]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - include/linux/skbuff.h
netfilter: bridge: add and use nf_bridge_info_get helper
[mirror_ubuntu-zesty-kernel.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>
5203cd28 37#include <net/flow_keys.h>
1da177e4 38
78ea85f1
DB
39/* A. Checksumming of received packets by device.
40 *
41 * CHECKSUM_NONE:
42 *
43 * Device failed to checksum this packet e.g. due to lack of capabilities.
44 * The packet contains full (though not verified) checksum in packet but
45 * not in skb->csum. Thus, skb->csum is undefined in this case.
46 *
47 * CHECKSUM_UNNECESSARY:
48 *
49 * The hardware you're dealing with doesn't calculate the full checksum
50 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
51 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
52 * if their checksums are okay. skb->csum is still undefined in this case
53 * though. It is a bad option, but, unfortunately, nowadays most vendors do
54 * this. Apparently with the secret goal to sell you new devices, when you
55 * will add new protocol to your host, f.e. IPv6 8)
56 *
57 * CHECKSUM_UNNECESSARY is applicable to following protocols:
58 * TCP: IPv6 and IPv4.
59 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
60 * zero UDP checksum for either IPv4 or IPv6, the networking stack
61 * may perform further validation in this case.
62 * GRE: only if the checksum is present in the header.
63 * SCTP: indicates the CRC in SCTP header has been validated.
64 *
65 * skb->csum_level indicates the number of consecutive checksums found in
66 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
67 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
68 * and a device is able to verify the checksums for UDP (possibly zero),
69 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
70 * two. If the device were only able to verify the UDP checksum and not
71 * GRE, either because it doesn't support GRE checksum of because GRE
72 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
73 * not considered in this case).
78ea85f1
DB
74 *
75 * CHECKSUM_COMPLETE:
76 *
77 * This is the most generic way. The device supplied checksum of the _whole_
78 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
79 * hardware doesn't need to parse L3/L4 headers to implement this.
80 *
81 * Note: Even if device supports only some protocols, but is able to produce
82 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
83 *
84 * CHECKSUM_PARTIAL:
85 *
6edec0e6
TH
86 * A checksum is set up to be offloaded to a device as described in the
87 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 88 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
89 * on the same host, or it may be set in the input path in GRO or remote
90 * checksum offload. For the purposes of checksum verification, the checksum
91 * referred to by skb->csum_start + skb->csum_offset and any preceding
92 * checksums in the packet are considered verified. Any checksums in the
93 * packet that are after the checksum being offloaded are not considered to
94 * be verified.
78ea85f1
DB
95 *
96 * B. Checksumming on output.
97 *
98 * CHECKSUM_NONE:
99 *
100 * The skb was already checksummed by the protocol, or a checksum is not
101 * required.
102 *
103 * CHECKSUM_PARTIAL:
104 *
105 * The device is required to checksum the packet as seen by hard_start_xmit()
106 * from skb->csum_start up to the end, and to record/write the checksum at
107 * offset skb->csum_start + skb->csum_offset.
108 *
109 * The device must show its capabilities in dev->features, set up at device
110 * setup time, e.g. netdev_features.h:
111 *
112 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
113 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
114 * IPv4. Sigh. Vendors like this way for an unknown reason.
115 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
116 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
117 * NETIF_F_... - Well, you get the picture.
118 *
119 * CHECKSUM_UNNECESSARY:
120 *
121 * Normally, the device will do per protocol specific checksumming. Protocol
122 * implementations that do not want the NIC to perform the checksum
123 * calculation should use this flag in their outgoing skbs.
124 *
125 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
126 * offload. Correspondingly, the FCoE protocol driver
127 * stack should use CHECKSUM_UNNECESSARY.
128 *
129 * Any questions? No questions, good. --ANK
130 */
131
60476372 132/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
133#define CHECKSUM_NONE 0
134#define CHECKSUM_UNNECESSARY 1
135#define CHECKSUM_COMPLETE 2
136#define CHECKSUM_PARTIAL 3
1da177e4 137
77cffe23
TH
138/* Maximum value in skb->csum_level */
139#define SKB_MAX_CSUM_LEVEL 3
140
0bec8c88 141#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 142#define SKB_WITH_OVERHEAD(X) \
deea84b0 143 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
144#define SKB_MAX_ORDER(X, ORDER) \
145 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
146#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
147#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
148
87fb4b7b
ED
149/* return minimum truesize of one skb containing X bytes of data */
150#define SKB_TRUESIZE(X) ((X) + \
151 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
152 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
153
1da177e4 154struct net_device;
716ea3a7 155struct scatterlist;
9c55e01c 156struct pipe_inode_info;
a8f820aa 157struct iov_iter;
fd11a83d 158struct napi_struct;
1da177e4 159
5f79e0f9 160#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
161struct nf_conntrack {
162 atomic_t use;
1da177e4 163};
5f79e0f9 164#endif
1da177e4 165
34666d46 166#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 167struct nf_bridge_info {
bf1ac5ca
ED
168 atomic_t use;
169 unsigned int mask;
170 struct net_device *physindev;
171 struct net_device *physoutdev;
e70deecb 172 char neigh_header[8];
1da177e4
LT
173};
174#endif
175
1da177e4
LT
176struct sk_buff_head {
177 /* These two members must be first. */
178 struct sk_buff *next;
179 struct sk_buff *prev;
180
181 __u32 qlen;
182 spinlock_t lock;
183};
184
185struct sk_buff;
186
9d4dde52
IC
187/* To allow 64K frame to be packed as single skb without frag_list we
188 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
189 * buffers which do not start on a page boundary.
190 *
191 * Since GRO uses frags we allocate at least 16 regardless of page
192 * size.
a715dea3 193 */
9d4dde52 194#if (65536/PAGE_SIZE + 1) < 16
eec00954 195#define MAX_SKB_FRAGS 16UL
a715dea3 196#else
9d4dde52 197#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 198#endif
1da177e4
LT
199
200typedef struct skb_frag_struct skb_frag_t;
201
202struct skb_frag_struct {
a8605c60
IC
203 struct {
204 struct page *p;
205 } page;
cb4dfe56 206#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
207 __u32 page_offset;
208 __u32 size;
cb4dfe56
ED
209#else
210 __u16 page_offset;
211 __u16 size;
212#endif
1da177e4
LT
213};
214
9e903e08
ED
215static inline unsigned int skb_frag_size(const skb_frag_t *frag)
216{
217 return frag->size;
218}
219
220static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
221{
222 frag->size = size;
223}
224
225static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
226{
227 frag->size += delta;
228}
229
230static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
231{
232 frag->size -= delta;
233}
234
ac45f602
PO
235#define HAVE_HW_TIME_STAMP
236
237/**
d3a21be8 238 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
239 * @hwtstamp: hardware time stamp transformed into duration
240 * since arbitrary point in time
ac45f602
PO
241 *
242 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 243 * skb->tstamp.
ac45f602
PO
244 *
245 * hwtstamps can only be compared against other hwtstamps from
246 * the same device.
247 *
248 * This structure is attached to packets as part of the
249 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
250 */
251struct skb_shared_hwtstamps {
252 ktime_t hwtstamp;
ac45f602
PO
253};
254
2244d07b
OH
255/* Definitions for tx_flags in struct skb_shared_info */
256enum {
257 /* generate hardware time stamp */
258 SKBTX_HW_TSTAMP = 1 << 0,
259
e7fd2885 260 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
261 SKBTX_SW_TSTAMP = 1 << 1,
262
263 /* device driver is going to provide hardware time stamp */
264 SKBTX_IN_PROGRESS = 1 << 2,
265
a6686f2f 266 /* device driver supports TX zero-copy buffers */
62b1a8ab 267 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
268
269 /* generate wifi status information (where possible) */
62b1a8ab 270 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
271
272 /* This indicates at least one fragment might be overwritten
273 * (as in vmsplice(), sendfile() ...)
274 * If we need to compute a TX checksum, we'll need to copy
275 * all frags to avoid possible bad checksum
276 */
277 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
278
279 /* generate software time stamp when entering packet scheduling */
280 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
281
282 /* generate software timestamp on peer data acknowledgment */
283 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
284};
285
e1c8a607
WB
286#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
287 SKBTX_SCHED_TSTAMP | \
288 SKBTX_ACK_TSTAMP)
f24b9be5
WB
289#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
290
a6686f2f
SM
291/*
292 * The callback notifies userspace to release buffers when skb DMA is done in
293 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
294 * The zerocopy_success argument is true if zero copy transmit occurred,
295 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
296 * The ctx field is used to track device context.
297 * The desc field is used to track userspace buffer index.
a6686f2f
SM
298 */
299struct ubuf_info {
e19d6763 300 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 301 void *ctx;
a6686f2f 302 unsigned long desc;
ac45f602
PO
303};
304
1da177e4
LT
305/* This data is invariant across clones and lives at
306 * the end of the header data, ie. at skb->end.
307 */
308struct skb_shared_info {
9f42f126
IC
309 unsigned char nr_frags;
310 __u8 tx_flags;
7967168c
HX
311 unsigned short gso_size;
312 /* Warning: this field is not always filled in (UFO)! */
313 unsigned short gso_segs;
314 unsigned short gso_type;
1da177e4 315 struct sk_buff *frag_list;
ac45f602 316 struct skb_shared_hwtstamps hwtstamps;
09c2d251 317 u32 tskey;
9f42f126 318 __be32 ip6_frag_id;
ec7d2f2c
ED
319
320 /*
321 * Warning : all fields before dataref are cleared in __alloc_skb()
322 */
323 atomic_t dataref;
324
69e3c75f
JB
325 /* Intermediate layers must ensure that destructor_arg
326 * remains valid until skb destructor */
327 void * destructor_arg;
a6686f2f 328
fed66381
ED
329 /* must be last field, see pskb_expand_head() */
330 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
331};
332
333/* We divide dataref into two halves. The higher 16 bits hold references
334 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
335 * the entire skb->data. A clone of a headerless skb holds the length of
336 * the header in skb->hdr_len.
1da177e4
LT
337 *
338 * All users must obey the rule that the skb->data reference count must be
339 * greater than or equal to the payload reference count.
340 *
341 * Holding a reference to the payload part means that the user does not
342 * care about modifications to the header part of skb->data.
343 */
344#define SKB_DATAREF_SHIFT 16
345#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
346
d179cd12
DM
347
348enum {
c8753d55
VS
349 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
350 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
351 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
352};
353
7967168c
HX
354enum {
355 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 356 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
357
358 /* This indicates the skb is from an untrusted source. */
359 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
360
361 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
362 SKB_GSO_TCP_ECN = 1 << 3,
363
364 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
365
366 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
367
368 SKB_GSO_GRE = 1 << 6,
73136267 369
4b28252c 370 SKB_GSO_GRE_CSUM = 1 << 7,
0d89d203 371
4b28252c 372 SKB_GSO_IPIP = 1 << 8,
cb32f511 373
4b28252c 374 SKB_GSO_SIT = 1 << 9,
61c1db7f 375
4b28252c 376 SKB_GSO_UDP_TUNNEL = 1 << 10,
0f4f4ffa
TH
377
378 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c 379
59b93b41 380 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
7967168c
HX
381};
382
2e07fa9c
ACM
383#if BITS_PER_LONG > 32
384#define NET_SKBUFF_DATA_USES_OFFSET 1
385#endif
386
387#ifdef NET_SKBUFF_DATA_USES_OFFSET
388typedef unsigned int sk_buff_data_t;
389#else
390typedef unsigned char *sk_buff_data_t;
391#endif
392
363ec392
ED
393/**
394 * struct skb_mstamp - multi resolution time stamps
395 * @stamp_us: timestamp in us resolution
396 * @stamp_jiffies: timestamp in jiffies
397 */
398struct skb_mstamp {
399 union {
400 u64 v64;
401 struct {
402 u32 stamp_us;
403 u32 stamp_jiffies;
404 };
405 };
406};
407
408/**
409 * skb_mstamp_get - get current timestamp
410 * @cl: place to store timestamps
411 */
412static inline void skb_mstamp_get(struct skb_mstamp *cl)
413{
414 u64 val = local_clock();
415
416 do_div(val, NSEC_PER_USEC);
417 cl->stamp_us = (u32)val;
418 cl->stamp_jiffies = (u32)jiffies;
419}
420
421/**
422 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
423 * @t1: pointer to newest sample
424 * @t0: pointer to oldest sample
425 */
426static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
427 const struct skb_mstamp *t0)
428{
429 s32 delta_us = t1->stamp_us - t0->stamp_us;
430 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
431
432 /* If delta_us is negative, this might be because interval is too big,
433 * or local_clock() drift is too big : fallback using jiffies.
