 464 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 465 * @sk: Socket we are owned by
1da177e4 466 * @dev: Device we arrived on/are leaving by
d84e0bd7 467 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 468 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 469 * @sp: the security path, used for xfrm
1da177e4
LT
470 * @len: Length of actual data
471 * @data_len: Data length
472 * @mac_len: Length of link layer header
334a8132 473 * @hdr_len: writable header length of cloned skb
663ead3b
HX
474 * @csum: Checksum (must include start/offset pair)
475 * @csum_start: Offset from skb->head where checksumming should start
476 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 477 * @priority: Packet queueing priority
60ff7467 478 * @ignore_df: allow local fragmentation
1da177e4 479 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 480 * @ip_summed: Driver fed us an IP checksum
1da177e4 481 * @nohdr: Payload reference only, must not modify header
d84e0bd7 482 * @nfctinfo: Relationship of this skb to the connection
1da177e4 483 * @pkt_type: Packet class
c83c2486 484 * @fclone: skbuff clone status
c83c2486 485 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
486 * @peeked: this packet has been seen already, so stats have been
487 * done for it, don't do them again
ba9dda3a 488 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
489 * @protocol: Packet protocol from driver
490 * @destructor: Destruct function
491 * @nfct: Associated connection, if any
1da177e4 492 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 493 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
494 * @tc_index: Traffic control index
495 * @tc_verd: traffic control verdict
61b905da 496 * @hash: the packet hash
d84e0bd7 497 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 498 * @xmit_more: More SKBs are pending for this queue
553a5672 499 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 500 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 501 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 502 * ports.
a3b18ddb 503 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
504 * @wifi_acked_valid: wifi_acked was set
505 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 506 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
06021292 507 * @napi_id: id of the NAPI struct this skb came from
984bc16c 508 * @secmark: security marking
d84e0bd7 509 * @mark: Generic packet mark
86a9bad3 510 * @vlan_proto: vlan encapsulation protocol
6aa895b0 511 * @vlan_tci: vlan tag control information
0d89d203 512 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
513 * @inner_transport_header: Inner transport layer header (encapsulation)
514 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 515 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
516 * @transport_header: Transport layer header
517 * @network_header: Network layer header
518 * @mac_header: Link layer header
519 * @tail: Tail pointer
520 * @end: End pointer
521 * @head: Head of buffer
522 * @data: Data head pointer
523 * @truesize: Buffer size
524 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
525 */
526
527struct sk_buff {
363ec392 528 union {
56b17425
ED
529 struct {
530 /* These two members must be first. */
531 struct sk_buff *next;
532 struct sk_buff *prev;
533
534 union {
535 ktime_t tstamp;
536 struct skb_mstamp skb_mstamp;
537 };
538 };
539 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 540 };
da3f5cf1 541 struct sock *sk;
1da177e4 542 struct net_device *dev;
1da177e4 543
1da177e4
LT
544 /*
545 * This is the control buffer. It is free to use for every
546 * layer. Please put your private variables there. If you
547 * want to keep them across layers you have to do a skb_clone()
548 * first. This is owned by whoever has the skb queued ATM.
549 */
da3f5cf1 550 char cb[48] __aligned(8);
1da177e4 551
7fee226a 552 unsigned long _skb_refdst;
b1937227 553 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
554#ifdef CONFIG_XFRM
555 struct sec_path *sp;
b1937227
ED
556#endif
557#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
558 struct nf_conntrack *nfct;
559#endif
85224844 560#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 561 struct nf_bridge_info *nf_bridge;
da3f5cf1 562#endif
1da177e4 563 unsigned int len,
334a8132
PM
564 data_len;
565 __u16 mac_len,
566 hdr_len;
b1937227
ED
567
568 /* Following fields are _not_ copied in __copy_skb_header()
569 * Note that queue_mapping is here mostly to fill a hole.
570 */
fe55f6d5 571 kmemcheck_bitfield_begin(flags1);
b1937227
ED
572 __u16 queue_mapping;
573 __u8 cloned:1,
6869c4d8 574 nohdr:1,
b84f4cc9 575 fclone:2,
a59322be 576 peeked:1,
b1937227
ED
577 head_frag:1,
578 xmit_more:1;
579 /* one bit hole */
fe55f6d5 580 kmemcheck_bitfield_end(flags1);
4031ae6e 581
b1937227
ED
582 /* fields enclosed in headers_start/headers_end are copied
583 * using a single memcpy() in __copy_skb_header()
584 */
ebcf34f3 585 /* private: */
b1937227 586 __u32 headers_start[0];
ebcf34f3 587 /* public: */
4031ae6e 588
233577a2
HFS
589/* if you move pkt_type around you also must adapt those constants */
590#ifdef __BIG_ENDIAN_BITFIELD
591#define PKT_TYPE_MAX (7 << 5)
592#else
593#define PKT_TYPE_MAX 7
1da177e4 594#endif
233577a2 595#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 596
233577a2 597 __u8 __pkt_type_offset[0];
b1937227 598 __u8 pkt_type:3;
c93bdd0e 599 __u8 pfmemalloc:1;
b1937227
ED
600 __u8 ignore_df:1;
601 __u8 nfctinfo:3;
602
603 __u8 nf_trace:1;
604 __u8 ip_summed:2;
3853b584 605 __u8 ooo_okay:1;
61b905da 606 __u8 l4_hash:1;
a3b18ddb 607 __u8 sw_hash:1;
6e3e939f
JB
608 __u8 wifi_acked_valid:1;
609 __u8 wifi_acked:1;
b1937227 610
3bdc0eba 611 __u8 no_fcs:1;
77cffe23 612 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 613 __u8 encapsulation:1;
7e2b10c1 614 __u8 encap_hdr_csum:1;
5d0c2b95 615 __u8 csum_valid:1;
7e3cead5 616 __u8 csum_complete_sw:1;
b1937227
ED
617 __u8 csum_level:2;
618 __u8 csum_bad:1;
fe55f6d5 619
b1937227
ED
620#ifdef CONFIG_IPV6_NDISC_NODETYPE
621 __u8 ndisc_nodetype:2;
622#endif
623 __u8 ipvs_property:1;
8bce6d7d 624 __u8 inner_protocol_type:1;
e585f236
TH
625 __u8 remcsum_offload:1;
626 /* 3 or 5 bit hole */
b1937227
ED
627
628#ifdef CONFIG_NET_SCHED
629 __u16 tc_index; /* traffic control index */
630#ifdef CONFIG_NET_CLS_ACT
631 __u16 tc_verd; /* traffic control verdict */
632#endif
633#endif
fe55f6d5 634
b1937227
ED
635 union {
636 __wsum csum;
637 struct {
638 __u16 csum_start;
639 __u16 csum_offset;
640 };
641 };
642 __u32 priority;
643 int skb_iif;
644 __u32 hash;
645 __be16 vlan_proto;
646 __u16 vlan_tci;
2bd82484
ED
647#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
648 union {
649 unsigned int napi_id;
650 unsigned int sender_cpu;
651 };
97fc2f08 652#endif
984bc16c
JM
653#ifdef CONFIG_NETWORK_SECMARK
654 __u32 secmark;
655#endif
3b885787
NH
656 union {
657 __u32 mark;
16fad69c 658 __u32 reserved_tailroom;
3b885787 659 };
1da177e4 660
8bce6d7d
TH
661 union {
662 __be16 inner_protocol;
663 __u8 inner_ipproto;
664 };
665
1a37e412
SH
666 __u16 inner_transport_header;
667 __u16 inner_network_header;
668 __u16 inner_mac_header;
b1937227
ED
669
670 __be16 protocol;
1a37e412
SH
671 __u16 transport_header;
672 __u16 network_header;
673 __u16 mac_header;
b1937227 674
ebcf34f3 675 /* private: */
b1937227 676 __u32 headers_end[0];
ebcf34f3 677 /* public: */
b1937227 678
1da177e4 679 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 680 sk_buff_data_t tail;
4305b541 681 sk_buff_data_t end;
1da177e4 682 unsigned char *head,
4305b541 683 *data;
27a884dc
ACM
684 unsigned int truesize;
685 atomic_t users;
1da177e4
LT
686};
687
688#ifdef __KERNEL__
689/*
690 * Handling routines are only of interest to the kernel
691 */
692#include <linux/slab.h>
693
1da177e4 694
c93bdd0e
MG
695#define SKB_ALLOC_FCLONE 0x01
696#define SKB_ALLOC_RX 0x02
fd11a83d 697#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
698
699/* Returns true if the skb was allocated from PFMEMALLOC reserves */
700static inline bool skb_pfmemalloc(const struct sk_buff *skb)
701{
702 return unlikely(skb->pfmemalloc);
703}
704
7fee226a
ED
705/*
706 * skb might have a dst pointer attached, refcounted or not.
707 * _skb_refdst low order bit is set if refcount was _not_ taken
708 */
709#define SKB_DST_NOREF 1UL
710#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
711
712/**
713 * skb_dst - returns skb dst_entry
714 * @skb: buffer
715 *
716 * Returns skb dst_entry, regardless of reference taken or not.
717 */
adf30907
ED
718static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
719{
7fee226a
ED
720 /* If refdst was not refcounted, check we still are in a
721 * rcu_read_lock section
722 */
723 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
724 !rcu_read_lock_held() &&
725 !rcu_read_lock_bh_held());
726 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
727}
728
7fee226a
ED
729/**
730 * skb_dst_set - sets skb dst
731 * @skb: buffer
732 * @dst: dst entry
733 *
734 * Sets skb dst, assuming a reference was taken on dst and should
735 * be released by skb_dst_drop()
736 */
adf30907
ED
737static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
738{
7fee226a
ED
739 skb->_skb_refdst = (unsigned long)dst;
740}
741
932bc4d7
JA
742/**
743 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
744 * @skb: buffer
745 * @dst: dst entry
746 *
747 * Sets skb dst, assuming a reference was not taken on dst.
748 * If dst entry is cached, we do not take reference and dst_release
749 * will be avoided by refdst_drop. If dst entry is not cached, we take
750 * reference, so that last dst_release can destroy the dst immediately.
751 */
752static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
753{
dbfc4fb7
HFS
754 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
755 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 756}
7fee226a
ED
757
758/**
25985edc 759 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
760 * @skb: buffer
761 */
762static inline bool skb_dst_is_noref(const struct sk_buff *skb)
763{
764 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
765}
766
511c3f92
ED
767static inline struct rtable *skb_rtable(const struct sk_buff *skb)
768{
adf30907 769 return (struct rtable *)skb_dst(skb);
511c3f92
ED
770}
771
7965bd4d
JP
772void kfree_skb(struct sk_buff *skb);
773void kfree_skb_list(struct sk_buff *segs);
774void skb_tx_error(struct sk_buff *skb);
775void consume_skb(struct sk_buff *skb);
776void __kfree_skb(struct sk_buff *skb);
d7e8883c 777extern struct kmem_cache *skbuff_head_cache;
bad43ca8 778
7965bd4d
JP
779void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
780bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
781 bool *fragstolen, int *delta_truesize);
bad43ca8 782
7965bd4d
JP
783struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
784 int node);
2ea2f62c 785struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 786struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 787static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 788 gfp_t priority)
d179cd12 789{
564824b0 790 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
791}
792
2e4e4410
ED
793struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
794 unsigned long data_len,
795 int max_page_order,
796 int *errcode,
797 gfp_t gfp_mask);
798
d0bf4a9e
ED
799/* Layout of fast clones : [skb1][skb2][fclone_ref] */
800struct sk_buff_fclones {
801 struct sk_buff skb1;
802
803 struct sk_buff skb2;
804
805 atomic_t fclone_ref;
806};
807
808/**
809 * skb_fclone_busy - check if fclone is busy
810 * @skb: buffer
811 *
812 * Returns true is skb is a fast clone, and its clone is not freed.
