]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/skbuff.h
netfilter: ipset: Use better include files in xt_set.c
[mirror_ubuntu-bionic-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 168 atomic_t use;
3eaf4025
FW
169 enum {
170 BRNF_PROTO_UNCHANGED,
171 BRNF_PROTO_8021Q,
172 BRNF_PROTO_PPPOE
173 } orig_proto;
a1e67951 174 bool pkt_otherhost;
bf1ac5ca
ED
175 unsigned int mask;
176 struct net_device *physindev;
177 struct net_device *physoutdev;
e70deecb 178 char neigh_header[8];
1da177e4
LT
179};
180#endif
181
1da177e4
LT
182struct sk_buff_head {
183 /* These two members must be first. */
184 struct sk_buff *next;
185 struct sk_buff *prev;
186
187 __u32 qlen;
188 spinlock_t lock;
189};
190
191struct sk_buff;
192
9d4dde52
IC
193/* To allow 64K frame to be packed as single skb without frag_list we
194 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
195 * buffers which do not start on a page boundary.
196 *
197 * Since GRO uses frags we allocate at least 16 regardless of page
198 * size.
a715dea3 199 */
9d4dde52 200#if (65536/PAGE_SIZE + 1) < 16
eec00954 201#define MAX_SKB_FRAGS 16UL
a715dea3 202#else
9d4dde52 203#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 204#endif
1da177e4
LT
205
206typedef struct skb_frag_struct skb_frag_t;
207
208struct skb_frag_struct {
a8605c60
IC
209 struct {
210 struct page *p;
211 } page;
cb4dfe56 212#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
213 __u32 page_offset;
214 __u32 size;
cb4dfe56
ED
215#else
216 __u16 page_offset;
217 __u16 size;
218#endif
1da177e4
LT
219};
220
9e903e08
ED
221static inline unsigned int skb_frag_size(const skb_frag_t *frag)
222{
223 return frag->size;
224}
225
226static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
227{
228 frag->size = size;
229}
230
231static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
232{
233 frag->size += delta;
234}
235
236static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
237{
238 frag->size -= delta;
239}
240
ac45f602
PO
241#define HAVE_HW_TIME_STAMP
242
243/**
d3a21be8 244 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
245 * @hwtstamp: hardware time stamp transformed into duration
246 * since arbitrary point in time
ac45f602
PO
247 *
248 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 249 * skb->tstamp.
ac45f602
PO
250 *
251 * hwtstamps can only be compared against other hwtstamps from
252 * the same device.
253 *
254 * This structure is attached to packets as part of the
255 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
256 */
257struct skb_shared_hwtstamps {
258 ktime_t hwtstamp;
ac45f602
PO
259};
260
2244d07b
OH
261/* Definitions for tx_flags in struct skb_shared_info */
262enum {
263 /* generate hardware time stamp */
264 SKBTX_HW_TSTAMP = 1 << 0,
265
e7fd2885 266 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
267 SKBTX_SW_TSTAMP = 1 << 1,
268
269 /* device driver is going to provide hardware time stamp */
270 SKBTX_IN_PROGRESS = 1 << 2,
271
a6686f2f 272 /* device driver supports TX zero-copy buffers */
62b1a8ab 273 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
274
275 /* generate wifi status information (where possible) */
62b1a8ab 276 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
277
278 /* This indicates at least one fragment might be overwritten
279 * (as in vmsplice(), sendfile() ...)
280 * If we need to compute a TX checksum, we'll need to copy
281 * all frags to avoid possible bad checksum
282 */
283 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
284
285 /* generate software time stamp when entering packet scheduling */
286 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
287
288 /* generate software timestamp on peer data acknowledgment */
289 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
290};
291
e1c8a607
WB
292#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
293 SKBTX_SCHED_TSTAMP | \
294 SKBTX_ACK_TSTAMP)
f24b9be5
WB
295#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
296
a6686f2f
SM
297/*
298 * The callback notifies userspace to release buffers when skb DMA is done in
299 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
300 * The zerocopy_success argument is true if zero copy transmit occurred,
301 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
302 * The ctx field is used to track device context.
303 * The desc field is used to track userspace buffer index.
a6686f2f
SM
304 */
305struct ubuf_info {
e19d6763 306 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 307 void *ctx;
a6686f2f 308 unsigned long desc;
ac45f602
PO
309};
310
1da177e4
LT
311/* This data is invariant across clones and lives at
312 * the end of the header data, ie. at skb->end.
313 */
314struct skb_shared_info {
9f42f126
IC
315 unsigned char nr_frags;
316 __u8 tx_flags;
7967168c
HX
317 unsigned short gso_size;
318 /* Warning: this field is not always filled in (UFO)! */
319 unsigned short gso_segs;
320 unsigned short gso_type;
1da177e4 321 struct sk_buff *frag_list;
ac45f602 322 struct skb_shared_hwtstamps hwtstamps;
09c2d251 323 u32 tskey;
9f42f126 324 __be32 ip6_frag_id;
ec7d2f2c
ED
325
326 /*
327 * Warning : all fields before dataref are cleared in __alloc_skb()
328 */
329 atomic_t dataref;
330
69e3c75f
JB
331 /* Intermediate layers must ensure that destructor_arg
332 * remains valid until skb destructor */
333 void * destructor_arg;
a6686f2f 334
fed66381
ED
335 /* must be last field, see pskb_expand_head() */
336 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
337};
338
339/* We divide dataref into two halves. The higher 16 bits hold references
340 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
341 * the entire skb->data. A clone of a headerless skb holds the length of
342 * the header in skb->hdr_len.
1da177e4
LT
343 *
344 * All users must obey the rule that the skb->data reference count must be
345 * greater than or equal to the payload reference count.
346 *
347 * Holding a reference to the payload part means that the user does not
348 * care about modifications to the header part of skb->data.
349 */
350#define SKB_DATAREF_SHIFT 16
351#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
352
d179cd12
DM
353
354enum {
c8753d55
VS
355 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
356 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
357 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
358};
359
7967168c
HX
360enum {
361 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 362 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
363
364 /* This indicates the skb is from an untrusted source. */
365 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
366
367 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
368 SKB_GSO_TCP_ECN = 1 << 3,
369
370 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
371
372 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
373
374 SKB_GSO_GRE = 1 << 6,
73136267 375
4b28252c 376 SKB_GSO_GRE_CSUM = 1 << 7,
0d89d203 377
4b28252c 378 SKB_GSO_IPIP = 1 << 8,
cb32f511 379
4b28252c 380 SKB_GSO_SIT = 1 << 9,
61c1db7f 381
4b28252c 382 SKB_GSO_UDP_TUNNEL = 1 << 10,
0f4f4ffa
TH
383
384 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c 385
59b93b41 386 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
7967168c
HX
387};
388
2e07fa9c
ACM
389#if BITS_PER_LONG > 32
390#define NET_SKBUFF_DATA_USES_OFFSET 1
391#endif
392
393#ifdef NET_SKBUFF_DATA_USES_OFFSET
394typedef unsigned int sk_buff_data_t;
395#else
396typedef unsigned char *sk_buff_data_t;
397#endif
398
363ec392
ED
399/**
400 * struct skb_mstamp - multi resolution time stamps
401 * @stamp_us: timestamp in us resolution
402 * @stamp_jiffies: timestamp in jiffies
403 */
404struct skb_mstamp {
405 union {
406 u64 v64;
407 struct {
408 u32 stamp_us;
409 u32 stamp_jiffies;
410 };
411 };
412};
413
414/**
415 * skb_mstamp_get - get current timestamp
416 * @cl: place to store timestamps
417 */
418static inline void skb_mstamp_get(struct skb_mstamp *cl)
419{
420 u64 val = local_clock();
421
422 do_div(val, NSEC_PER_USEC);
423 cl->stamp_us = (u32)val;
424 cl->stamp_jiffies = (u32)jiffies;
425}
426
427/**
428 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
429 * @t1: pointer to newest sample
430 * @t0: pointer to oldest sample
431 */
432static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
433 const struct skb_mstamp *t0)
434{
435 s32 delta_us = t1->stamp_us - t0->stamp_us;
436 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
437
438 /* If delta_us is negative, this might be because interval is too big,
439 * or local_clock() drift is too big : fallback using jiffies.
440 */
441 if (delta_us <= 0 ||
442 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
443
444 delta_us = jiffies_to_usecs(delta_jiffies);
445
446 return delta_us;
447}
448
449
1da177e4
LT
450/**
451 * struct sk_buff - socket buffer
452 * @next: Next buffer in list
453 * @prev: Previous buffer in list
363ec392 454 * @tstamp: Time we arrived/left
56b17425 455 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 456 * @sk: Socket we are owned by
1da177e4 457 * @dev: Device we arrived on/are leaving by
d84e0bd7 458 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 459 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 460 * @sp: the security path, used for xfrm
1da177e4
LT
461 * @len: Length of actual data
462 * @data_len: Data length
463 * @mac_len: Length of link layer header
334a8132 464 * @hdr_len: writable header length of cloned skb
663ead3b
HX
465 * @csum: Checksum (must include start/offset pair)
466 * @csum_start: Offset from skb->head where checksumming should start
467 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 468 * @priority: Packet queueing priority
60ff7467 469 * @ignore_df: allow local fragmentation
1da177e4 470 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 471 * @ip_summed: Driver fed us an IP checksum
1da177e4 472 * @nohdr: Payload reference only, must not modify header
d84e0bd7 473 * @nfctinfo: Relationship of this skb to the connection
1da177e4 474 * @pkt_type: Packet class
c83c2486 475 * @fclone: skbuff clone status
c83c2486 476 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
477 * @peeked: this packet has been seen already, so stats have been
478 * done for it, don't do them again
ba9dda3a 479 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
480 * @protocol: Packet protocol from driver
481 * @destructor: Destruct function
482 * @nfct: Associated connection, if any
1da177e4 483 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 484 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
485 * @tc_index: Traffic control index
486 * @tc_verd: traffic control verdict
61b905da 487 * @hash: the packet hash
d84e0bd7 488 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 489 * @xmit_more: More SKBs are pending for this queue
553a5672 490 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 491 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 492 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 493 * ports.
a3b18ddb 494 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
495 * @wifi_acked_valid: wifi_acked was set
496 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 497 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
06021292 498 * @napi_id: id of the NAPI struct this skb came from
984bc16c 499 * @secmark: security marking
d84e0bd7 500 * @mark: Generic packet mark
86a9bad3 501 * @vlan_proto: vlan encapsulation protocol
6aa895b0 502 * @vlan_tci: vlan tag control information
0d89d203 503 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
504 * @inner_transport_header: Inner transport layer header (encapsulation)
505 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 506 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
507 * @transport_header: Transport layer header
508 * @network_header: Network layer header
509 * @mac_header: Link layer header
510 * @tail: Tail pointer
511 * @end: End pointer
512 * @head: Head of buffer
513 * @data: Data head pointer
514 * @truesize: Buffer size
515 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
516 */
517
518struct sk_buff {
363ec392 519 union {
56b17425
ED
520 struct {
521 /* These two members must be first. */
522 struct sk_buff *next;
523 struct sk_buff *prev;
524
525 union {
526 ktime_t tstamp;
527 struct skb_mstamp skb_mstamp;
528 };
529 };
530 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 531 };
da3f5cf1 532 struct sock *sk;
1da177e4 533 struct net_device *dev;
1da177e4 534
1da177e4
LT
535 /*
536 * This is the control buffer. It is free to use for every
537 * layer. Please put your private variables there. If you
538 * want to keep them across layers you have to do a skb_clone()
539 * first. This is owned by whoever has the skb queued ATM.
540 */
da3f5cf1 541 char cb[48] __aligned(8);
1da177e4 542
7fee226a 543 unsigned long _skb_refdst;
b1937227 544 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
545#ifdef CONFIG_XFRM
546 struct sec_path *sp;
b1937227
ED
547#endif
548#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
549 struct nf_conntrack *nfct;
550#endif
85224844 551#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 552 struct nf_bridge_info *nf_bridge;
da3f5cf1 553#endif
1da177e4 554 unsigned int len,
334a8132
PM
555 data_len;
556 __u16 mac_len,
557 hdr_len;
b1937227
ED
558
559 /* Following fields are _not_ copied in __copy_skb_header()
560 * Note that queue_mapping is here mostly to fill a hole.
