]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/skbuff.h
fou: Fix typo in returning flags in netlink
[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 *
86 * This is identical to the case for output below. This may occur on a packet
87 * received directly from another Linux OS, e.g., a virtualized Linux kernel
88 * on the same host. The packet can be treated in the same way as
89 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
90 * checksum must be filled in by the OS or the hardware.
91 *
92 * B. Checksumming on output.
93 *
94 * CHECKSUM_NONE:
95 *
96 * The skb was already checksummed by the protocol, or a checksum is not
97 * required.
98 *
99 * CHECKSUM_PARTIAL:
100 *
101 * The device is required to checksum the packet as seen by hard_start_xmit()
102 * from skb->csum_start up to the end, and to record/write the checksum at
103 * offset skb->csum_start + skb->csum_offset.
104 *
105 * The device must show its capabilities in dev->features, set up at device
106 * setup time, e.g. netdev_features.h:
107 *
108 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
109 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
110 * IPv4. Sigh. Vendors like this way for an unknown reason.
111 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
112 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
113 * NETIF_F_... - Well, you get the picture.
114 *
115 * CHECKSUM_UNNECESSARY:
116 *
117 * Normally, the device will do per protocol specific checksumming. Protocol
118 * implementations that do not want the NIC to perform the checksum
119 * calculation should use this flag in their outgoing skbs.
120 *
121 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
122 * offload. Correspondingly, the FCoE protocol driver
123 * stack should use CHECKSUM_UNNECESSARY.
124 *
125 * Any questions? No questions, good. --ANK
126 */
127
60476372 128/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
129#define CHECKSUM_NONE 0
130#define CHECKSUM_UNNECESSARY 1
131#define CHECKSUM_COMPLETE 2
132#define CHECKSUM_PARTIAL 3
1da177e4 133
77cffe23
TH
134/* Maximum value in skb->csum_level */
135#define SKB_MAX_CSUM_LEVEL 3
136
0bec8c88 137#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 138#define SKB_WITH_OVERHEAD(X) \
deea84b0 139 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
140#define SKB_MAX_ORDER(X, ORDER) \
141 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
142#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
143#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
144
87fb4b7b
ED
145/* return minimum truesize of one skb containing X bytes of data */
146#define SKB_TRUESIZE(X) ((X) + \
147 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
148 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
149
1da177e4 150struct net_device;
716ea3a7 151struct scatterlist;
9c55e01c 152struct pipe_inode_info;
1da177e4 153
5f79e0f9 154#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
155struct nf_conntrack {
156 atomic_t use;
1da177e4 157};
5f79e0f9 158#endif
1da177e4 159
34666d46 160#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 161struct nf_bridge_info {
bf1ac5ca
ED
162 atomic_t use;
163 unsigned int mask;
164 struct net_device *physindev;
165 struct net_device *physoutdev;
166 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
167};
168#endif
169
1da177e4
LT
170struct sk_buff_head {
171 /* These two members must be first. */
172 struct sk_buff *next;
173 struct sk_buff *prev;
174
175 __u32 qlen;
176 spinlock_t lock;
177};
178
179struct sk_buff;
180
9d4dde52
IC
181/* To allow 64K frame to be packed as single skb without frag_list we
182 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
183 * buffers which do not start on a page boundary.
184 *
185 * Since GRO uses frags we allocate at least 16 regardless of page
186 * size.
a715dea3 187 */
9d4dde52 188#if (65536/PAGE_SIZE + 1) < 16
eec00954 189#define MAX_SKB_FRAGS 16UL
a715dea3 190#else
9d4dde52 191#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 192#endif
1da177e4
LT
193
194typedef struct skb_frag_struct skb_frag_t;
195
196struct skb_frag_struct {
a8605c60
IC
197 struct {
198 struct page *p;
199 } page;
cb4dfe56 200#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
201 __u32 page_offset;
202 __u32 size;
cb4dfe56
ED
203#else
204 __u16 page_offset;
205 __u16 size;
206#endif
1da177e4
LT
207};
208
9e903e08
ED
209static inline unsigned int skb_frag_size(const skb_frag_t *frag)
210{
211 return frag->size;
212}
213
214static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
215{
216 frag->size = size;
217}
218
219static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
220{
221 frag->size += delta;
222}
223
224static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
225{
226 frag->size -= delta;
227}
228
ac45f602
PO
229#define HAVE_HW_TIME_STAMP
230
231/**
d3a21be8 232 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
233 * @hwtstamp: hardware time stamp transformed into duration
234 * since arbitrary point in time
ac45f602
PO
235 *
236 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 237 * skb->tstamp.
ac45f602
PO
238 *
239 * hwtstamps can only be compared against other hwtstamps from
240 * the same device.
241 *
242 * This structure is attached to packets as part of the
243 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
244 */
245struct skb_shared_hwtstamps {
246 ktime_t hwtstamp;
ac45f602
PO
247};
248
2244d07b
OH
249/* Definitions for tx_flags in struct skb_shared_info */
250enum {
251 /* generate hardware time stamp */
252 SKBTX_HW_TSTAMP = 1 << 0,
253
e7fd2885 254 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
255 SKBTX_SW_TSTAMP = 1 << 1,
256
257 /* device driver is going to provide hardware time stamp */
258 SKBTX_IN_PROGRESS = 1 << 2,
259
a6686f2f 260 /* device driver supports TX zero-copy buffers */
62b1a8ab 261 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
262
263 /* generate wifi status information (where possible) */
62b1a8ab 264 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
265
266 /* This indicates at least one fragment might be overwritten
267 * (as in vmsplice(), sendfile() ...)
268 * If we need to compute a TX checksum, we'll need to copy
269 * all frags to avoid possible bad checksum
270 */
271 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
272
273 /* generate software time stamp when entering packet scheduling */
274 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
275
276 /* generate software timestamp on peer data acknowledgment */
277 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
278};
279
e1c8a607
WB
280#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
281 SKBTX_SCHED_TSTAMP | \
282 SKBTX_ACK_TSTAMP)
f24b9be5
WB
283#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
284
a6686f2f
SM
285/*
286 * The callback notifies userspace to release buffers when skb DMA is done in
287 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
288 * The zerocopy_success argument is true if zero copy transmit occurred,
289 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
290 * The ctx field is used to track device context.