39bb5e62
ED
813 * Some drivers call skb_orphan() in their ndo_start_xmit(),
814 * so we also check that this didnt happen.
d0bf4a9e 815 */
39bb5e62
ED
816static inline bool skb_fclone_busy(const struct sock *sk,
817 const struct sk_buff *skb)
d0bf4a9e
ED
818{
819 const struct sk_buff_fclones *fclones;
820
821 fclones = container_of(skb, struct sk_buff_fclones, skb1);
822
823 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 824 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 825 fclones->skb2.sk == sk;
d0bf4a9e
ED
826}
827
d179cd12 828static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 829 gfp_t priority)
d179cd12 830{
c93bdd0e 831 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
832}
833
7965bd4d 834struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
835static inline struct sk_buff *alloc_skb_head(gfp_t priority)
836{
837 return __alloc_skb_head(priority, -1);
838}
839
7965bd4d
JP
840struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
841int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
842struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
843struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
844struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
845 gfp_t gfp_mask, bool fclone);
846static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
847 gfp_t gfp_mask)
848{
849 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
850}
7965bd4d
JP
851
852int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
853struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
854 unsigned int headroom);
855struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
856 int newtailroom, gfp_t priority);
25a91d8d
FD
857int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
858 int offset, int len);
7965bd4d
JP
859int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
860 int len);
861int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
862int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 863#define dev_kfree_skb(a) consume_skb(a)
1da177e4 864
7965bd4d
JP
865int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
866 int getfrag(void *from, char *to, int offset,
867 int len, int odd, struct sk_buff *skb),
868 void *from, int length);
e89e9cf5 869
be12a1fe
HFS
870int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
871 int offset, size_t size);
872
d94d9fee 873struct skb_seq_state {
677e90ed
TG
874 __u32 lower_offset;
875 __u32 upper_offset;
876 __u32 frag_idx;
877 __u32 stepped_offset;
878 struct sk_buff *root_skb;
879 struct sk_buff *cur_skb;
880 __u8 *frag_data;
881};
882
7965bd4d
JP
883void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
884 unsigned int to, struct skb_seq_state *st);
885unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
886 struct skb_seq_state *st);
887void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 888
7965bd4d 889unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 890 unsigned int to, struct ts_config *config);
3fc7e8a6 891
09323cc4
TH
892/*
893 * Packet hash types specify the type of hash in skb_set_hash.
894 *
895 * Hash types refer to the protocol layer addresses which are used to
896 * construct a packet's hash. The hashes are used to differentiate or identify
897 * flows of the protocol layer for the hash type. Hash types are either
898 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
899 *
900 * Properties of hashes:
901 *
902 * 1) Two packets in different flows have different hash values
903 * 2) Two packets in the same flow should have the same hash value
904 *
905 * A hash at a higher layer is considered to be more specific. A driver should
906 * set the most specific hash possible.
907 *
908 * A driver cannot indicate a more specific hash than the layer at which a hash
909 * was computed. For instance an L3 hash cannot be set as an L4 hash.
910 *
911 * A driver may indicate a hash level which is less specific than the
912 * actual layer the hash was computed on. For instance, a hash computed
913 * at L4 may be considered an L3 hash. This should only be done if the
914 * driver can't unambiguously determine that the HW computed the hash at
915 * the higher layer. Note that the "should" in the second property above
916 * permits this.
917 */
918enum pkt_hash_types {
919 PKT_HASH_TYPE_NONE, /* Undefined type */
920 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
921 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
922 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
923};
924
925static inline void
926skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
927{
61b905da 928 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 929 skb->sw_hash = 0;
61b905da 930 skb->hash = hash;
09323cc4
TH
931}
932
3958afa1 933static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 934{
a3b18ddb 935 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 936 __skb_get_hash(skb);
bfb564e7 937
61b905da 938 return skb->hash;
bfb564e7
KK
939}
940
50fb7992
TH
941__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
942
57bdf7f4
TH
943static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
944{
61b905da 945 return skb->hash;
57bdf7f4
TH
946}
947
7539fadc
TH
948static inline void skb_clear_hash(struct sk_buff *skb)
949{
61b905da 950 skb->hash = 0;
a3b18ddb 951 skb->sw_hash = 0;
61b905da 952 skb->l4_hash = 0;
7539fadc
TH
953}
954
955static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
956{
61b905da 957 if (!skb->l4_hash)
7539fadc
TH
958 skb_clear_hash(skb);
959}
960
3df7a74e
TH
961static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
962{
61b905da 963 to->hash = from->hash;
a3b18ddb 964 to->sw_hash = from->sw_hash;
61b905da 965 to->l4_hash = from->l4_hash;
3df7a74e
TH
966};
967
c29390c6
ED
968static inline void skb_sender_cpu_clear(struct sk_buff *skb)
969{
970#ifdef CONFIG_XPS
971 skb->sender_cpu = 0;
972#endif
973}
974
4305b541
ACM
975#ifdef NET_SKBUFF_DATA_USES_OFFSET
976static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
977{
978 return skb->head + skb->end;
979}
ec47ea82
AD
980
981static inline unsigned int skb_end_offset(const struct sk_buff *skb)
982{
983 return skb->end;
984}
4305b541
ACM
985#else
986static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
987{
988 return skb->end;
989}
ec47ea82
AD
990
991static inline unsigned int skb_end_offset(const struct sk_buff *skb)
992{
993 return skb->end - skb->head;
994}
4305b541
ACM
995#endif
996
1da177e4 997/* Internal */
4305b541 998#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 999
ac45f602
PO
1000static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1001{
1002 return &skb_shinfo(skb)->hwtstamps;
1003}
1004
1da177e4
LT
1005/**
1006 * skb_queue_empty - check if a queue is empty
1007 * @list: queue head
1008 *
1009 * Returns true if the queue is empty, false otherwise.
1010 */
1011static inline int skb_queue_empty(const struct sk_buff_head *list)
1012{
fd44b93c 1013 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1014}
1015
fc7ebb21
DM
1016/**
1017 * skb_queue_is_last - check if skb is the last entry in the queue
1018 * @list: queue head
1019 * @skb: buffer
1020 *
1021 * Returns true if @skb is the last buffer on the list.
1022 */
1023static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1024 const struct sk_buff *skb)
1025{
fd44b93c 1026 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1027}
1028
832d11c5
IJ
1029/**
1030 * skb_queue_is_first - check if skb is the first entry in the queue
1031 * @list: queue head
1032 * @skb: buffer
1033 *
1034 * Returns true if @skb is the first buffer on the list.
1035 */
1036static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1037 const struct sk_buff *skb)
1038{
fd44b93c 1039 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1040}
1041
249c8b42
DM
1042/**
1043 * skb_queue_next - return the next packet in the queue
1044 * @list: queue head
1045 * @skb: current buffer
1046 *
1047 * Return the next packet in @list after @skb. It is only valid to
1048 * call this if skb_queue_is_last() evaluates to false.
1049 */
1050static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1051 const struct sk_buff *skb)
1052{
1053 /* This BUG_ON may seem severe, but if we just return then we
1054 * are going to dereference garbage.
1055 */
1056 BUG_ON(skb_queue_is_last(list, skb));
1057 return skb->next;
1058}
1059
832d11c5
IJ
1060/**
1061 * skb_queue_prev - return the prev packet in the queue
1062 * @list: queue head
1063 * @skb: current buffer
1064 *
1065 * Return the prev packet in @list before @skb. It is only valid to
1066 * call this if skb_queue_is_first() evaluates to false.
1067 */
1068static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1069 const struct sk_buff *skb)
1070{
1071 /* This BUG_ON may seem severe, but if we just return then we
1072 * are going to dereference garbage.
1073 */
1074 BUG_ON(skb_queue_is_first(list, skb));
1075 return skb->prev;
1076}
1077
1da177e4
LT
1078/**
1079 * skb_get - reference buffer
1080 * @skb: buffer to reference
1081 *
1082 * Makes another reference to a socket buffer and returns a pointer
1083 * to the buffer.
1084 */
1085static inline struct sk_buff *skb_get(struct sk_buff *skb)
1086{
1087 atomic_inc(&skb->users);
1088 return skb;
1089}
1090
1091/*
1092 * If users == 1, we are the only owner and are can avoid redundant
1093 * atomic change.
1094 */
1095
1da177e4
LT
1096/**
1097 * skb_cloned - is the buffer a clone
1098 * @skb: buffer to check
1099 *
1100 * Returns true if the buffer was generated with skb_clone() and is
1101 * one of multiple shared copies of the buffer. Cloned buffers are
1102 * shared data so must not be written to under normal circumstances.
1103 */
1104static inline int skb_cloned(const struct sk_buff *skb)
1105{
1106 return skb->cloned &&
1107 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1108}
1109
14bbd6a5
PS
1110static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1111{
1112 might_sleep_if(pri & __GFP_WAIT);
1113
1114 if (skb_cloned(skb))
1115 return pskb_expand_head(skb, 0, 0, pri);
1116
1117 return 0;
1118}
1119
1da177e4
LT
1120/**
1121 * skb_header_cloned - is the header a clone
1122 * @skb: buffer to check
1123 *
1124 * Returns true if modifying the header part of the buffer requires
1125 * the data to be copied.
1126 */
1127static inline int skb_header_cloned(const struct sk_buff *skb)
1128{
1129 int dataref;
1130
1131 if (!skb->cloned)
1132 return 0;
1133
1134 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1135 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1136 return dataref != 1;
1137}
1138
1139/**
1140 * skb_header_release - release reference to header
1141 * @skb: buffer to operate on
1142 *
1143 * Drop a reference to the header part of the buffer. This is done
1144 * by acquiring a payload reference. You must not read from the header
1145 * part of skb->data after this.
f4a775d1 1146 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1147 */
1148static inline void skb_header_release(struct sk_buff *skb)
1149{
1150 BUG_ON(skb->nohdr);
1151 skb->nohdr = 1;
1152 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1153}
1154
f4a775d1
ED
1155/**
1156 * __skb_header_release - release reference to header
1157 * @skb: buffer to operate on
1158 *
1159 * Variant of skb_header_release() assuming skb is private to caller.
1160 * We can avoid one atomic operation.
1161 */
1162static inline void __skb_header_release(struct sk_buff *skb)
1163{
1164 skb->nohdr = 1;
1165 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1166}
1167
1168
1da177e4
LT
1169/**
1170 * skb_shared - is the buffer shared
1171 * @skb: buffer to check
1172 *
1173 * Returns true if more than one person has a reference to this
1174 * buffer.