561 */
fe55f6d5 562 kmemcheck_bitfield_begin(flags1);
b1937227
ED
563 __u16 queue_mapping;
564 __u8 cloned:1,
6869c4d8 565 nohdr:1,
b84f4cc9 566 fclone:2,
a59322be 567 peeked:1,
b1937227
ED
568 head_frag:1,
569 xmit_more:1;
570 /* one bit hole */
fe55f6d5 571 kmemcheck_bitfield_end(flags1);
4031ae6e 572
b1937227
ED
573 /* fields enclosed in headers_start/headers_end are copied
574 * using a single memcpy() in __copy_skb_header()
575 */
ebcf34f3 576 /* private: */
b1937227 577 __u32 headers_start[0];
ebcf34f3 578 /* public: */
4031ae6e 579
233577a2
HFS
580/* if you move pkt_type around you also must adapt those constants */
581#ifdef __BIG_ENDIAN_BITFIELD
582#define PKT_TYPE_MAX (7 << 5)
583#else
584#define PKT_TYPE_MAX 7
1da177e4 585#endif
233577a2 586#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 587
233577a2 588 __u8 __pkt_type_offset[0];
b1937227 589 __u8 pkt_type:3;
c93bdd0e 590 __u8 pfmemalloc:1;
b1937227
ED
591 __u8 ignore_df:1;
592 __u8 nfctinfo:3;
593
594 __u8 nf_trace:1;
595 __u8 ip_summed:2;
3853b584 596 __u8 ooo_okay:1;
61b905da 597 __u8 l4_hash:1;
a3b18ddb 598 __u8 sw_hash:1;
6e3e939f
JB
599 __u8 wifi_acked_valid:1;
600 __u8 wifi_acked:1;
b1937227 601
3bdc0eba 602 __u8 no_fcs:1;
77cffe23 603 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 604 __u8 encapsulation:1;
7e2b10c1 605 __u8 encap_hdr_csum:1;
5d0c2b95 606 __u8 csum_valid:1;
7e3cead5 607 __u8 csum_complete_sw:1;
b1937227
ED
608 __u8 csum_level:2;
609 __u8 csum_bad:1;
fe55f6d5 610
b1937227
ED
611#ifdef CONFIG_IPV6_NDISC_NODETYPE
612 __u8 ndisc_nodetype:2;
613#endif
614 __u8 ipvs_property:1;
8bce6d7d 615 __u8 inner_protocol_type:1;
e585f236
TH
616 __u8 remcsum_offload:1;
617 /* 3 or 5 bit hole */
b1937227
ED
618
619#ifdef CONFIG_NET_SCHED
620 __u16 tc_index; /* traffic control index */
621#ifdef CONFIG_NET_CLS_ACT
622 __u16 tc_verd; /* traffic control verdict */
623#endif
624#endif
fe55f6d5 625
b1937227
ED
626 union {
627 __wsum csum;
628 struct {
629 __u16 csum_start;
630 __u16 csum_offset;
631 };
632 };
633 __u32 priority;
634 int skb_iif;
635 __u32 hash;
636 __be16 vlan_proto;
637 __u16 vlan_tci;
2bd82484
ED
638#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
639 union {
640 unsigned int napi_id;
641 unsigned int sender_cpu;
642 };
97fc2f08 643#endif
984bc16c
JM
644#ifdef CONFIG_NETWORK_SECMARK
645 __u32 secmark;
646#endif
3b885787
NH
647 union {
648 __u32 mark;
16fad69c 649 __u32 reserved_tailroom;
3b885787 650 };
1da177e4 651
8bce6d7d
TH
652 union {
653 __be16 inner_protocol;
654 __u8 inner_ipproto;
655 };
656
1a37e412
SH
657 __u16 inner_transport_header;
658 __u16 inner_network_header;
659 __u16 inner_mac_header;
b1937227
ED
660
661 __be16 protocol;
1a37e412
SH
662 __u16 transport_header;
663 __u16 network_header;
664 __u16 mac_header;
b1937227 665
ebcf34f3 666 /* private: */
b1937227 667 __u32 headers_end[0];
ebcf34f3 668 /* public: */
b1937227 669
1da177e4 670 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 671 sk_buff_data_t tail;
4305b541 672 sk_buff_data_t end;
1da177e4 673 unsigned char *head,
4305b541 674 *data;
27a884dc
ACM
675 unsigned int truesize;
676 atomic_t users;
1da177e4
LT
677};
678
679#ifdef __KERNEL__
680/*
681 * Handling routines are only of interest to the kernel
682 */
683#include <linux/slab.h>
684
1da177e4 685
c93bdd0e
MG
686#define SKB_ALLOC_FCLONE 0x01
687#define SKB_ALLOC_RX 0x02
fd11a83d 688#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
689
690/* Returns true if the skb was allocated from PFMEMALLOC reserves */
691static inline bool skb_pfmemalloc(const struct sk_buff *skb)
692{
693 return unlikely(skb->pfmemalloc);
694}
695
7fee226a
ED
696/*
697 * skb might have a dst pointer attached, refcounted or not.
698 * _skb_refdst low order bit is set if refcount was _not_ taken
699 */
700#define SKB_DST_NOREF 1UL
701#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
702
703/**
704 * skb_dst - returns skb dst_entry
705 * @skb: buffer
706 *
707 * Returns skb dst_entry, regardless of reference taken or not.
708 */
adf30907
ED
709static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
710{
7fee226a
ED
711 /* If refdst was not refcounted, check we still are in a
712 * rcu_read_lock section
713 */
714 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
715 !rcu_read_lock_held() &&
716 !rcu_read_lock_bh_held());
717 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
718}
719
7fee226a
ED
720/**
721 * skb_dst_set - sets skb dst
722 * @skb: buffer
723 * @dst: dst entry
724 *
725 * Sets skb dst, assuming a reference was taken on dst and should
726 * be released by skb_dst_drop()
727 */
adf30907
ED
728static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
729{
7fee226a
ED
730 skb->_skb_refdst = (unsigned long)dst;
731}
732
932bc4d7
JA
733/**
734 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
735 * @skb: buffer
736 * @dst: dst entry
737 *
738 * Sets skb dst, assuming a reference was not taken on dst.
739 * If dst entry is cached, we do not take reference and dst_release
740 * will be avoided by refdst_drop. If dst entry is not cached, we take
741 * reference, so that last dst_release can destroy the dst immediately.
742 */
743static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
744{
dbfc4fb7
HFS
745 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
746 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 747}
7fee226a
ED
748
749/**
25985edc 750 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
751 * @skb: buffer
752 */
753static inline bool skb_dst_is_noref(const struct sk_buff *skb)
754{
755 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
756}
757
511c3f92
ED
758static inline struct rtable *skb_rtable(const struct sk_buff *skb)
759{
adf30907 760 return (struct rtable *)skb_dst(skb);
511c3f92
ED
761}
762
7965bd4d
JP
763void kfree_skb(struct sk_buff *skb);
764void kfree_skb_list(struct sk_buff *segs);
765void skb_tx_error(struct sk_buff *skb);
766void consume_skb(struct sk_buff *skb);
767void __kfree_skb(struct sk_buff *skb);
d7e8883c 768extern struct kmem_cache *skbuff_head_cache;
bad43ca8 769
7965bd4d
JP
770void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
771bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
772 bool *fragstolen, int *delta_truesize);
bad43ca8 773
7965bd4d
JP
774struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
775 int node);
2ea2f62c 776struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 777struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 778static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 779 gfp_t priority)
d179cd12 780{
564824b0 781 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
782}
783
2e4e4410
ED
784struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
785 unsigned long data_len,
786 int max_page_order,
787 int *errcode,
788 gfp_t gfp_mask);
789
d0bf4a9e
ED
790/* Layout of fast clones : [skb1][skb2][fclone_ref] */
791struct sk_buff_fclones {
792 struct sk_buff skb1;
793
794 struct sk_buff skb2;
795
796 atomic_t fclone_ref;
797};
798
799/**
800 * skb_fclone_busy - check if fclone is busy
801 * @skb: buffer
802 *
803 * Returns true is skb is a fast clone, and its clone is not freed.
39bb5e62
ED
804 * Some drivers call skb_orphan() in their ndo_start_xmit(),
805 * so we also check that this didnt happen.
d0bf4a9e 806 */
39bb5e62
ED
807static inline bool skb_fclone_busy(const struct sock *sk,
808 const struct sk_buff *skb)
d0bf4a9e
ED
809{
810 const struct sk_buff_fclones *fclones;
811
812 fclones = container_of(skb, struct sk_buff_fclones, skb1);
813
814 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 815 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 816 fclones->skb2.sk == sk;
d0bf4a9e
ED
817}
818
d179cd12 819static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 820 gfp_t priority)
d179cd12 821{
c93bdd0e 822 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
823}
824
7965bd4d 825struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
826static inline struct sk_buff *alloc_skb_head(gfp_t priority)
827{
828 return __alloc_skb_head(priority, -1);
829}
830
7965bd4d
JP
831struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
832int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
833struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
834struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
835struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
836 gfp_t gfp_mask, bool fclone);
837static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
838 gfp_t gfp_mask)
839{
840 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
841}
7965bd4d
JP
842
843int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
844struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
845 unsigned int headroom);
846struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
847 int newtailroom, gfp_t priority);
25a91d8d
FD
848int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
849 int offset, int len);
7965bd4d
JP
850int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
851 int len);
852int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
853int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 854#define dev_kfree_skb(a) consume_skb(a)
1da177e4 855
7965bd4d
JP
856int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
857 int getfrag(void *from, char *to, int offset,
858 int len, int odd, struct sk_buff *skb),
859 void *from, int length);
e89e9cf5 860
d94d9fee 861struct skb_seq_state {
677e90ed
TG
862 __u32 lower_offset;
863 __u32 upper_offset;
864 __u32 frag_idx;
865 __u32 stepped_offset;
866 struct sk_buff *root_skb;
867 struct sk_buff *cur_skb;
868 __u8 *frag_data;
869};
870
7965bd4d
JP
871void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
872 unsigned int to, struct skb_seq_state *st);
873unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
874 struct skb_seq_state *st);
875void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 876
7965bd4d 877unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 878 unsigned int to, struct ts_config *config);
3fc7e8a6 879
09323cc4
TH
880/*
881 * Packet hash types specify the type of hash in skb_set_hash.
882 *
883 * Hash types refer to the protocol layer addresses which are used to
884 * construct a packet's hash. The hashes are used to differentiate or identify
885 * flows of the protocol layer for the hash type. Hash types are either
886 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
887 *
888 * Properties of hashes:
889 *
890 * 1) Two packets in different flows have different hash values
891 * 2) Two packets in the same flow should have the same hash value
892 *
893 * A hash at a higher layer is considered to be more specific. A driver should
894 * set the most specific hash possible.
895 *
896 * A driver cannot indicate a more specific hash than the layer at which a hash
897 * was computed. For instance an L3 hash cannot be set as an L4 hash.
898 *
899 * A driver may indicate a hash level which is less specific than the
900 * actual layer the hash was computed on. For instance, a hash computed
901 * at L4 may be considered an L3 hash. This should only be done if the
902 * driver can't unambiguously determine that the HW computed the hash at
903 * the higher layer. Note that the "should" in the second property above
904 * permits this.