291 * The desc field is used to track userspace buffer index.
a6686f2f
SM
292 */
293struct ubuf_info {
e19d6763 294 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 295 void *ctx;
a6686f2f 296 unsigned long desc;
ac45f602
PO
297};
298
1da177e4
LT
299/* This data is invariant across clones and lives at
300 * the end of the header data, ie. at skb->end.
301 */
302struct skb_shared_info {
9f42f126
IC
303 unsigned char nr_frags;
304 __u8 tx_flags;
7967168c
HX
305 unsigned short gso_size;
306 /* Warning: this field is not always filled in (UFO)! */
307 unsigned short gso_segs;
308 unsigned short gso_type;
1da177e4 309 struct sk_buff *frag_list;
ac45f602 310 struct skb_shared_hwtstamps hwtstamps;
09c2d251 311 u32 tskey;
9f42f126 312 __be32 ip6_frag_id;
ec7d2f2c
ED
313
314 /*
315 * Warning : all fields before dataref are cleared in __alloc_skb()
316 */
317 atomic_t dataref;
318
69e3c75f
JB
319 /* Intermediate layers must ensure that destructor_arg
320 * remains valid until skb destructor */
321 void * destructor_arg;
a6686f2f 322
fed66381
ED
323 /* must be last field, see pskb_expand_head() */
324 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
325};
326
327/* We divide dataref into two halves. The higher 16 bits hold references
328 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
329 * the entire skb->data. A clone of a headerless skb holds the length of
330 * the header in skb->hdr_len.
1da177e4
LT
331 *
332 * All users must obey the rule that the skb->data reference count must be
333 * greater than or equal to the payload reference count.
334 *
335 * Holding a reference to the payload part means that the user does not
336 * care about modifications to the header part of skb->data.
337 */
338#define SKB_DATAREF_SHIFT 16
339#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
340
d179cd12
DM
341
342enum {
c8753d55
VS
343 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
344 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
345 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
346 SKB_FCLONE_FREE, /* this companion fclone skb is available */
d179cd12
DM
347};
348
7967168c
HX
349enum {
350 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 351 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
352
353 /* This indicates the skb is from an untrusted source. */
354 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
355
356 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
357 SKB_GSO_TCP_ECN = 1 << 3,
358
359 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
360
361 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
362
363 SKB_GSO_GRE = 1 << 6,
73136267 364
4b28252c 365 SKB_GSO_GRE_CSUM = 1 << 7,
0d89d203 366
4b28252c 367 SKB_GSO_IPIP = 1 << 8,
cb32f511 368
4b28252c 369 SKB_GSO_SIT = 1 << 9,
61c1db7f 370
4b28252c 371 SKB_GSO_UDP_TUNNEL = 1 << 10,
0f4f4ffa
TH
372
373 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c 374
4b28252c
TH
375 SKB_GSO_MPLS = 1 << 12,
376
e585f236 377 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
7967168c
HX
378};
379
2e07fa9c
ACM
380#if BITS_PER_LONG > 32
381#define NET_SKBUFF_DATA_USES_OFFSET 1
382#endif
383
384#ifdef NET_SKBUFF_DATA_USES_OFFSET
385typedef unsigned int sk_buff_data_t;
386#else
387typedef unsigned char *sk_buff_data_t;
388#endif
389
363ec392
ED
390/**
391 * struct skb_mstamp - multi resolution time stamps
392 * @stamp_us: timestamp in us resolution
393 * @stamp_jiffies: timestamp in jiffies
394 */
395struct skb_mstamp {
396 union {
397 u64 v64;
398 struct {
399 u32 stamp_us;
400 u32 stamp_jiffies;
401 };
402 };
403};
404
405/**
406 * skb_mstamp_get - get current timestamp
407 * @cl: place to store timestamps
408 */
409static inline void skb_mstamp_get(struct skb_mstamp *cl)
410{
411 u64 val = local_clock();
412
413 do_div(val, NSEC_PER_USEC);
414 cl->stamp_us = (u32)val;
415 cl->stamp_jiffies = (u32)jiffies;
416}
417
418/**
419 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
420 * @t1: pointer to newest sample
421 * @t0: pointer to oldest sample
422 */
423static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
424 const struct skb_mstamp *t0)
425{
426 s32 delta_us = t1->stamp_us - t0->stamp_us;
427 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
428
429 /* If delta_us is negative, this might be because interval is too big,
430 * or local_clock() drift is too big : fallback using jiffies.
431 */
432 if (delta_us <= 0 ||
433 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
434
435 delta_us = jiffies_to_usecs(delta_jiffies);
436
437 return delta_us;
438}
439
440
1da177e4
LT
441/**
442 * struct sk_buff - socket buffer
443 * @next: Next buffer in list
444 * @prev: Previous buffer in list
363ec392 445 * @tstamp: Time we arrived/left
56b17425 446 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 447 * @sk: Socket we are owned by
1da177e4 448 * @dev: Device we arrived on/are leaving by
d84e0bd7 449 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 450 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 451 * @sp: the security path, used for xfrm
1da177e4
LT
452 * @len: Length of actual data
453 * @data_len: Data length
454 * @mac_len: Length of link layer header
334a8132 455 * @hdr_len: writable header length of cloned skb
663ead3b
HX
456 * @csum: Checksum (must include start/offset pair)
457 * @csum_start: Offset from skb->head where checksumming should start
458 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 459 * @priority: Packet queueing priority
60ff7467 460 * @ignore_df: allow local fragmentation
1da177e4 461 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 462 * @ip_summed: Driver fed us an IP checksum
1da177e4 463 * @nohdr: Payload reference only, must not modify header
d84e0bd7 464 * @nfctinfo: Relationship of this skb to the connection
1da177e4 465 * @pkt_type: Packet class
c83c2486 466 * @fclone: skbuff clone status
c83c2486 467 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
468 * @peeked: this packet has been seen already, so stats have been
469 * done for it, don't do them again
ba9dda3a 470 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
471 * @protocol: Packet protocol from driver
472 * @destructor: Destruct function
473 * @nfct: Associated connection, if any
1da177e4 474 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 475 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
476 * @tc_index: Traffic control index
477 * @tc_verd: traffic control verdict
61b905da 478 * @hash: the packet hash
d84e0bd7 479 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 480 * @xmit_more: More SKBs are pending for this queue
553a5672 481 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 482 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 483 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 484 * ports.