1175 */
1176static inline int skb_shared(const struct sk_buff *skb)
1177{
1178 return atomic_read(&skb->users) != 1;
1179}
1180
1181/**
1182 * skb_share_check - check if buffer is shared and if so clone it
1183 * @skb: buffer to check
1184 * @pri: priority for memory allocation
1185 *
1186 * If the buffer is shared the buffer is cloned and the old copy
1187 * drops a reference. A new clone with a single reference is returned.
1188 * If the buffer is not shared the original buffer is returned. When
1189 * being called from interrupt status or with spinlocks held pri must
1190 * be GFP_ATOMIC.
1191 *
1192 * NULL is returned on a memory allocation failure.
1193 */
47061bc4 1194static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1195{
1196 might_sleep_if(pri & __GFP_WAIT);
1197 if (skb_shared(skb)) {
1198 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1199
1200 if (likely(nskb))
1201 consume_skb(skb);
1202 else
1203 kfree_skb(skb);
1da177e4
LT
1204 skb = nskb;
1205 }
1206 return skb;
1207}
1208
1209/*
1210 * Copy shared buffers into a new sk_buff. We effectively do COW on
1211 * packets to handle cases where we have a local reader and forward
1212 * and a couple of other messy ones. The normal one is tcpdumping
1213 * a packet thats being forwarded.
1214 */
1215
1216/**
1217 * skb_unshare - make a copy of a shared buffer
1218 * @skb: buffer to check
1219 * @pri: priority for memory allocation
1220 *
1221 * If the socket buffer is a clone then this function creates a new
1222 * copy of the data, drops a reference count on the old copy and returns
1223 * the new copy with the reference count at 1. If the buffer is not a clone
1224 * the original buffer is returned. When called with a spinlock held or
1225 * from interrupt state @pri must be %GFP_ATOMIC
1226 *
1227 * %NULL is returned on a memory allocation failure.
1228 */
e2bf521d 1229static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1230 gfp_t pri)
1da177e4
LT
1231{
1232 might_sleep_if(pri & __GFP_WAIT);
1233 if (skb_cloned(skb)) {
1234 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1235
1236 /* Free our shared copy */
1237 if (likely(nskb))
1238 consume_skb(skb);
1239 else
1240 kfree_skb(skb);
1da177e4
LT
1241 skb = nskb;
1242 }
1243 return skb;
1244}
1245
1246/**
1a5778aa 1247 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1248 * @list_: list to peek at
1249 *
1250 * Peek an &sk_buff. Unlike most other operations you _MUST_
1251 * be careful with this one. A peek leaves the buffer on the
1252 * list and someone else may run off with it. You must hold
1253 * the appropriate locks or have a private queue to do this.
1254 *
1255 * Returns %NULL for an empty list or a pointer to the head element.
1256 * The reference count is not incremented and the reference is therefore
1257 * volatile. Use with caution.
1258 */
05bdd2f1 1259static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1260{
18d07000
ED
1261 struct sk_buff *skb = list_->next;
1262
1263 if (skb == (struct sk_buff *)list_)
1264 skb = NULL;
1265 return skb;
1da177e4
LT
1266}
1267
da5ef6e5
PE
1268/**
1269 * skb_peek_next - peek skb following the given one from a queue
1270 * @skb: skb to start from
1271 * @list_: list to peek at
1272 *
1273 * Returns %NULL when the end of the list is met or a pointer to the
1274 * next element. The reference count is not incremented and the
1275 * reference is therefore volatile. Use with caution.
1276 */
1277static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1278 const struct sk_buff_head *list_)
1279{
1280 struct sk_buff *next = skb->next;
18d07000 1281
da5ef6e5
PE
1282 if (next == (struct sk_buff *)list_)
1283 next = NULL;
1284 return next;
1285}
1286
1da177e4 1287/**
1a5778aa 1288 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1289 * @list_: list to peek at
1290 *
1291 * Peek an &sk_buff. Unlike most other operations you _MUST_
1292 * be careful with this one. A peek leaves the buffer on the
1293 * list and someone else may run off with it. You must hold
1294 * the appropriate locks or have a private queue to do this.
1295 *
1296 * Returns %NULL for an empty list or a pointer to the tail element.
1297 * The reference count is not incremented and the reference is therefore
1298 * volatile. Use with caution.
1299 */
05bdd2f1 1300static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1301{
18d07000
ED
1302 struct sk_buff *skb = list_->prev;
1303
1304 if (skb == (struct sk_buff *)list_)
1305 skb = NULL;
1306 return skb;
1307
1da177e4
LT
1308}
1309
1310/**
1311 * skb_queue_len - get queue length
1312 * @list_: list to measure
1313 *
1314 * Return the length of an &sk_buff queue.
1315 */
1316static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1317{
1318 return list_->qlen;
1319}
1320
67fed459
DM
1321/**
1322 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1323 * @list: queue to initialize
1324 *
1325 * This initializes only the list and queue length aspects of
1326 * an sk_buff_head object. This allows to initialize the list
1327 * aspects of an sk_buff_head without reinitializing things like
1328 * the spinlock. It can also be used for on-stack sk_buff_head
1329 * objects where the spinlock is known to not be used.
1330 */
1331static inline void __skb_queue_head_init(struct sk_buff_head *list)
1332{
1333 list->prev = list->next = (struct sk_buff *)list;
1334 list->qlen = 0;
1335}
1336
76f10ad0
AV
1337/*
1338 * This function creates a split out lock class for each invocation;
1339 * this is needed for now since a whole lot of users of the skb-queue
1340 * infrastructure in drivers have different locking usage (in hardirq)
1341 * than the networking core (in softirq only). In the long run either the
1342 * network layer or drivers should need annotation to consolidate the
1343 * main types of usage into 3 classes.
1344 */
1da177e4
LT
1345static inline void skb_queue_head_init(struct sk_buff_head *list)
1346{
1347 spin_lock_init(&list->lock);
67fed459 1348 __skb_queue_head_init(list);
1da177e4
LT
1349}
1350
c2ecba71
PE
1351static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1352 struct lock_class_key *class)
1353{
1354 skb_queue_head_init(list);
1355 lockdep_set_class(&list->lock, class);
1356}
1357
1da177e4 1358/*
bf299275 1359 * Insert an sk_buff on a list.
1da177e4
LT
1360 *
1361 * The "__skb_xxxx()" functions are the non-atomic ones that
1362 * can only be called with interrupts disabled.
1363 */
7965bd4d
JP
1364void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1365 struct sk_buff_head *list);
bf299275
GR
1366static inline void __skb_insert(struct sk_buff *newsk,
1367 struct sk_buff *prev, struct sk_buff *next,
1368 struct sk_buff_head *list)
1369{
1370 newsk->next = next;
1371 newsk->prev = prev;
1372 next->prev = prev->next = newsk;
1373 list->qlen++;
1374}
1da177e4 1375
67fed459
DM
1376static inline void __skb_queue_splice(const struct sk_buff_head *list,
1377 struct sk_buff *prev,
1378 struct sk_buff *next)
1379{
1380 struct sk_buff *first = list->next;
1381 struct sk_buff *last = list->prev;
1382
1383 first->prev = prev;
1384 prev->next = first;
1385
1386 last->next = next;
1387 next->prev = last;
1388}
1389
1390/**
1391 * skb_queue_splice - join two skb lists, this is designed for stacks
1392 * @list: the new list to add
1393 * @head: the place to add it in the first list
1394 */
1395static inline void skb_queue_splice(const struct sk_buff_head *list,
1396 struct sk_buff_head *head)
1397{
1398 if (!skb_queue_empty(list)) {
1399 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1400 head->qlen += list->qlen;
67fed459
DM
1401 }
1402}
1403
1404/**
d9619496 1405 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1406 * @list: the new list to add
1407 * @head: the place to add it in the first list
1408 *
1409 * The list at @list is reinitialised
1410 */
1411static inline void skb_queue_splice_init(struct sk_buff_head *list,
1412 struct sk_buff_head *head)
1413{
1414 if (!skb_queue_empty(list)) {
1415 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1416 head->qlen += list->qlen;
67fed459
DM
1417 __skb_queue_head_init(list);
1418 }
1419}
1420
1421/**
1422 * skb_queue_splice_tail - join two skb lists, each list being a queue
1423 * @list: the new list to add
1424 * @head: the place to add it in the first list
1425 */
1426static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1427 struct sk_buff_head *head)
1428{
1429 if (!skb_queue_empty(list)) {
1430 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1431 head->qlen += list->qlen;
67fed459
DM
1432 }
1433}
1434
1435/**
d9619496 1436 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1437 * @list: the new list to add
1438 * @head: the place to add it in the first list
1439 *
1440 * Each of the lists is a queue.
1441 * The list at @list is reinitialised
1442 */
1443static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1444 struct sk_buff_head *head)
1445{
1446 if (!skb_queue_empty(list)) {
1447 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1448 head->qlen += list->qlen;
67fed459
DM
1449 __skb_queue_head_init(list);
1450 }
1451}
1452
1da177e4 1453/**
300ce174 1454 * __skb_queue_after - queue a buffer at the list head
1da177e4 1455 * @list: list to use
300ce174 1456 * @prev: place after this buffer
1da177e4
LT
1457 * @newsk: buffer to queue
1458 *
300ce174 1459 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1460 * and you must therefore hold required locks before calling it.
1461 *
1462 * A buffer cannot be placed on two lists at the same time.
1463 */
300ce174
SH
1464static inline void __skb_queue_after(struct sk_buff_head *list,
1465 struct sk_buff *prev,
1466 struct sk_buff *newsk)
1da177e4 1467{
bf299275 1468 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1469}
1470
7965bd4d
JP
1471void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1472 struct sk_buff_head *list);
7de6c033 1473
f5572855
GR
1474static inline void __skb_queue_before(struct sk_buff_head *list,
1475 struct sk_buff *next,
1476 struct sk_buff *newsk)
1477{
1478 __skb_insert(newsk, next->prev, next, list);
1479}
1480
300ce174
SH
1481/**
1482 * __skb_queue_head - queue a buffer at the list head
1483 * @list: list to use
1484 * @newsk: buffer to queue
1485 *
1486 * Queue a buffer at the start of a list. This function takes no locks
1487 * and you must therefore hold required locks before calling it.
1488 *
1489 * A buffer cannot be placed on two lists at the same time.
1490 */
7965bd4d 1491void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1492static inline void __skb_queue_head(struct sk_buff_head *list,
1493 struct sk_buff *newsk)
1494{
1495 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1496}
1497
1da177e4
LT
1498/**
1499 * __skb_queue_tail - queue a buffer at the list tail
1500 * @list: list to use
1501 * @newsk: buffer to queue
1502 *
1503 * Queue a buffer at the end of a list. This function takes no locks
1504 * and you must therefore hold required locks before calling it.
1505 *
1506 * A buffer cannot be placed on two lists at the same time.
1507 */
7965bd4d 1508void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1509static inline void __skb_queue_tail(struct sk_buff_head *list,
1510 struct sk_buff *newsk)
1511{
f5572855 1512 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1513}
1514
1da177e4
LT
1515/*
1516 * remove sk_buff from list. _Must_ be called atomically, and with
1517 * the list known..