905 */
906enum pkt_hash_types {
907 PKT_HASH_TYPE_NONE, /* Undefined type */
908 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
909 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
910 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
911};
912
913static inline void
914skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
915{
61b905da 916 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 917 skb->sw_hash = 0;
61b905da 918 skb->hash = hash;
09323cc4
TH
919}
920
3958afa1
TH
921void __skb_get_hash(struct sk_buff *skb);
922static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 923{
a3b18ddb 924 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 925 __skb_get_hash(skb);
bfb564e7 926
61b905da 927 return skb->hash;
bfb564e7
KK
928}
929
50fb7992
TH
930__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
931
57bdf7f4
TH
932static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
933{
61b905da 934 return skb->hash;
57bdf7f4
TH
935}
936
7539fadc
TH
937static inline void skb_clear_hash(struct sk_buff *skb)
938{
61b905da 939 skb->hash = 0;
a3b18ddb 940 skb->sw_hash = 0;
61b905da 941 skb->l4_hash = 0;
7539fadc
TH
942}
943
944static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
945{
61b905da 946 if (!skb->l4_hash)
7539fadc
TH
947 skb_clear_hash(skb);
948}
949
3df7a74e
TH
950static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
951{
61b905da 952 to->hash = from->hash;
a3b18ddb 953 to->sw_hash = from->sw_hash;
61b905da 954 to->l4_hash = from->l4_hash;
3df7a74e
TH
955};
956
c29390c6
ED
957static inline void skb_sender_cpu_clear(struct sk_buff *skb)
958{
959#ifdef CONFIG_XPS
960 skb->sender_cpu = 0;
961#endif
962}
963
4305b541
ACM
964#ifdef NET_SKBUFF_DATA_USES_OFFSET
965static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
966{
967 return skb->head + skb->end;
968}
ec47ea82
AD
969
970static inline unsigned int skb_end_offset(const struct sk_buff *skb)
971{
972 return skb->end;
973}
4305b541
ACM
974#else
975static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
976{
977 return skb->end;
978}
ec47ea82
AD
979
980static inline unsigned int skb_end_offset(const struct sk_buff *skb)
981{
982 return skb->end - skb->head;
983}
4305b541
ACM
984#endif
985
1da177e4 986/* Internal */
4305b541 987#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 988
ac45f602
PO
989static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
990{
991 return &skb_shinfo(skb)->hwtstamps;
992}
993
1da177e4
LT
994/**
995 * skb_queue_empty - check if a queue is empty
996 * @list: queue head
997 *
998 * Returns true if the queue is empty, false otherwise.
999 */
1000static inline int skb_queue_empty(const struct sk_buff_head *list)
1001{
fd44b93c 1002 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1003}
1004
fc7ebb21
DM
1005/**
1006 * skb_queue_is_last - check if skb is the last entry in the queue
1007 * @list: queue head
1008 * @skb: buffer
1009 *
1010 * Returns true if @skb is the last buffer on the list.
1011 */
1012static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1013 const struct sk_buff *skb)
1014{
fd44b93c 1015 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1016}
1017
832d11c5
IJ
1018/**
1019 * skb_queue_is_first - check if skb is the first entry in the queue
1020 * @list: queue head
1021 * @skb: buffer
1022 *
1023 * Returns true if @skb is the first buffer on the list.
1024 */
1025static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1026 const struct sk_buff *skb)
1027{
fd44b93c 1028 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1029}
1030
249c8b42
DM
1031/**
1032 * skb_queue_next - return the next packet in the queue
1033 * @list: queue head
1034 * @skb: current buffer
1035 *
1036 * Return the next packet in @list after @skb. It is only valid to
1037 * call this if skb_queue_is_last() evaluates to false.
1038 */
1039static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1040 const struct sk_buff *skb)
1041{
1042 /* This BUG_ON may seem severe, but if we just return then we
1043 * are going to dereference garbage.
1044 */
1045 BUG_ON(skb_queue_is_last(list, skb));
1046 return skb->next;
1047}
1048
832d11c5
IJ
1049/**
1050 * skb_queue_prev - return the prev packet in the queue
1051 * @list: queue head
1052 * @skb: current buffer
1053 *
1054 * Return the prev packet in @list before @skb. It is only valid to
1055 * call this if skb_queue_is_first() evaluates to false.
1056 */
1057static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1058 const struct sk_buff *skb)
1059{
1060 /* This BUG_ON may seem severe, but if we just return then we
1061 * are going to dereference garbage.
1062 */
1063 BUG_ON(skb_queue_is_first(list, skb));
1064 return skb->prev;
1065}
1066
1da177e4
LT
1067/**
1068 * skb_get - reference buffer
1069 * @skb: buffer to reference
1070 *
1071 * Makes another reference to a socket buffer and returns a pointer
1072 * to the buffer.
1073 */
1074static inline struct sk_buff *skb_get(struct sk_buff *skb)
1075{
1076 atomic_inc(&skb->users);
1077 return skb;
1078}
1079
1080/*
1081 * If users == 1, we are the only owner and are can avoid redundant
1082 * atomic change.
1083 */
1084
1da177e4
LT
1085/**
1086 * skb_cloned - is the buffer a clone
1087 * @skb: buffer to check
1088 *
1089 * Returns true if the buffer was generated with skb_clone() and is
1090 * one of multiple shared copies of the buffer. Cloned buffers are
1091 * shared data so must not be written to under normal circumstances.
1092 */
1093static inline int skb_cloned(const struct sk_buff *skb)
1094{
1095 return skb->cloned &&
1096 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1097}
1098
14bbd6a5
PS
1099static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1100{
1101 might_sleep_if(pri & __GFP_WAIT);
1102
1103 if (skb_cloned(skb))
1104 return pskb_expand_head(skb, 0, 0, pri);
1105
1106 return 0;
1107}
1108
1da177e4
LT
1109/**
1110 * skb_header_cloned - is the header a clone
1111 * @skb: buffer to check
1112 *
1113 * Returns true if modifying the header part of the buffer requires
1114 * the data to be copied.
1115 */
1116static inline int skb_header_cloned(const struct sk_buff *skb)
1117{
1118 int dataref;
1119
1120 if (!skb->cloned)
1121 return 0;
1122
1123 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1124 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1125 return dataref != 1;
1126}
1127
1128/**
1129 * skb_header_release - release reference to header
1130 * @skb: buffer to operate on
1131 *
1132 * Drop a reference to the header part of the buffer. This is done
1133 * by acquiring a payload reference. You must not read from the header
1134 * part of skb->data after this.
f4a775d1 1135 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1136 */
1137static inline void skb_header_release(struct sk_buff *skb)
1138{
1139 BUG_ON(skb->nohdr);
1140 skb->nohdr = 1;
1141 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1142}
1143
f4a775d1
ED
1144/**
1145 * __skb_header_release - release reference to header
1146 * @skb: buffer to operate on
1147 *
1148 * Variant of skb_header_release() assuming skb is private to caller.
1149 * We can avoid one atomic operation.
1150 */
1151static inline void __skb_header_release(struct sk_buff *skb)
1152{
1153 skb->nohdr = 1;
1154 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1155}
1156
1157
1da177e4
LT
1158/**
1159 * skb_shared - is the buffer shared
1160 * @skb: buffer to check
1161 *
1162 * Returns true if more than one person has a reference to this
1163 * buffer.
1164 */
1165static inline int skb_shared(const struct sk_buff *skb)
1166{
1167 return atomic_read(&skb->users) != 1;
1168}
1169
1170/**
1171 * skb_share_check - check if buffer is shared and if so clone it
1172 * @skb: buffer to check
1173 * @pri: priority for memory allocation
1174 *
1175 * If the buffer is shared the buffer is cloned and the old copy
1176 * drops a reference. A new clone with a single reference is returned.
1177 * If the buffer is not shared the original buffer is returned. When
1178 * being called from interrupt status or with spinlocks held pri must
1179 * be GFP_ATOMIC.
1180 *
1181 * NULL is returned on a memory allocation failure.
1182 */
47061bc4 1183static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1184{
1185 might_sleep_if(pri & __GFP_WAIT);
1186 if (skb_shared(skb)) {
1187 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1188
1189 if (likely(nskb))
1190 consume_skb(skb);
1191 else
1192 kfree_skb(skb);
1da177e4
LT
1193 skb = nskb;
1194 }
1195 return skb;
1196}
1197
1198/*
1199 * Copy shared buffers into a new sk_buff. We effectively do COW on
1200 * packets to handle cases where we have a local reader and forward
1201 * and a couple of other messy ones. The normal one is tcpdumping
1202 * a packet thats being forwarded.
1203 */
1204
1205/**
1206 * skb_unshare - make a copy of a shared buffer
1207 * @skb: buffer to check
1208 * @pri: priority for memory allocation
1209 *
1210 * If the socket buffer is a clone then this function creates a new
1211 * copy of the data, drops a reference count on the old copy and returns
1212 * the new copy with the reference count at 1. If the buffer is not a clone
1213 * the original buffer is returned. When called with a spinlock held or
1214 * from interrupt state @pri must be %GFP_ATOMIC
1215 *
1216 * %NULL is returned on a memory allocation failure.
1217 */
e2bf521d 1218static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1219 gfp_t pri)
1da177e4
LT
1220{
1221 might_sleep_if(pri & __GFP_WAIT);
1222 if (skb_cloned(skb)) {
1223 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1224
1225 /* Free our shared copy */
1226 if (likely(nskb))
1227 consume_skb(skb);
1228 else
1229 kfree_skb(skb);
1da177e4
LT
1230 skb = nskb;
1231 }
1232 return skb;
1233}
1234
1235/**
1a5778aa 1236 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1237 * @list_: list to peek at
1238 *
1239 * Peek an &sk_buff. Unlike most other operations you _MUST_
1240 * be careful with this one. A peek leaves the buffer on the
1241 * list and someone else may run off with it. You must hold
1242 * the appropriate locks or have a private queue to do this.
1243 *
1244 * Returns %NULL for an empty list or a pointer to the head element.
1245 * The reference count is not incremented and the reference is therefore
1246 * volatile. Use with caution.
1247 */
05bdd2f1 1248static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1249{
18d07000
ED
1250 struct sk_buff *skb = list_->next;
1251
1252 if (skb == (struct sk_buff *)list_)
1253 skb = NULL;
1254 return skb;
1da177e4
LT
1255}
1256
da5ef6e5
PE
1257/**
1258 * skb_peek_next - peek skb following the given one from a queue
1259 * @skb: skb to start from
1260 * @list_: list to peek at
1261 *
1262 * Returns %NULL when the end of the list is met or a pointer to the
1263 * next element. The reference count is not incremented and the
1264 * reference is therefore volatile. Use with caution.
1265 */
1266static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1267 const struct sk_buff_head *list_)
1268{
1269 struct sk_buff *next = skb->next;
18d07000 1270
da5ef6e5
PE
1271 if (next == (struct sk_buff *)list_)
1272 next = NULL;
1273 return next;
1274}
1275
1da177e4 1276/**
1a5778aa 1277 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1278 * @list_: list to peek at
1279 *
1280 * Peek an &sk_buff. Unlike most other operations you _MUST_
1281 * be careful with this one. A peek leaves the buffer on the
1282 * list and someone else may run off with it. You must hold
1283 * the appropriate locks or have a private queue to do this.
1284 *
1285 * Returns %NULL for an empty list or a pointer to the tail element.
1286 * The reference count is not incremented and the reference is therefore
1287 * volatile. Use with caution.
1288 */
05bdd2f1 1289static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1290{
18d07000
ED
1291 struct sk_buff *skb = list_->prev;
1292
1293 if (skb == (struct sk_buff *)list_)
1294 skb = NULL;
1295 return skb;
1296
1da177e4
LT
1297}
1298
1299/**
1300 * skb_queue_len - get queue length
1301 * @list_: list to measure
1302 *
1303 * Return the length of an &sk_buff queue.
1304 */
1305static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1306{
1307 return list_->qlen;
1308}
1309
67fed459
DM
1310/**
1311 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1312 * @list: queue to initialize
1313 *
1314 * This initializes only the list and queue length aspects of
1315 * an sk_buff_head object. This allows to initialize the list
1316 * aspects of an sk_buff_head without reinitializing things like
1317 * the spinlock. It can also be used for on-stack sk_buff_head
1318 * objects where the spinlock is known to not be used.
1319 */
1320static inline void __skb_queue_head_init(struct sk_buff_head *list)
1321{
1322 list->prev = list->next = (struct sk_buff *)list;
1323 list->qlen = 0;
1324}
1325
76f10ad0
AV
1326/*
1327 * This function creates a split out lock class for each invocation;
1328 * this is needed for now since a whole lot of users of the skb-queue
1329 * infrastructure in drivers have different locking usage (in hardirq)
1330 * than the networking core (in softirq only). In the long run either the
1331 * network layer or drivers should need annotation to consolidate the
1332 * main types of usage into 3 classes.