a3b18ddb 485 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
486 * @wifi_acked_valid: wifi_acked was set
487 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 488 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
06021292 489 * @napi_id: id of the NAPI struct this skb came from
984bc16c 490 * @secmark: security marking
d84e0bd7
DB
491 * @mark: Generic packet mark
492 * @dropcount: total number of sk_receive_queue overflows
86a9bad3 493 * @vlan_proto: vlan encapsulation protocol
6aa895b0 494 * @vlan_tci: vlan tag control information
0d89d203 495 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
496 * @inner_transport_header: Inner transport layer header (encapsulation)
497 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 498 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
499 * @transport_header: Transport layer header
500 * @network_header: Network layer header
501 * @mac_header: Link layer header
502 * @tail: Tail pointer
503 * @end: End pointer
504 * @head: Head of buffer
505 * @data: Data head pointer
506 * @truesize: Buffer size
507 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
508 */
509
510struct sk_buff {
363ec392 511 union {
56b17425
ED
512 struct {
513 /* These two members must be first. */
514 struct sk_buff *next;
515 struct sk_buff *prev;
516
517 union {
518 ktime_t tstamp;
519 struct skb_mstamp skb_mstamp;
520 };
521 };
522 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 523 };
da3f5cf1 524 struct sock *sk;
1da177e4 525 struct net_device *dev;
1da177e4 526
1da177e4
LT
527 /*
528 * This is the control buffer. It is free to use for every
529 * layer. Please put your private variables there. If you
530 * want to keep them across layers you have to do a skb_clone()
531 * first. This is owned by whoever has the skb queued ATM.
532 */
da3f5cf1 533 char cb[48] __aligned(8);
1da177e4 534
7fee226a 535 unsigned long _skb_refdst;
b1937227 536 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
537#ifdef CONFIG_XFRM
538 struct sec_path *sp;
b1937227
ED
539#endif
540#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
541 struct nf_conntrack *nfct;
542#endif
85224844 543#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 544 struct nf_bridge_info *nf_bridge;
da3f5cf1 545#endif
1da177e4 546 unsigned int len,
334a8132
PM
547 data_len;
548 __u16 mac_len,
549 hdr_len;
b1937227
ED
550
551 /* Following fields are _not_ copied in __copy_skb_header()
552 * Note that queue_mapping is here mostly to fill a hole.
553 */
fe55f6d5 554 kmemcheck_bitfield_begin(flags1);
b1937227
ED
555 __u16 queue_mapping;
556 __u8 cloned:1,
6869c4d8 557 nohdr:1,
b84f4cc9 558 fclone:2,
a59322be 559 peeked:1,
b1937227
ED
560 head_frag:1,
561 xmit_more:1;
562 /* one bit hole */
fe55f6d5 563 kmemcheck_bitfield_end(flags1);
4031ae6e 564
b1937227
ED
565 /* fields enclosed in headers_start/headers_end are copied
566 * using a single memcpy() in __copy_skb_header()
567 */
ebcf34f3 568 /* private: */
b1937227 569 __u32 headers_start[0];
ebcf34f3 570 /* public: */
4031ae6e 571
233577a2
HFS
572/* if you move pkt_type around you also must adapt those constants */
573#ifdef __BIG_ENDIAN_BITFIELD
574#define PKT_TYPE_MAX (7 << 5)
575#else
576#define PKT_TYPE_MAX 7
1da177e4 577#endif
233577a2 578#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 579
233577a2 580 __u8 __pkt_type_offset[0];
b1937227 581 __u8 pkt_type:3;
c93bdd0e 582 __u8 pfmemalloc:1;
b1937227
ED
583 __u8 ignore_df:1;
584 __u8 nfctinfo:3;
585
586 __u8 nf_trace:1;
587 __u8 ip_summed:2;
3853b584 588 __u8 ooo_okay:1;
61b905da 589 __u8 l4_hash:1;
a3b18ddb 590 __u8 sw_hash:1;
6e3e939f
JB
591 __u8 wifi_acked_valid:1;
592 __u8 wifi_acked:1;
b1937227 593
3bdc0eba 594 __u8 no_fcs:1;
77cffe23 595 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 596 __u8 encapsulation:1;
7e2b10c1 597 __u8 encap_hdr_csum:1;
5d0c2b95 598 __u8 csum_valid:1;
7e3cead5 599 __u8 csum_complete_sw:1;
b1937227
ED
600 __u8 csum_level:2;
601 __u8 csum_bad:1;
fe55f6d5 602
b1937227
ED
603#ifdef CONFIG_IPV6_NDISC_NODETYPE
604 __u8 ndisc_nodetype:2;
605#endif
606 __u8 ipvs_property:1;
8bce6d7d 607 __u8 inner_protocol_type:1;
e585f236
TH
608 __u8 remcsum_offload:1;
609 /* 3 or 5 bit hole */
b1937227
ED
610
611#ifdef CONFIG_NET_SCHED
612 __u16 tc_index; /* traffic control index */
613#ifdef CONFIG_NET_CLS_ACT
614 __u16 tc_verd; /* traffic control verdict */
615#endif
616#endif
fe55f6d5 617
b1937227
ED
618 union {
619 __wsum csum;
620 struct {
621 __u16 csum_start;
622 __u16 csum_offset;
623 };
624 };
625 __u32 priority;
626 int skb_iif;
627 __u32 hash;
628 __be16 vlan_proto;
629 __u16 vlan_tci;
7bced397
DW
630#ifdef CONFIG_NET_RX_BUSY_POLL
631 unsigned int napi_id;
97fc2f08 632#endif
984bc16c
JM
633#ifdef CONFIG_NETWORK_SECMARK
634 __u32 secmark;
635#endif
3b885787
NH
636 union {
637 __u32 mark;
638 __u32 dropcount;
16fad69c 639 __u32 reserved_tailroom;
3b885787 640 };
1da177e4 641
8bce6d7d
TH
642 union {
643 __be16 inner_protocol;
644 __u8 inner_ipproto;
645 };
646
1a37e412
SH
647 __u16 inner_transport_header;
648 __u16 inner_network_header;
649 __u16 inner_mac_header;
b1937227
ED
650
651 __be16 protocol;
1a37e412
SH
652 __u16 transport_header;
653 __u16 network_header;
654 __u16 mac_header;
b1937227 655
ebcf34f3 656 /* private: */
b1937227 657 __u32 headers_end[0];
ebcf34f3 658 /* public: */
b1937227 659
1da177e4 660 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 661 sk_buff_data_t tail;
4305b541 662 sk_buff_data_t end;
1da177e4 663 unsigned char *head,
4305b541 664 *data;
27a884dc
ACM
665 unsigned int truesize;
666 atomic_t users;
1da177e4
LT
667};
668
669#ifdef __KERNEL__
670/*
671 * Handling routines are only of interest to the kernel
672 */
673#include <linux/slab.h>
674
1da177e4 675
c93bdd0e
MG
676#define SKB_ALLOC_FCLONE 0x01
677#define SKB_ALLOC_RX 0x02
678
679/* Returns true if the skb was allocated from PFMEMALLOC reserves */
680static inline bool skb_pfmemalloc(const struct sk_buff *skb)
681{
682 return unlikely(skb->pfmemalloc);
683}
684
7fee226a
ED
685/*
686 * skb might have a dst pointer attached, refcounted or not.