1518 */
7965bd4d 1519void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1520static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1521{
1522 struct sk_buff *next, *prev;
1523
1524 list->qlen--;
1525 next = skb->next;
1526 prev = skb->prev;
1527 skb->next = skb->prev = NULL;
1da177e4
LT
1528 next->prev = prev;
1529 prev->next = next;
1530}
1531
f525c06d
GR
1532/**
1533 * __skb_dequeue - remove from the head of the queue
1534 * @list: list to dequeue from
1535 *
1536 * Remove the head of the list. This function does not take any locks
1537 * so must be used with appropriate locks held only. The head item is
1538 * returned or %NULL if the list is empty.
1539 */
7965bd4d 1540struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1541static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1542{
1543 struct sk_buff *skb = skb_peek(list);
1544 if (skb)
1545 __skb_unlink(skb, list);
1546 return skb;
1547}
1da177e4
LT
1548
1549/**
1550 * __skb_dequeue_tail - remove from the tail of the queue
1551 * @list: list to dequeue from
1552 *
1553 * Remove the tail of the list. This function does not take any locks
1554 * so must be used with appropriate locks held only. The tail item is
1555 * returned or %NULL if the list is empty.
1556 */
7965bd4d 1557struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1558static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1559{
1560 struct sk_buff *skb = skb_peek_tail(list);
1561 if (skb)
1562 __skb_unlink(skb, list);
1563 return skb;
1564}
1565
1566
bdcc0924 1567static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1568{
1569 return skb->data_len;
1570}
1571
1572static inline unsigned int skb_headlen(const struct sk_buff *skb)
1573{
1574 return skb->len - skb->data_len;
1575}
1576
1577static inline int skb_pagelen(const struct sk_buff *skb)
1578{
1579 int i, len = 0;
1580
1581 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1582 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1583 return len + skb_headlen(skb);
1584}
1585
131ea667
IC
1586/**
1587 * __skb_fill_page_desc - initialise a paged fragment in an skb
1588 * @skb: buffer containing fragment to be initialised
1589 * @i: paged fragment index to initialise
1590 * @page: the page to use for this fragment
1591 * @off: the offset to the data with @page
1592 * @size: the length of the data
1593 *
1594 * Initialises the @i'th fragment of @skb to point to &size bytes at
1595 * offset @off within @page.
1596 *
1597 * Does not take any additional reference on the fragment.
1598 */
1599static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1600 struct page *page, int off, int size)
1da177e4
LT
1601{
1602 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1603
c48a11c7
MG
1604 /*
1605 * Propagate page->pfmemalloc to the skb if we can. The problem is
1606 * that not all callers have unique ownership of the page. If
1607 * pfmemalloc is set, we check the mapping as a mapping implies
1608 * page->index is set (index and pfmemalloc share space).
1609 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1610 * do not lose pfmemalloc information as the pages would not be
1611 * allocated using __GFP_MEMALLOC.
1612 */
a8605c60 1613 frag->page.p = page;
1da177e4 1614 frag->page_offset = off;
9e903e08 1615 skb_frag_size_set(frag, size);
cca7af38
PE
1616
1617 page = compound_head(page);
1618 if (page->pfmemalloc && !page->mapping)
1619 skb->pfmemalloc = true;
131ea667
IC
1620}
1621
1622/**
1623 * skb_fill_page_desc - initialise a paged fragment in an skb
1624 * @skb: buffer containing fragment to be initialised
1625 * @i: paged fragment index to initialise
1626 * @page: the page to use for this fragment
1627 * @off: the offset to the data with @page
1628 * @size: the length of the data
1629 *
1630 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1631 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1632 * addition updates @skb such that @i is the last fragment.
1633 *
1634 * Does not take any additional reference on the fragment.
1635 */
1636static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1637 struct page *page, int off, int size)
1638{
1639 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1640 skb_shinfo(skb)->nr_frags = i + 1;
1641}
1642
7965bd4d
JP
1643void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1644 int size, unsigned int truesize);
654bed16 1645
f8e617e1
JW
1646void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1647 unsigned int truesize);
1648
1da177e4 1649#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1650#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1651#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1652
27a884dc
ACM
1653#ifdef NET_SKBUFF_DATA_USES_OFFSET
1654static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1655{
1656 return skb->head + skb->tail;
1657}
1658
1659static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1660{
1661 skb->tail = skb->data - skb->head;
1662}
1663
1664static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1665{
1666 skb_reset_tail_pointer(skb);
1667 skb->tail += offset;
1668}
7cc46190 1669
27a884dc
ACM
1670#else /* NET_SKBUFF_DATA_USES_OFFSET */
1671static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1672{
1673 return skb->tail;
1674}
1675
1676static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1677{
1678 skb->tail = skb->data;
1679}
1680
1681static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1682{
1683 skb->tail = skb->data + offset;
1684}
4305b541 1685
27a884dc
ACM
1686#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1687
1da177e4
LT
1688/*
1689 * Add data to an sk_buff
1690 */
0c7ddf36 1691unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1692unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1693static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1694{
27a884dc 1695 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1696 SKB_LINEAR_ASSERT(skb);
1697 skb->tail += len;
1698 skb->len += len;
1699 return tmp;
1700}
1701
7965bd4d 1702unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1703static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1704{
1705 skb->data -= len;
1706 skb->len += len;
1707 return skb->data;
1708}
1709
7965bd4d 1710unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1711static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1712{
1713 skb->len -= len;
1714 BUG_ON(skb->len < skb->data_len);
1715 return skb->data += len;
1716}
1717
47d29646
DM
1718static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1719{
1720 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1721}
1722
7965bd4d 1723unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1724
1725static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1726{
1727 if (len > skb_headlen(skb) &&
987c402a 1728 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1729 return NULL;
1730 skb->len -= len;
1731 return skb->data += len;
1732}
1733
1734static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1735{
1736 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1737}
1738
1739static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1740{
1741 if (likely(len <= skb_headlen(skb)))
1742 return 1;
1743 if (unlikely(len > skb->len))
1744 return 0;
987c402a 1745 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1746}
1747
1748/**
1749 * skb_headroom - bytes at buffer head
1750 * @skb: buffer to check
1751 *
1752 * Return the number of bytes of free space at the head of an &sk_buff.
1753 */
c2636b4d 1754static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1755{
1756 return skb->data - skb->head;
1757}
1758
1759/**
1760 * skb_tailroom - bytes at buffer end
1761 * @skb: buffer to check
1762 *
1763 * Return the number of bytes of free space at the tail of an sk_buff
1764 */
1765static inline int skb_tailroom(const struct sk_buff *skb)
1766{
4305b541 1767 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1768}
1769
a21d4572
ED
1770/**
1771 * skb_availroom - bytes at buffer end
1772 * @skb: buffer to check
1773 *
1774 * Return the number of bytes of free space at the tail of an sk_buff
1775 * allocated by sk_stream_alloc()
1776 */
1777static inline int skb_availroom(const struct sk_buff *skb)
1778{
16fad69c
ED
1779 if (skb_is_nonlinear(skb))
1780 return 0;
1781
1782 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1783}
1784
1da177e4
LT
1785/**
1786 * skb_reserve - adjust headroom
1787 * @skb: buffer to alter
1788 * @len: bytes to move
1789 *
1790 * Increase the headroom of an empty &sk_buff by reducing the tail
1791 * room. This is only allowed for an empty buffer.
1792 */
8243126c 1793static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1794{
1795 skb->data += len;
1796 skb->tail += len;
1797}
1798
8bce6d7d
TH
1799#define ENCAP_TYPE_ETHER 0
1800#define ENCAP_TYPE_IPPROTO 1
1801
1802static inline void skb_set_inner_protocol(struct sk_buff *skb,
1803 __be16 protocol)
1804{
1805 skb->inner_protocol = protocol;
1806 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1807}
1808
1809static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1810 __u8 ipproto)
1811{
1812 skb->inner_ipproto = ipproto;
1813 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1814}
1815
6a674e9c
JG
1816static inline void skb_reset_inner_headers(struct sk_buff *skb)
1817{
aefbd2b3 1818 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1819 skb->inner_network_header = skb->network_header;
1820 skb->inner_transport_header = skb->transport_header;
1821}
1822
0b5c9db1
JP
1823static inline void skb_reset_mac_len(struct sk_buff *skb)
1824{
1825 skb->mac_len = skb->network_header - skb->mac_header;
1826}
1827
6a674e9c
JG
1828static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1829 *skb)
1830{
1831 return skb->head + skb->inner_transport_header;
1832}
1833
1834static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1835{
1836 skb->inner_transport_header = skb->data - skb->head;
1837}
1838
1839static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1840 const int offset)
1841{
1842 skb_reset_inner_transport_header(skb);
1843 skb->inner_transport_header += offset;
1844}
1845
1846static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1847{
1848 return skb->head + skb->inner_network_header;
1849}
1850
1851static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1852{
1853 skb->inner_network_header = skb->data - skb->head;
1854}
1855
1856static inline void skb_set_inner_network_header(struct sk_buff *skb,
1857 const int offset)
1858{
1859 skb_reset_inner_network_header(skb);
1860 skb->inner_network_header += offset;
1861}
1862
aefbd2b3
PS
1863static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1864{
1865 return skb->head + skb->inner_mac_header;
1866}
1867
1868static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1869{
1870 skb->inner_mac_header = skb->data - skb->head;
1871}
1872
1873static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1874 const int offset)
1875{
1876 skb_reset_inner_mac_header(skb);
1877 skb->inner_mac_header += offset;
1878}
fda55eca
ED
1879static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1880{
35d04610 1881 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1882}
1883
9c70220b
ACM
1884static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1885{
2e07fa9c 1886 return skb->head + skb->transport_header;
9c70220b
ACM
1887}
1888
badff6d0
ACM
1889static inline void skb_reset_transport_header(struct sk_buff *skb)
1890{
2e07fa9c 1891 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1892}
1893
967b05f6
ACM
1894static inline void skb_set_transport_header(struct sk_buff *skb,
1895 const int offset)
1896{
2e07fa9c
ACM
1897 skb_reset_transport_header(skb);
1898 skb->transport_header += offset;
ea2ae17d
ACM
1899}
1900
d56f90a7
ACM
1901static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1902{
2e07fa9c 1903 return skb->head + skb->network_header;
d56f90a7
ACM
1904}
1905
c1d2bbe1
ACM
1906static inline void skb_reset_network_header(struct sk_buff *skb)
1907{
2e07fa9c 1908 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1909}
1910
c14d2450
ACM
1911static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1912{
2e07fa9c
ACM
1913 skb_reset_network_header(skb);
1914 skb->network_header += offset;
c14d2450
ACM
1915}
1916
2e07fa9c 1917static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1918{
2e07fa9c 1919 return skb->head + skb->mac_header;
bbe735e4
ACM
1920}
1921
2e07fa9c 1922static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1923{
35d04610 1924 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1925}
1926
1927static inline void skb_reset_mac_header(struct sk_buff *skb)
1928{
1929 skb->mac_header = skb->data - skb->head;
1930}
1931
1932static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1933{
1934 skb_reset_mac_header(skb);
1935 skb->mac_header += offset;
1936}
1937
0e3da5bb
TT
1938static inline void skb_pop_mac_header(struct sk_buff *skb)
1939{
1940 skb->mac_header = skb->network_header;
1941}
1942
fbbdb8f0
YX
1943static inline void skb_probe_transport_header(struct sk_buff *skb,
1944 const int offset_hint)
1945{
1946 struct flow_keys keys;
1947
1948 if (skb_transport_header_was_set(skb))
1949 return;
06635a35 1950 else if (skb_flow_dissect_flow_keys(skb, &keys))
42aecaa9 1951 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
1952 else
1953 skb_set_transport_header(skb, offset_hint);
1954}
1955
03606895
ED
1956static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1957{
1958 if (skb_mac_header_was_set(skb)) {
1959 const unsigned char *old_mac = skb_mac_header(skb);
1960
1961 skb_set_mac_header(skb, -skb->mac_len);
1962 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1963 }
1964}
1965
04fb451e
MM
1966static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1967{
1968 return skb->csum_start - skb_headroom(skb);
1969}
1970
2e07fa9c
ACM
1971static inline int skb_transport_offset(const struct sk_buff *skb)
1972{
1973 return skb_transport_header(skb) - skb->data;
1974}
1975
1976static inline u32 skb_network_header_len(const struct sk_buff *skb)
1977{
1978 return skb->transport_header - skb->network_header;
1979}
1980
6a674e9c
JG
1981static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1982{
1983 return skb->inner_transport_header - skb->inner_network_header;
1984}
1985
2e07fa9c
ACM
1986static inline int skb_network_offset(const struct sk_buff *skb)
1987{
1988 return skb_network_header(skb) - skb->data;
1989}
48d49d0c 1990
6a674e9c
JG
1991static inline int skb_inner_network_offset(const struct sk_buff *skb)
1992{
1993 return skb_inner_network_header(skb) - skb->data;
1994}
1995
f9599ce1
CG
1996static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1997{
1998 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1999}
2000
1da177e4
LT
2001/*
2002 * CPUs often take a performance hit when accessing unaligned memory
2003 * locations. The actual performance hit varies, it can be small if the
2004 * hardware handles it or large if we have to take an exception and fix it
2005 * in software.