1333 */
1da177e4
LT
1334static inline void skb_queue_head_init(struct sk_buff_head *list)
1335{
1336 spin_lock_init(&list->lock);
67fed459 1337 __skb_queue_head_init(list);
1da177e4
LT
1338}
1339
c2ecba71
PE
1340static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1341 struct lock_class_key *class)
1342{
1343 skb_queue_head_init(list);
1344 lockdep_set_class(&list->lock, class);
1345}
1346
1da177e4 1347/*
bf299275 1348 * Insert an sk_buff on a list.
1da177e4
LT
1349 *
1350 * The "__skb_xxxx()" functions are the non-atomic ones that
1351 * can only be called with interrupts disabled.
1352 */
7965bd4d
JP
1353void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1354 struct sk_buff_head *list);
bf299275
GR
1355static inline void __skb_insert(struct sk_buff *newsk,
1356 struct sk_buff *prev, struct sk_buff *next,
1357 struct sk_buff_head *list)
1358{
1359 newsk->next = next;
1360 newsk->prev = prev;
1361 next->prev = prev->next = newsk;
1362 list->qlen++;
1363}
1da177e4 1364
67fed459
DM
1365static inline void __skb_queue_splice(const struct sk_buff_head *list,
1366 struct sk_buff *prev,
1367 struct sk_buff *next)
1368{
1369 struct sk_buff *first = list->next;
1370 struct sk_buff *last = list->prev;
1371
1372 first->prev = prev;
1373 prev->next = first;
1374
1375 last->next = next;
1376 next->prev = last;
1377}
1378
1379/**
1380 * skb_queue_splice - join two skb lists, this is designed for stacks
1381 * @list: the new list to add
1382 * @head: the place to add it in the first list
1383 */
1384static inline void skb_queue_splice(const struct sk_buff_head *list,
1385 struct sk_buff_head *head)
1386{
1387 if (!skb_queue_empty(list)) {
1388 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1389 head->qlen += list->qlen;
67fed459
DM
1390 }
1391}
1392
1393/**
d9619496 1394 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1395 * @list: the new list to add
1396 * @head: the place to add it in the first list
1397 *
1398 * The list at @list is reinitialised
1399 */
1400static inline void skb_queue_splice_init(struct sk_buff_head *list,
1401 struct sk_buff_head *head)
1402{
1403 if (!skb_queue_empty(list)) {
1404 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1405 head->qlen += list->qlen;
67fed459
DM
1406 __skb_queue_head_init(list);
1407 }
1408}
1409
1410/**
1411 * skb_queue_splice_tail - join two skb lists, each list being a queue
1412 * @list: the new list to add
1413 * @head: the place to add it in the first list
1414 */
1415static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1416 struct sk_buff_head *head)
1417{
1418 if (!skb_queue_empty(list)) {
1419 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1420 head->qlen += list->qlen;
67fed459
DM
1421 }
1422}
1423
1424/**
d9619496 1425 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1426 * @list: the new list to add
1427 * @head: the place to add it in the first list
1428 *
1429 * Each of the lists is a queue.
1430 * The list at @list is reinitialised
1431 */
1432static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1433 struct sk_buff_head *head)
1434{
1435 if (!skb_queue_empty(list)) {
1436 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1437 head->qlen += list->qlen;
67fed459
DM
1438 __skb_queue_head_init(list);
1439 }
1440}
1441
1da177e4 1442/**
300ce174 1443 * __skb_queue_after - queue a buffer at the list head
1da177e4 1444 * @list: list to use
300ce174 1445 * @prev: place after this buffer
1da177e4
LT
1446 * @newsk: buffer to queue
1447 *
300ce174 1448 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1449 * and you must therefore hold required locks before calling it.
1450 *
1451 * A buffer cannot be placed on two lists at the same time.
1452 */
300ce174
SH
1453static inline void __skb_queue_after(struct sk_buff_head *list,
1454 struct sk_buff *prev,
1455 struct sk_buff *newsk)
1da177e4 1456{
bf299275 1457 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1458}
1459
7965bd4d
JP
1460void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1461 struct sk_buff_head *list);
7de6c033 1462
f5572855
GR
1463static inline void __skb_queue_before(struct sk_buff_head *list,
1464 struct sk_buff *next,
1465 struct sk_buff *newsk)
1466{
1467 __skb_insert(newsk, next->prev, next, list);
1468}
1469
300ce174
SH
1470/**
1471 * __skb_queue_head - queue a buffer at the list head
1472 * @list: list to use
1473 * @newsk: buffer to queue
1474 *
1475 * Queue a buffer at the start of a list. This function takes no locks
1476 * and you must therefore hold required locks before calling it.
1477 *
1478 * A buffer cannot be placed on two lists at the same time.
1479 */
7965bd4d 1480void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1481static inline void __skb_queue_head(struct sk_buff_head *list,
1482 struct sk_buff *newsk)
1483{
1484 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1485}
1486
1da177e4
LT
1487/**
1488 * __skb_queue_tail - queue a buffer at the list tail
1489 * @list: list to use
1490 * @newsk: buffer to queue
1491 *
1492 * Queue a buffer at the end of a list. This function takes no locks
1493 * and you must therefore hold required locks before calling it.
1494 *
1495 * A buffer cannot be placed on two lists at the same time.
1496 */
7965bd4d 1497void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1498static inline void __skb_queue_tail(struct sk_buff_head *list,
1499 struct sk_buff *newsk)
1500{
f5572855 1501 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1502}
1503
1da177e4
LT
1504/*
1505 * remove sk_buff from list. _Must_ be called atomically, and with
1506 * the list known..
1507 */
7965bd4d 1508void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1509static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1510{
1511 struct sk_buff *next, *prev;
1512
1513 list->qlen--;
1514 next = skb->next;
1515 prev = skb->prev;
1516 skb->next = skb->prev = NULL;
1da177e4
LT
1517 next->prev = prev;
1518 prev->next = next;
1519}
1520
f525c06d
GR
1521/**
1522 * __skb_dequeue - remove from the head of the queue
1523 * @list: list to dequeue from
1524 *
1525 * Remove the head of the list. This function does not take any locks
1526 * so must be used with appropriate locks held only. The head item is
1527 * returned or %NULL if the list is empty.
1528 */
7965bd4d 1529struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1530static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1531{
1532 struct sk_buff *skb = skb_peek(list);
1533 if (skb)
1534 __skb_unlink(skb, list);
1535 return skb;
1536}
1da177e4
LT
1537
1538/**
1539 * __skb_dequeue_tail - remove from the tail of the queue
1540 * @list: list to dequeue from
1541 *
1542 * Remove the tail of the list. This function does not take any locks
1543 * so must be used with appropriate locks held only. The tail item is
1544 * returned or %NULL if the list is empty.
1545 */
7965bd4d 1546struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1547static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1548{
1549 struct sk_buff *skb = skb_peek_tail(list);
1550 if (skb)
1551 __skb_unlink(skb, list);
1552 return skb;
1553}
1554
1555
bdcc0924 1556static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1557{
1558 return skb->data_len;
1559}
1560
1561static inline unsigned int skb_headlen(const struct sk_buff *skb)
1562{
1563 return skb->len - skb->data_len;
1564}
1565
1566static inline int skb_pagelen(const struct sk_buff *skb)
1567{
1568 int i, len = 0;
1569
1570 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1571 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1572 return len + skb_headlen(skb);
1573}
1574
131ea667
IC
1575/**
1576 * __skb_fill_page_desc - initialise a paged fragment in an skb
1577 * @skb: buffer containing fragment to be initialised
1578 * @i: paged fragment index to initialise
1579 * @page: the page to use for this fragment
1580 * @off: the offset to the data with @page
1581 * @size: the length of the data
1582 *
1583 * Initialises the @i'th fragment of @skb to point to &size bytes at
1584 * offset @off within @page.
1585 *
1586 * Does not take any additional reference on the fragment.
1587 */
1588static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1589 struct page *page, int off, int size)
1da177e4
LT
1590{
1591 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1592
c48a11c7
MG
1593 /*
1594 * Propagate page->pfmemalloc to the skb if we can. The problem is
1595 * that not all callers have unique ownership of the page. If
1596 * pfmemalloc is set, we check the mapping as a mapping implies
1597 * page->index is set (index and pfmemalloc share space).
1598 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1599 * do not lose pfmemalloc information as the pages would not be
1600 * allocated using __GFP_MEMALLOC.
1601 */
a8605c60 1602 frag->page.p = page;
1da177e4 1603 frag->page_offset = off;
9e903e08 1604 skb_frag_size_set(frag, size);
cca7af38
PE
1605
1606 page = compound_head(page);
1607 if (page->pfmemalloc && !page->mapping)
1608 skb->pfmemalloc = true;
131ea667
IC
1609}
1610
1611/**
1612 * skb_fill_page_desc - initialise a paged fragment in an skb
1613 * @skb: buffer containing fragment to be initialised
1614 * @i: paged fragment index to initialise
1615 * @page: the page to use for this fragment
1616 * @off: the offset to the data with @page
1617 * @size: the length of the data
1618 *
1619 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1620 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1621 * addition updates @skb such that @i is the last fragment.
1622 *
1623 * Does not take any additional reference on the fragment.
1624 */
1625static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1626 struct page *page, int off, int size)
1627{
1628 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1629 skb_shinfo(skb)->nr_frags = i + 1;
1630}
1631
7965bd4d
JP
1632void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1633 int size, unsigned int truesize);
654bed16 1634
f8e617e1
JW
1635void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1636 unsigned int truesize);
1637
1da177e4 1638#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1639#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1640#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1641
27a884dc
ACM
1642#ifdef NET_SKBUFF_DATA_USES_OFFSET
1643static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1644{
1645 return skb->head + skb->tail;
1646}
1647
1648static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1649{
1650 skb->tail = skb->data - skb->head;
1651}
1652
1653static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1654{
1655 skb_reset_tail_pointer(skb);
1656 skb->tail += offset;
1657}
7cc46190 1658
27a884dc
ACM
1659#else /* NET_SKBUFF_DATA_USES_OFFSET */
1660static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1661{
1662 return skb->tail;
1663}
1664
1665static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1666{
1667 skb->tail = skb->data;
1668}
1669
1670static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1671{
1672 skb->tail = skb->data + offset;
1673}
4305b541 1674
27a884dc
ACM
1675#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1676
1da177e4
LT
1677/*
1678 * Add data to an sk_buff
1679 */
0c7ddf36 1680unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1681unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1682static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1683{
27a884dc 1684 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1685 SKB_LINEAR_ASSERT(skb);
1686 skb->tail += len;
1687 skb->len += len;
1688 return tmp;
1689}
1690
7965bd4d 1691unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1692static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1693{
1694 skb->data -= len;
1695 skb->len += len;
1696 return skb->data;
1697}
1698
7965bd4d 1699unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1700static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1701{
1702 skb->len -= len;
1703 BUG_ON(skb->len < skb->data_len);
1704 return skb->data += len;
1705}
1706
47d29646
DM
1707static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1708{
1709 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1710}
1711
7965bd4d 1712unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1713
1714static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1715{
1716 if (len > skb_headlen(skb) &&
987c402a 1717 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1718 return NULL;
1719 skb->len -= len;
1720 return skb->data += len;
1721}
1722
1723static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1724{
1725 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1726}
1727
1728static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1729{
1730 if (likely(len <= skb_headlen(skb)))
1731 return 1;
1732 if (unlikely(len > skb->len))
1733 return 0;
987c402a 1734 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1735}
1736
1737/**
1738 * skb_headroom - bytes at buffer head
1739 * @skb: buffer to check
1740 *
1741 * Return the number of bytes of free space at the head of an &sk_buff.
1742 */
c2636b4d 1743static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1744{
1745 return skb->data - skb->head;
1746}
1747
1748/**
1749 * skb_tailroom - bytes at buffer end
1750 * @skb: buffer to check
1751 *
1752 * Return the number of bytes of free space at the tail of an sk_buff
1753 */
1754static inline int skb_tailroom(const struct sk_buff *skb)
1755{
4305b541 1756 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1757}
1758
a21d4572
ED
1759/**
1760 * skb_availroom - bytes at buffer end
1761 * @skb: buffer to check
1762 *
1763 * Return the number of bytes of free space at the tail of an sk_buff
1764 * allocated by sk_stream_alloc()
1765 */
1766static inline int skb_availroom(const struct sk_buff *skb)
1767{
16fad69c
ED
1768 if (skb_is_nonlinear(skb))
1769 return 0;
1770
1771 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1772}
1773
1da177e4
LT
1774/**
1775 * skb_reserve - adjust headroom
1776 * @skb: buffer to alter
1777 * @len: bytes to move
1778 *
1779 * Increase the headroom of an empty &sk_buff by reducing the tail
1780 * room. This is only allowed for an empty buffer.