434 */
435 if (delta_us <= 0 ||
436 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
437
438 delta_us = jiffies_to_usecs(delta_jiffies);
439
440 return delta_us;
441}
442
443
1da177e4
LT
444/**
445 * struct sk_buff - socket buffer
446 * @next: Next buffer in list
447 * @prev: Previous buffer in list
363ec392 448 * @tstamp: Time we arrived/left
56b17425 449 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 450 * @sk: Socket we are owned by
1da177e4 451 * @dev: Device we arrived on/are leaving by
d84e0bd7 452 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 453 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 454 * @sp: the security path, used for xfrm
1da177e4
LT
455 * @len: Length of actual data
456 * @data_len: Data length
457 * @mac_len: Length of link layer header
334a8132 458 * @hdr_len: writable header length of cloned skb
663ead3b
HX
459 * @csum: Checksum (must include start/offset pair)
460 * @csum_start: Offset from skb->head where checksumming should start
461 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 462 * @priority: Packet queueing priority
60ff7467 463 * @ignore_df: allow local fragmentation
1da177e4 464 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 465 * @ip_summed: Driver fed us an IP checksum
1da177e4 466 * @nohdr: Payload reference only, must not modify header
d84e0bd7 467 * @nfctinfo: Relationship of this skb to the connection
1da177e4 468 * @pkt_type: Packet class
c83c2486 469 * @fclone: skbuff clone status
c83c2486 470 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
471 * @peeked: this packet has been seen already, so stats have been
472 * done for it, don't do them again
ba9dda3a 473 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
474 * @protocol: Packet protocol from driver
475 * @destructor: Destruct function
476 * @nfct: Associated connection, if any
1da177e4 477 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 478 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
479 * @tc_index: Traffic control index
480 * @tc_verd: traffic control verdict
61b905da 481 * @hash: the packet hash
d84e0bd7 482 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 483 * @xmit_more: More SKBs are pending for this queue
553a5672 484 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 485 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 486 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 487 * ports.
a3b18ddb 488 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
489 * @wifi_acked_valid: wifi_acked was set
490 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 491 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
06021292 492 * @napi_id: id of the NAPI struct this skb came from
984bc16c 493 * @secmark: security marking
d84e0bd7 494 * @mark: Generic packet mark
86a9bad3 495 * @vlan_proto: vlan encapsulation protocol
6aa895b0 496 * @vlan_tci: vlan tag control information
0d89d203 497 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
498 * @inner_transport_header: Inner transport layer header (encapsulation)
499 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 500 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
501 * @transport_header: Transport layer header
502 * @network_header: Network layer header
503 * @mac_header: Link layer header
504 * @tail: Tail pointer
505 * @end: End pointer
506 * @head: Head of buffer
507 * @data: Data head pointer
508 * @truesize: Buffer size
509 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
510 */
511
512struct sk_buff {
363ec392 513 union {
56b17425
ED
514 struct {
515 /* These two members must be first. */
516 struct sk_buff *next;
517 struct sk_buff *prev;
518
519 union {
520 ktime_t tstamp;
521 struct skb_mstamp skb_mstamp;
522 };
523 };
524 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 525 };
da3f5cf1 526 struct sock *sk;
1da177e4 527 struct net_device *dev;
1da177e4 528
1da177e4
LT
529 /*
530 * This is the control buffer. It is free to use for every
531 * layer. Please put your private variables there. If you
532 * want to keep them across layers you have to do a skb_clone()
533 * first. This is owned by whoever has the skb queued ATM.
534 */
da3f5cf1 535 char cb[48] __aligned(8);
1da177e4 536
7fee226a 537 unsigned long _skb_refdst;
b1937227 538 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
539#ifdef CONFIG_XFRM
540 struct sec_path *sp;
b1937227
ED
541#endif
542#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
543 struct nf_conntrack *nfct;
544#endif
85224844 545#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 546 struct nf_bridge_info *nf_bridge;
da3f5cf1 547#endif
1da177e4 548 unsigned int len,
334a8132
PM
549 data_len;
550 __u16 mac_len,
551 hdr_len;
b1937227
ED
552
553 /* Following fields are _not_ copied in __copy_skb_header()
554 * Note that queue_mapping is here mostly to fill a hole.
555 */
fe55f6d5 556 kmemcheck_bitfield_begin(flags1);
b1937227
ED
557 __u16 queue_mapping;
558 __u8 cloned:1,
6869c4d8 559 nohdr:1,
b84f4cc9 560 fclone:2,
a59322be 561 peeked:1,
b1937227
ED
562 head_frag:1,
563 xmit_more:1;
564 /* one bit hole */
fe55f6d5 565 kmemcheck_bitfield_end(flags1);
4031ae6e 566
b1937227
ED
567 /* fields enclosed in headers_start/headers_end are copied
568 * using a single memcpy() in __copy_skb_header()
569 */
ebcf34f3 570 /* private: */
b1937227 571 __u32 headers_start[0];
ebcf34f3 572 /* public: */
4031ae6e 573
233577a2
HFS
574/* if you move pkt_type around you also must adapt those constants */
575#ifdef __BIG_ENDIAN_BITFIELD
576#define PKT_TYPE_MAX (7 << 5)
577#else
578#define PKT_TYPE_MAX 7
1da177e4 579#endif
233577a2 580#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 581
233577a2 582 __u8 __pkt_type_offset[0];
b1937227 583 __u8 pkt_type:3;
c93bdd0e 584 __u8 pfmemalloc:1;
b1937227
ED
585 __u8 ignore_df:1;
586 __u8 nfctinfo:3;
587
588 __u8 nf_trace:1;
589 __u8 ip_summed:2;
3853b584 590 __u8 ooo_okay:1;
61b905da 591 __u8 l4_hash:1;
a3b18ddb 592 __u8 sw_hash:1;
6e3e939f
JB
593 __u8 wifi_acked_valid:1;
594 __u8 wifi_acked:1;
b1937227 595
3bdc0eba 596 __u8 no_fcs:1;
77cffe23 597 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 598 __u8 encapsulation:1;
7e2b10c1 599 __u8 encap_hdr_csum:1;
5d0c2b95 600 __u8 csum_valid:1;
7e3cead5 601 __u8 csum_complete_sw:1;
b1937227
ED
602 __u8 csum_level:2;
603 __u8 csum_bad:1;
fe55f6d5 604
b1937227
ED
605#ifdef CONFIG_IPV6_NDISC_NODETYPE
606 __u8 ndisc_nodetype:2;
607#endif
608 __u8 ipvs_property:1;
8bce6d7d 609 __u8 inner_protocol_type:1;
e585f236
TH
610 __u8 remcsum_offload:1;
611 /* 3 or 5 bit hole */
b1937227
ED
612
613#ifdef CONFIG_NET_SCHED
614 __u16 tc_index; /* traffic control index */
615#ifdef CONFIG_NET_CLS_ACT
616 __u16 tc_verd; /* traffic control verdict */
617#endif
618#endif
fe55f6d5 619
b1937227
ED
620 union {
621 __wsum csum;
622 struct {
623 __u16 csum_start;
624 __u16 csum_offset;
625 };
626 };
627 __u32 priority;
628 int skb_iif;
629 __u32 hash;
630 __be16 vlan_proto;
631 __u16 vlan_tci;
2bd82484
ED
632#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
633 union {
634 unsigned int napi_id;
635 unsigned int sender_cpu;
636 };
97fc2f08 637#endif
984bc16c
JM
638#ifdef CONFIG_NETWORK_SECMARK
639 __u32 secmark;
640#endif
3b885787
NH
641 union {
642 __u32 mark;
16fad69c 643 __u32 reserved_tailroom;
3b885787 644 };
1da177e4 645
8bce6d7d
TH
646 union {
647 __be16 inner_protocol;
648 __u8 inner_ipproto;
649 };
650
1a37e412
SH
651 __u16 inner_transport_header;
652 __u16 inner_network_header;
653 __u16 inner_mac_header;
b1937227
ED
654
655 __be16 protocol;
1a37e412
SH
656 __u16 transport_header;
657 __u16 network_header;
658 __u16 mac_header;
b1937227 659
ebcf34f3 660 /* private: */
b1937227 661 __u32 headers_end[0];
ebcf34f3 662 /* public: */
b1937227 663
1da177e4 664 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 665 sk_buff_data_t tail;
4305b541 666 sk_buff_data_t end;
1da177e4 667 unsigned char *head,
4305b541 668 *data;
27a884dc
ACM
669 unsigned int truesize;
670 atomic_t users;
1da177e4
LT
671};
672
673#ifdef __KERNEL__
674/*
675 * Handling routines are only of interest to the kernel
676 */
677#include <linux/slab.h>
678
1da177e4 679
c93bdd0e
MG
680#define SKB_ALLOC_FCLONE 0x01
681#define SKB_ALLOC_RX 0x02
fd11a83d 682#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
683
684/* Returns true if the skb was allocated from PFMEMALLOC reserves */
685static inline bool skb_pfmemalloc(const struct sk_buff *skb)
686{
687 return unlikely(skb->pfmemalloc);
688}
689
7fee226a
ED
690/*
691 * skb might have a dst pointer attached, refcounted or not.
692 * _skb_refdst low order bit is set if refcount was _not_ taken
693 */
694#define SKB_DST_NOREF 1UL
695#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
696
697/**
698 * skb_dst - returns skb dst_entry
699 * @skb: buffer
700 *
701 * Returns skb dst_entry, regardless of reference taken or not.
702 */
adf30907
ED
703static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
704{
7fee226a
ED
705 /* If refdst was not refcounted, check we still are in a
706 * rcu_read_lock section
707 */
708 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
709 !rcu_read_lock_held() &&
710 !rcu_read_lock_bh_held());
711 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
712}
713
7fee226a
ED
714/**
715 * skb_dst_set - sets skb dst
716 * @skb: buffer
717 * @dst: dst entry
718 *
719 * Sets skb dst, assuming a reference was taken on dst and should
720 * be released by skb_dst_drop()
721 */
adf30907
ED
722static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
723{
7fee226a
ED
724 skb->_skb_refdst = (unsigned long)dst;
725}
726
932bc4d7
JA
727/**
728 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
729 * @skb: buffer
730 * @dst: dst entry
731 *
732 * Sets skb dst, assuming a reference was not taken on dst.
733 * If dst entry is cached, we do not take reference and dst_release
734 * will be avoided by refdst_drop. If dst entry is not cached, we take
735 * reference, so that last dst_release can destroy the dst immediately.
736 */
737static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
738{
dbfc4fb7
HFS
739 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
740 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 741}
7fee226a
ED
742
743/**
25985edc 744 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
745 * @skb: buffer
746 */
747static inline bool skb_dst_is_noref(const struct sk_buff *skb)
748{
749 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
750}
751
511c3f92
ED
752static inline struct rtable *skb_rtable(const struct sk_buff *skb)
753{
adf30907 754 return (struct rtable *)skb_dst(skb);
511c3f92
ED
755}
756
7965bd4d
JP
757void kfree_skb(struct sk_buff *skb);
758void kfree_skb_list(struct sk_buff *segs);
759void skb_tx_error(struct sk_buff *skb);
760void consume_skb(struct sk_buff *skb);
761void __kfree_skb(struct sk_buff *skb);
d7e8883c 762extern struct kmem_cache *skbuff_head_cache;
bad43ca8 763
7965bd4d
JP
764void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
765bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
766 bool *fragstolen, int *delta_truesize);
bad43ca8 767
7965bd4d
JP
768struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
769 int node);
770struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 771static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 772 gfp_t priority)
d179cd12 773{
564824b0 774 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
775}
776
2e4e4410
ED
777struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
778 unsigned long data_len,
779 int max_page_order,
780 int *errcode,
781 gfp_t gfp_mask);
782
d0bf4a9e
ED
783/* Layout of fast clones : [skb1][skb2][fclone_ref] */
784struct sk_buff_fclones {
785 struct sk_buff skb1;
786
787 struct sk_buff skb2;
788
789 atomic_t fclone_ref;
790};
791
792/**
793 * skb_fclone_busy - check if fclone is busy
794 * @skb: buffer
795 *
796 * Returns true is skb is a fast clone, and its clone is not freed.