2006 *
2007 * Since an ethernet header is 14 bytes network drivers often end up with
2008 * the IP header at an unaligned offset. The IP header can be aligned by
2009 * shifting the start of the packet by 2 bytes. Drivers should do this
2010 * with:
2011 *
8660c124 2012 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2013 *
2014 * The downside to this alignment of the IP header is that the DMA is now
2015 * unaligned. On some architectures the cost of an unaligned DMA is high
2016 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2017 *
1da177e4
LT
2018 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2019 * to be overridden.
2020 */
2021#ifndef NET_IP_ALIGN
2022#define NET_IP_ALIGN 2
2023#endif
2024
025be81e
AB
2025/*
2026 * The networking layer reserves some headroom in skb data (via
2027 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2028 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2029 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2030 *
2031 * Unfortunately this headroom changes the DMA alignment of the resulting
2032 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2033 * on some architectures. An architecture can override this value,
2034 * perhaps setting it to a cacheline in size (since that will maintain
2035 * cacheline alignment of the DMA). It must be a power of 2.
2036 *
d6301d3d 2037 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2038 * headroom, you should not reduce this.
5933dd2f
ED
2039 *
2040 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2041 * to reduce average number of cache lines per packet.
2042 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2043 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2044 */
2045#ifndef NET_SKB_PAD
5933dd2f 2046#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2047#endif
2048
7965bd4d 2049int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2050
2051static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2052{
c4264f27 2053 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2054 WARN_ON(1);
2055 return;
2056 }
27a884dc
ACM
2057 skb->len = len;
2058 skb_set_tail_pointer(skb, len);
1da177e4
LT
2059}
2060
7965bd4d 2061void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2062
2063static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2064{
3cc0e873
HX
2065 if (skb->data_len)
2066 return ___pskb_trim(skb, len);
2067 __skb_trim(skb, len);
2068 return 0;
1da177e4
LT
2069}
2070
2071static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2072{
2073 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2074}
2075
e9fa4f7b
HX
2076/**
2077 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2078 * @skb: buffer to alter
2079 * @len: new length
2080 *
2081 * This is identical to pskb_trim except that the caller knows that
2082 * the skb is not cloned so we should never get an error due to out-
2083 * of-memory.
2084 */
2085static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2086{
2087 int err = pskb_trim(skb, len);
2088 BUG_ON(err);
2089}
2090
1da177e4
LT
2091/**
2092 * skb_orphan - orphan a buffer
2093 * @skb: buffer to orphan
2094 *
2095 * If a buffer currently has an owner then we call the owner's
2096 * destructor function and make the @skb unowned. The buffer continues
2097 * to exist but is no longer charged to its former owner.
2098 */
2099static inline void skb_orphan(struct sk_buff *skb)
2100{
c34a7612 2101 if (skb->destructor) {
1da177e4 2102 skb->destructor(skb);
c34a7612
ED
2103 skb->destructor = NULL;
2104 skb->sk = NULL;
376c7311
ED
2105 } else {
2106 BUG_ON(skb->sk);
c34a7612 2107 }
1da177e4
LT
2108}
2109
a353e0ce
MT
2110/**
2111 * skb_orphan_frags - orphan the frags contained in a buffer
2112 * @skb: buffer to orphan frags from
2113 * @gfp_mask: allocation mask for replacement pages
2114 *
2115 * For each frag in the SKB which needs a destructor (i.e. has an
2116 * owner) create a copy of that frag and release the original
2117 * page by calling the destructor.
2118 */
2119static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2120{
2121 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2122 return 0;
2123 return skb_copy_ubufs(skb, gfp_mask);
2124}
2125
1da177e4
LT
2126/**
2127 * __skb_queue_purge - empty a list
2128 * @list: list to empty
2129 *
2130 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2131 * the list and one reference dropped. This function does not take the
2132 * list lock and the caller must hold the relevant locks to use it.
2133 */
7965bd4d 2134void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2135static inline void __skb_queue_purge(struct sk_buff_head *list)
2136{
2137 struct sk_buff *skb;
2138 while ((skb = __skb_dequeue(list)) != NULL)
2139 kfree_skb(skb);
2140}
2141
7965bd4d 2142void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2143
7965bd4d
JP
2144struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2145 gfp_t gfp_mask);
8af27456
CH
2146
2147/**
2148 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2149 * @dev: network device to receive on
2150 * @length: length to allocate
2151 *
2152 * Allocate a new &sk_buff and assign it a usage count of one. The
2153 * buffer has unspecified headroom built in. Users should allocate
2154 * the headroom they think they need without accounting for the
2155 * built in space. The built in space is used for optimisations.
2156 *
2157 * %NULL is returned if there is no free memory. Although this function
2158 * allocates memory it can be called from an interrupt.
2159 */
2160static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2161 unsigned int length)
8af27456
CH
2162{
2163 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2164}
2165
6f532612
ED
2166/* legacy helper around __netdev_alloc_skb() */
2167static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2168 gfp_t gfp_mask)
2169{
2170 return __netdev_alloc_skb(NULL, length, gfp_mask);
2171}
2172
2173/* legacy helper around netdev_alloc_skb() */
2174static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2175{
2176 return netdev_alloc_skb(NULL, length);
2177}
2178
2179
4915a0de
ED
2180static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2181 unsigned int length, gfp_t gfp)
61321bbd 2182{
4915a0de 2183 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2184
2185 if (NET_IP_ALIGN && skb)
2186 skb_reserve(skb, NET_IP_ALIGN);
2187 return skb;
2188}
2189
4915a0de
ED
2190static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2191 unsigned int length)
2192{
2193 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2194}
2195
181edb2b
AD
2196static inline void skb_free_frag(void *addr)
2197{
2198 __free_page_frag(addr);
2199}
2200
ffde7328 2201void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2202struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2203 unsigned int length, gfp_t gfp_mask);
2204static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2205 unsigned int length)
2206{
2207 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2208}
ffde7328 2209
71dfda58
AD
2210/**
2211 * __dev_alloc_pages - allocate page for network Rx
2212 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2213 * @order: size of the allocation
2214 *
2215 * Allocate a new page.
2216 *
2217 * %NULL is returned if there is no free memory.
2218*/
2219static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2220 unsigned int order)
2221{
2222 /* This piece of code contains several assumptions.
2223 * 1. This is for device Rx, therefor a cold page is preferred.
2224 * 2. The expectation is the user wants a compound page.
2225 * 3. If requesting a order 0 page it will not be compound
2226 * due to the check to see if order has a value in prep_new_page
2227 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2228 * code in gfp_to_alloc_flags that should be enforcing this.
2229 */
2230 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2231
2232 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2233}
2234
2235static inline struct page *dev_alloc_pages(unsigned int order)
2236{
2237 return __dev_alloc_pages(GFP_ATOMIC, order);
2238}
2239
2240/**
2241 * __dev_alloc_page - allocate a page for network Rx
2242 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2243 *
2244 * Allocate a new page.
2245 *
2246 * %NULL is returned if there is no free memory.
2247 */
2248static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2249{
2250 return __dev_alloc_pages(gfp_mask, 0);
2251}
2252
2253static inline struct page *dev_alloc_page(void)
2254{
2255 return __dev_alloc_page(GFP_ATOMIC);
2256}
2257
0614002b
MG
2258/**
2259 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2260 * @page: The page that was allocated from skb_alloc_page
2261 * @skb: The skb that may need pfmemalloc set
2262 */
2263static inline void skb_propagate_pfmemalloc(struct page *page,
2264 struct sk_buff *skb)
2265{
2266 if (page && page->pfmemalloc)
2267 skb->pfmemalloc = true;
2268}
2269
131ea667 2270/**
e227867f 2271 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2272 * @frag: the paged fragment
2273 *
2274 * Returns the &struct page associated with @frag.
2275 */
2276static inline struct page *skb_frag_page(const skb_frag_t *frag)
2277{
a8605c60 2278 return frag->page.p;
131ea667
IC
2279}
2280
2281/**
2282 * __skb_frag_ref - take an addition reference on a paged fragment.
2283 * @frag: the paged fragment
2284 *
2285 * Takes an additional reference on the paged fragment @frag.
2286 */
2287static inline void __skb_frag_ref(skb_frag_t *frag)
2288{
2289 get_page(skb_frag_page(frag));
2290}
2291
2292/**
2293 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2294 * @skb: the buffer
2295 * @f: the fragment offset.
2296 *
2297 * Takes an additional reference on the @f'th paged fragment of @skb.
2298 */
2299static inline void skb_frag_ref(struct sk_buff *skb, int f)
2300{
2301 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2302}
2303
2304/**
2305 * __skb_frag_unref - release a reference on a paged fragment.
2306 * @frag: the paged fragment
2307 *
2308 * Releases a reference on the paged fragment @frag.
2309 */
2310static inline void __skb_frag_unref(skb_frag_t *frag)
2311{
2312 put_page(skb_frag_page(frag));
2313}
2314
2315/**
2316 * skb_frag_unref - release a reference on a paged fragment of an skb.
2317 * @skb: the buffer
2318 * @f: the fragment offset
2319 *
2320 * Releases a reference on the @f'th paged fragment of @skb.