1781 */
8243126c 1782static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1783{
1784 skb->data += len;
1785 skb->tail += len;
1786}
1787
8bce6d7d
TH
1788#define ENCAP_TYPE_ETHER 0
1789#define ENCAP_TYPE_IPPROTO 1
1790
1791static inline void skb_set_inner_protocol(struct sk_buff *skb,
1792 __be16 protocol)
1793{
1794 skb->inner_protocol = protocol;
1795 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1796}
1797
1798static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1799 __u8 ipproto)
1800{
1801 skb->inner_ipproto = ipproto;
1802 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1803}
1804
6a674e9c
JG
1805static inline void skb_reset_inner_headers(struct sk_buff *skb)
1806{
aefbd2b3 1807 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1808 skb->inner_network_header = skb->network_header;
1809 skb->inner_transport_header = skb->transport_header;
1810}
1811
0b5c9db1
JP
1812static inline void skb_reset_mac_len(struct sk_buff *skb)
1813{
1814 skb->mac_len = skb->network_header - skb->mac_header;
1815}
1816
6a674e9c
JG
1817static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1818 *skb)
1819{
1820 return skb->head + skb->inner_transport_header;
1821}
1822
1823static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1824{
1825 skb->inner_transport_header = skb->data - skb->head;
1826}
1827
1828static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1829 const int offset)
1830{
1831 skb_reset_inner_transport_header(skb);
1832 skb->inner_transport_header += offset;
1833}
1834
1835static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1836{
1837 return skb->head + skb->inner_network_header;
1838}
1839
1840static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1841{
1842 skb->inner_network_header = skb->data - skb->head;
1843}
1844
1845static inline void skb_set_inner_network_header(struct sk_buff *skb,
1846 const int offset)
1847{
1848 skb_reset_inner_network_header(skb);
1849 skb->inner_network_header += offset;
1850}
1851
aefbd2b3
PS
1852static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1853{
1854 return skb->head + skb->inner_mac_header;
1855}
1856
1857static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1858{
1859 skb->inner_mac_header = skb->data - skb->head;
1860}
1861
1862static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1863 const int offset)
1864{
1865 skb_reset_inner_mac_header(skb);
1866 skb->inner_mac_header += offset;
1867}
fda55eca
ED
1868static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1869{
35d04610 1870 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1871}
1872
9c70220b
ACM
1873static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1874{
2e07fa9c 1875 return skb->head + skb->transport_header;
9c70220b
ACM
1876}
1877
badff6d0
ACM
1878static inline void skb_reset_transport_header(struct sk_buff *skb)
1879{
2e07fa9c 1880 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1881}
1882
967b05f6
ACM
1883static inline void skb_set_transport_header(struct sk_buff *skb,
1884 const int offset)
1885{
2e07fa9c
ACM
1886 skb_reset_transport_header(skb);
1887 skb->transport_header += offset;
ea2ae17d
ACM
1888}
1889
d56f90a7
ACM
1890static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1891{
2e07fa9c 1892 return skb->head + skb->network_header;
d56f90a7
ACM
1893}
1894
c1d2bbe1
ACM
1895static inline void skb_reset_network_header(struct sk_buff *skb)
1896{
2e07fa9c 1897 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1898}
1899
c14d2450
ACM
1900static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1901{
2e07fa9c
ACM
1902 skb_reset_network_header(skb);
1903 skb->network_header += offset;
c14d2450
ACM
1904}
1905
2e07fa9c 1906static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1907{
2e07fa9c 1908 return skb->head + skb->mac_header;
bbe735e4
ACM
1909}
1910
2e07fa9c 1911static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1912{
35d04610 1913 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1914}
1915
1916static inline void skb_reset_mac_header(struct sk_buff *skb)
1917{
1918 skb->mac_header = skb->data - skb->head;
1919}
1920
1921static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1922{
1923 skb_reset_mac_header(skb);
1924 skb->mac_header += offset;
1925}
1926
0e3da5bb
TT
1927static inline void skb_pop_mac_header(struct sk_buff *skb)
1928{
1929 skb->mac_header = skb->network_header;
1930}
1931
fbbdb8f0
YX
1932static inline void skb_probe_transport_header(struct sk_buff *skb,
1933 const int offset_hint)
1934{
1935 struct flow_keys keys;
1936
1937 if (skb_transport_header_was_set(skb))
1938 return;
1939 else if (skb_flow_dissect(skb, &keys))
1940 skb_set_transport_header(skb, keys.thoff);
1941 else
1942 skb_set_transport_header(skb, offset_hint);
1943}
1944
03606895
ED
1945static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1946{
1947 if (skb_mac_header_was_set(skb)) {
1948 const unsigned char *old_mac = skb_mac_header(skb);
1949
1950 skb_set_mac_header(skb, -skb->mac_len);
1951 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1952 }
1953}
1954
04fb451e
MM
1955static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1956{
1957 return skb->csum_start - skb_headroom(skb);
1958}
1959
2e07fa9c
ACM
1960static inline int skb_transport_offset(const struct sk_buff *skb)
1961{
1962 return skb_transport_header(skb) - skb->data;
1963}
1964
1965static inline u32 skb_network_header_len(const struct sk_buff *skb)
1966{
1967 return skb->transport_header - skb->network_header;
1968}
1969
6a674e9c
JG
1970static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1971{
1972 return skb->inner_transport_header - skb->inner_network_header;
1973}
1974
2e07fa9c
ACM
1975static inline int skb_network_offset(const struct sk_buff *skb)
1976{
1977 return skb_network_header(skb) - skb->data;
1978}
48d49d0c 1979
6a674e9c
JG
1980static inline int skb_inner_network_offset(const struct sk_buff *skb)
1981{
1982 return skb_inner_network_header(skb) - skb->data;
1983}
1984
f9599ce1
CG
1985static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1986{
1987 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1988}
1989
1da177e4
LT
1990/*
1991 * CPUs often take a performance hit when accessing unaligned memory
1992 * locations. The actual performance hit varies, it can be small if the
1993 * hardware handles it or large if we have to take an exception and fix it
1994 * in software.
1995 *
1996 * Since an ethernet header is 14 bytes network drivers often end up with
1997 * the IP header at an unaligned offset. The IP header can be aligned by
1998 * shifting the start of the packet by 2 bytes. Drivers should do this
1999 * with:
2000 *
8660c124 2001 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2002 *
2003 * The downside to this alignment of the IP header is that the DMA is now
2004 * unaligned. On some architectures the cost of an unaligned DMA is high
2005 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2006 *
1da177e4
LT
2007 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2008 * to be overridden.
2009 */
2010#ifndef NET_IP_ALIGN
2011#define NET_IP_ALIGN 2
2012#endif
2013
025be81e
AB
2014/*
2015 * The networking layer reserves some headroom in skb data (via
2016 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2017 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2018 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2019 *
2020 * Unfortunately this headroom changes the DMA alignment of the resulting
2021 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2022 * on some architectures. An architecture can override this value,
2023 * perhaps setting it to a cacheline in size (since that will maintain
2024 * cacheline alignment of the DMA). It must be a power of 2.
2025 *
d6301d3d 2026 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2027 * headroom, you should not reduce this.
5933dd2f
ED
2028 *
2029 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2030 * to reduce average number of cache lines per packet.
2031 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2032 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2033 */
2034#ifndef NET_SKB_PAD
5933dd2f 2035#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2036#endif
2037
7965bd4d 2038int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2039
2040static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2041{
c4264f27 2042 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2043 WARN_ON(1);
2044 return;
2045 }
27a884dc
ACM
2046 skb->len = len;
2047 skb_set_tail_pointer(skb, len);
1da177e4
LT
2048}
2049
7965bd4d 2050void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2051
2052static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2053{
3cc0e873
HX
2054 if (skb->data_len)
2055 return ___pskb_trim(skb, len);
2056 __skb_trim(skb, len);
2057 return 0;
1da177e4
LT
2058}
2059
2060static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2061{
2062 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2063}
2064
e9fa4f7b
HX
2065/**
2066 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2067 * @skb: buffer to alter
2068 * @len: new length
2069 *
2070 * This is identical to pskb_trim except that the caller knows that
2071 * the skb is not cloned so we should never get an error due to out-
2072 * of-memory.
2073 */
2074static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2075{
2076 int err = pskb_trim(skb, len);
2077 BUG_ON(err);
2078}
2079
1da177e4
LT
2080/**
2081 * skb_orphan - orphan a buffer
2082 * @skb: buffer to orphan
2083 *
2084 * If a buffer currently has an owner then we call the owner's
2085 * destructor function and make the @skb unowned. The buffer continues
2086 * to exist but is no longer charged to its former owner.
2087 */
2088static inline void skb_orphan(struct sk_buff *skb)
2089{
c34a7612 2090 if (skb->destructor) {
1da177e4 2091 skb->destructor(skb);
c34a7612
ED
2092 skb->destructor = NULL;
2093 skb->sk = NULL;
376c7311
ED
2094 } else {
2095 BUG_ON(skb->sk);
c34a7612 2096 }
1da177e4
LT
2097}
2098
a353e0ce
MT
2099/**
2100 * skb_orphan_frags - orphan the frags contained in a buffer
2101 * @skb: buffer to orphan frags from
2102 * @gfp_mask: allocation mask for replacement pages
2103 *
2104 * For each frag in the SKB which needs a destructor (i.e. has an
2105 * owner) create a copy of that frag and release the original
2106 * page by calling the destructor.
2107 */
2108static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2109{
2110 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2111 return 0;
2112 return skb_copy_ubufs(skb, gfp_mask);
2113}
2114
1da177e4
LT
2115/**
2116 * __skb_queue_purge - empty a list
2117 * @list: list to empty
2118 *
2119 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2120 * the list and one reference dropped. This function does not take the
2121 * list lock and the caller must hold the relevant locks to use it.
2122 */
7965bd4d 2123void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2124static inline void __skb_queue_purge(struct sk_buff_head *list)
2125{
2126 struct sk_buff *skb;
2127 while ((skb = __skb_dequeue(list)) != NULL)
2128 kfree_skb(skb);
2129}
2130
7965bd4d 2131void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2132
7965bd4d
JP
2133struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2134 gfp_t gfp_mask);
8af27456
CH
2135
2136/**
2137 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2138 * @dev: network device to receive on
2139 * @length: length to allocate
2140 *
2141 * Allocate a new &sk_buff and assign it a usage count of one. The
2142 * buffer has unspecified headroom built in. Users should allocate
2143 * the headroom they think they need without accounting for the
2144 * built in space. The built in space is used for optimisations.
2145 *
2146 * %NULL is returned if there is no free memory. Although this function
2147 * allocates memory it can be called from an interrupt.
2148 */
2149static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2150 unsigned int length)
8af27456
CH
2151{
2152 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2153}
2154
6f532612
ED
2155/* legacy helper around __netdev_alloc_skb() */
2156static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2157 gfp_t gfp_mask)
2158{
2159 return __netdev_alloc_skb(NULL, length, gfp_mask);
2160}
2161
2162/* legacy helper around netdev_alloc_skb() */
2163static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2164{
2165 return netdev_alloc_skb(NULL, length);
2166}
2167
2168
4915a0de
ED
2169static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2170 unsigned int length, gfp_t gfp)
61321bbd 2171{
4915a0de 2172 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2173
2174 if (NET_IP_ALIGN && skb)
2175 skb_reserve(skb, NET_IP_ALIGN);
2176 return skb;
2177}
2178
4915a0de
ED
2179static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2180 unsigned int length)
2181{
2182 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2183}
2184
181edb2b
AD
2185static inline void skb_free_frag(void *addr)
2186{
2187 __free_page_frag(addr);
2188}
2189
ffde7328 2190void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2191struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2192 unsigned int length, gfp_t gfp_mask);
2193static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2194 unsigned int length)
2195{
2196 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2197}
ffde7328 2198
71dfda58
AD
2199/**
2200 * __dev_alloc_pages - allocate page for network Rx
2201 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2202 * @order: size of the allocation
2203 *
2204 * Allocate a new page.