39bb5e62
ED
797 * Some drivers call skb_orphan() in their ndo_start_xmit(),
798 * so we also check that this didnt happen.
d0bf4a9e 799 */
39bb5e62
ED
800static inline bool skb_fclone_busy(const struct sock *sk,
801 const struct sk_buff *skb)
d0bf4a9e
ED
802{
803 const struct sk_buff_fclones *fclones;
804
805 fclones = container_of(skb, struct sk_buff_fclones, skb1);
806
807 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 808 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 809 fclones->skb2.sk == sk;
d0bf4a9e
ED
810}
811
d179cd12 812static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 813 gfp_t priority)
d179cd12 814{
c93bdd0e 815 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
816}
817
7965bd4d 818struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
819static inline struct sk_buff *alloc_skb_head(gfp_t priority)
820{
821 return __alloc_skb_head(priority, -1);
822}
823
7965bd4d
JP
824struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
825int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
826struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
827struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
828struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
829 gfp_t gfp_mask, bool fclone);
830static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
831 gfp_t gfp_mask)
832{
833 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
834}
7965bd4d
JP
835
836int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
837struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
838 unsigned int headroom);
839struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
840 int newtailroom, gfp_t priority);
25a91d8d
FD
841int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
842 int offset, int len);
7965bd4d
JP
843int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
844 int len);
845int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
846int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 847#define dev_kfree_skb(a) consume_skb(a)
1da177e4 848
7965bd4d
JP
849int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
850 int getfrag(void *from, char *to, int offset,
851 int len, int odd, struct sk_buff *skb),
852 void *from, int length);
e89e9cf5 853
d94d9fee 854struct skb_seq_state {
677e90ed
TG
855 __u32 lower_offset;
856 __u32 upper_offset;
857 __u32 frag_idx;
858 __u32 stepped_offset;
859 struct sk_buff *root_skb;
860 struct sk_buff *cur_skb;
861 __u8 *frag_data;
862};
863
7965bd4d
JP
864void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
865 unsigned int to, struct skb_seq_state *st);
866unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
867 struct skb_seq_state *st);
868void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 869
7965bd4d 870unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 871 unsigned int to, struct ts_config *config);
3fc7e8a6 872
09323cc4
TH
873/*
874 * Packet hash types specify the type of hash in skb_set_hash.
875 *
876 * Hash types refer to the protocol layer addresses which are used to
877 * construct a packet's hash. The hashes are used to differentiate or identify
878 * flows of the protocol layer for the hash type. Hash types are either
879 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
880 *
881 * Properties of hashes:
882 *
883 * 1) Two packets in different flows have different hash values
884 * 2) Two packets in the same flow should have the same hash value
885 *
886 * A hash at a higher layer is considered to be more specific. A driver should
887 * set the most specific hash possible.
888 *
889 * A driver cannot indicate a more specific hash than the layer at which a hash
890 * was computed. For instance an L3 hash cannot be set as an L4 hash.
891 *
892 * A driver may indicate a hash level which is less specific than the
893 * actual layer the hash was computed on. For instance, a hash computed
894 * at L4 may be considered an L3 hash. This should only be done if the
895 * driver can't unambiguously determine that the HW computed the hash at
896 * the higher layer. Note that the "should" in the second property above
897 * permits this.
898 */
899enum pkt_hash_types {
900 PKT_HASH_TYPE_NONE, /* Undefined type */
901 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
902 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
903 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
904};
905
906static inline void
907skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
908{
61b905da 909 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 910 skb->sw_hash = 0;
61b905da 911 skb->hash = hash;
09323cc4
TH
912}
913
3958afa1
TH
914void __skb_get_hash(struct sk_buff *skb);
915static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 916{
a3b18ddb 917 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 918 __skb_get_hash(skb);
bfb564e7 919
61b905da 920 return skb->hash;
bfb564e7
KK
921}
922
57bdf7f4
TH
923static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
924{
61b905da 925 return skb->hash;
57bdf7f4
TH
926}
927
7539fadc
TH
928static inline void skb_clear_hash(struct sk_buff *skb)
929{
61b905da 930 skb->hash = 0;
a3b18ddb 931 skb->sw_hash = 0;
61b905da 932 skb->l4_hash = 0;
7539fadc
TH
933}
934
935static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
936{
61b905da 937 if (!skb->l4_hash)
7539fadc
TH
938 skb_clear_hash(skb);
939}
940
3df7a74e
TH
941static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
942{
61b905da 943 to->hash = from->hash;
a3b18ddb 944 to->sw_hash = from->sw_hash;
61b905da 945 to->l4_hash = from->l4_hash;
3df7a74e
TH
946};
947
c29390c6
ED
948static inline void skb_sender_cpu_clear(struct sk_buff *skb)
949{
950#ifdef CONFIG_XPS
951 skb->sender_cpu = 0;
952#endif
953}
954
4305b541
ACM
955#ifdef NET_SKBUFF_DATA_USES_OFFSET
956static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
957{
958 return skb->head + skb->end;
959}
ec47ea82
AD
960
961static inline unsigned int skb_end_offset(const struct sk_buff *skb)
962{
963 return skb->end;
964}
4305b541
ACM
965#else
966static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
967{
968 return skb->end;
969}
ec47ea82
AD
970
971static inline unsigned int skb_end_offset(const struct sk_buff *skb)
972{
973 return skb->end - skb->head;
974}
4305b541
ACM
975#endif
976
1da177e4 977/* Internal */
4305b541 978#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 979
ac45f602
PO
980static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
981{
982 return &skb_shinfo(skb)->hwtstamps;
983}
984
1da177e4
LT
985/**
986 * skb_queue_empty - check if a queue is empty
987 * @list: queue head
988 *
989 * Returns true if the queue is empty, false otherwise.
990 */
991static inline int skb_queue_empty(const struct sk_buff_head *list)
992{
fd44b93c 993 return list->next == (const struct sk_buff *) list;
1da177e4
LT
994}
995
fc7ebb21
DM
996/**
997 * skb_queue_is_last - check if skb is the last entry in the queue
998 * @list: queue head
999 * @skb: buffer
1000 *
1001 * Returns true if @skb is the last buffer on the list.
1002 */
1003static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1004 const struct sk_buff *skb)
1005{
fd44b93c 1006 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1007}
1008
832d11c5
IJ
1009/**
1010 * skb_queue_is_first - check if skb is the first entry in the queue
1011 * @list: queue head
1012 * @skb: buffer
1013 *
1014 * Returns true if @skb is the first buffer on the list.
1015 */
1016static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1017 const struct sk_buff *skb)
1018{
fd44b93c 1019 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1020}
1021
249c8b42
DM
1022/**
1023 * skb_queue_next - return the next packet in the queue
1024 * @list: queue head
1025 * @skb: current buffer
1026 *
1027 * Return the next packet in @list after @skb. It is only valid to
1028 * call this if skb_queue_is_last() evaluates to false.
1029 */
1030static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1031 const struct sk_buff *skb)
1032{
1033 /* This BUG_ON may seem severe, but if we just return then we
1034 * are going to dereference garbage.
1035 */
1036 BUG_ON(skb_queue_is_last(list, skb));
1037 return skb->next;
1038}
1039
832d11c5
IJ
1040/**
1041 * skb_queue_prev - return the prev packet in the queue
1042 * @list: queue head
1043 * @skb: current buffer
1044 *
1045 * Return the prev packet in @list before @skb. It is only valid to
1046 * call this if skb_queue_is_first() evaluates to false.
1047 */
1048static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1049 const struct sk_buff *skb)
1050{
1051 /* This BUG_ON may seem severe, but if we just return then we
1052 * are going to dereference garbage.
1053 */
1054 BUG_ON(skb_queue_is_first(list, skb));
1055 return skb->prev;
1056}
1057
1da177e4
LT
1058/**
1059 * skb_get - reference buffer
1060 * @skb: buffer to reference
1061 *
1062 * Makes another reference to a socket buffer and returns a pointer
1063 * to the buffer.
1064 */
1065static inline struct sk_buff *skb_get(struct sk_buff *skb)
1066{
1067 atomic_inc(&skb->users);
1068 return skb;
1069}
1070
1071/*
1072 * If users == 1, we are the only owner and are can avoid redundant
1073 * atomic change.
1074 */
1075
1da177e4
LT
1076/**
1077 * skb_cloned - is the buffer a clone
1078 * @skb: buffer to check
1079 *
1080 * Returns true if the buffer was generated with skb_clone() and is
1081 * one of multiple shared copies of the buffer. Cloned buffers are
1082 * shared data so must not be written to under normal circumstances.
1083 */
1084static inline int skb_cloned(const struct sk_buff *skb)
1085{
1086 return skb->cloned &&
1087 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1088}
1089
14bbd6a5
PS
1090static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1091{
1092 might_sleep_if(pri & __GFP_WAIT);
1093
1094 if (skb_cloned(skb))
1095 return pskb_expand_head(skb, 0, 0, pri);
1096
1097 return 0;
1098}
1099
1da177e4
LT
1100/**
1101 * skb_header_cloned - is the header a clone
1102 * @skb: buffer to check
1103 *
1104 * Returns true if modifying the header part of the buffer requires
1105 * the data to be copied.
1106 */
1107static inline int skb_header_cloned(const struct sk_buff *skb)
1108{
1109 int dataref;
1110
1111 if (!skb->cloned)
1112 return 0;
1113
1114 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1115 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1116 return dataref != 1;
1117}
1118
1119/**
1120 * skb_header_release - release reference to header
1121 * @skb: buffer to operate on
1122 *
1123 * Drop a reference to the header part of the buffer. This is done
1124 * by acquiring a payload reference. You must not read from the header
1125 * part of skb->data after this.
f4a775d1 1126 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1127 */
1128static inline void skb_header_release(struct sk_buff *skb)
1129{
1130 BUG_ON(skb->nohdr);
1131 skb->nohdr = 1;
1132 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1133}
1134
f4a775d1
ED
1135/**
1136 * __skb_header_release - release reference to header
1137 * @skb: buffer to operate on
1138 *
1139 * Variant of skb_header_release() assuming skb is private to caller.
1140 * We can avoid one atomic operation.
1141 */
1142static inline void __skb_header_release(struct sk_buff *skb)
1143{
1144 skb->nohdr = 1;
1145 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1146}
1147
1148
1da177e4
LT
1149/**
1150 * skb_shared - is the buffer shared
1151 * @skb: buffer to check
1152 *
1153 * Returns true if more than one person has a reference to this
1154 * buffer.
1155 */
1156static inline int skb_shared(const struct sk_buff *skb)
1157{
1158 return atomic_read(&skb->users) != 1;
1159}
1160
1161/**
1162 * skb_share_check - check if buffer is shared and if so clone it
1163 * @skb: buffer to check
1164 * @pri: priority for memory allocation
1165 *
1166 * If the buffer is shared the buffer is cloned and the old copy
1167 * drops a reference. A new clone with a single reference is returned.
1168 * If the buffer is not shared the original buffer is returned. When
1169 * being called from interrupt status or with spinlocks held pri must
1170 * be GFP_ATOMIC.
1171 *
1172 * NULL is returned on a memory allocation failure.
1173 */
47061bc4 1174static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1175{
1176 might_sleep_if(pri & __GFP_WAIT);
1177 if (skb_shared(skb)) {
1178 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1179
1180 if (likely(nskb))
1181 consume_skb(skb);
1182 else
1183 kfree_skb(skb);
1da177e4
LT
1184 skb = nskb;
1185 }
1186 return skb;
1187}
1188
1189/*
1190 * Copy shared buffers into a new sk_buff. We effectively do COW on
1191 * packets to handle cases where we have a local reader and forward
1192 * and a couple of other messy ones. The normal one is tcpdumping
1193 * a packet thats being forwarded.
1194 */
1195
1196/**
1197 * skb_unshare - make a copy of a shared buffer
1198 * @skb: buffer to check
1199 * @pri: priority for memory allocation
1200 *
1201 * If the socket buffer is a clone then this function creates a new
1202 * copy of the data, drops a reference count on the old copy and returns
1203 * the new copy with the reference count at 1. If the buffer is not a clone
1204 * the original buffer is returned. When called with a spinlock held or
1205 * from interrupt state @pri must be %GFP_ATOMIC
1206 *
1207 * %NULL is returned on a memory allocation failure.
1208 */
e2bf521d 1209static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1210 gfp_t pri)
1da177e4
LT
1211{
1212 might_sleep_if(pri & __GFP_WAIT);
1213 if (skb_cloned(skb)) {
1214 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1215
1216 /* Free our shared copy */
1217 if (likely(nskb))
1218 consume_skb(skb);
1219 else
1220 kfree_skb(skb);
1da177e4
LT
1221 skb = nskb;
1222 }
1223 return skb;
1224}
1225
1226/**
1a5778aa 1227 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1228 * @list_: list to peek at
1229 *
1230 * Peek an &sk_buff. Unlike most other operations you _MUST_
1231 * be careful with this one. A peek leaves the buffer on the
1232 * list and someone else may run off with it. You must hold
1233 * the appropriate locks or have a private queue to do this.
1234 *
1235 * Returns %NULL for an empty list or a pointer to the head element.
1236 * The reference count is not incremented and the reference is therefore
1237 * volatile. Use with caution.