2321 */
2322static inline void skb_frag_unref(struct sk_buff *skb, int f)
2323{
2324 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2325}
2326
2327/**
2328 * skb_frag_address - gets the address of the data contained in a paged fragment
2329 * @frag: the paged fragment buffer
2330 *
2331 * Returns the address of the data within @frag. The page must already
2332 * be mapped.
2333 */
2334static inline void *skb_frag_address(const skb_frag_t *frag)
2335{
2336 return page_address(skb_frag_page(frag)) + frag->page_offset;
2337}
2338
2339/**
2340 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2341 * @frag: the paged fragment buffer
2342 *
2343 * Returns the address of the data within @frag. Checks that the page
2344 * is mapped and returns %NULL otherwise.
2345 */
2346static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2347{
2348 void *ptr = page_address(skb_frag_page(frag));
2349 if (unlikely(!ptr))
2350 return NULL;
2351
2352 return ptr + frag->page_offset;
2353}
2354
2355/**
2356 * __skb_frag_set_page - sets the page contained in a paged fragment
2357 * @frag: the paged fragment
2358 * @page: the page to set
2359 *
2360 * Sets the fragment @frag to contain @page.
2361 */
2362static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2363{
a8605c60 2364 frag->page.p = page;
131ea667
IC
2365}
2366
2367/**
2368 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2369 * @skb: the buffer
2370 * @f: the fragment offset
2371 * @page: the page to set
2372 *
2373 * Sets the @f'th fragment of @skb to contain @page.
2374 */
2375static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2376 struct page *page)
2377{
2378 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2379}
2380
400dfd3a
ED
2381bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2382
131ea667
IC
2383/**
2384 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2385 * @dev: the device to map the fragment to
131ea667
IC
2386 * @frag: the paged fragment to map
2387 * @offset: the offset within the fragment (starting at the
2388 * fragment's own offset)
2389 * @size: the number of bytes to map
f83347df 2390 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2391 *
2392 * Maps the page associated with @frag to @device.
2393 */
2394static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2395 const skb_frag_t *frag,
2396 size_t offset, size_t size,
2397 enum dma_data_direction dir)
2398{
2399 return dma_map_page(dev, skb_frag_page(frag),
2400 frag->page_offset + offset, size, dir);
2401}
2402
117632e6
ED
2403static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2404 gfp_t gfp_mask)
2405{
2406 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2407}
2408
bad93e9d
OP
2409
2410static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2411 gfp_t gfp_mask)
2412{
2413 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2414}
2415
2416
334a8132
PM
2417/**
2418 * skb_clone_writable - is the header of a clone writable
2419 * @skb: buffer to check
2420 * @len: length up to which to write
2421 *
2422 * Returns true if modifying the header part of the cloned buffer
2423 * does not requires the data to be copied.
2424 */
05bdd2f1 2425static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2426{
2427 return !skb_header_cloned(skb) &&
2428 skb_headroom(skb) + len <= skb->hdr_len;
2429}
2430
d9cc2048
HX
2431static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2432 int cloned)
2433{
2434 int delta = 0;
2435
d9cc2048
HX
2436 if (headroom > skb_headroom(skb))
2437 delta = headroom - skb_headroom(skb);
2438
2439 if (delta || cloned)
2440 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2441 GFP_ATOMIC);
2442 return 0;
2443}
2444
1da177e4
LT
2445/**
2446 * skb_cow - copy header of skb when it is required
2447 * @skb: buffer to cow
2448 * @headroom: needed headroom
2449 *
2450 * If the skb passed lacks sufficient headroom or its data part
2451 * is shared, data is reallocated. If reallocation fails, an error
2452 * is returned and original skb is not changed.
2453 *
2454 * The result is skb with writable area skb->head...skb->tail
2455 * and at least @headroom of space at head.
2456 */
2457static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2458{
d9cc2048
HX
2459 return __skb_cow(skb, headroom, skb_cloned(skb));
2460}
1da177e4 2461
d9cc2048
HX
2462/**
2463 * skb_cow_head - skb_cow but only making the head writable
2464 * @skb: buffer to cow
2465 * @headroom: needed headroom
2466 *
2467 * This function is identical to skb_cow except that we replace the
2468 * skb_cloned check by skb_header_cloned. It should be used when
2469 * you only need to push on some header and do not need to modify
2470 * the data.
2471 */
2472static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2473{
2474 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2475}
2476
2477/**
2478 * skb_padto - pad an skbuff up to a minimal size
2479 * @skb: buffer to pad
2480 * @len: minimal length
2481 *
2482 * Pads up a buffer to ensure the trailing bytes exist and are
2483 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2484 * is untouched. Otherwise it is extended. Returns zero on
2485 * success. The skb is freed on error.
1da177e4 2486 */
5b057c6b 2487static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2488{
2489 unsigned int size = skb->len;
2490 if (likely(size >= len))
5b057c6b 2491 return 0;
987c402a 2492 return skb_pad(skb, len - size);
1da177e4
LT
2493}
2494
9c0c1124
AD
2495/**
2496 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2497 * @skb: buffer to pad
2498 * @len: minimal length
2499 *
2500 * Pads up a buffer to ensure the trailing bytes exist and are
2501 * blanked. If the buffer already contains sufficient data it
2502 * is untouched. Otherwise it is extended. Returns zero on
2503 * success. The skb is freed on error.
2504 */
2505static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2506{
2507 unsigned int size = skb->len;
2508
2509 if (unlikely(size < len)) {
2510 len -= size;
2511 if (skb_pad(skb, len))
2512 return -ENOMEM;
2513 __skb_put(skb, len);
2514 }
2515 return 0;
2516}
2517
1da177e4 2518static inline int skb_add_data(struct sk_buff *skb,
af2b040e 2519 struct iov_iter *from, int copy)
1da177e4
LT
2520{
2521 const int off = skb->len;
2522
2523 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e
AV
2524 __wsum csum = 0;
2525 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2526 &csum, from) == copy) {
1da177e4
LT
2527 skb->csum = csum_block_add(skb->csum, csum, off);
2528 return 0;
2529 }
af2b040e 2530 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
1da177e4
LT
2531 return 0;
2532
2533 __skb_trim(skb, off);
2534 return -EFAULT;
2535}
2536
38ba0a65
ED
2537static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2538 const struct page *page, int off)
1da177e4
LT
2539{
2540 if (i) {
9e903e08 2541 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2542
ea2ab693 2543 return page == skb_frag_page(frag) &&
9e903e08 2544 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2545 }
38ba0a65 2546 return false;
1da177e4
LT
2547}
2548
364c6bad
HX
2549static inline int __skb_linearize(struct sk_buff *skb)
2550{
2551 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2552}
2553
1da177e4
LT
2554/**
2555 * skb_linearize - convert paged skb to linear one
2556 * @skb: buffer to linarize
1da177e4
LT
2557 *
2558 * If there is no free memory -ENOMEM is returned, otherwise zero
2559 * is returned and the old skb data released.
2560 */
364c6bad
HX
2561static inline int skb_linearize(struct sk_buff *skb)
2562{
2563 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2564}
2565
cef401de
ED
2566/**
2567 * skb_has_shared_frag - can any frag be overwritten
2568 * @skb: buffer to test
2569 *
2570 * Return true if the skb has at least one frag that might be modified
2571 * by an external entity (as in vmsplice()/sendfile())
2572 */
2573static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2574{
c9af6db4
PS
2575 return skb_is_nonlinear(skb) &&
2576 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2577}
2578
364c6bad
HX
2579/**
2580 * skb_linearize_cow - make sure skb is linear and writable
2581 * @skb: buffer to process
2582 *
2583 * If there is no free memory -ENOMEM is returned, otherwise zero
2584 * is returned and the old skb data released.
2585 */
2586static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2587{
364c6bad
HX
2588 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2589 __skb_linearize(skb) : 0;
1da177e4
LT
2590}
2591
2592/**
2593 * skb_postpull_rcsum - update checksum for received skb after pull
2594 * @skb: buffer to update
2595 * @start: start of data before pull
2596 * @len: length of data pulled
2597 *
2598 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2599 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2600 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2601 */
2602
2603static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2604 const void *start, unsigned int len)
1da177e4 2605{
84fa7933 2606 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2607 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2608}
2609
cbb042f9
HX
2610unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2611
7ce5a27f
DM
2612/**
2613 * pskb_trim_rcsum - trim received skb and update checksum
2614 * @skb: buffer to trim
2615 * @len: new length
2616 *
2617 * This is exactly the same as pskb_trim except that it ensures the
2618 * checksum of received packets are still valid after the operation.
2619 */
2620
2621static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2622{
2623 if (likely(len >= skb->len))
2624 return 0;
2625 if (skb->ip_summed == CHECKSUM_COMPLETE)
2626 skb->ip_summed = CHECKSUM_NONE;
2627 return __pskb_trim(skb, len);
2628}
2629
1da177e4
LT
2630#define skb_queue_walk(queue, skb) \
2631 for (skb = (queue)->next; \
a1e4891f 2632 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2633 skb = skb->next)
2634
46f8914e
JC
2635#define skb_queue_walk_safe(queue, skb, tmp) \
2636 for (skb = (queue)->next, tmp = skb->next; \
2637 skb != (struct sk_buff *)(queue); \
2638 skb = tmp, tmp = skb->next)
2639
1164f52a 2640#define skb_queue_walk_from(queue, skb) \
a1e4891f 2641 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2642 skb = skb->next)
2643
2644#define skb_queue_walk_from_safe(queue, skb, tmp) \
2645 for (tmp = skb->next; \
2646 skb != (struct sk_buff *)(queue); \
2647 skb = tmp, tmp = skb->next)
2648
300ce174
SH
2649#define skb_queue_reverse_walk(queue, skb) \
2650 for (skb = (queue)->prev; \
a1e4891f 2651 skb != (struct sk_buff *)(queue); \
300ce174
SH
2652 skb = skb->prev)
2653
686a2955
DM
2654#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2655 for (skb = (queue)->prev, tmp = skb->prev; \
2656 skb != (struct sk_buff *)(queue); \
2657 skb = tmp, tmp = skb->prev)
2658
2659#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2660 for (tmp = skb->prev; \
2661 skb != (struct sk_buff *)(queue); \
2662 skb = tmp, tmp = skb->prev)
1da177e4 2663
21dc3301 2664static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2665{
2666 return skb_shinfo(skb)->frag_list != NULL;
2667}
2668
2669static inline void skb_frag_list_init(struct sk_buff *skb)
2670{
2671 skb_shinfo(skb)->frag_list = NULL;
2672}
2673
2674static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2675{
2676 frag->next = skb_shinfo(skb)->frag_list;
2677 skb_shinfo(skb)->frag_list = frag;
2678}
2679
2680#define skb_walk_frags(skb, iter) \
2681 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2682
7965bd4d
JP
2683struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2684 int *peeked, int *off, int *err);
2685struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2686 int *err);
2687unsigned int datagram_poll(struct file *file, struct socket *sock,
2688 struct poll_table_struct *wait);
c0371da6
AV
2689int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2690 struct iov_iter *to, int size);
51f3d02b
DM
2691static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2692 struct msghdr *msg, int size)
2693{
e5a4b0bb 2694 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 2695}
e5a4b0bb
AV
2696int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2697 struct msghdr *msg);
3a654f97
AV
2698int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2699 struct iov_iter *from, int len);
3a654f97 2700int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d
JP
2701void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2702void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2703int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2704int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2705int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2706__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2707 int len, __wsum csum);
a60e3cc7
HFS
2708ssize_t skb_socket_splice(struct sock *sk,
2709 struct pipe_inode_info *pipe,
2710 struct splice_pipe_desc *spd);
2711int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 2712 struct pipe_inode_info *pipe, unsigned int len,
a60e3cc7
HFS
2713 unsigned int flags,
2714 ssize_t (*splice_cb)(struct sock *,
2715 struct pipe_inode_info *,
2716 struct splice_pipe_desc *));
7965bd4d 2717void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2718unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2719int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2720 int len, int hlen);
7965bd4d
JP
2721void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2722int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2723void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2724unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2725struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2726struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 2727int skb_ensure_writable(struct sk_buff *skb, int write_len);
93515d53
JP
2728int skb_vlan_pop(struct sk_buff *skb);
2729int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
20380731 2730
6ce8e9ce
AV
2731static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2732{
21226abb 2733 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
6ce8e9ce
AV
2734}
2735
7eab8d9e
AV
2736static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2737{
e5a4b0bb 2738 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
2739}
2740
2817a336
DB
2741struct skb_checksum_ops {
2742 __wsum (*update)(const void *mem, int len, __wsum wsum);
2743 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2744};
2745
2746__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2747 __wsum csum, const struct skb_checksum_ops *ops);
2748__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2749 __wsum csum);
2750
1e98a0f0
ED
2751static inline void * __must_check
2752__skb_header_pointer(const struct sk_buff *skb, int offset,
2753 int len, void *data, int hlen, void *buffer)
1da177e4 2754{
55820ee2 2755 if (hlen - offset >= len)
690e36e7 2756 return data + offset;
1da177e4 2757
690e36e7
DM
2758 if (!skb ||
2759 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2760 return NULL;
2761
2762 return buffer;
2763}
2764
1e98a0f0
ED
2765static inline void * __must_check
2766skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
2767{
2768 return __skb_header_pointer(skb, offset, len, skb->data,
2769 skb_headlen(skb), buffer);
2770}
2771
4262e5cc
DB
2772/**
2773 * skb_needs_linearize - check if we need to linearize a given skb
2774 * depending on the given device features.