687 * _skb_refdst low order bit is set if refcount was _not_ taken
688 */
689#define SKB_DST_NOREF 1UL
690#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
691
692/**
693 * skb_dst - returns skb dst_entry
694 * @skb: buffer
695 *
696 * Returns skb dst_entry, regardless of reference taken or not.
697 */
adf30907
ED
698static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
699{
7fee226a
ED
700 /* If refdst was not refcounted, check we still are in a
701 * rcu_read_lock section
702 */
703 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
704 !rcu_read_lock_held() &&
705 !rcu_read_lock_bh_held());
706 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
707}
708
7fee226a
ED
709/**
710 * skb_dst_set - sets skb dst
711 * @skb: buffer
712 * @dst: dst entry
713 *
714 * Sets skb dst, assuming a reference was taken on dst and should
715 * be released by skb_dst_drop()
716 */
adf30907
ED
717static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
718{
7fee226a
ED
719 skb->_skb_refdst = (unsigned long)dst;
720}
721
7965bd4d
JP
722void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
723 bool force);
932bc4d7
JA
724
725/**
726 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
727 * @skb: buffer
728 * @dst: dst entry
729 *
730 * Sets skb dst, assuming a reference was not taken on dst.
731 * If dst entry is cached, we do not take reference and dst_release
732 * will be avoided by refdst_drop. If dst entry is not cached, we take
733 * reference, so that last dst_release can destroy the dst immediately.
734 */
735static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
736{
737 __skb_dst_set_noref(skb, dst, false);
738}
739
740/**
741 * skb_dst_set_noref_force - sets skb dst, without taking reference
742 * @skb: buffer
743 * @dst: dst entry
744 *
745 * Sets skb dst, assuming a reference was not taken on dst.
746 * No reference is taken and no dst_release will be called. While for
747 * cached dsts deferred reclaim is a basic feature, for entries that are
748 * not cached it is caller's job to guarantee that last dst_release for
749 * provided dst happens when nobody uses it, eg. after a RCU grace period.
750 */
751static inline void skb_dst_set_noref_force(struct sk_buff *skb,
752 struct dst_entry *dst)
753{
754 __skb_dst_set_noref(skb, dst, true);
755}
7fee226a
ED
756
757/**
25985edc 758 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
759 * @skb: buffer
760 */
761static inline bool skb_dst_is_noref(const struct sk_buff *skb)
762{
763 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
764}
765
511c3f92
ED
766static inline struct rtable *skb_rtable(const struct sk_buff *skb)
767{
adf30907 768 return (struct rtable *)skb_dst(skb);
511c3f92
ED
769}
770
7965bd4d
JP
771void kfree_skb(struct sk_buff *skb);
772void kfree_skb_list(struct sk_buff *segs);
773void skb_tx_error(struct sk_buff *skb);
774void consume_skb(struct sk_buff *skb);
775void __kfree_skb(struct sk_buff *skb);
d7e8883c 776extern struct kmem_cache *skbuff_head_cache;
bad43ca8 777
7965bd4d
JP
778void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
779bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
780 bool *fragstolen, int *delta_truesize);
bad43ca8 781
7965bd4d
JP
782struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
783 int node);
784struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 785static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 786 gfp_t priority)
d179cd12 787{
564824b0 788 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
789}
790
2e4e4410
ED
791struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
792 unsigned long data_len,
793 int max_page_order,
794 int *errcode,
795 gfp_t gfp_mask);
796
d0bf4a9e
ED
797/* Layout of fast clones : [skb1][skb2][fclone_ref] */
798struct sk_buff_fclones {
799 struct sk_buff skb1;
800
801 struct sk_buff skb2;
802
803 atomic_t fclone_ref;
804};
805
806/**
807 * skb_fclone_busy - check if fclone is busy
808 * @skb: buffer
809 *
810 * Returns true is skb is a fast clone, and its clone is not freed.
39bb5e62
ED
811 * Some drivers call skb_orphan() in their ndo_start_xmit(),
812 * so we also check that this didnt happen.