2205 *
2206 * %NULL is returned if there is no free memory.
2207*/
2208static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2209 unsigned int order)
2210{
2211 /* This piece of code contains several assumptions.
2212 * 1. This is for device Rx, therefor a cold page is preferred.
2213 * 2. The expectation is the user wants a compound page.
2214 * 3. If requesting a order 0 page it will not be compound
2215 * due to the check to see if order has a value in prep_new_page
2216 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2217 * code in gfp_to_alloc_flags that should be enforcing this.
2218 */
2219 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2220
2221 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2222}
2223
2224static inline struct page *dev_alloc_pages(unsigned int order)
2225{
2226 return __dev_alloc_pages(GFP_ATOMIC, order);
2227}
2228
2229/**
2230 * __dev_alloc_page - allocate a page for network Rx
2231 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2232 *
2233 * Allocate a new page.
2234 *
2235 * %NULL is returned if there is no free memory.
2236 */
2237static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2238{
2239 return __dev_alloc_pages(gfp_mask, 0);
2240}
2241
2242static inline struct page *dev_alloc_page(void)
2243{
2244 return __dev_alloc_page(GFP_ATOMIC);
2245}
2246
0614002b
MG
2247/**
2248 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2249 * @page: The page that was allocated from skb_alloc_page
2250 * @skb: The skb that may need pfmemalloc set
2251 */
2252static inline void skb_propagate_pfmemalloc(struct page *page,
2253 struct sk_buff *skb)
2254{
2255 if (page && page->pfmemalloc)
2256 skb->pfmemalloc = true;
2257}
2258
131ea667 2259/**
e227867f 2260 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2261 * @frag: the paged fragment
2262 *
2263 * Returns the &struct page associated with @frag.
2264 */
2265static inline struct page *skb_frag_page(const skb_frag_t *frag)
2266{
a8605c60 2267 return frag->page.p;
131ea667
IC
2268}
2269
2270/**
2271 * __skb_frag_ref - take an addition reference on a paged fragment.
2272 * @frag: the paged fragment
2273 *
2274 * Takes an additional reference on the paged fragment @frag.
2275 */
2276static inline void __skb_frag_ref(skb_frag_t *frag)
2277{
2278 get_page(skb_frag_page(frag));
2279}
2280
2281/**
2282 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2283 * @skb: the buffer
2284 * @f: the fragment offset.
2285 *
2286 * Takes an additional reference on the @f'th paged fragment of @skb.
2287 */
2288static inline void skb_frag_ref(struct sk_buff *skb, int f)
2289{
2290 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2291}
2292
2293/**
2294 * __skb_frag_unref - release a reference on a paged fragment.
2295 * @frag: the paged fragment
2296 *
2297 * Releases a reference on the paged fragment @frag.
2298 */
2299static inline void __skb_frag_unref(skb_frag_t *frag)
2300{
2301 put_page(skb_frag_page(frag));
2302}
2303
2304/**
2305 * skb_frag_unref - release a reference on a paged fragment of an skb.
2306 * @skb: the buffer
2307 * @f: the fragment offset
2308 *
2309 * Releases a reference on the @f'th paged fragment of @skb.
2310 */
2311static inline void skb_frag_unref(struct sk_buff *skb, int f)
2312{
2313 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2314}
2315
2316/**
2317 * skb_frag_address - gets the address of the data contained in a paged fragment
2318 * @frag: the paged fragment buffer
2319 *
2320 * Returns the address of the data within @frag. The page must already
2321 * be mapped.
2322 */
2323static inline void *skb_frag_address(const skb_frag_t *frag)
2324{
2325 return page_address(skb_frag_page(frag)) + frag->page_offset;
2326}
2327
2328/**
2329 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2330 * @frag: the paged fragment buffer
2331 *
2332 * Returns the address of the data within @frag. Checks that the page
2333 * is mapped and returns %NULL otherwise.
2334 */
2335static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2336{
2337 void *ptr = page_address(skb_frag_page(frag));
2338 if (unlikely(!ptr))
2339 return NULL;
2340
2341 return ptr + frag->page_offset;
2342}
2343
2344/**
2345 * __skb_frag_set_page - sets the page contained in a paged fragment
2346 * @frag: the paged fragment
2347 * @page: the page to set
2348 *
2349 * Sets the fragment @frag to contain @page.
2350 */
2351static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2352{
a8605c60 2353 frag->page.p = page;
131ea667
IC
2354}
2355
2356/**
2357 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2358 * @skb: the buffer
2359 * @f: the fragment offset
2360 * @page: the page to set
2361 *
2362 * Sets the @f'th fragment of @skb to contain @page.
2363 */
2364static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2365 struct page *page)
2366{
2367 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2368}
2369
400dfd3a
ED
2370bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2371
131ea667
IC
2372/**
2373 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2374 * @dev: the device to map the fragment to
131ea667
IC
2375 * @frag: the paged fragment to map
2376 * @offset: the offset within the fragment (starting at the
2377 * fragment's own offset)
2378 * @size: the number of bytes to map
f83347df 2379 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2380 *
2381 * Maps the page associated with @frag to @device.
2382 */
2383static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2384 const skb_frag_t *frag,
2385 size_t offset, size_t size,
2386 enum dma_data_direction dir)
2387{
2388 return dma_map_page(dev, skb_frag_page(frag),
2389 frag->page_offset + offset, size, dir);
2390}
2391
117632e6
ED
2392static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2393 gfp_t gfp_mask)
2394{
2395 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2396}
2397
bad93e9d
OP
2398
2399static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2400 gfp_t gfp_mask)
2401{
2402 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2403}
2404
2405
334a8132
PM
2406/**
2407 * skb_clone_writable - is the header of a clone writable
2408 * @skb: buffer to check
2409 * @len: length up to which to write
2410 *
2411 * Returns true if modifying the header part of the cloned buffer
2412 * does not requires the data to be copied.
2413 */
05bdd2f1 2414static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2415{
2416 return !skb_header_cloned(skb) &&
2417 skb_headroom(skb) + len <= skb->hdr_len;
2418}
2419
d9cc2048
HX
2420static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2421 int cloned)
2422{
2423 int delta = 0;
2424
d9cc2048
HX
2425 if (headroom > skb_headroom(skb))
2426 delta = headroom - skb_headroom(skb);
2427
2428 if (delta || cloned)
2429 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2430 GFP_ATOMIC);
2431 return 0;
2432}
2433
1da177e4
LT
2434/**
2435 * skb_cow - copy header of skb when it is required
2436 * @skb: buffer to cow
2437 * @headroom: needed headroom
2438 *
2439 * If the skb passed lacks sufficient headroom or its data part
2440 * is shared, data is reallocated. If reallocation fails, an error
2441 * is returned and original skb is not changed.
2442 *
2443 * The result is skb with writable area skb->head...skb->tail
2444 * and at least @headroom of space at head.
2445 */
2446static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2447{
d9cc2048
HX
2448 return __skb_cow(skb, headroom, skb_cloned(skb));
2449}
1da177e4 2450
d9cc2048
HX
2451/**
2452 * skb_cow_head - skb_cow but only making the head writable
2453 * @skb: buffer to cow
2454 * @headroom: needed headroom
2455 *
2456 * This function is identical to skb_cow except that we replace the
2457 * skb_cloned check by skb_header_cloned. It should be used when
2458 * you only need to push on some header and do not need to modify
2459 * the data.
2460 */
2461static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2462{
2463 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2464}
2465
2466/**
2467 * skb_padto - pad an skbuff up to a minimal size
2468 * @skb: buffer to pad
2469 * @len: minimal length
2470 *
2471 * Pads up a buffer to ensure the trailing bytes exist and are
2472 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2473 * is untouched. Otherwise it is extended. Returns zero on
2474 * success. The skb is freed on error.
1da177e4 2475 */
5b057c6b 2476static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2477{
2478 unsigned int size = skb->len;
2479 if (likely(size >= len))
5b057c6b 2480 return 0;
987c402a 2481 return skb_pad(skb, len - size);
1da177e4
LT
2482}
2483
9c0c1124
AD
2484/**
2485 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2486 * @skb: buffer to pad
2487 * @len: minimal length
2488 *
2489 * Pads up a buffer to ensure the trailing bytes exist and are
2490 * blanked. If the buffer already contains sufficient data it
2491 * is untouched. Otherwise it is extended. Returns zero on
2492 * success. The skb is freed on error.
2493 */
2494static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2495{
2496 unsigned int size = skb->len;
2497
2498 if (unlikely(size < len)) {
2499 len -= size;
2500 if (skb_pad(skb, len))
2501 return -ENOMEM;
2502 __skb_put(skb, len);
2503 }
2504 return 0;
2505}
2506
1da177e4 2507static inline int skb_add_data(struct sk_buff *skb,
af2b040e 2508 struct iov_iter *from, int copy)
1da177e4
LT
2509{
2510 const int off = skb->len;
2511
2512 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e
AV
2513 __wsum csum = 0;
2514 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2515 &csum, from) == copy) {
1da177e4
LT
2516 skb->csum = csum_block_add(skb->csum, csum, off);
2517 return 0;
2518 }
af2b040e 2519 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
1da177e4
LT
2520 return 0;
2521
2522 __skb_trim(skb, off);
2523 return -EFAULT;
2524}
2525
38ba0a65
ED
2526static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2527 const struct page *page, int off)
1da177e4
LT
2528{
2529 if (i) {
9e903e08 2530 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2531
ea2ab693 2532 return page == skb_frag_page(frag) &&
9e903e08 2533 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2534 }
38ba0a65 2535 return false;
1da177e4
LT
2536}
2537
364c6bad
HX
2538static inline int __skb_linearize(struct sk_buff *skb)
2539{
2540 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2541}
2542
1da177e4
LT
2543/**
2544 * skb_linearize - convert paged skb to linear one
2545 * @skb: buffer to linarize
1da177e4
LT
2546 *
2547 * If there is no free memory -ENOMEM is returned, otherwise zero
2548 * is returned and the old skb data released.
2549 */
364c6bad
HX
2550static inline int skb_linearize(struct sk_buff *skb)
2551{
2552 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2553}
2554
cef401de
ED
2555/**
2556 * skb_has_shared_frag - can any frag be overwritten
2557 * @skb: buffer to test
2558 *
2559 * Return true if the skb has at least one frag that might be modified
2560 * by an external entity (as in vmsplice()/sendfile())
2561 */
2562static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2563{
c9af6db4
PS
2564 return skb_is_nonlinear(skb) &&
2565 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2566}
2567
364c6bad
HX
2568/**
2569 * skb_linearize_cow - make sure skb is linear and writable
2570 * @skb: buffer to process
2571 *
2572 * If there is no free memory -ENOMEM is returned, otherwise zero
2573 * is returned and the old skb data released.
2574 */
2575static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2576{
364c6bad
HX
2577 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2578 __skb_linearize(skb) : 0;
1da177e4
LT
2579}
2580
2581/**
2582 * skb_postpull_rcsum - update checksum for received skb after pull
2583 * @skb: buffer to update
2584 * @start: start of data before pull
2585 * @len: length of data pulled
2586 *
2587 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2588 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2589 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2590 */
2591
2592static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2593 const void *start, unsigned int len)
1da177e4 2594{
84fa7933 2595 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2596 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2597}
2598
cbb042f9
HX
2599unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2600
7ce5a27f
DM
2601/**
2602 * pskb_trim_rcsum - trim received skb and update checksum
2603 * @skb: buffer to trim
2604 * @len: new length
2605 *
2606 * This is exactly the same as pskb_trim except that it ensures the
2607 * checksum of received packets are still valid after the operation.