2775 * @skb: socket buffer to check
2776 * @features: net device features
2777 *
2778 * Returns true if either:
2779 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2780 * 2. skb is fragmented and the device does not support SG.
2781 */
2782static inline bool skb_needs_linearize(struct sk_buff *skb,
2783 netdev_features_t features)
2784{
2785 return skb_is_nonlinear(skb) &&
2786 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2787 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2788}
2789
d626f62b
ACM
2790static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2791 void *to,
2792 const unsigned int len)
2793{
2794 memcpy(to, skb->data, len);
2795}
2796
2797static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2798 const int offset, void *to,
2799 const unsigned int len)
2800{
2801 memcpy(to, skb->data + offset, len);
2802}
2803
27d7ff46
ACM
2804static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2805 const void *from,
2806 const unsigned int len)
2807{
2808 memcpy(skb->data, from, len);
2809}
2810
2811static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2812 const int offset,
2813 const void *from,
2814 const unsigned int len)
2815{
2816 memcpy(skb->data + offset, from, len);
2817}
2818
7965bd4d 2819void skb_init(void);
1da177e4 2820
ac45f602
PO
2821static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2822{
2823 return skb->tstamp;
2824}
2825
a61bbcf2
PM
2826/**
2827 * skb_get_timestamp - get timestamp from a skb
2828 * @skb: skb to get stamp from
2829 * @stamp: pointer to struct timeval to store stamp in
2830 *
2831 * Timestamps are stored in the skb as offsets to a base timestamp.
2832 * This function converts the offset back to a struct timeval and stores
2833 * it in stamp.
2834 */
ac45f602
PO
2835static inline void skb_get_timestamp(const struct sk_buff *skb,
2836 struct timeval *stamp)
a61bbcf2 2837{
b7aa0bf7 2838 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2839}
2840
ac45f602
PO
2841static inline void skb_get_timestampns(const struct sk_buff *skb,
2842 struct timespec *stamp)
2843{
2844 *stamp = ktime_to_timespec(skb->tstamp);
2845}
2846
b7aa0bf7 2847static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2848{
b7aa0bf7 2849 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2850}
2851
164891aa
SH
2852static inline ktime_t net_timedelta(ktime_t t)
2853{
2854 return ktime_sub(ktime_get_real(), t);
2855}
2856
b9ce204f
IJ
2857static inline ktime_t net_invalid_timestamp(void)
2858{
2859 return ktime_set(0, 0);
2860}
a61bbcf2 2861
62bccb8c
AD
2862struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2863
c1f19b51
RC
2864#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2865
7965bd4d
JP
2866void skb_clone_tx_timestamp(struct sk_buff *skb);
2867bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2868
2869#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2870
2871static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2872{
2873}
2874
2875static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2876{
2877 return false;
2878}
2879
2880#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2881
2882/**
2883 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2884 *
da92b194
RC
2885 * PHY drivers may accept clones of transmitted packets for
2886 * timestamping via their phy_driver.txtstamp method. These drivers
2887 * must call this function to return the skb back to the stack, with
2888 * or without a timestamp.
2889 *
c1f19b51 2890 * @skb: clone of the the original outgoing packet
da92b194 2891 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2892 *
2893 */
2894void skb_complete_tx_timestamp(struct sk_buff *skb,
2895 struct skb_shared_hwtstamps *hwtstamps);
2896
e7fd2885
WB
2897void __skb_tstamp_tx(struct sk_buff *orig_skb,
2898 struct skb_shared_hwtstamps *hwtstamps,
2899 struct sock *sk, int tstype);
2900
ac45f602
PO
2901/**
2902 * skb_tstamp_tx - queue clone of skb with send time stamps
2903 * @orig_skb: the original outgoing packet
2904 * @hwtstamps: hardware time stamps, may be NULL if not available
2905 *
2906 * If the skb has a socket associated, then this function clones the
2907 * skb (thus sharing the actual data and optional structures), stores
2908 * the optional hardware time stamping information (if non NULL) or
2909 * generates a software time stamp (otherwise), then queues the clone
2910 * to the error queue of the socket. Errors are silently ignored.
2911 */
7965bd4d
JP
2912void skb_tstamp_tx(struct sk_buff *orig_skb,
2913 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2914
4507a715
RC
2915static inline void sw_tx_timestamp(struct sk_buff *skb)
2916{
2244d07b
OH
2917 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2918 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2919 skb_tstamp_tx(skb, NULL);
2920}
2921
2922/**
2923 * skb_tx_timestamp() - Driver hook for transmit timestamping
2924 *
2925 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2926 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2927 *
73409f3b
DM
2928 * Specifically, one should make absolutely sure that this function is
2929 * called before TX completion of this packet can trigger. Otherwise
2930 * the packet could potentially already be freed.
2931 *
4507a715
RC
2932 * @skb: A socket buffer.
2933 */
2934static inline void skb_tx_timestamp(struct sk_buff *skb)
2935{
c1f19b51 2936 skb_clone_tx_timestamp(skb);
4507a715
RC
2937 sw_tx_timestamp(skb);
2938}
2939
6e3e939f
JB
2940/**
2941 * skb_complete_wifi_ack - deliver skb with wifi status
2942 *
2943 * @skb: the original outgoing packet
2944 * @acked: ack status
2945 *
2946 */
2947void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2948
7965bd4d
JP
2949__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2950__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2951
60476372
HX
2952static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2953{
6edec0e6
TH
2954 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
2955 skb->csum_valid ||
2956 (skb->ip_summed == CHECKSUM_PARTIAL &&
2957 skb_checksum_start_offset(skb) >= 0));
60476372
HX
2958}
2959
fb286bb2
HX
2960/**
2961 * skb_checksum_complete - Calculate checksum of an entire packet
2962 * @skb: packet to process
2963 *
2964 * This function calculates the checksum over the entire packet plus
2965 * the value of skb->csum. The latter can be used to supply the
2966 * checksum of a pseudo header as used by TCP/UDP. It returns the
2967 * checksum.
2968 *
2969 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2970 * this function can be used to verify that checksum on received
2971 * packets. In that case the function should return zero if the
2972 * checksum is correct. In particular, this function will return zero
2973 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2974 * hardware has already verified the correctness of the checksum.
2975 */
4381ca3c 2976static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2977{
60476372
HX
2978 return skb_csum_unnecessary(skb) ?
2979 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2980}
2981
77cffe23
TH
2982static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2983{
2984 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2985 if (skb->csum_level == 0)
2986 skb->ip_summed = CHECKSUM_NONE;
2987 else
2988 skb->csum_level--;
2989 }
2990}
2991
2992static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2993{
2994 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2995 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2996 skb->csum_level++;
2997 } else if (skb->ip_summed == CHECKSUM_NONE) {
2998 skb->ip_summed = CHECKSUM_UNNECESSARY;
2999 skb->csum_level = 0;
3000 }
3001}
3002
5a212329
TH
3003static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3004{
3005 /* Mark current checksum as bad (typically called from GRO
3006 * path). In the case that ip_summed is CHECKSUM_NONE
3007 * this must be the first checksum encountered in the packet.
3008 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3009 * checksum after the last one validated. For UDP, a zero
3010 * checksum can not be marked as bad.
3011 */
3012
3013 if (skb->ip_summed == CHECKSUM_NONE ||
3014 skb->ip_summed == CHECKSUM_UNNECESSARY)
3015 skb->csum_bad = 1;
3016}
3017
76ba0aae
TH
3018/* Check if we need to perform checksum complete validation.
3019 *
3020 * Returns true if checksum complete is needed, false otherwise
3021 * (either checksum is unnecessary or zero checksum is allowed).
3022 */
3023static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3024 bool zero_okay,
3025 __sum16 check)
3026{
5d0c2b95
TH
3027 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3028 skb->csum_valid = 1;
77cffe23 3029 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3030 return false;
3031 }
3032
3033 return true;
3034}
3035
3036/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3037 * in checksum_init.
3038 */
3039#define CHECKSUM_BREAK 76
3040
4e18b9ad
TH
3041/* Unset checksum-complete
3042 *
3043 * Unset checksum complete can be done when packet is being modified
3044 * (uncompressed for instance) and checksum-complete value is
3045 * invalidated.
3046 */
3047static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3048{
3049 if (skb->ip_summed == CHECKSUM_COMPLETE)
3050 skb->ip_summed = CHECKSUM_NONE;
3051}
3052
76ba0aae
TH
3053/* Validate (init) checksum based on checksum complete.
3054 *
3055 * Return values:
3056 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3057 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3058 * checksum is stored in skb->csum for use in __skb_checksum_complete
3059 * non-zero: value of invalid checksum
3060 *
3061 */
3062static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3063 bool complete,
3064 __wsum psum)
3065{
3066 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3067 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3068 skb->csum_valid = 1;
76ba0aae
TH
3069 return 0;
3070 }
5a212329
TH
3071 } else if (skb->csum_bad) {
3072 /* ip_summed == CHECKSUM_NONE in this case */
c91d4606 3073 return (__force __sum16)1;
76ba0aae
TH
3074 }
3075
3076 skb->csum = psum;
3077
5d0c2b95
TH
3078 if (complete || skb->len <= CHECKSUM_BREAK) {
3079 __sum16 csum;
3080
3081 csum = __skb_checksum_complete(skb);
3082 skb->csum_valid = !csum;
3083 return csum;
3084 }
76ba0aae
TH
3085
3086 return 0;
3087}
3088
3089static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3090{
3091 return 0;
3092}
3093
3094/* Perform checksum validate (init). Note that this is a macro since we only
3095 * want to calculate the pseudo header which is an input function if necessary.