1238 */
05bdd2f1 1239static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1240{
18d07000
ED
1241 struct sk_buff *skb = list_->next;
1242
1243 if (skb == (struct sk_buff *)list_)
1244 skb = NULL;
1245 return skb;
1da177e4
LT
1246}
1247
da5ef6e5
PE
1248/**
1249 * skb_peek_next - peek skb following the given one from a queue
1250 * @skb: skb to start from
1251 * @list_: list to peek at
1252 *
1253 * Returns %NULL when the end of the list is met or a pointer to the
1254 * next element. The reference count is not incremented and the
1255 * reference is therefore volatile. Use with caution.
1256 */
1257static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1258 const struct sk_buff_head *list_)
1259{
1260 struct sk_buff *next = skb->next;
18d07000 1261
da5ef6e5
PE
1262 if (next == (struct sk_buff *)list_)
1263 next = NULL;
1264 return next;
1265}
1266
1da177e4 1267/**
1a5778aa 1268 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1269 * @list_: list to peek at
1270 *
1271 * Peek an &sk_buff. Unlike most other operations you _MUST_
1272 * be careful with this one. A peek leaves the buffer on the
1273 * list and someone else may run off with it. You must hold
1274 * the appropriate locks or have a private queue to do this.
1275 *
1276 * Returns %NULL for an empty list or a pointer to the tail element.
1277 * The reference count is not incremented and the reference is therefore
1278 * volatile. Use with caution.
1279 */
05bdd2f1 1280static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1281{
18d07000
ED
1282 struct sk_buff *skb = list_->prev;
1283
1284 if (skb == (struct sk_buff *)list_)
1285 skb = NULL;
1286 return skb;
1287
1da177e4
LT
1288}
1289
1290/**
1291 * skb_queue_len - get queue length
1292 * @list_: list to measure
1293 *
1294 * Return the length of an &sk_buff queue.
1295 */
1296static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1297{
1298 return list_->qlen;
1299}
1300
67fed459
DM
1301/**
1302 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1303 * @list: queue to initialize
1304 *
1305 * This initializes only the list and queue length aspects of
1306 * an sk_buff_head object. This allows to initialize the list
1307 * aspects of an sk_buff_head without reinitializing things like
1308 * the spinlock. It can also be used for on-stack sk_buff_head
1309 * objects where the spinlock is known to not be used.
1310 */
1311static inline void __skb_queue_head_init(struct sk_buff_head *list)
1312{
1313 list->prev = list->next = (struct sk_buff *)list;
1314 list->qlen = 0;
1315}
1316
76f10ad0
AV
1317/*
1318 * This function creates a split out lock class for each invocation;
1319 * this is needed for now since a whole lot of users of the skb-queue
1320 * infrastructure in drivers have different locking usage (in hardirq)
1321 * than the networking core (in softirq only). In the long run either the
1322 * network layer or drivers should need annotation to consolidate the
1323 * main types of usage into 3 classes.
1324 */
1da177e4
LT
1325static inline void skb_queue_head_init(struct sk_buff_head *list)
1326{
1327 spin_lock_init(&list->lock);
67fed459 1328 __skb_queue_head_init(list);
1da177e4
LT
1329}
1330
c2ecba71
PE
1331static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1332 struct lock_class_key *class)
1333{
1334 skb_queue_head_init(list);
1335 lockdep_set_class(&list->lock, class);
1336}
1337
1da177e4 1338/*
bf299275 1339 * Insert an sk_buff on a list.
1da177e4
LT
1340 *
1341 * The "__skb_xxxx()" functions are the non-atomic ones that
1342 * can only be called with interrupts disabled.
1343 */
7965bd4d
JP
1344void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1345 struct sk_buff_head *list);
bf299275
GR
1346static inline void __skb_insert(struct sk_buff *newsk,
1347 struct sk_buff *prev, struct sk_buff *next,
1348 struct sk_buff_head *list)
1349{
1350 newsk->next = next;
1351 newsk->prev = prev;
1352 next->prev = prev->next = newsk;
1353 list->qlen++;
1354}
1da177e4 1355
67fed459
DM
1356static inline void __skb_queue_splice(const struct sk_buff_head *list,
1357 struct sk_buff *prev,
1358 struct sk_buff *next)
1359{
1360 struct sk_buff *first = list->next;
1361 struct sk_buff *last = list->prev;
1362
1363 first->prev = prev;
1364 prev->next = first;
1365
1366 last->next = next;
1367 next->prev = last;
1368}
1369
1370/**
1371 * skb_queue_splice - join two skb lists, this is designed for stacks
1372 * @list: the new list to add
1373 * @head: the place to add it in the first list
1374 */
1375static inline void skb_queue_splice(const struct sk_buff_head *list,
1376 struct sk_buff_head *head)
1377{
1378 if (!skb_queue_empty(list)) {
1379 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1380 head->qlen += list->qlen;
67fed459
DM
1381 }
1382}
1383
1384/**
d9619496 1385 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1386 * @list: the new list to add
1387 * @head: the place to add it in the first list
1388 *
1389 * The list at @list is reinitialised
1390 */
1391static inline void skb_queue_splice_init(struct sk_buff_head *list,
1392 struct sk_buff_head *head)
1393{
1394 if (!skb_queue_empty(list)) {
1395 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1396 head->qlen += list->qlen;
67fed459
DM
1397 __skb_queue_head_init(list);
1398 }
1399}
1400
1401/**
1402 * skb_queue_splice_tail - join two skb lists, each list being a queue
1403 * @list: the new list to add
1404 * @head: the place to add it in the first list
1405 */
1406static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1407 struct sk_buff_head *head)
1408{
1409 if (!skb_queue_empty(list)) {
1410 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1411 head->qlen += list->qlen;
67fed459
DM
1412 }
1413}
1414
1415/**
d9619496 1416 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1417 * @list: the new list to add
1418 * @head: the place to add it in the first list
1419 *
1420 * Each of the lists is a queue.
1421 * The list at @list is reinitialised
1422 */
1423static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1424 struct sk_buff_head *head)
1425{
1426 if (!skb_queue_empty(list)) {
1427 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1428 head->qlen += list->qlen;
67fed459
DM
1429 __skb_queue_head_init(list);
1430 }
1431}
1432
1da177e4 1433/**
300ce174 1434 * __skb_queue_after - queue a buffer at the list head
1da177e4 1435 * @list: list to use
300ce174 1436 * @prev: place after this buffer
1da177e4
LT
1437 * @newsk: buffer to queue
1438 *
300ce174 1439 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1440 * and you must therefore hold required locks before calling it.
1441 *
1442 * A buffer cannot be placed on two lists at the same time.
1443 */
300ce174
SH
1444static inline void __skb_queue_after(struct sk_buff_head *list,
1445 struct sk_buff *prev,
1446 struct sk_buff *newsk)
1da177e4 1447{
bf299275 1448 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1449}
1450
7965bd4d
JP
1451void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1452 struct sk_buff_head *list);
7de6c033 1453
f5572855
GR
1454static inline void __skb_queue_before(struct sk_buff_head *list,
1455 struct sk_buff *next,
1456 struct sk_buff *newsk)
1457{
1458 __skb_insert(newsk, next->prev, next, list);
1459}
1460
300ce174
SH
1461/**
1462 * __skb_queue_head - queue a buffer at the list head
1463 * @list: list to use
1464 * @newsk: buffer to queue
1465 *
1466 * Queue a buffer at the start of a list. This function takes no locks
1467 * and you must therefore hold required locks before calling it.
1468 *
1469 * A buffer cannot be placed on two lists at the same time.
1470 */
7965bd4d 1471void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1472static inline void __skb_queue_head(struct sk_buff_head *list,
1473 struct sk_buff *newsk)
1474{
1475 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1476}
1477
1da177e4
LT
1478/**
1479 * __skb_queue_tail - queue a buffer at the list tail
1480 * @list: list to use
1481 * @newsk: buffer to queue
1482 *
1483 * Queue a buffer at the end of a list. This function takes no locks
1484 * and you must therefore hold required locks before calling it.
1485 *
1486 * A buffer cannot be placed on two lists at the same time.
1487 */
7965bd4d 1488void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1489static inline void __skb_queue_tail(struct sk_buff_head *list,
1490 struct sk_buff *newsk)
1491{
f5572855 1492 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1493}
1494
1da177e4
LT
1495/*
1496 * remove sk_buff from list. _Must_ be called atomically, and with
1497 * the list known..
1498 */
7965bd4d 1499void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1500static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1501{
1502 struct sk_buff *next, *prev;
1503
1504 list->qlen--;
1505 next = skb->next;
1506 prev = skb->prev;
1507 skb->next = skb->prev = NULL;
1da177e4
LT
1508 next->prev = prev;
1509 prev->next = next;
1510}
1511
f525c06d
GR
1512/**
1513 * __skb_dequeue - remove from the head of the queue
1514 * @list: list to dequeue from
1515 *
1516 * Remove the head of the list. This function does not take any locks
1517 * so must be used with appropriate locks held only. The head item is
1518 * returned or %NULL if the list is empty.
1519 */
7965bd4d 1520struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1521static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1522{
1523 struct sk_buff *skb = skb_peek(list);
1524 if (skb)
1525 __skb_unlink(skb, list);
1526 return skb;
1527}
1da177e4
LT
1528
1529/**
1530 * __skb_dequeue_tail - remove from the tail of the queue
1531 * @list: list to dequeue from
1532 *
1533 * Remove the tail of the list. This function does not take any locks
1534 * so must be used with appropriate locks held only. The tail item is
1535 * returned or %NULL if the list is empty.
1536 */
7965bd4d 1537struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1538static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1539{
1540 struct sk_buff *skb = skb_peek_tail(list);
1541 if (skb)
1542 __skb_unlink(skb, list);
1543 return skb;
1544}
1545
1546
bdcc0924 1547static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1548{
1549 return skb->data_len;
1550}
1551
1552static inline unsigned int skb_headlen(const struct sk_buff *skb)
1553{
1554 return skb->len - skb->data_len;
1555}
1556
1557static inline int skb_pagelen(const struct sk_buff *skb)
1558{
1559 int i, len = 0;
1560
1561 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1562 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1563 return len + skb_headlen(skb);
1564}
1565
131ea667
IC
1566/**
1567 * __skb_fill_page_desc - initialise a paged fragment in an skb
1568 * @skb: buffer containing fragment to be initialised
1569 * @i: paged fragment index to initialise
1570 * @page: the page to use for this fragment
1571 * @off: the offset to the data with @page
1572 * @size: the length of the data
1573 *
1574 * Initialises the @i'th fragment of @skb to point to &size bytes at
1575 * offset @off within @page.
1576 *
1577 * Does not take any additional reference on the fragment.
1578 */
1579static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1580 struct page *page, int off, int size)
1da177e4
LT
1581{
1582 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1583
c48a11c7
MG
1584 /*
1585 * Propagate page->pfmemalloc to the skb if we can. The problem is
1586 * that not all callers have unique ownership of the page. If
1587 * pfmemalloc is set, we check the mapping as a mapping implies
1588 * page->index is set (index and pfmemalloc share space).
1589 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1590 * do not lose pfmemalloc information as the pages would not be
1591 * allocated using __GFP_MEMALLOC.
1592 */
a8605c60 1593 frag->page.p = page;
1da177e4 1594 frag->page_offset = off;
9e903e08 1595 skb_frag_size_set(frag, size);
cca7af38
PE
1596
1597 page = compound_head(page);
1598 if (page->pfmemalloc && !page->mapping)
1599 skb->pfmemalloc = true;
131ea667
IC
1600}
1601
1602/**
1603 * skb_fill_page_desc - initialise a paged fragment in an skb
1604 * @skb: buffer containing fragment to be initialised
1605 * @i: paged fragment index to initialise
1606 * @page: the page to use for this fragment
1607 * @off: the offset to the data with @page
1608 * @size: the length of the data
1609 *
1610 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1611 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1612 * addition updates @skb such that @i is the last fragment.
1613 *
1614 * Does not take any additional reference on the fragment.