d0bf4a9e 813 */
39bb5e62
ED
814static inline bool skb_fclone_busy(const struct sock *sk,
815 const struct sk_buff *skb)
d0bf4a9e
ED
816{
817 const struct sk_buff_fclones *fclones;
818
819 fclones = container_of(skb, struct sk_buff_fclones, skb1);
820
821 return skb->fclone == SKB_FCLONE_ORIG &&
39bb5e62
ED
822 fclones->skb2.fclone == SKB_FCLONE_CLONE &&
823 fclones->skb2.sk == sk;
d0bf4a9e
ED
824}
825
d179cd12 826static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 827 gfp_t priority)
d179cd12 828{
c93bdd0e 829 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
830}
831
7965bd4d 832struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
833static inline struct sk_buff *alloc_skb_head(gfp_t priority)
834{
835 return __alloc_skb_head(priority, -1);
836}
837
7965bd4d
JP
838struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
839int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
840struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
841struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
842struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
843 gfp_t gfp_mask, bool fclone);
844static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
845 gfp_t gfp_mask)
846{
847 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
848}
7965bd4d
JP
849
850int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
851struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
852 unsigned int headroom);
853struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
854 int newtailroom, gfp_t priority);
25a91d8d
FD
855int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
856 int offset, int len);
7965bd4d
JP
857int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
858 int len);
859int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
860int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 861#define dev_kfree_skb(a) consume_skb(a)
1da177e4 862
7965bd4d
JP
863int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
864 int getfrag(void *from, char *to, int offset,
865 int len, int odd, struct sk_buff *skb),
866 void *from, int length);
e89e9cf5 867
d94d9fee 868struct skb_seq_state {
677e90ed
TG
869 __u32 lower_offset;
870 __u32 upper_offset;
871 __u32 frag_idx;
872 __u32 stepped_offset;
873 struct sk_buff *root_skb;
874 struct sk_buff *cur_skb;
875 __u8 *frag_data;
876};
877
7965bd4d
JP
878void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
879 unsigned int to, struct skb_seq_state *st);
880unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
881 struct skb_seq_state *st);
882void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 883
7965bd4d
JP
884unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
885 unsigned int to, struct ts_config *config,
886 struct ts_state *state);
3fc7e8a6 887
09323cc4
TH
888/*
889 * Packet hash types specify the type of hash in skb_set_hash.
890 *
891 * Hash types refer to the protocol layer addresses which are used to
892 * construct a packet's hash. The hashes are used to differentiate or identify
893 * flows of the protocol layer for the hash type. Hash types are either
894 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
895 *
896 * Properties of hashes:
897 *
898 * 1) Two packets in different flows have different hash values
899 * 2) Two packets in the same flow should have the same hash value
900 *
901 * A hash at a higher layer is considered to be more specific. A driver should
902 * set the most specific hash possible.
903 *
904 * A driver cannot indicate a more specific hash than the layer at which a hash
905 * was computed. For instance an L3 hash cannot be set as an L4 hash.
906 *
907 * A driver may indicate a hash level which is less specific than the
908 * actual layer the hash was computed on. For instance, a hash computed
909 * at L4 may be considered an L3 hash. This should only be done if the
910 * driver can't unambiguously determine that the HW computed the hash at
911 * the higher layer. Note that the "should" in the second property above
912 * permits this.
913 */
914enum pkt_hash_types {
915 PKT_HASH_TYPE_NONE, /* Undefined type */
916 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
917 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
918 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
919};
920
921static inline void
922skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
923{
61b905da 924 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 925 skb->sw_hash = 0;
61b905da 926 skb->hash = hash;
09323cc4
TH
927}
928
3958afa1
TH
929void __skb_get_hash(struct sk_buff *skb);
930static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 931{
a3b18ddb 932 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 933 __skb_get_hash(skb);
bfb564e7 934
61b905da 935 return skb->hash;
bfb564e7
KK
936}
937
57bdf7f4
TH
938static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
939{
61b905da 940 return skb->hash;
57bdf7f4
TH
941}
942
7539fadc
TH
943static inline void skb_clear_hash(struct sk_buff *skb)
944{
61b905da 945 skb->hash = 0;
a3b18ddb 946 skb->sw_hash = 0;
61b905da 947 skb->l4_hash = 0;
7539fadc
TH
948}
949
950static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
951{
61b905da 952 if (!skb->l4_hash)
7539fadc
TH
953 skb_clear_hash(skb);
954}
955
3df7a74e
TH
956static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
957{
61b905da 958 to->hash = from->hash;
a3b18ddb 959 to->sw_hash = from->sw_hash;
61b905da 960 to->l4_hash = from->l4_hash;
3df7a74e
TH
961};
962
4305b541
ACM
963#ifdef NET_SKBUFF_DATA_USES_OFFSET
964static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
965{
966 return skb->head + skb->end;
967}
ec47ea82
AD
968
969static inline unsigned int skb_end_offset(const struct sk_buff *skb)
970{
971 return skb->end;
972}
4305b541
ACM
973#else
974static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
975{
976 return skb->end;
977}
ec47ea82
AD
978
979static inline unsigned int skb_end_offset(const struct sk_buff *skb)
980{
981 return skb->end - skb->head;
982}
4305b541
ACM
983#endif
984
1da177e4 985/* Internal */
4305b541 986#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 987
ac45f602
PO
988static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
989{
990 return &skb_shinfo(skb)->hwtstamps;
991}
992
1da177e4
LT
993/**
994 * skb_queue_empty - check if a queue is empty
995 * @list: queue head
996 *
997 * Returns true if the queue is empty, false otherwise.
998 */
999static inline int skb_queue_empty(const struct sk_buff_head *list)
1000{
fd44b93c 1001 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1002}
1003
fc7ebb21
DM
1004/**
1005 * skb_queue_is_last - check if skb is the last entry in the queue
1006 * @list: queue head
1007 * @skb: buffer
1008 *
1009 * Returns true if @skb is the last buffer on the list.
1010 */
1011static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1012 const struct sk_buff *skb)
1013{
fd44b93c 1014 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1015}
1016
832d11c5
IJ
1017/**
1018 * skb_queue_is_first - check if skb is the first entry in the queue
1019 * @list: queue head
1020 * @skb: buffer
1021 *
1022 * Returns true if @skb is the first buffer on the list.
1023 */
1024static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1025 const struct sk_buff *skb)
1026{
fd44b93c 1027 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1028}
1029
249c8b42
DM
1030/**
1031 * skb_queue_next - return the next packet in the queue
1032 * @list: queue head
1033 * @skb: current buffer
1034 *
1035 * Return the next packet in @list after @skb. It is only valid to
1036 * call this if skb_queue_is_last() evaluates to false.
1037 */
1038static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1039 const struct sk_buff *skb)
1040{
1041 /* This BUG_ON may seem severe, but if we just return then we
1042 * are going to dereference garbage.
1043 */
1044 BUG_ON(skb_queue_is_last(list, skb));
1045 return skb->next;
1046}
1047
832d11c5
IJ
1048/**
1049 * skb_queue_prev - return the prev packet in the queue
1050 * @list: queue head
1051 * @skb: current buffer
1052 *
1053 * Return the prev packet in @list before @skb. It is only valid to
1054 * call this if skb_queue_is_first() evaluates to false.