2608 */
2609
2610static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2611{
2612 if (likely(len >= skb->len))
2613 return 0;
2614 if (skb->ip_summed == CHECKSUM_COMPLETE)
2615 skb->ip_summed = CHECKSUM_NONE;
2616 return __pskb_trim(skb, len);
2617}
2618
1da177e4
LT
2619#define skb_queue_walk(queue, skb) \
2620 for (skb = (queue)->next; \
a1e4891f 2621 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2622 skb = skb->next)
2623
46f8914e
JC
2624#define skb_queue_walk_safe(queue, skb, tmp) \
2625 for (skb = (queue)->next, tmp = skb->next; \
2626 skb != (struct sk_buff *)(queue); \
2627 skb = tmp, tmp = skb->next)
2628
1164f52a 2629#define skb_queue_walk_from(queue, skb) \
a1e4891f 2630 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2631 skb = skb->next)
2632
2633#define skb_queue_walk_from_safe(queue, skb, tmp) \
2634 for (tmp = skb->next; \
2635 skb != (struct sk_buff *)(queue); \
2636 skb = tmp, tmp = skb->next)
2637
300ce174
SH
2638#define skb_queue_reverse_walk(queue, skb) \
2639 for (skb = (queue)->prev; \
a1e4891f 2640 skb != (struct sk_buff *)(queue); \
300ce174
SH
2641 skb = skb->prev)
2642
686a2955
DM
2643#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2644 for (skb = (queue)->prev, tmp = skb->prev; \
2645 skb != (struct sk_buff *)(queue); \
2646 skb = tmp, tmp = skb->prev)
2647
2648#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2649 for (tmp = skb->prev; \
2650 skb != (struct sk_buff *)(queue); \
2651 skb = tmp, tmp = skb->prev)
1da177e4 2652
21dc3301 2653static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2654{
2655 return skb_shinfo(skb)->frag_list != NULL;
2656}
2657
2658static inline void skb_frag_list_init(struct sk_buff *skb)
2659{
2660 skb_shinfo(skb)->frag_list = NULL;
2661}
2662
2663static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2664{
2665 frag->next = skb_shinfo(skb)->frag_list;
2666 skb_shinfo(skb)->frag_list = frag;
2667}
2668
2669#define skb_walk_frags(skb, iter) \
2670 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2671
7965bd4d
JP
2672struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2673 int *peeked, int *off, int *err);
2674struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2675 int *err);
2676unsigned int datagram_poll(struct file *file, struct socket *sock,
2677 struct poll_table_struct *wait);
c0371da6
AV
2678int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2679 struct iov_iter *to, int size);
51f3d02b
DM
2680static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2681 struct msghdr *msg, int size)
2682{
e5a4b0bb 2683 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 2684}
e5a4b0bb
AV
2685int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2686 struct msghdr *msg);
3a654f97
AV
2687int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2688 struct iov_iter *from, int len);
3a654f97 2689int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d
JP
2690void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2691void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2692int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2693int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2694int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2695__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2696 int len, __wsum csum);
2697int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2698 struct pipe_inode_info *pipe, unsigned int len,
2699 unsigned int flags);
2700void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2701unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2702int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2703 int len, int hlen);
7965bd4d
JP
2704void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2705int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2706void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2707unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2708struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2709struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 2710int skb_ensure_writable(struct sk_buff *skb, int write_len);
93515d53
JP
2711int skb_vlan_pop(struct sk_buff *skb);
2712int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
20380731 2713
6ce8e9ce
AV
2714static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2715{
21226abb 2716 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
6ce8e9ce
AV
2717}
2718
7eab8d9e
AV
2719static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2720{
e5a4b0bb 2721 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
2722}
2723
2817a336
DB
2724struct skb_checksum_ops {
2725 __wsum (*update)(const void *mem, int len, __wsum wsum);
2726 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2727};
2728
2729__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2730 __wsum csum, const struct skb_checksum_ops *ops);
2731__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2732 __wsum csum);
2733
690e36e7
DM
2734static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2735 int len, void *data, int hlen, void *buffer)
1da177e4 2736{
55820ee2 2737 if (hlen - offset >= len)
690e36e7 2738 return data + offset;
1da177e4 2739
690e36e7
DM
2740 if (!skb ||
2741 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2742 return NULL;
2743
2744 return buffer;
2745}
2746
690e36e7
DM
2747static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2748 int len, void *buffer)
2749{
2750 return __skb_header_pointer(skb, offset, len, skb->data,
2751 skb_headlen(skb), buffer);
2752}
2753
4262e5cc
DB
2754/**
2755 * skb_needs_linearize - check if we need to linearize a given skb
2756 * depending on the given device features.
2757 * @skb: socket buffer to check
2758 * @features: net device features
2759 *
2760 * Returns true if either:
2761 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2762 * 2. skb is fragmented and the device does not support SG.
2763 */
2764static inline bool skb_needs_linearize(struct sk_buff *skb,
2765 netdev_features_t features)
2766{
2767 return skb_is_nonlinear(skb) &&
2768 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2769 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2770}
2771
d626f62b
ACM
2772static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2773 void *to,
2774 const unsigned int len)
2775{
2776 memcpy(to, skb->data, len);
2777}
2778
2779static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2780 const int offset, void *to,
2781 const unsigned int len)
2782{
2783 memcpy(to, skb->data + offset, len);
2784}
2785
27d7ff46
ACM
2786static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2787 const void *from,
2788 const unsigned int len)
2789{
2790 memcpy(skb->data, from, len);
2791}
2792
2793static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2794 const int offset,
2795 const void *from,
2796 const unsigned int len)
2797{
2798 memcpy(skb->data + offset, from, len);
2799}
2800
7965bd4d 2801void skb_init(void);
1da177e4 2802
ac45f602
PO
2803static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2804{
2805 return skb->tstamp;
2806}
2807
a61bbcf2
PM
2808/**
2809 * skb_get_timestamp - get timestamp from a skb
2810 * @skb: skb to get stamp from
2811 * @stamp: pointer to struct timeval to store stamp in
2812 *
2813 * Timestamps are stored in the skb as offsets to a base timestamp.
2814 * This function converts the offset back to a struct timeval and stores
2815 * it in stamp.
2816 */
ac45f602
PO
2817static inline void skb_get_timestamp(const struct sk_buff *skb,
2818 struct timeval *stamp)
a61bbcf2 2819{
b7aa0bf7 2820 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2821}
2822
ac45f602
PO
2823static inline void skb_get_timestampns(const struct sk_buff *skb,
2824 struct timespec *stamp)
2825{
2826 *stamp = ktime_to_timespec(skb->tstamp);
2827}
2828
b7aa0bf7 2829static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2830{
b7aa0bf7 2831 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2832}
2833
164891aa
SH
2834static inline ktime_t net_timedelta(ktime_t t)
2835{
2836 return ktime_sub(ktime_get_real(), t);
2837}
2838
b9ce204f
IJ
2839static inline ktime_t net_invalid_timestamp(void)
2840{
2841 return ktime_set(0, 0);
2842}
a61bbcf2 2843
62bccb8c
AD
2844struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2845
c1f19b51
RC
2846#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2847
7965bd4d
JP
2848void skb_clone_tx_timestamp(struct sk_buff *skb);
2849bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2850
2851#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2852
2853static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2854{
2855}
2856
2857static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2858{
2859 return false;
2860}
2861
2862#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2863
2864/**
2865 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2866 *
da92b194
RC
2867 * PHY drivers may accept clones of transmitted packets for
2868 * timestamping via their phy_driver.txtstamp method. These drivers
2869 * must call this function to return the skb back to the stack, with
2870 * or without a timestamp.
2871 *
c1f19b51 2872 * @skb: clone of the the original outgoing packet
da92b194 2873 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2874 *
2875 */
2876void skb_complete_tx_timestamp(struct sk_buff *skb,
2877 struct skb_shared_hwtstamps *hwtstamps);
2878
e7fd2885
WB
2879void __skb_tstamp_tx(struct sk_buff *orig_skb,
2880 struct skb_shared_hwtstamps *hwtstamps,
2881 struct sock *sk, int tstype);
2882
ac45f602
PO
2883/**
2884 * skb_tstamp_tx - queue clone of skb with send time stamps
2885 * @orig_skb: the original outgoing packet
2886 * @hwtstamps: hardware time stamps, may be NULL if not available
2887 *
2888 * If the skb has a socket associated, then this function clones the
2889 * skb (thus sharing the actual data and optional structures), stores
2890 * the optional hardware time stamping information (if non NULL) or
2891 * generates a software time stamp (otherwise), then queues the clone
2892 * to the error queue of the socket. Errors are silently ignored.
2893 */
7965bd4d
JP
2894void skb_tstamp_tx(struct sk_buff *orig_skb,
2895 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2896
4507a715
RC
2897static inline void sw_tx_timestamp(struct sk_buff *skb)
2898{
2244d07b
OH
2899 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2900 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2901 skb_tstamp_tx(skb, NULL);
2902}
2903
2904/**
2905 * skb_tx_timestamp() - Driver hook for transmit timestamping
2906 *
2907 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2908 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2909 *
73409f3b
DM
2910 * Specifically, one should make absolutely sure that this function is
2911 * called before TX completion of this packet can trigger. Otherwise
2912 * the packet could potentially already be freed.
2913 *
4507a715
RC
2914 * @skb: A socket buffer.
2915 */
2916static inline void skb_tx_timestamp(struct sk_buff *skb)
2917{
c1f19b51 2918 skb_clone_tx_timestamp(skb);
4507a715
RC
2919 sw_tx_timestamp(skb);
2920}
2921
6e3e939f
JB
2922/**
2923 * skb_complete_wifi_ack - deliver skb with wifi status
2924 *
2925 * @skb: the original outgoing packet
2926 * @acked: ack status
2927 *
2928 */
2929void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2930
7965bd4d
JP
2931__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2932__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2933
60476372
HX
2934static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2935{
6edec0e6
TH
2936 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
2937 skb->csum_valid ||
2938 (skb->ip_summed == CHECKSUM_PARTIAL &&
2939 skb_checksum_start_offset(skb) >= 0));
60476372
HX
2940}
2941
fb286bb2
HX
2942/**
2943 * skb_checksum_complete - Calculate checksum of an entire packet
2944 * @skb: packet to process
2945 *
2946 * This function calculates the checksum over the entire packet plus
2947 * the value of skb->csum. The latter can be used to supply the
2948 * checksum of a pseudo header as used by TCP/UDP. It returns the
2949 * checksum.
2950 *
2951 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2952 * this function can be used to verify that checksum on received
2953 * packets. In that case the function should return zero if the
2954 * checksum is correct. In particular, this function will return zero
2955 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2956 * hardware has already verified the correctness of the checksum.
2957 */
4381ca3c 2958static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2959{
60476372
HX
2960 return skb_csum_unnecessary(skb) ?
2961 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2962}
2963
77cffe23
TH
2964static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2965{
2966 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2967 if (skb->csum_level == 0)
2968 skb->ip_summed = CHECKSUM_NONE;
2969 else
2970 skb->csum_level--;
2971 }
2972}
2973
2974static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2975{
2976 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2977 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2978 skb->csum_level++;
2979 } else if (skb->ip_summed == CHECKSUM_NONE) {
2980 skb->ip_summed = CHECKSUM_UNNECESSARY;
2981 skb->csum_level = 0;
2982 }
2983}
2984
5a212329
TH
2985static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2986{
2987 /* Mark current checksum as bad (typically called from GRO
2988 * path). In the case that ip_summed is CHECKSUM_NONE
2989 * this must be the first checksum encountered in the packet.
2990 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2991 * checksum after the last one validated. For UDP, a zero
2992 * checksum can not be marked as bad.
2993 */
2994
2995 if (skb->ip_summed == CHECKSUM_NONE ||
2996 skb->ip_summed == CHECKSUM_UNNECESSARY)
2997 skb->csum_bad = 1;
2998}
2999
76ba0aae
TH
3000/* Check if we need to perform checksum complete validation.
3001 *
3002 * Returns true if checksum complete is needed, false otherwise
3003 * (either checksum is unnecessary or zero checksum is allowed).
3004 */
3005static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3006 bool zero_okay,
3007 __sum16 check)
3008{
5d0c2b95
TH
3009 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3010 skb->csum_valid = 1;
77cffe23 3011 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3012 return false;
3013 }
3014
3015 return true;
3016}
3017
3018/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3019 * in checksum_init.
3020 */
3021#define CHECKSUM_BREAK 76
3022
4e18b9ad
TH
3023/* Unset checksum-complete
3024 *
3025 * Unset checksum complete can be done when packet is being modified
3026 * (uncompressed for instance) and checksum-complete value is
3027 * invalidated.