1615 */
1616static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1617 struct page *page, int off, int size)
1618{
1619 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1620 skb_shinfo(skb)->nr_frags = i + 1;
1621}
1622
7965bd4d
JP
1623void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1624 int size, unsigned int truesize);
654bed16 1625
f8e617e1
JW
1626void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1627 unsigned int truesize);
1628
1da177e4 1629#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1630#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1631#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1632
27a884dc
ACM
1633#ifdef NET_SKBUFF_DATA_USES_OFFSET
1634static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1635{
1636 return skb->head + skb->tail;
1637}
1638
1639static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1640{
1641 skb->tail = skb->data - skb->head;
1642}
1643
1644static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1645{
1646 skb_reset_tail_pointer(skb);
1647 skb->tail += offset;
1648}
7cc46190 1649
27a884dc
ACM
1650#else /* NET_SKBUFF_DATA_USES_OFFSET */
1651static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1652{
1653 return skb->tail;
1654}
1655
1656static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1657{
1658 skb->tail = skb->data;
1659}
1660
1661static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1662{
1663 skb->tail = skb->data + offset;
1664}
4305b541 1665
27a884dc
ACM
1666#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1667
1da177e4
LT
1668/*
1669 * Add data to an sk_buff
1670 */
0c7ddf36 1671unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1672unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1673static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1674{
27a884dc 1675 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1676 SKB_LINEAR_ASSERT(skb);
1677 skb->tail += len;
1678 skb->len += len;
1679 return tmp;
1680}
1681
7965bd4d 1682unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1683static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1684{
1685 skb->data -= len;
1686 skb->len += len;
1687 return skb->data;
1688}
1689
7965bd4d 1690unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1691static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1692{
1693 skb->len -= len;
1694 BUG_ON(skb->len < skb->data_len);
1695 return skb->data += len;
1696}
1697
47d29646
DM
1698static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1699{
1700 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1701}
1702
7965bd4d 1703unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1704
1705static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1706{
1707 if (len > skb_headlen(skb) &&
987c402a 1708 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1709 return NULL;
1710 skb->len -= len;
1711 return skb->data += len;
1712}
1713
1714static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1715{
1716 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1717}
1718
1719static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1720{
1721 if (likely(len <= skb_headlen(skb)))
1722 return 1;
1723 if (unlikely(len > skb->len))
1724 return 0;
987c402a 1725 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1726}
1727
1728/**
1729 * skb_headroom - bytes at buffer head
1730 * @skb: buffer to check
1731 *
1732 * Return the number of bytes of free space at the head of an &sk_buff.
1733 */
c2636b4d 1734static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1735{
1736 return skb->data - skb->head;
1737}
1738
1739/**
1740 * skb_tailroom - bytes at buffer end
1741 * @skb: buffer to check
1742 *
1743 * Return the number of bytes of free space at the tail of an sk_buff
1744 */
1745static inline int skb_tailroom(const struct sk_buff *skb)
1746{
4305b541 1747 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1748}
1749
a21d4572
ED
1750/**
1751 * skb_availroom - bytes at buffer end
1752 * @skb: buffer to check
1753 *
1754 * Return the number of bytes of free space at the tail of an sk_buff
1755 * allocated by sk_stream_alloc()
1756 */
1757static inline int skb_availroom(const struct sk_buff *skb)
1758{
16fad69c
ED
1759 if (skb_is_nonlinear(skb))
1760 return 0;
1761
1762 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1763}
1764
1da177e4
LT
1765/**
1766 * skb_reserve - adjust headroom
1767 * @skb: buffer to alter
1768 * @len: bytes to move
1769 *
1770 * Increase the headroom of an empty &sk_buff by reducing the tail
1771 * room. This is only allowed for an empty buffer.
1772 */
8243126c 1773static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1774{
1775 skb->data += len;
1776 skb->tail += len;
1777}
1778
8bce6d7d
TH
1779#define ENCAP_TYPE_ETHER 0
1780#define ENCAP_TYPE_IPPROTO 1
1781
1782static inline void skb_set_inner_protocol(struct sk_buff *skb,
1783 __be16 protocol)
1784{
1785 skb->inner_protocol = protocol;
1786 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1787}
1788
1789static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1790 __u8 ipproto)
1791{
1792 skb->inner_ipproto = ipproto;
1793 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1794}
1795
6a674e9c
JG
1796static inline void skb_reset_inner_headers(struct sk_buff *skb)
1797{
aefbd2b3 1798 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1799 skb->inner_network_header = skb->network_header;
1800 skb->inner_transport_header = skb->transport_header;
1801}
1802
0b5c9db1
JP
1803static inline void skb_reset_mac_len(struct sk_buff *skb)
1804{
1805 skb->mac_len = skb->network_header - skb->mac_header;
1806}
1807
6a674e9c
JG
1808static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1809 *skb)
1810{
1811 return skb->head + skb->inner_transport_header;
1812}
1813
1814static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1815{
1816 skb->inner_transport_header = skb->data - skb->head;
1817}
1818
1819static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1820 const int offset)
1821{
1822 skb_reset_inner_transport_header(skb);
1823 skb->inner_transport_header += offset;
1824}
1825
1826static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1827{
1828 return skb->head + skb->inner_network_header;
1829}
1830
1831static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1832{
1833 skb->inner_network_header = skb->data - skb->head;
1834}
1835
1836static inline void skb_set_inner_network_header(struct sk_buff *skb,
1837 const int offset)
1838{
1839 skb_reset_inner_network_header(skb);
1840 skb->inner_network_header += offset;
1841}
1842
aefbd2b3
PS
1843static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1844{
1845 return skb->head + skb->inner_mac_header;
1846}
1847
1848static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1849{
1850 skb->inner_mac_header = skb->data - skb->head;
1851}
1852
1853static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1854 const int offset)
1855{
1856 skb_reset_inner_mac_header(skb);
1857 skb->inner_mac_header += offset;
1858}
fda55eca
ED
1859static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1860{
35d04610 1861 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1862}
1863
9c70220b
ACM
1864static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1865{
2e07fa9c 1866 return skb->head + skb->transport_header;
9c70220b
ACM
1867}
1868
badff6d0
ACM
1869static inline void skb_reset_transport_header(struct sk_buff *skb)
1870{
2e07fa9c 1871 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1872}
1873
967b05f6
ACM
1874static inline void skb_set_transport_header(struct sk_buff *skb,
1875 const int offset)
1876{
2e07fa9c
ACM
1877 skb_reset_transport_header(skb);
1878 skb->transport_header += offset;
ea2ae17d
ACM
1879}
1880
d56f90a7
ACM
1881static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1882{
2e07fa9c 1883 return skb->head + skb->network_header;
d56f90a7
ACM
1884}
1885
c1d2bbe1
ACM
1886static inline void skb_reset_network_header(struct sk_buff *skb)
1887{
2e07fa9c 1888 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1889}
1890
c14d2450
ACM
1891static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1892{
2e07fa9c
ACM
1893 skb_reset_network_header(skb);
1894 skb->network_header += offset;
c14d2450
ACM
1895}
1896
2e07fa9c 1897static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1898{
2e07fa9c 1899 return skb->head + skb->mac_header;
bbe735e4
ACM
1900}
1901
2e07fa9c 1902static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1903{
35d04610 1904 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1905}
1906
1907static inline void skb_reset_mac_header(struct sk_buff *skb)
1908{
1909 skb->mac_header = skb->data - skb->head;
1910}
1911
1912static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1913{
1914 skb_reset_mac_header(skb);
1915 skb->mac_header += offset;
1916}
1917
0e3da5bb
TT
1918static inline void skb_pop_mac_header(struct sk_buff *skb)
1919{
1920 skb->mac_header = skb->network_header;
1921}
1922
fbbdb8f0
YX
1923static inline void skb_probe_transport_header(struct sk_buff *skb,
1924 const int offset_hint)
1925{
1926 struct flow_keys keys;
1927
1928 if (skb_transport_header_was_set(skb))
1929 return;
1930 else if (skb_flow_dissect(skb, &keys))
1931 skb_set_transport_header(skb, keys.thoff);
1932 else
1933 skb_set_transport_header(skb, offset_hint);
1934}
1935
03606895
ED
1936static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1937{
1938 if (skb_mac_header_was_set(skb)) {
1939 const unsigned char *old_mac = skb_mac_header(skb);
1940
1941 skb_set_mac_header(skb, -skb->mac_len);
1942 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1943 }
1944}
1945
04fb451e
MM
1946static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1947{
1948 return skb->csum_start - skb_headroom(skb);
1949}
1950
2e07fa9c
ACM
1951static inline int skb_transport_offset(const struct sk_buff *skb)
1952{
1953 return skb_transport_header(skb) - skb->data;
1954}
1955
1956static inline u32 skb_network_header_len(const struct sk_buff *skb)
1957{
1958 return skb->transport_header - skb->network_header;
1959}
1960
6a674e9c
JG
1961static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1962{
1963 return skb->inner_transport_header - skb->inner_network_header;
1964}
1965
2e07fa9c
ACM
1966static inline int skb_network_offset(const struct sk_buff *skb)
1967{
1968 return skb_network_header(skb) - skb->data;
1969}
48d49d0c 1970
6a674e9c
JG
1971static inline int skb_inner_network_offset(const struct sk_buff *skb)
1972{
1973 return skb_inner_network_header(skb) - skb->data;
1974}
1975
f9599ce1
CG
1976static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1977{
1978 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1979}
1980
1da177e4
LT
1981/*
1982 * CPUs often take a performance hit when accessing unaligned memory
1983 * locations. The actual performance hit varies, it can be small if the
1984 * hardware handles it or large if we have to take an exception and fix it
1985 * in software.
1986 *
1987 * Since an ethernet header is 14 bytes network drivers often end up with
1988 * the IP header at an unaligned offset. The IP header can be aligned by
1989 * shifting the start of the packet by 2 bytes. Drivers should do this
1990 * with:
1991 *
8660c124 1992 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1993 *
1994 * The downside to this alignment of the IP header is that the DMA is now
1995 * unaligned. On some architectures the cost of an unaligned DMA is high
1996 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1997 *
1da177e4
LT
1998 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1999 * to be overridden.
2000 */
2001#ifndef NET_IP_ALIGN
2002#define NET_IP_ALIGN 2
2003#endif
2004
025be81e
AB
2005/*
2006 * The networking layer reserves some headroom in skb data (via
2007 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2008 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2009 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2010 *
2011 * Unfortunately this headroom changes the DMA alignment of the resulting
2012 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2013 * on some architectures. An architecture can override this value,
2014 * perhaps setting it to a cacheline in size (since that will maintain
2015 * cacheline alignment of the DMA). It must be a power of 2.
2016 *
d6301d3d 2017 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2018 * headroom, you should not reduce this.
5933dd2f
ED
2019 *
2020 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2021 * to reduce average number of cache lines per packet.
2022 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2023 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2024 */
2025#ifndef NET_SKB_PAD
5933dd2f 2026#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2027#endif
2028
7965bd4d 2029int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2030
2031static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2032{
c4264f27 2033 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2034 WARN_ON(1);
2035 return;
2036 }
27a884dc
ACM
2037 skb->len = len;
2038 skb_set_tail_pointer(skb, len);
1da177e4
LT
2039}
2040
7965bd4d 2041void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2042
2043static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2044{
3cc0e873
HX
2045 if (skb->data_len)
2046 return ___pskb_trim(skb, len);
2047 __skb_trim(skb, len);
2048 return 0;
1da177e4
LT
2049}
2050
2051static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2052{
2053 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2054}
2055
e9fa4f7b
HX
2056/**
2057 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2058 * @skb: buffer to alter
2059 * @len: new length
2060 *
2061 * This is identical to pskb_trim except that the caller knows that
2062 * the skb is not cloned so we should never get an error due to out-
2063 * of-memory.
2064 */
2065static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2066{
2067 int err = pskb_trim(skb, len);
2068 BUG_ON(err);
2069}
2070
1da177e4
LT
2071/**
2072 * skb_orphan - orphan a buffer
2073 * @skb: buffer to orphan
2074 *
2075 * If a buffer currently has an owner then we call the owner's
2076 * destructor function and make the @skb unowned. The buffer continues
2077 * to exist but is no longer charged to its former owner.
2078 */
2079static inline void skb_orphan(struct sk_buff *skb)
2080{
c34a7612 2081 if (skb->destructor) {
1da177e4 2082 skb->destructor(skb);
c34a7612
ED
2083 skb->destructor = NULL;
2084 skb->sk = NULL;
376c7311
ED
2085 } else {
2086 BUG_ON(skb->sk);
c34a7612 2087 }
1da177e4
LT
2088}
2089
a353e0ce
MT
2090/**
2091 * skb_orphan_frags - orphan the frags contained in a buffer
2092 * @skb: buffer to orphan frags from
2093 * @gfp_mask: allocation mask for replacement pages
2094 *
2095 * For each frag in the SKB which needs a destructor (i.e. has an
2096 * owner) create a copy of that frag and release the original
2097 * page by calling the destructor.
2098 */
2099static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2100{
2101 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2102 return 0;
2103 return skb_copy_ubufs(skb, gfp_mask);
2104}
2105
1da177e4
LT
2106/**
2107 * __skb_queue_purge - empty a list
2108 * @list: list to empty
2109 *
2110 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2111 * the list and one reference dropped. This function does not take the
2112 * list lock and the caller must hold the relevant locks to use it.