1055 */
1056static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1057 const struct sk_buff *skb)
1058{
1059 /* This BUG_ON may seem severe, but if we just return then we
1060 * are going to dereference garbage.
1061 */
1062 BUG_ON(skb_queue_is_first(list, skb));
1063 return skb->prev;
1064}
1065
1da177e4
LT
1066/**
1067 * skb_get - reference buffer
1068 * @skb: buffer to reference
1069 *
1070 * Makes another reference to a socket buffer and returns a pointer
1071 * to the buffer.
1072 */
1073static inline struct sk_buff *skb_get(struct sk_buff *skb)
1074{
1075 atomic_inc(&skb->users);
1076 return skb;
1077}
1078
1079/*
1080 * If users == 1, we are the only owner and are can avoid redundant
1081 * atomic change.
1082 */
1083
1da177e4
LT
1084/**
1085 * skb_cloned - is the buffer a clone
1086 * @skb: buffer to check
1087 *
1088 * Returns true if the buffer was generated with skb_clone() and is
1089 * one of multiple shared copies of the buffer. Cloned buffers are
1090 * shared data so must not be written to under normal circumstances.
1091 */
1092static inline int skb_cloned(const struct sk_buff *skb)
1093{
1094 return skb->cloned &&
1095 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1096}
1097
14bbd6a5
PS
1098static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1099{
1100 might_sleep_if(pri & __GFP_WAIT);
1101
1102 if (skb_cloned(skb))
1103 return pskb_expand_head(skb, 0, 0, pri);
1104
1105 return 0;
1106}
1107
1da177e4
LT
1108/**
1109 * skb_header_cloned - is the header a clone
1110 * @skb: buffer to check
1111 *
1112 * Returns true if modifying the header part of the buffer requires
1113 * the data to be copied.
1114 */
1115static inline int skb_header_cloned(const struct sk_buff *skb)
1116{
1117 int dataref;
1118
1119 if (!skb->cloned)
1120 return 0;
1121
1122 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1123 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1124 return dataref != 1;
1125}
1126
1127/**
1128 * skb_header_release - release reference to header
1129 * @skb: buffer to operate on
1130 *
1131 * Drop a reference to the header part of the buffer. This is done
1132 * by acquiring a payload reference. You must not read from the header
1133 * part of skb->data after this.
f4a775d1 1134 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1135 */
1136static inline void skb_header_release(struct sk_buff *skb)
1137{
1138 BUG_ON(skb->nohdr);
1139 skb->nohdr = 1;
1140 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1141}
1142
f4a775d1
ED
1143/**
1144 * __skb_header_release - release reference to header
1145 * @skb: buffer to operate on
1146 *
1147 * Variant of skb_header_release() assuming skb is private to caller.
1148 * We can avoid one atomic operation.
1149 */
1150static inline void __skb_header_release(struct sk_buff *skb)
1151{
1152 skb->nohdr = 1;
1153 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1154}
1155
1156
1da177e4
LT
1157/**
1158 * skb_shared - is the buffer shared
1159 * @skb: buffer to check
1160 *
1161 * Returns true if more than one person has a reference to this
1162 * buffer.
1163 */
1164static inline int skb_shared(const struct sk_buff *skb)
1165{
1166 return atomic_read(&skb->users) != 1;
1167}
1168
1169/**
1170 * skb_share_check - check if buffer is shared and if so clone it
1171 * @skb: buffer to check
1172 * @pri: priority for memory allocation
1173 *
1174 * If the buffer is shared the buffer is cloned and the old copy
1175 * drops a reference. A new clone with a single reference is returned.
1176 * If the buffer is not shared the original buffer is returned. When
1177 * being called from interrupt status or with spinlocks held pri must
1178 * be GFP_ATOMIC.
1179 *
1180 * NULL is returned on a memory allocation failure.
1181 */
47061bc4 1182static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1183{
1184 might_sleep_if(pri & __GFP_WAIT);
1185 if (skb_shared(skb)) {
1186 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1187
1188 if (likely(nskb))
1189 consume_skb(skb);
1190 else
1191 kfree_skb(skb);
1da177e4
LT
1192 skb = nskb;
1193 }
1194 return skb;
1195}
1196
1197/*
1198 * Copy shared buffers into a new sk_buff. We effectively do COW on
1199 * packets to handle cases where we have a local reader and forward
1200 * and a couple of other messy ones. The normal one is tcpdumping
1201 * a packet thats being forwarded.
1202 */
1203
1204/**
1205 * skb_unshare - make a copy of a shared buffer
1206 * @skb: buffer to check
1207 * @pri: priority for memory allocation
1208 *
1209 * If the socket buffer is a clone then this function creates a new
1210 * copy of the data, drops a reference count on the old copy and returns
1211 * the new copy with the reference count at 1. If the buffer is not a clone
1212 * the original buffer is returned. When called with a spinlock held or
1213 * from interrupt state @pri must be %GFP_ATOMIC
1214 *
1215 * %NULL is returned on a memory allocation failure.
1216 */
e2bf521d 1217static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1218 gfp_t pri)
1da177e4
LT
1219{
1220 might_sleep_if(pri & __GFP_WAIT);
1221 if (skb_cloned(skb)) {
1222 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1223
1224 /* Free our shared copy */
1225 if (likely(nskb))
1226 consume_skb(skb);
1227 else
1228 kfree_skb(skb);
1da177e4
LT
1229 skb = nskb;
1230 }
1231 return skb;
1232}
1233
1234/**
1a5778aa 1235 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1236 * @list_: list to peek at
1237 *
1238 * Peek an &sk_buff. Unlike most other operations you _MUST_
1239 * be careful with this one. A peek leaves the buffer on the
1240 * list and someone else may run off with it. You must hold
1241 * the appropriate locks or have a private queue to do this.
1242 *
1243 * Returns %NULL for an empty list or a pointer to the head element.
1244 * The reference count is not incremented and the reference is therefore
1245 * volatile. Use with caution.