3096 * First we try to validate without any computation (checksum unnecessary) and
3097 * then calculate based on checksum complete calling the function to compute
3098 * pseudo header.
3099 *
3100 * Return values:
3101 * 0: checksum is validated or try to in skb_checksum_complete
3102 * non-zero: value of invalid checksum
3103 */
3104#define __skb_checksum_validate(skb, proto, complete, \
3105 zero_okay, check, compute_pseudo) \
3106({ \
3107 __sum16 __ret = 0; \
5d0c2b95 3108 skb->csum_valid = 0; \
76ba0aae
TH
3109 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3110 __ret = __skb_checksum_validate_complete(skb, \
3111 complete, compute_pseudo(skb, proto)); \
3112 __ret; \
3113})
3114
3115#define skb_checksum_init(skb, proto, compute_pseudo) \
3116 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3117
3118#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3119 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3120
3121#define skb_checksum_validate(skb, proto, compute_pseudo) \
3122 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3123
3124#define skb_checksum_validate_zero_check(skb, proto, check, \
3125 compute_pseudo) \
096a4cfa 3126 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3127
3128#define skb_checksum_simple_validate(skb) \
3129 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3130
d96535a1
TH
3131static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3132{
3133 return (skb->ip_summed == CHECKSUM_NONE &&
3134 skb->csum_valid && !skb->csum_bad);
3135}
3136
3137static inline void __skb_checksum_convert(struct sk_buff *skb,
3138 __sum16 check, __wsum pseudo)
3139{
3140 skb->csum = ~pseudo;
3141 skb->ip_summed = CHECKSUM_COMPLETE;
3142}
3143
3144#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3145do { \
3146 if (__skb_checksum_convert_check(skb)) \
3147 __skb_checksum_convert(skb, check, \
3148 compute_pseudo(skb, proto)); \
3149} while (0)
3150
15e2396d
TH
3151static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3152 u16 start, u16 offset)
3153{
3154 skb->ip_summed = CHECKSUM_PARTIAL;
3155 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3156 skb->csum_offset = offset - start;
3157}
3158
dcdc8994
TH
3159/* Update skbuf and packet to reflect the remote checksum offload operation.
3160 * When called, ptr indicates the starting point for skb->csum when
3161 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3162 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3163 */
3164static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3165 int start, int offset, bool nopartial)
dcdc8994
TH
3166{
3167 __wsum delta;
3168
15e2396d
TH
3169 if (!nopartial) {
3170 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3171 return;
3172 }
3173
dcdc8994
TH
3174 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3175 __skb_checksum_complete(skb);
3176 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3177 }
3178
3179 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3180
3181 /* Adjust skb->csum since we changed the packet */
3182 skb->csum = csum_add(skb->csum, delta);
3183}
3184
5f79e0f9 3185#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3186void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3187static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3188{
3189 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3190 nf_conntrack_destroy(nfct);
1da177e4
LT
3191}
3192static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3193{
3194 if (nfct)
3195 atomic_inc(&nfct->use);
3196}
2fc72c7b 3197#endif
34666d46 3198#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3199static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3200{
3201 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3202 kfree(nf_bridge);
3203}
3204static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3205{
3206 if (nf_bridge)
3207 atomic_inc(&nf_bridge->use);
3208}
3209#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3210static inline void nf_reset(struct sk_buff *skb)
3211{
5f79e0f9 3212#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3213 nf_conntrack_put(skb->nfct);
3214 skb->nfct = NULL;
2fc72c7b 3215#endif
34666d46 3216#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3217 nf_bridge_put(skb->nf_bridge);
3218 skb->nf_bridge = NULL;
3219#endif
3220}
3221
124dff01
PM
3222static inline void nf_reset_trace(struct sk_buff *skb)
3223{
478b360a 3224#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3225 skb->nf_trace = 0;
3226#endif
a193a4ab
PM
3227}
3228
edda553c 3229/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3230static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3231 bool copy)
edda553c 3232{
5f79e0f9 3233#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3234 dst->nfct = src->nfct;
3235 nf_conntrack_get(src->nfct);
b1937227
ED
3236 if (copy)
3237 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3238#endif
34666d46 3239#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3240 dst->nf_bridge = src->nf_bridge;
3241 nf_bridge_get(src->nf_bridge);
3242#endif
478b360a 3243#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3244 if (copy)
3245 dst->nf_trace = src->nf_trace;
478b360a 3246#endif
edda553c
YK
3247}
3248
e7ac05f3
YK
3249static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3250{
e7ac05f3 3251#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3252 nf_conntrack_put(dst->nfct);
2fc72c7b 3253#endif
34666d46 3254#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3255 nf_bridge_put(dst->nf_bridge);
3256#endif
b1937227 3257 __nf_copy(dst, src, true);
e7ac05f3
YK
3258}
3259
984bc16c
JM
3260#ifdef CONFIG_NETWORK_SECMARK
3261static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3262{
3263 to->secmark = from->secmark;
3264}
3265
3266static inline void skb_init_secmark(struct sk_buff *skb)
3267{
3268 skb->secmark = 0;
3269}
3270#else
3271static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3272{ }
3273
3274static inline void skb_init_secmark(struct sk_buff *skb)
3275{ }
3276#endif
3277
574f7194
EB
3278static inline bool skb_irq_freeable(const struct sk_buff *skb)
3279{
3280 return !skb->destructor &&
3281#if IS_ENABLED(CONFIG_XFRM)
3282 !skb->sp &&
3283#endif
3284#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3285 !skb->nfct &&
3286#endif
3287 !skb->_skb_refdst &&
3288 !skb_has_frag_list(skb);
3289}
3290
f25f4e44
PWJ
3291static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3292{
f25f4e44 3293 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3294}
3295
9247744e 3296static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3297{
4e3ab47a 3298 return skb->queue_mapping;
4e3ab47a
PE
3299}
3300
f25f4e44
PWJ
3301static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3302{
f25f4e44 3303 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3304}
3305
d5a9e24a
DM
3306static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3307{
3308 skb->queue_mapping = rx_queue + 1;
3309}
3310
9247744e 3311static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3312{
3313 return skb->queue_mapping - 1;
3314}
3315
9247744e 3316static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3317{
a02cec21 3318 return skb->queue_mapping != 0;
d5a9e24a
DM
3319}
3320
def8b4fa
AD
3321static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3322{
0b3d8e08 3323#ifdef CONFIG_XFRM
def8b4fa 3324 return skb->sp;
def8b4fa 3325#else
def8b4fa 3326 return NULL;
def8b4fa 3327#endif
0b3d8e08 3328}
def8b4fa 3329
68c33163
PS
3330/* Keeps track of mac header offset relative to skb->head.
3331 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3332 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3333 * tunnel skb it points to outer mac header.
3334 * Keeps track of level of encapsulation of network headers.
3335 */
68c33163 3336struct skb_gso_cb {
3347c960
ED
3337 int mac_offset;
3338 int encap_level;
7e2b10c1 3339 __u16 csum_start;
68c33163
PS
3340};
3341#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3342
3343static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3344{
3345 return (skb_mac_header(inner_skb) - inner_skb->head) -
3346 SKB_GSO_CB(inner_skb)->mac_offset;
3347}
3348
1e2bd517
PS
3349static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3350{
3351 int new_headroom, headroom;
3352 int ret;
3353
3354 headroom = skb_headroom(skb);
3355 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3356 if (ret)
3357 return ret;
3358
3359 new_headroom = skb_headroom(skb);
3360 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3361 return 0;
3362}
3363
7e2b10c1
TH
3364/* Compute the checksum for a gso segment. First compute the checksum value
3365 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3366 * then add in skb->csum (checksum from csum_start to end of packet).
3367 * skb->csum and csum_start are then updated to reflect the checksum of the
3368 * resultant packet starting from the transport header-- the resultant checksum
3369 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3370 * header.
3371 */
3372static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3373{
3374 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
c91d4606
ED
3375 skb_transport_offset(skb);
3376 __wsum partial;
7e2b10c1 3377
c91d4606 3378 partial = csum_partial(skb_transport_header(skb), plen, skb->csum);
7e2b10c1
TH
3379 skb->csum = res;
3380 SKB_GSO_CB(skb)->csum_start -= plen;
3381
c91d4606 3382 return csum_fold(partial);
7e2b10c1
TH
3383}
3384
bdcc0924 3385static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3386{
3387 return skb_shinfo(skb)->gso_size;
3388}
3389
36a8f39e 3390/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3391static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3392{
3393 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3394}
3395
7965bd4d 3396void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3397
3398static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3399{
3400 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3401 * wanted then gso_type will be set. */
05bdd2f1
ED
3402 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3403
b78462eb
AD
3404 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3405 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3406 __skb_warn_lro_forwarding(skb);
3407 return true;
3408 }
3409 return false;
3410}
3411
35fc92a9
HX
3412static inline void skb_forward_csum(struct sk_buff *skb)
3413{
3414 /* Unfortunately we don't support this one. Any brave souls? */
3415 if (skb->ip_summed == CHECKSUM_COMPLETE)
3416 skb->ip_summed = CHECKSUM_NONE;
3417}
3418
bc8acf2c
ED
3419/**
3420 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3421 * @skb: skb to check
3422 *
3423 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3424 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3425 * use this helper, to document places where we make this assertion.
3426 */
05bdd2f1 3427static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3428{
3429#ifdef DEBUG
3430 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3431#endif
3432}
3433
f35d9d8a 3434bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3435
ed1f50c3 3436int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
3437struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3438 unsigned int transport_len,
3439 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 3440
3a7c1ee4
AD
3441/**
3442 * skb_head_is_locked - Determine if the skb->head is locked down
3443 * @skb: skb to check
3444 *
3445 * The head on skbs build around a head frag can be removed if they are
3446 * not cloned. This function returns true if the skb head is locked down
3447 * due to either being allocated via kmalloc, or by being a clone with
3448 * multiple references to the head.
3449 */
3450static inline bool skb_head_is_locked(const struct sk_buff *skb)
3451{
3452 return !skb->head_frag || skb_cloned(skb);
3453}
fe6cc55f
FW
3454
3455/**
3456 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3457 *
3458 * @skb: GSO skb
3459 *
3460 * skb_gso_network_seglen is used to determine the real size of the
3461 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3462 *
3463 * The MAC/L2 header is not accounted for.
3464 */
3465static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3466{
3467 unsigned int hdr_len = skb_transport_header(skb) -
3468 skb_network_header(skb);
3469 return hdr_len + skb_gso_transport_seglen(skb);
3470}
1da177e4
LT
3471#endif /* __KERNEL__ */
3472#endif /* _LINUX_SKBUFF_H */
This page took 2.092465 seconds and 5 git commands to generate.