2113 */
7965bd4d 2114void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2115static inline void __skb_queue_purge(struct sk_buff_head *list)
2116{
2117 struct sk_buff *skb;
2118 while ((skb = __skb_dequeue(list)) != NULL)
2119 kfree_skb(skb);
2120}
2121
e5e67305
AD
2122#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2123#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2124#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2125
7965bd4d 2126void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2127
7965bd4d
JP
2128struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2129 gfp_t gfp_mask);
8af27456
CH
2130
2131/**
2132 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2133 * @dev: network device to receive on
2134 * @length: length to allocate
2135 *
2136 * Allocate a new &sk_buff and assign it a usage count of one. The
2137 * buffer has unspecified headroom built in. Users should allocate
2138 * the headroom they think they need without accounting for the
2139 * built in space. The built in space is used for optimisations.
2140 *
2141 * %NULL is returned if there is no free memory. Although this function
2142 * allocates memory it can be called from an interrupt.
2143 */
2144static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2145 unsigned int length)
8af27456
CH
2146{
2147 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2148}
2149
6f532612
ED
2150/* legacy helper around __netdev_alloc_skb() */
2151static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2152 gfp_t gfp_mask)
2153{
2154 return __netdev_alloc_skb(NULL, length, gfp_mask);
2155}
2156
2157/* legacy helper around netdev_alloc_skb() */
2158static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2159{
2160 return netdev_alloc_skb(NULL, length);
2161}
2162
2163
4915a0de
ED
2164static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2165 unsigned int length, gfp_t gfp)
61321bbd 2166{
4915a0de 2167 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2168
2169 if (NET_IP_ALIGN && skb)
2170 skb_reserve(skb, NET_IP_ALIGN);
2171 return skb;
2172}
2173
4915a0de
ED
2174static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2175 unsigned int length)
2176{
2177 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2178}
2179
ffde7328 2180void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2181struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2182 unsigned int length, gfp_t gfp_mask);
2183static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2184 unsigned int length)
2185{
2186 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2187}
ffde7328 2188
71dfda58
AD
2189/**
2190 * __dev_alloc_pages - allocate page for network Rx
2191 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2192 * @order: size of the allocation
2193 *
2194 * Allocate a new page.
2195 *
2196 * %NULL is returned if there is no free memory.
2197*/
2198static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2199 unsigned int order)
2200{
2201 /* This piece of code contains several assumptions.
2202 * 1. This is for device Rx, therefor a cold page is preferred.
2203 * 2. The expectation is the user wants a compound page.
2204 * 3. If requesting a order 0 page it will not be compound
2205 * due to the check to see if order has a value in prep_new_page
2206 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2207 * code in gfp_to_alloc_flags that should be enforcing this.
2208 */
2209 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2210
2211 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2212}
2213
2214static inline struct page *dev_alloc_pages(unsigned int order)
2215{
2216 return __dev_alloc_pages(GFP_ATOMIC, order);
2217}
2218
2219/**
2220 * __dev_alloc_page - allocate a page for network Rx
2221 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2222 *
2223 * Allocate a new page.
2224 *
2225 * %NULL is returned if there is no free memory.
2226 */
2227static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2228{
2229 return __dev_alloc_pages(gfp_mask, 0);
2230}
2231
2232static inline struct page *dev_alloc_page(void)
2233{
2234 return __dev_alloc_page(GFP_ATOMIC);
2235}
2236
0614002b
MG
2237/**
2238 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2239 * @page: The page that was allocated from skb_alloc_page
2240 * @skb: The skb that may need pfmemalloc set
2241 */
2242static inline void skb_propagate_pfmemalloc(struct page *page,
2243 struct sk_buff *skb)
2244{
2245 if (page && page->pfmemalloc)
2246 skb->pfmemalloc = true;
2247}
2248
131ea667 2249/**
e227867f 2250 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2251 * @frag: the paged fragment
2252 *
2253 * Returns the &struct page associated with @frag.
2254 */
2255static inline struct page *skb_frag_page(const skb_frag_t *frag)
2256{
a8605c60 2257 return frag->page.p;
131ea667
IC
2258}
2259
2260/**
2261 * __skb_frag_ref - take an addition reference on a paged fragment.
2262 * @frag: the paged fragment
2263 *
2264 * Takes an additional reference on the paged fragment @frag.
2265 */
2266static inline void __skb_frag_ref(skb_frag_t *frag)
2267{
2268 get_page(skb_frag_page(frag));
2269}
2270
2271/**
2272 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2273 * @skb: the buffer
2274 * @f: the fragment offset.
2275 *
2276 * Takes an additional reference on the @f'th paged fragment of @skb.
2277 */
2278static inline void skb_frag_ref(struct sk_buff *skb, int f)
2279{
2280 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2281}
2282
2283/**
2284 * __skb_frag_unref - release a reference on a paged fragment.
2285 * @frag: the paged fragment
2286 *
2287 * Releases a reference on the paged fragment @frag.
2288 */
2289static inline void __skb_frag_unref(skb_frag_t *frag)
2290{
2291 put_page(skb_frag_page(frag));
2292}
2293
2294/**
2295 * skb_frag_unref - release a reference on a paged fragment of an skb.
2296 * @skb: the buffer
2297 * @f: the fragment offset
2298 *
2299 * Releases a reference on the @f'th paged fragment of @skb.
2300 */
2301static inline void skb_frag_unref(struct sk_buff *skb, int f)
2302{
2303 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2304}
2305
2306/**
2307 * skb_frag_address - gets the address of the data contained in a paged fragment
2308 * @frag: the paged fragment buffer
2309 *
2310 * Returns the address of the data within @frag. The page must already
2311 * be mapped.
2312 */
2313static inline void *skb_frag_address(const skb_frag_t *frag)
2314{
2315 return page_address(skb_frag_page(frag)) + frag->page_offset;
2316}
2317
2318/**
2319 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2320 * @frag: the paged fragment buffer
2321 *
2322 * Returns the address of the data within @frag. Checks that the page
2323 * is mapped and returns %NULL otherwise.
2324 */
2325static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2326{
2327 void *ptr = page_address(skb_frag_page(frag));
2328 if (unlikely(!ptr))
2329 return NULL;
2330
2331 return ptr + frag->page_offset;
2332}
2333
2334/**
2335 * __skb_frag_set_page - sets the page contained in a paged fragment
2336 * @frag: the paged fragment
2337 * @page: the page to set
2338 *
2339 * Sets the fragment @frag to contain @page.
2340 */
2341static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2342{
a8605c60 2343 frag->page.p = page;
131ea667
IC
2344}
2345
2346/**
2347 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2348 * @skb: the buffer
2349 * @f: the fragment offset
2350 * @page: the page to set
2351 *
2352 * Sets the @f'th fragment of @skb to contain @page.
2353 */
2354static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2355 struct page *page)
2356{
2357 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2358}
2359
400dfd3a
ED
2360bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2361
131ea667
IC
2362/**
2363 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2364 * @dev: the device to map the fragment to
131ea667
IC
2365 * @frag: the paged fragment to map
2366 * @offset: the offset within the fragment (starting at the
2367 * fragment's own offset)
2368 * @size: the number of bytes to map
f83347df 2369 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2370 *
2371 * Maps the page associated with @frag to @device.
2372 */
2373static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2374 const skb_frag_t *frag,
2375 size_t offset, size_t size,
2376 enum dma_data_direction dir)
2377{
2378 return dma_map_page(dev, skb_frag_page(frag),
2379 frag->page_offset + offset, size, dir);
2380}
2381
117632e6
ED
2382static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2383 gfp_t gfp_mask)
2384{
2385 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2386}
2387
bad93e9d
OP
2388
2389static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2390 gfp_t gfp_mask)
2391{
2392 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2393}
2394
2395
334a8132
PM
2396/**
2397 * skb_clone_writable - is the header of a clone writable
2398 * @skb: buffer to check
2399 * @len: length up to which to write
2400 *
2401 * Returns true if modifying the header part of the cloned buffer
2402 * does not requires the data to be copied.
2403 */
05bdd2f1 2404static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2405{
2406 return !skb_header_cloned(skb) &&
2407 skb_headroom(skb) + len <= skb->hdr_len;
2408}
2409
d9cc2048
HX
2410static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2411 int cloned)
2412{
2413 int delta = 0;
2414
d9cc2048
HX
2415 if (headroom > skb_headroom(skb))
2416 delta = headroom - skb_headroom(skb);
2417
2418 if (delta || cloned)
2419 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2420 GFP_ATOMIC);
2421 return 0;
2422}
2423
1da177e4
LT
2424/**
2425 * skb_cow - copy header of skb when it is required
2426 * @skb: buffer to cow
2427 * @headroom: needed headroom
2428 *
2429 * If the skb passed lacks sufficient headroom or its data part
2430 * is shared, data is reallocated. If reallocation fails, an error
2431 * is returned and original skb is not changed.
2432 *
2433 * The result is skb with writable area skb->head...skb->tail
2434 * and at least @headroom of space at head.
2435 */
2436static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2437{
d9cc2048
HX
2438 return __skb_cow(skb, headroom, skb_cloned(skb));
2439}
1da177e4 2440
d9cc2048
HX
2441/**
2442 * skb_cow_head - skb_cow but only making the head writable
2443 * @skb: buffer to cow
2444 * @headroom: needed headroom
2445 *
2446 * This function is identical to skb_cow except that we replace the
2447 * skb_cloned check by skb_header_cloned. It should be used when
2448 * you only need to push on some header and do not need to modify
2449 * the data.
2450 */
2451static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2452{
2453 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2454}
2455
2456/**
2457 * skb_padto - pad an skbuff up to a minimal size
2458 * @skb: buffer to pad
2459 * @len: minimal length
2460 *
2461 * Pads up a buffer to ensure the trailing bytes exist and are
2462 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2463 * is untouched. Otherwise it is extended. Returns zero on
2464 * success. The skb is freed on error.
1da177e4 2465 */
5b057c6b 2466static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2467{
2468 unsigned int size = skb->len;
2469 if (likely(size >= len))
5b057c6b 2470 return 0;
987c402a 2471 return skb_pad(skb, len - size);
1da177e4
LT
2472}
2473
9c0c1124
AD
2474/**
2475 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2476 * @skb: buffer to pad
2477 * @len: minimal length
2478 *
2479 * Pads up a buffer to ensure the trailing bytes exist and are
2480 * blanked. If the buffer already contains sufficient data it
2481 * is untouched. Otherwise it is extended. Returns zero on
2482 * success. The skb is freed on error.
2483 */
2484static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2485{
2486 unsigned int size = skb->len;
2487
2488 if (unlikely(size < len)) {
2489 len -= size;
2490 if (skb_pad(skb, len))
2491 return -ENOMEM;
2492 __skb_put(skb, len);
2493 }
2494 return 0;
2495}
2496
1da177e4 2497static inline int skb_add_data(struct sk_buff *skb,
af2b040e 2498 struct iov_iter *from, int copy)
1da177e4
LT
2499{
2500 const int off = skb->len;
2501
2502 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e
AV
2503 __wsum csum = 0;
2504 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2505 &csum, from) == copy) {
1da177e4
LT
2506 skb->csum = csum_block_add(skb->csum, csum, off);
2507 return 0;
2508 }
af2b040e 2509 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
1da177e4
LT
2510 return 0;
2511
2512 __skb_trim(skb, off);
2513 return -EFAULT;
2514}
2515
38ba0a65
ED
2516static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2517 const struct page *page, int off)
1da177e4
LT
2518{
2519 if (i) {
9e903e08 2520 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2521
ea2ab693 2522 return page == skb_frag_page(frag) &&
9e903e08 2523 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2524 }
38ba0a65 2525 return false;
1da177e4
LT
2526}
2527
364c6bad
HX
2528static inline int __skb_linearize(struct sk_buff *skb)
2529{
2530 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2531}
2532
1da177e4
LT
2533/**
2534 * skb_linearize - convert paged skb to linear one
2535 * @skb: buffer to linarize
1da177e4
LT
2536 *
2537 * If there is no free memory -ENOMEM is returned, otherwise zero
2538 * is returned and the old skb data released.
2539 */
364c6bad
HX
2540static inline int skb_linearize(struct sk_buff *skb)
2541{
2542 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2543}
2544
cef401de
ED
2545/**
2546 * skb_has_shared_frag - can any frag be overwritten
2547 * @skb: buffer to test
2548 *
2549 * Return true if the skb has at least one frag that might be modified
2550 * by an external entity (as in vmsplice()/sendfile())
2551 */
2552static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2553{
c9af6db4
PS
2554 return skb_is_nonlinear(skb) &&
2555 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2556}
2557
364c6bad
HX
2558/**
2559 * skb_linearize_cow - make sure skb is linear and writable
2560 * @skb: buffer to process
2561 *
2562 * If there is no free memory -ENOMEM is returned, otherwise zero
2563 * is returned and the old skb data released.