1246 */
05bdd2f1 1247static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1248{
18d07000
ED
1249 struct sk_buff *skb = list_->next;
1250
1251 if (skb == (struct sk_buff *)list_)
1252 skb = NULL;
1253 return skb;
1da177e4
LT
1254}
1255
da5ef6e5
PE
1256/**
1257 * skb_peek_next - peek skb following the given one from a queue
1258 * @skb: skb to start from
1259 * @list_: list to peek at
1260 *
1261 * Returns %NULL when the end of the list is met or a pointer to the
1262 * next element. The reference count is not incremented and the
1263 * reference is therefore volatile. Use with caution.
1264 */
1265static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1266 const struct sk_buff_head *list_)
1267{
1268 struct sk_buff *next = skb->next;
18d07000 1269
da5ef6e5
PE
1270 if (next == (struct sk_buff *)list_)
1271 next = NULL;
1272 return next;
1273}
1274
1da177e4 1275/**
1a5778aa 1276 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1277 * @list_: list to peek at
1278 *
1279 * Peek an &sk_buff. Unlike most other operations you _MUST_
1280 * be careful with this one. A peek leaves the buffer on the
1281 * list and someone else may run off with it. You must hold
1282 * the appropriate locks or have a private queue to do this.
1283 *
1284 * Returns %NULL for an empty list or a pointer to the tail element.
1285 * The reference count is not incremented and the reference is therefore
1286 * volatile. Use with caution.
1287 */
05bdd2f1 1288static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1289{
18d07000
ED
1290 struct sk_buff *skb = list_->prev;
1291
1292 if (skb == (struct sk_buff *)list_)
1293 skb = NULL;
1294 return skb;
1295
1da177e4
LT
1296}
1297
1298/**
1299 * skb_queue_len - get queue length
1300 * @list_: list to measure
1301 *
1302 * Return the length of an &sk_buff queue.
1303 */
1304static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1305{
1306 return list_->qlen;
1307}
1308
67fed459
DM
1309/**
1310 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1311 * @list: queue to initialize
1312 *
1313 * This initializes only the list and queue length aspects of
1314 * an sk_buff_head object. This allows to initialize the list
1315 * aspects of an sk_buff_head without reinitializing things like
1316 * the spinlock. It can also be used for on-stack sk_buff_head
1317 * objects where the spinlock is known to not be used.
1318 */
1319static inline void __skb_queue_head_init(struct sk_buff_head *list)
1320{
1321 list->prev = list->next = (struct sk_buff *)list;
1322 list->qlen = 0;
1323}
1324
76f10ad0
AV
1325/*
1326 * This function creates a split out lock class for each invocation;
1327 * this is needed for now since a whole lot of users of the skb-queue
1328 * infrastructure in drivers have different locking usage (in hardirq)
1329 * than the networking core (in softirq only). In the long run either the
1330 * network layer or drivers should need annotation to consolidate the
1331 * main types of usage into 3 classes.
1332 */
1da177e4
LT
1333static inline void skb_queue_head_init(struct sk_buff_head *list)
1334{
1335 spin_lock_init(&list->lock);
67fed459 1336 __skb_queue_head_init(list);
1da177e4
LT
1337}
1338
c2ecba71
PE
1339static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1340 struct lock_class_key *class)
1341{
1342 skb_queue_head_init(list);
1343 lockdep_set_class(&list->lock, class);
1344}
1345
1da177e4 1346/*
bf299275 1347 * Insert an sk_buff on a list.
1da177e4
LT
1348 *
1349 * The "__skb_xxxx()" functions are the non-atomic ones that
1350 * can only be called with interrupts disabled.
1351 */
7965bd4d
JP
1352void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1353 struct sk_buff_head *list);
bf299275
GR
1354static inline void __skb_insert(struct sk_buff *newsk,
1355 struct sk_buff *prev, struct sk_buff *next,
1356 struct sk_buff_head *list)
1357{
1358 newsk->next = next;
1359 newsk->prev = prev;
1360 next->prev = prev->next = newsk;
1361 list->qlen++;
1362}
1da177e4 1363
67fed459
DM
1364static inline void __skb_queue_splice(const struct sk_buff_head *list,
1365 struct sk_buff *prev,
1366 struct sk_buff *next)
1367{
1368 struct sk_buff *first = list->next;
1369 struct sk_buff *last = list->prev;
1370
1371 first->prev = prev;
1372 prev->next = first;
1373
1374 last->next = next;
1375 next->prev = last;
1376}
1377
1378/**
1379 * skb_queue_splice - join two skb lists, this is designed for stacks
1380 * @list: the new list to add
1381 * @head: the place to add it in the first list
1382 */
1383static inline void skb_queue_splice(const struct sk_buff_head *list,
1384 struct sk_buff_head *head)
1385{
1386 if (!skb_queue_empty(list)) {
1387 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1388 head->qlen += list->qlen;
67fed459
DM
1389 }
1390}
1391
1392/**
d9619496 1393 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1394 * @list: the new list to add
1395 * @head: the place to add it in the first list
1396 *
1397 * The list at @list is reinitialised
1398 */
1399static inline void skb_queue_splice_init(struct sk_buff_head *list,
1400 struct sk_buff_head *head)
1401{
1402 if (!skb_queue_empty(list)) {
1403 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1404 head->qlen += list->qlen;
67fed459
DM
1405 __skb_queue_head_init(list);
1406 }
1407}
1408
1409/**
1410 * skb_queue_splice_tail - join two skb lists, each list being a queue
1411 * @list: the new list to add
1412 * @head: the place to add it in the first list
1413 */
1414static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1415 struct sk_buff_head *head)
1416{
1417 if (!skb_queue_empty(list)) {
1418 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1419 head->qlen += list->qlen;
67fed459
DM
1420 }
1421}
1422
1423/**
d9619496 1424 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1425 * @list: the new list to add
1426 * @head: the place to add it in the first list
1427 *
1428 * Each of the lists is a queue.
1429 * The list at @list is reinitialised
1430 */
1431static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1432 struct sk_buff_head *head)
1433{
1434 if (!skb_queue_empty(list)) {
1435 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1436 head->qlen += list->qlen;
67fed459
DM
1437 __skb_queue_head_init(list);
1438 }
1439}
1440
1da177e4 1441/**
300ce174 1442 * __skb_queue_after - queue a buffer at the list head
1da177e4 1443 * @list: list to use
300ce174 1444 * @prev: place after this buffer
1da177e4
LT
1445 * @newsk: buffer to queue
1446 *
300ce174 1447 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1448 * and you must therefore hold required locks before calling it.