3028 */
3029static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3030{
3031 if (skb->ip_summed == CHECKSUM_COMPLETE)
3032 skb->ip_summed = CHECKSUM_NONE;
3033}
3034
76ba0aae
TH
3035/* Validate (init) checksum based on checksum complete.
3036 *
3037 * Return values:
3038 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3039 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3040 * checksum is stored in skb->csum for use in __skb_checksum_complete
3041 * non-zero: value of invalid checksum
3042 *
3043 */
3044static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3045 bool complete,
3046 __wsum psum)
3047{
3048 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3049 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3050 skb->csum_valid = 1;
76ba0aae
TH
3051 return 0;
3052 }
5a212329
TH
3053 } else if (skb->csum_bad) {
3054 /* ip_summed == CHECKSUM_NONE in this case */
3055 return 1;
76ba0aae
TH
3056 }
3057
3058 skb->csum = psum;
3059
5d0c2b95
TH
3060 if (complete || skb->len <= CHECKSUM_BREAK) {
3061 __sum16 csum;
3062
3063 csum = __skb_checksum_complete(skb);
3064 skb->csum_valid = !csum;
3065 return csum;
3066 }
76ba0aae
TH
3067
3068 return 0;
3069}
3070
3071static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3072{
3073 return 0;
3074}
3075
3076/* Perform checksum validate (init). Note that this is a macro since we only
3077 * want to calculate the pseudo header which is an input function if necessary.
3078 * First we try to validate without any computation (checksum unnecessary) and
3079 * then calculate based on checksum complete calling the function to compute
3080 * pseudo header.
3081 *
3082 * Return values:
3083 * 0: checksum is validated or try to in skb_checksum_complete
3084 * non-zero: value of invalid checksum
3085 */
3086#define __skb_checksum_validate(skb, proto, complete, \
3087 zero_okay, check, compute_pseudo) \
3088({ \
3089 __sum16 __ret = 0; \
5d0c2b95 3090 skb->csum_valid = 0; \
76ba0aae
TH
3091 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3092 __ret = __skb_checksum_validate_complete(skb, \
3093 complete, compute_pseudo(skb, proto)); \
3094 __ret; \
3095})
3096
3097#define skb_checksum_init(skb, proto, compute_pseudo) \
3098 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3099
3100#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3101 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3102
3103#define skb_checksum_validate(skb, proto, compute_pseudo) \
3104 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3105
3106#define skb_checksum_validate_zero_check(skb, proto, check, \
3107 compute_pseudo) \
096a4cfa 3108 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3109
3110#define skb_checksum_simple_validate(skb) \
3111 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3112
d96535a1
TH
3113static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3114{
3115 return (skb->ip_summed == CHECKSUM_NONE &&
3116 skb->csum_valid && !skb->csum_bad);
3117}
3118
3119static inline void __skb_checksum_convert(struct sk_buff *skb,
3120 __sum16 check, __wsum pseudo)
3121{
3122 skb->csum = ~pseudo;
3123 skb->ip_summed = CHECKSUM_COMPLETE;
3124}
3125
3126#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3127do { \
3128 if (__skb_checksum_convert_check(skb)) \
3129 __skb_checksum_convert(skb, check, \
3130 compute_pseudo(skb, proto)); \
3131} while (0)
3132
15e2396d
TH
3133static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3134 u16 start, u16 offset)
3135{
3136 skb->ip_summed = CHECKSUM_PARTIAL;
3137 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3138 skb->csum_offset = offset - start;
3139}
3140
dcdc8994
TH
3141/* Update skbuf and packet to reflect the remote checksum offload operation.
3142 * When called, ptr indicates the starting point for skb->csum when
3143 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3144 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3145 */
3146static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3147 int start, int offset, bool nopartial)
dcdc8994
TH
3148{
3149 __wsum delta;
3150
15e2396d
TH
3151 if (!nopartial) {
3152 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3153 return;
3154 }
3155
dcdc8994
TH
3156 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3157 __skb_checksum_complete(skb);
3158 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3159 }
3160
3161 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3162
3163 /* Adjust skb->csum since we changed the packet */
3164 skb->csum = csum_add(skb->csum, delta);
3165}
3166
5f79e0f9 3167#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3168void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3169static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3170{
3171 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3172 nf_conntrack_destroy(nfct);
1da177e4
LT
3173}
3174static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3175{
3176 if (nfct)
3177 atomic_inc(&nfct->use);
3178}
2fc72c7b 3179#endif
34666d46 3180#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3181static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3182{
3183 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3184 kfree(nf_bridge);
3185}
3186static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3187{
3188 if (nf_bridge)
3189 atomic_inc(&nf_bridge->use);
3190}
3191#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3192static inline void nf_reset(struct sk_buff *skb)
3193{
5f79e0f9 3194#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3195 nf_conntrack_put(skb->nfct);
3196 skb->nfct = NULL;
2fc72c7b 3197#endif
34666d46 3198#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3199 nf_bridge_put(skb->nf_bridge);
3200 skb->nf_bridge = NULL;
3201#endif
3202}
3203
124dff01
PM
3204static inline void nf_reset_trace(struct sk_buff *skb)
3205{
478b360a 3206#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3207 skb->nf_trace = 0;
3208#endif
a193a4ab
PM
3209}
3210
edda553c 3211/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3212static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3213 bool copy)
edda553c 3214{
5f79e0f9 3215#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3216 dst->nfct = src->nfct;
3217 nf_conntrack_get(src->nfct);
b1937227
ED
3218 if (copy)
3219 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3220#endif
34666d46 3221#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3222 dst->nf_bridge = src->nf_bridge;
3223 nf_bridge_get(src->nf_bridge);
3224#endif
478b360a 3225#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3226 if (copy)
3227 dst->nf_trace = src->nf_trace;
478b360a 3228#endif
edda553c
YK
3229}
3230
e7ac05f3
YK
3231static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3232{
e7ac05f3 3233#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3234 nf_conntrack_put(dst->nfct);
2fc72c7b 3235#endif
34666d46 3236#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3237 nf_bridge_put(dst->nf_bridge);
3238#endif
b1937227 3239 __nf_copy(dst, src, true);
e7ac05f3
YK
3240}
3241
984bc16c
JM
3242#ifdef CONFIG_NETWORK_SECMARK
3243static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3244{
3245 to->secmark = from->secmark;
3246}
3247
3248static inline void skb_init_secmark(struct sk_buff *skb)
3249{
3250 skb->secmark = 0;
3251}
3252#else
3253static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3254{ }
3255
3256static inline void skb_init_secmark(struct sk_buff *skb)
3257{ }
3258#endif
3259
574f7194
EB
3260static inline bool skb_irq_freeable(const struct sk_buff *skb)
3261{
3262 return !skb->destructor &&
3263#if IS_ENABLED(CONFIG_XFRM)
3264 !skb->sp &&
3265#endif
3266#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3267 !skb->nfct &&
3268#endif
3269 !skb->_skb_refdst &&
3270 !skb_has_frag_list(skb);
3271}
3272
f25f4e44
PWJ
3273static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3274{
f25f4e44 3275 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3276}
3277
9247744e 3278static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3279{
4e3ab47a 3280 return skb->queue_mapping;
4e3ab47a
PE
3281}
3282
f25f4e44
PWJ
3283static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3284{
f25f4e44 3285 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3286}
3287
d5a9e24a
DM
3288static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3289{
3290 skb->queue_mapping = rx_queue + 1;
3291}
3292
9247744e 3293static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3294{
3295 return skb->queue_mapping - 1;
3296}
3297
9247744e 3298static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3299{
a02cec21 3300 return skb->queue_mapping != 0;
d5a9e24a
DM
3301}
3302
0e001614 3303u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
7965bd4d 3304 unsigned int num_tx_queues);
9247744e 3305
def8b4fa
AD
3306static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3307{
0b3d8e08 3308#ifdef CONFIG_XFRM
def8b4fa 3309 return skb->sp;
def8b4fa 3310#else
def8b4fa 3311 return NULL;
def8b4fa 3312#endif
0b3d8e08 3313}
def8b4fa 3314
68c33163
PS
3315/* Keeps track of mac header offset relative to skb->head.
3316 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3317 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3318 * tunnel skb it points to outer mac header.
3319 * Keeps track of level of encapsulation of network headers.
3320 */
68c33163 3321struct skb_gso_cb {
3347c960
ED
3322 int mac_offset;
3323 int encap_level;
7e2b10c1 3324 __u16 csum_start;
68c33163
PS
3325};
3326#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3327
3328static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3329{
3330 return (skb_mac_header(inner_skb) - inner_skb->head) -
3331 SKB_GSO_CB(inner_skb)->mac_offset;
3332}
3333
1e2bd517
PS
3334static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3335{
3336 int new_headroom, headroom;
3337 int ret;
3338
3339 headroom = skb_headroom(skb);
3340 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3341 if (ret)
3342 return ret;
3343
3344 new_headroom = skb_headroom(skb);
3345 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3346 return 0;
3347}
3348
7e2b10c1
TH
3349/* Compute the checksum for a gso segment. First compute the checksum value
3350 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3351 * then add in skb->csum (checksum from csum_start to end of packet).
3352 * skb->csum and csum_start are then updated to reflect the checksum of the
3353 * resultant packet starting from the transport header-- the resultant checksum
3354 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3355 * header.
3356 */
3357static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3358{
3359 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3360 skb_transport_offset(skb);
3361 __u16 csum;
3362
3363 csum = csum_fold(csum_partial(skb_transport_header(skb),
3364 plen, skb->csum));
3365 skb->csum = res;
3366 SKB_GSO_CB(skb)->csum_start -= plen;
3367
3368 return csum;
3369}
3370
bdcc0924 3371static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3372{
3373 return skb_shinfo(skb)->gso_size;
3374}
3375
36a8f39e 3376/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3377static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3378{
3379 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3380}
3381
7965bd4d 3382void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3383
3384static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3385{
3386 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3387 * wanted then gso_type will be set. */
05bdd2f1
ED
3388 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3389
b78462eb
AD
3390 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3391 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3392 __skb_warn_lro_forwarding(skb);
3393 return true;
3394 }
3395 return false;
3396}
3397
35fc92a9
HX
3398static inline void skb_forward_csum(struct sk_buff *skb)
3399{
3400 /* Unfortunately we don't support this one. Any brave souls? */
3401 if (skb->ip_summed == CHECKSUM_COMPLETE)
3402 skb->ip_summed = CHECKSUM_NONE;
3403}
3404
bc8acf2c
ED
3405/**
3406 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3407 * @skb: skb to check
3408 *
3409 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3410 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3411 * use this helper, to document places where we make this assertion.
3412 */
05bdd2f1 3413static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3414{
3415#ifdef DEBUG
3416 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3417#endif
3418}
3419
f35d9d8a 3420bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3421
ed1f50c3 3422int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
3423struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3424 unsigned int transport_len,
3425 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 3426
56193d1b
AD
3427u32 skb_get_poff(const struct sk_buff *skb);
3428u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3429 const struct flow_keys *keys, int hlen);
f77668dc 3430
3a7c1ee4
AD
3431/**
3432 * skb_head_is_locked - Determine if the skb->head is locked down
3433 * @skb: skb to check
3434 *
3435 * The head on skbs build around a head frag can be removed if they are
3436 * not cloned. This function returns true if the skb head is locked down
3437 * due to either being allocated via kmalloc, or by being a clone with
3438 * multiple references to the head.
3439 */
3440static inline bool skb_head_is_locked(const struct sk_buff *skb)
3441{
3442 return !skb->head_frag || skb_cloned(skb);
3443}
fe6cc55f
FW
3444
3445/**
3446 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3447 *
3448 * @skb: GSO skb
3449 *
3450 * skb_gso_network_seglen is used to determine the real size of the
3451 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3452 *
3453 * The MAC/L2 header is not accounted for.
3454 */
3455static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3456{
3457 unsigned int hdr_len = skb_transport_header(skb) -
3458 skb_network_header(skb);
3459 return hdr_len + skb_gso_transport_seglen(skb);
3460}
1da177e4
LT
3461#endif /* __KERNEL__ */
3462#endif /* _LINUX_SKBUFF_H */