2564 */
2565static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2566{
364c6bad
HX
2567 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2568 __skb_linearize(skb) : 0;
1da177e4
LT
2569}
2570
2571/**
2572 * skb_postpull_rcsum - update checksum for received skb after pull
2573 * @skb: buffer to update
2574 * @start: start of data before pull
2575 * @len: length of data pulled
2576 *
2577 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2578 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2579 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2580 */
2581
2582static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2583 const void *start, unsigned int len)
1da177e4 2584{
84fa7933 2585 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2586 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2587}
2588
cbb042f9
HX
2589unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2590
7ce5a27f
DM
2591/**
2592 * pskb_trim_rcsum - trim received skb and update checksum
2593 * @skb: buffer to trim
2594 * @len: new length
2595 *
2596 * This is exactly the same as pskb_trim except that it ensures the
2597 * checksum of received packets are still valid after the operation.
2598 */
2599
2600static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2601{
2602 if (likely(len >= skb->len))
2603 return 0;
2604 if (skb->ip_summed == CHECKSUM_COMPLETE)
2605 skb->ip_summed = CHECKSUM_NONE;
2606 return __pskb_trim(skb, len);
2607}
2608
1da177e4
LT
2609#define skb_queue_walk(queue, skb) \
2610 for (skb = (queue)->next; \
a1e4891f 2611 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2612 skb = skb->next)
2613
46f8914e
JC
2614#define skb_queue_walk_safe(queue, skb, tmp) \
2615 for (skb = (queue)->next, tmp = skb->next; \
2616 skb != (struct sk_buff *)(queue); \
2617 skb = tmp, tmp = skb->next)
2618
1164f52a 2619#define skb_queue_walk_from(queue, skb) \
a1e4891f 2620 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2621 skb = skb->next)
2622
2623#define skb_queue_walk_from_safe(queue, skb, tmp) \
2624 for (tmp = skb->next; \
2625 skb != (struct sk_buff *)(queue); \
2626 skb = tmp, tmp = skb->next)
2627
300ce174
SH
2628#define skb_queue_reverse_walk(queue, skb) \
2629 for (skb = (queue)->prev; \
a1e4891f 2630 skb != (struct sk_buff *)(queue); \
300ce174
SH
2631 skb = skb->prev)
2632
686a2955
DM
2633#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2634 for (skb = (queue)->prev, tmp = skb->prev; \
2635 skb != (struct sk_buff *)(queue); \
2636 skb = tmp, tmp = skb->prev)
2637
2638#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2639 for (tmp = skb->prev; \
2640 skb != (struct sk_buff *)(queue); \
2641 skb = tmp, tmp = skb->prev)
1da177e4 2642
21dc3301 2643static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2644{
2645 return skb_shinfo(skb)->frag_list != NULL;
2646}
2647
2648static inline void skb_frag_list_init(struct sk_buff *skb)
2649{
2650 skb_shinfo(skb)->frag_list = NULL;
2651}
2652
2653static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2654{
2655 frag->next = skb_shinfo(skb)->frag_list;
2656 skb_shinfo(skb)->frag_list = frag;
2657}
2658
2659#define skb_walk_frags(skb, iter) \
2660 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2661
7965bd4d
JP
2662struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2663 int *peeked, int *off, int *err);
2664struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2665 int *err);
2666unsigned int datagram_poll(struct file *file, struct socket *sock,
2667 struct poll_table_struct *wait);
c0371da6
AV
2668int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2669 struct iov_iter *to, int size);
51f3d02b
DM
2670static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2671 struct msghdr *msg, int size)
2672{
e5a4b0bb 2673 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 2674}
e5a4b0bb
AV
2675int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2676 struct msghdr *msg);
3a654f97
AV
2677int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2678 struct iov_iter *from, int len);
3a654f97 2679int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d
JP
2680void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2681void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2682int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2683int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2684int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2685__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2686 int len, __wsum csum);
2687int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2688 struct pipe_inode_info *pipe, unsigned int len,
2689 unsigned int flags);
2690void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2691unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2692int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2693 int len, int hlen);
7965bd4d
JP
2694void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2695int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2696void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2697unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2698struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2699struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 2700int skb_ensure_writable(struct sk_buff *skb, int write_len);
93515d53
JP
2701int skb_vlan_pop(struct sk_buff *skb);
2702int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
20380731 2703
6ce8e9ce
AV
2704static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2705{
21226abb 2706 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
6ce8e9ce
AV
2707}
2708
7eab8d9e
AV
2709static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2710{
e5a4b0bb 2711 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
2712}
2713
2817a336
DB
2714struct skb_checksum_ops {
2715 __wsum (*update)(const void *mem, int len, __wsum wsum);
2716 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2717};
2718
2719__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2720 __wsum csum, const struct skb_checksum_ops *ops);
2721__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2722 __wsum csum);
2723
690e36e7
DM
2724static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2725 int len, void *data, int hlen, void *buffer)
1da177e4 2726{
55820ee2 2727 if (hlen - offset >= len)
690e36e7 2728 return data + offset;
1da177e4 2729
690e36e7
DM
2730 if (!skb ||
2731 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2732 return NULL;
2733
2734 return buffer;
2735}
2736
690e36e7
DM
2737static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2738 int len, void *buffer)
2739{
2740 return __skb_header_pointer(skb, offset, len, skb->data,
2741 skb_headlen(skb), buffer);
2742}
2743
4262e5cc
DB
2744/**
2745 * skb_needs_linearize - check if we need to linearize a given skb
2746 * depending on the given device features.
2747 * @skb: socket buffer to check
2748 * @features: net device features
2749 *
2750 * Returns true if either:
2751 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2752 * 2. skb is fragmented and the device does not support SG.
2753 */
2754static inline bool skb_needs_linearize(struct sk_buff *skb,
2755 netdev_features_t features)
2756{
2757 return skb_is_nonlinear(skb) &&
2758 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2759 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2760}
2761
d626f62b
ACM
2762static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2763 void *to,
2764 const unsigned int len)
2765{
2766 memcpy(to, skb->data, len);
2767}
2768
2769static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2770 const int offset, void *to,
2771 const unsigned int len)
2772{
2773 memcpy(to, skb->data + offset, len);
2774}
2775
27d7ff46
ACM
2776static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2777 const void *from,
2778 const unsigned int len)
2779{
2780 memcpy(skb->data, from, len);
2781}
2782
2783static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2784 const int offset,
2785 const void *from,
2786 const unsigned int len)
2787{
2788 memcpy(skb->data + offset, from, len);
2789}
2790
7965bd4d 2791void skb_init(void);
1da177e4 2792
ac45f602
PO
2793static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2794{
2795 return skb->tstamp;
2796}
2797
a61bbcf2
PM
2798/**
2799 * skb_get_timestamp - get timestamp from a skb
2800 * @skb: skb to get stamp from
2801 * @stamp: pointer to struct timeval to store stamp in
2802 *
2803 * Timestamps are stored in the skb as offsets to a base timestamp.
2804 * This function converts the offset back to a struct timeval and stores
2805 * it in stamp.
2806 */
ac45f602
PO
2807static inline void skb_get_timestamp(const struct sk_buff *skb,
2808 struct timeval *stamp)
a61bbcf2 2809{
b7aa0bf7 2810 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2811}
2812
ac45f602
PO
2813static inline void skb_get_timestampns(const struct sk_buff *skb,
2814 struct timespec *stamp)
2815{
2816 *stamp = ktime_to_timespec(skb->tstamp);
2817}
2818
b7aa0bf7 2819static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2820{
b7aa0bf7 2821 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2822}
2823
164891aa
SH
2824static inline ktime_t net_timedelta(ktime_t t)
2825{
2826 return ktime_sub(ktime_get_real(), t);
2827}
2828
b9ce204f
IJ
2829static inline ktime_t net_invalid_timestamp(void)
2830{
2831 return ktime_set(0, 0);
2832}
a61bbcf2 2833
62bccb8c
AD
2834struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2835
c1f19b51
RC
2836#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2837
7965bd4d
JP
2838void skb_clone_tx_timestamp(struct sk_buff *skb);
2839bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2840
2841#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2842
2843static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2844{
2845}
2846
2847static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2848{
2849 return false;
2850}
2851
2852#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2853
2854/**
2855 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2856 *
da92b194
RC
2857 * PHY drivers may accept clones of transmitted packets for
2858 * timestamping via their phy_driver.txtstamp method. These drivers
2859 * must call this function to return the skb back to the stack, with
2860 * or without a timestamp.
2861 *
c1f19b51 2862 * @skb: clone of the the original outgoing packet
da92b194 2863 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2864 *
2865 */
2866void skb_complete_tx_timestamp(struct sk_buff *skb,
2867 struct skb_shared_hwtstamps *hwtstamps);
2868
e7fd2885
WB
2869void __skb_tstamp_tx(struct sk_buff *orig_skb,
2870 struct skb_shared_hwtstamps *hwtstamps,
2871 struct sock *sk, int tstype);
2872
ac45f602
PO
2873/**
2874 * skb_tstamp_tx - queue clone of skb with send time stamps
2875 * @orig_skb: the original outgoing packet
2876 * @hwtstamps: hardware time stamps, may be NULL if not available
2877 *
2878 * If the skb has a socket associated, then this function clones the
2879 * skb (thus sharing the actual data and optional structures), stores
2880 * the optional hardware time stamping information (if non NULL) or
2881 * generates a software time stamp (otherwise), then queues the clone
2882 * to the error queue of the socket. Errors are silently ignored.
2883 */
7965bd4d
JP
2884void skb_tstamp_tx(struct sk_buff *orig_skb,
2885 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2886
4507a715
RC
2887static inline void sw_tx_timestamp(struct sk_buff *skb)
2888{
2244d07b
OH
2889 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2890 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2891 skb_tstamp_tx(skb, NULL);
2892}
2893
2894/**
2895 * skb_tx_timestamp() - Driver hook for transmit timestamping
2896 *
2897 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2898 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2899 *
73409f3b
DM
2900 * Specifically, one should make absolutely sure that this function is
2901 * called before TX completion of this packet can trigger. Otherwise
2902 * the packet could potentially already be freed.
2903 *
4507a715
RC
2904 * @skb: A socket buffer.
2905 */
2906static inline void skb_tx_timestamp(struct sk_buff *skb)
2907{
c1f19b51 2908 skb_clone_tx_timestamp(skb);
4507a715
RC
2909 sw_tx_timestamp(skb);
2910}
2911
6e3e939f
JB
2912/**
2913 * skb_complete_wifi_ack - deliver skb with wifi status
2914 *
2915 * @skb: the original outgoing packet
2916 * @acked: ack status
2917 *
2918 */
2919void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2920
7965bd4d
JP
2921__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2922__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2923
60476372
HX
2924static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2925{
6edec0e6
TH
2926 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
2927 skb->csum_valid ||
2928 (skb->ip_summed == CHECKSUM_PARTIAL &&
2929 skb_checksum_start_offset(skb) >= 0));
60476372
HX
2930}
2931
fb286bb2
HX
2932/**
2933 * skb_checksum_complete - Calculate checksum of an entire packet
2934 * @skb: packet to process
2935 *
2936 * This function calculates the checksum over the entire packet plus
2937 * the value of skb->csum. The latter can be used to supply the
2938 * checksum of a pseudo header as used by TCP/UDP. It returns the
2939 * checksum.
2940 *
2941 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2942 * this function can be used to verify that checksum on received
2943 * packets. In that case the function should return zero if the
2944 * checksum is correct. In particular, this function will return zero
2945 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2946 * hardware has already verified the correctness of the checksum.
2947 */
4381ca3c 2948static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2949{
60476372
HX
2950 return skb_csum_unnecessary(skb) ?
2951 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2952}
2953
77cffe23
TH
2954static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2955{
2956 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2957 if (skb->csum_level == 0)
2958 skb->ip_summed = CHECKSUM_NONE;
2959 else
2960 skb->csum_level--;
2961 }
2962}
2963
2964static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2965{
2966 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2967 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2968 skb->csum_level++;
2969 } else if (skb->ip_summed == CHECKSUM_NONE) {
2970 skb->ip_summed = CHECKSUM_UNNECESSARY;
2971 skb->csum_level = 0;
2972 }
2973}
2974
5a212329
TH
2975static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2976{
2977 /* Mark current checksum as bad (typically called from GRO
2978 * path). In the case that ip_summed is CHECKSUM_NONE
2979 * this must be the first checksum encountered in the packet.
2980 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2981 * checksum after the last one validated. For UDP, a zero
2982 * checksum can not be marked as bad.
2983 */
2984
2985 if (skb->ip_summed == CHECKSUM_NONE ||
2986 skb->ip_summed == CHECKSUM_UNNECESSARY)
2987 skb->csum_bad = 1;
2988}
2989
76ba0aae
TH
2990/* Check if we need to perform checksum complete validation.
2991 *
2992 * Returns true if checksum complete is needed, false otherwise
2993 * (either checksum is unnecessary or zero checksum is allowed).
2994 */
2995static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2996 bool zero_okay,
2997 __sum16 check)
2998{
5d0c2b95
TH
2999 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3000 skb->csum_valid = 1;
77cffe23 3001 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3002 return false;
3003 }
3004
3005 return true;
3006}
3007
3008/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3009 * in checksum_init.
3010 */
3011#define CHECKSUM_BREAK 76
3012
3013/* Validate (init) checksum based on checksum complete.
3014 *
3015 * Return values:
3016 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3017 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3018 * checksum is stored in skb->csum for use in __skb_checksum_complete
3019 * non-zero: value of invalid checksum
3020 *
3021 */
3022static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3023 bool complete,
3024 __wsum psum)
3025{
3026 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3027 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3028 skb->csum_valid = 1;
76ba0aae
TH
3029 return 0;
3030 }
5a212329
TH
3031 } else if (skb->csum_bad) {
3032 /* ip_summed == CHECKSUM_NONE in this case */
3033 return 1;
76ba0aae
TH
3034 }
3035
3036 skb->csum = psum;
3037
5d0c2b95
TH
3038 if (complete || skb->len <= CHECKSUM_BREAK) {
3039 __sum16 csum;
3040
3041 csum = __skb_checksum_complete(skb);
3042 skb->csum_valid = !csum;
3043 return csum;
3044 }
76ba0aae
TH
3045
3046 return 0;
3047}
3048
3049static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3050{
3051 return 0;
3052}
3053
3054/* Perform checksum validate (init). Note that this is a macro since we only
3055 * want to calculate the pseudo header which is an input function if necessary.
3056 * First we try to validate without any computation (checksum unnecessary) and
3057 * then calculate based on checksum complete calling the function to compute
3058 * pseudo header.
3059 *
3060 * Return values:
3061 * 0: checksum is validated or try to in skb_checksum_complete
3062 * non-zero: value of invalid checksum
3063 */
3064#define __skb_checksum_validate(skb, proto, complete, \
3065 zero_okay, check, compute_pseudo) \
3066({ \
3067 __sum16 __ret = 0; \
5d0c2b95 3068 skb->csum_valid = 0; \
76ba0aae
TH
3069 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3070 __ret = __skb_checksum_validate_complete(skb, \
3071 complete, compute_pseudo(skb, proto)); \
3072 __ret; \
3073})
3074
3075#define skb_checksum_init(skb, proto, compute_pseudo) \
3076 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3077
3078#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3079 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3080
3081#define skb_checksum_validate(skb, proto, compute_pseudo) \
3082 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3083
3084#define skb_checksum_validate_zero_check(skb, proto, check, \
3085 compute_pseudo) \
096a4cfa 3086 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3087
3088#define skb_checksum_simple_validate(skb) \
3089 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3090
d96535a1
TH
3091static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3092{
3093 return (skb->ip_summed == CHECKSUM_NONE &&
3094 skb->csum_valid && !skb->csum_bad);
3095}
3096
3097static inline void __skb_checksum_convert(struct sk_buff *skb,
3098 __sum16 check, __wsum pseudo)
3099{
3100 skb->csum = ~pseudo;
3101 skb->ip_summed = CHECKSUM_COMPLETE;
3102}
3103
3104#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3105do { \
3106 if (__skb_checksum_convert_check(skb)) \
3107 __skb_checksum_convert(skb, check, \
3108 compute_pseudo(skb, proto)); \
3109} while (0)
3110
15e2396d
TH
3111static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3112 u16 start, u16 offset)
3113{
3114 skb->ip_summed = CHECKSUM_PARTIAL;
3115 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3116 skb->csum_offset = offset - start;
3117}
3118
dcdc8994
TH
3119/* Update skbuf and packet to reflect the remote checksum offload operation.
3120 * When called, ptr indicates the starting point for skb->csum when
3121 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3122 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3123 */
3124static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3125 int start, int offset, bool nopartial)
dcdc8994
TH
3126{
3127 __wsum delta;
3128
15e2396d
TH
3129 if (!nopartial) {
3130 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3131 return;
3132 }
3133
dcdc8994
TH
3134 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3135 __skb_checksum_complete(skb);
3136 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3137 }
3138
3139 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3140
3141 /* Adjust skb->csum since we changed the packet */
3142 skb->csum = csum_add(skb->csum, delta);
3143}
3144
5f79e0f9 3145#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3146void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3147static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3148{
3149 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3150 nf_conntrack_destroy(nfct);
1da177e4
LT
3151}
3152static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3153{
3154 if (nfct)
3155 atomic_inc(&nfct->use);
3156}
2fc72c7b 3157#endif
34666d46 3158#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3159static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3160{
3161 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3162 kfree(nf_bridge);
3163}
3164static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3165{
3166 if (nf_bridge)
3167 atomic_inc(&nf_bridge->use);
3168}
3169#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3170static inline void nf_reset(struct sk_buff *skb)
3171{
5f79e0f9 3172#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3173 nf_conntrack_put(skb->nfct);
3174 skb->nfct = NULL;
2fc72c7b 3175#endif
34666d46 3176#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3177 nf_bridge_put(skb->nf_bridge);
3178 skb->nf_bridge = NULL;
3179#endif
3180}
3181
124dff01
PM
3182static inline void nf_reset_trace(struct sk_buff *skb)
3183{
478b360a 3184#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3185 skb->nf_trace = 0;
3186#endif
a193a4ab
PM
3187}
3188
edda553c 3189/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3190static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3191 bool copy)
edda553c 3192{
5f79e0f9 3193#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3194 dst->nfct = src->nfct;
3195 nf_conntrack_get(src->nfct);
b1937227
ED
3196 if (copy)
3197 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3198#endif
34666d46 3199#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3200 dst->nf_bridge = src->nf_bridge;
3201 nf_bridge_get(src->nf_bridge);
3202#endif
478b360a 3203#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3204 if (copy)
3205 dst->nf_trace = src->nf_trace;
478b360a 3206#endif
edda553c
YK
3207}
3208
e7ac05f3
YK
3209static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3210{
e7ac05f3 3211#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3212 nf_conntrack_put(dst->nfct);
2fc72c7b 3213#endif
34666d46 3214#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3215 nf_bridge_put(dst->nf_bridge);
3216#endif
b1937227 3217 __nf_copy(dst, src, true);
e7ac05f3
YK
3218}
3219
984bc16c
JM
3220#ifdef CONFIG_NETWORK_SECMARK
3221static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3222{
3223 to->secmark = from->secmark;
3224}
3225
3226static inline void skb_init_secmark(struct sk_buff *skb)
3227{
3228 skb->secmark = 0;
3229}
3230#else
3231static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3232{ }
3233
3234static inline void skb_init_secmark(struct sk_buff *skb)
3235{ }
3236#endif
3237
574f7194
EB
3238static inline bool skb_irq_freeable(const struct sk_buff *skb)
3239{
3240 return !skb->destructor &&
3241#if IS_ENABLED(CONFIG_XFRM)
3242 !skb->sp &&
3243#endif
3244#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3245 !skb->nfct &&
3246#endif
3247 !skb->_skb_refdst &&
3248 !skb_has_frag_list(skb);
3249}
3250
f25f4e44
PWJ
3251static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3252{
f25f4e44 3253 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3254}
3255
9247744e 3256static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3257{
4e3ab47a 3258 return skb->queue_mapping;
4e3ab47a
PE
3259}
3260
f25f4e44
PWJ
3261static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3262{
f25f4e44 3263 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3264}
3265
d5a9e24a
DM
3266static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3267{
3268 skb->queue_mapping = rx_queue + 1;
3269}
3270
9247744e 3271static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3272{
3273 return skb->queue_mapping - 1;
3274}
3275
9247744e 3276static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3277{
a02cec21 3278 return skb->queue_mapping != 0;
d5a9e24a
DM
3279}
3280
0e001614 3281u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
7965bd4d 3282 unsigned int num_tx_queues);
9247744e 3283
def8b4fa
AD
3284static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3285{
0b3d8e08 3286#ifdef CONFIG_XFRM
def8b4fa 3287 return skb->sp;
def8b4fa 3288#else
def8b4fa 3289 return NULL;
def8b4fa 3290#endif
0b3d8e08 3291}
def8b4fa 3292
68c33163
PS
3293/* Keeps track of mac header offset relative to skb->head.
3294 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3295 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3296 * tunnel skb it points to outer mac header.
3297 * Keeps track of level of encapsulation of network headers.
3298 */
68c33163 3299struct skb_gso_cb {
3347c960
ED
3300 int mac_offset;
3301 int encap_level;
7e2b10c1 3302 __u16 csum_start;
68c33163
PS
3303};
3304#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3305
3306static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3307{
3308 return (skb_mac_header(inner_skb) - inner_skb->head) -
3309 SKB_GSO_CB(inner_skb)->mac_offset;
3310}
3311
1e2bd517
PS
3312static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3313{
3314 int new_headroom, headroom;
3315 int ret;
3316
3317 headroom = skb_headroom(skb);
3318 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3319 if (ret)
3320 return ret;
3321
3322 new_headroom = skb_headroom(skb);
3323 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3324 return 0;
3325}
3326
7e2b10c1
TH
3327/* Compute the checksum for a gso segment. First compute the checksum value
3328 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3329 * then add in skb->csum (checksum from csum_start to end of packet).
3330 * skb->csum and csum_start are then updated to reflect the checksum of the
3331 * resultant packet starting from the transport header-- the resultant checksum
3332 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3333 * header.
3334 */
3335static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3336{
3337 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3338 skb_transport_offset(skb);
3339 __u16 csum;
3340
3341 csum = csum_fold(csum_partial(skb_transport_header(skb),
3342 plen, skb->csum));
3343 skb->csum = res;
3344 SKB_GSO_CB(skb)->csum_start -= plen;
3345
3346 return csum;
3347}
3348
bdcc0924 3349static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3350{
3351 return skb_shinfo(skb)->gso_size;
3352}
3353
36a8f39e 3354/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3355static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3356{
3357 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3358}
3359
7965bd4d 3360void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3361
3362static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3363{
3364 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3365 * wanted then gso_type will be set. */
05bdd2f1
ED
3366 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3367
b78462eb
AD
3368 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3369 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3370 __skb_warn_lro_forwarding(skb);
3371 return true;
3372 }
3373 return false;
3374}
3375
35fc92a9
HX
3376static inline void skb_forward_csum(struct sk_buff *skb)
3377{
3378 /* Unfortunately we don't support this one. Any brave souls? */
3379 if (skb->ip_summed == CHECKSUM_COMPLETE)
3380 skb->ip_summed = CHECKSUM_NONE;
3381}
3382
bc8acf2c
ED
3383/**
3384 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3385 * @skb: skb to check
3386 *
3387 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3388 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3389 * use this helper, to document places where we make this assertion.
3390 */
05bdd2f1 3391static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3392{
3393#ifdef DEBUG
3394 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3395#endif
3396}
3397
f35d9d8a 3398bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3399
ed1f50c3
PD
3400int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3401
56193d1b
AD
3402u32 skb_get_poff(const struct sk_buff *skb);
3403u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3404 const struct flow_keys *keys, int hlen);
f77668dc 3405
3a7c1ee4
AD
3406/**
3407 * skb_head_is_locked - Determine if the skb->head is locked down
3408 * @skb: skb to check
3409 *
3410 * The head on skbs build around a head frag can be removed if they are
3411 * not cloned. This function returns true if the skb head is locked down
3412 * due to either being allocated via kmalloc, or by being a clone with
3413 * multiple references to the head.
3414 */
3415static inline bool skb_head_is_locked(const struct sk_buff *skb)
3416{
3417 return !skb->head_frag || skb_cloned(skb);
3418}
fe6cc55f
FW
3419
3420/**
3421 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3422 *
3423 * @skb: GSO skb
3424 *
3425 * skb_gso_network_seglen is used to determine the real size of the
3426 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3427 *
3428 * The MAC/L2 header is not accounted for.
3429 */
3430static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3431{
3432 unsigned int hdr_len = skb_transport_header(skb) -
3433 skb_network_header(skb);
3434 return hdr_len + skb_gso_transport_seglen(skb);
3435}
1da177e4
LT
3436#endif /* __KERNEL__ */
3437#endif /* _LINUX_SKBUFF_H */