1449 *
1450 * A buffer cannot be placed on two lists at the same time.
1451 */
300ce174
SH
1452static inline void __skb_queue_after(struct sk_buff_head *list,
1453 struct sk_buff *prev,
1454 struct sk_buff *newsk)
1da177e4 1455{
bf299275 1456 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1457}
1458
7965bd4d
JP
1459void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1460 struct sk_buff_head *list);
7de6c033 1461
f5572855
GR
1462static inline void __skb_queue_before(struct sk_buff_head *list,
1463 struct sk_buff *next,
1464 struct sk_buff *newsk)
1465{
1466 __skb_insert(newsk, next->prev, next, list);
1467}
1468
300ce174
SH
1469/**
1470 * __skb_queue_head - queue a buffer at the list head
1471 * @list: list to use
1472 * @newsk: buffer to queue
1473 *
1474 * Queue a buffer at the start of a list. This function takes no locks
1475 * and you must therefore hold required locks before calling it.
1476 *
1477 * A buffer cannot be placed on two lists at the same time.
1478 */
7965bd4d 1479void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1480static inline void __skb_queue_head(struct sk_buff_head *list,
1481 struct sk_buff *newsk)
1482{
1483 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1484}
1485
1da177e4
LT
1486/**
1487 * __skb_queue_tail - queue a buffer at the list tail
1488 * @list: list to use
1489 * @newsk: buffer to queue
1490 *
1491 * Queue a buffer at the end of a list. This function takes no locks
1492 * and you must therefore hold required locks before calling it.
1493 *
1494 * A buffer cannot be placed on two lists at the same time.
1495 */
7965bd4d 1496void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1497static inline void __skb_queue_tail(struct sk_buff_head *list,
1498 struct sk_buff *newsk)
1499{
f5572855 1500 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1501}
1502
1da177e4
LT
1503/*
1504 * remove sk_buff from list. _Must_ be called atomically, and with
1505 * the list known..
1506 */
7965bd4d 1507void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1508static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1509{
1510 struct sk_buff *next, *prev;
1511
1512 list->qlen--;
1513 next = skb->next;
1514 prev = skb->prev;
1515 skb->next = skb->prev = NULL;
1da177e4
LT
1516 next->prev = prev;
1517 prev->next = next;
1518}
1519
f525c06d
GR
1520/**
1521 * __skb_dequeue - remove from the head of the queue
1522 * @list: list to dequeue from
1523 *
1524 * Remove the head of the list. This function does not take any locks
1525 * so must be used with appropriate locks held only. The head item is
1526 * returned or %NULL if the list is empty.
1527 */
7965bd4d 1528struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1529static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1530{
1531 struct sk_buff *skb = skb_peek(list);
1532 if (skb)
1533 __skb_unlink(skb, list);
1534 return skb;
1535}
1da177e4
LT
1536
1537/**
1538 * __skb_dequeue_tail - remove from the tail of the queue
1539 * @list: list to dequeue from
1540 *
1541 * Remove the tail of the list. This function does not take any locks
1542 * so must be used with appropriate locks held only. The tail item is
1543 * returned or %NULL if the list is empty.
1544 */
7965bd4d 1545struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1546static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1547{
1548 struct sk_buff *skb = skb_peek_tail(list);
1549 if (skb)
1550 __skb_unlink(skb, list);
1551 return skb;
1552}
1553
1554
bdcc0924 1555static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1556{
1557 return skb->data_len;
1558}
1559
1560static inline unsigned int skb_headlen(const struct sk_buff *skb)
1561{
1562 return skb->len - skb->data_len;
1563}
1564
1565static inline int skb_pagelen(const struct sk_buff *skb)
1566{
1567 int i, len = 0;
1568
1569 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1570 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1571 return len + skb_headlen(skb);
1572}
1573
131ea667
IC
1574/**
1575 * __skb_fill_page_desc - initialise a paged fragment in an skb
1576 * @skb: buffer containing fragment to be initialised
1577 * @i: paged fragment index to initialise
1578 * @page: the page to use for this fragment
1579 * @off: the offset to the data with @page
1580 * @size: the length of the data
1581 *
1582 * Initialises the @i'th fragment of @skb to point to &size bytes at
1583 * offset @off within @page.
1584 *
1585 * Does not take any additional reference on the fragment.
1586 */
1587static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1588 struct page *page, int off, int size)
1da177e4
LT
1589{
1590 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1591
c48a11c7
MG
1592 /*
1593 * Propagate page->pfmemalloc to the skb if we can. The problem is
1594 * that not all callers have unique ownership of the page. If
1595 * pfmemalloc is set, we check the mapping as a mapping implies
1596 * page->index is set (index and pfmemalloc share space).
1597 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1598 * do not lose pfmemalloc information as the pages would not be
1599 * allocated using __GFP_MEMALLOC.
1600 */
a8605c60 1601 frag->page.p = page;
1da177e4 1602 frag->page_offset = off;
9e903e08 1603 skb_frag_size_set(frag, size);
cca7af38
PE
1604
1605 page = compound_head(page);
1606 if (page->pfmemalloc && !page->mapping)
1607 skb->pfmemalloc = true;
131ea667
IC
1608}
1609
1610/**
1611 * skb_fill_page_desc - initialise a paged fragment in an skb
1612 * @skb: buffer containing fragment to be initialised
1613 * @i: paged fragment index to initialise
1614 * @page: the page to use for this fragment
1615 * @off: the offset to the data with @page
1616 * @size: the length of the data
1617 *
1618 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1619 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1620 * addition updates @skb such that @i is the last fragment.
1621 *
1622 * Does not take any additional reference on the fragment.
1623 */
1624static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1625 struct page *page, int off, int size)
1626{
1627 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1628 skb_shinfo(skb)->nr_frags = i + 1;
1629}
1630
7965bd4d
JP
1631void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1632 int size, unsigned int truesize);
654bed16 1633
f8e617e1
JW
1634void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1635 unsigned int truesize);
1636
1da177e4 1637#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1638#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1639#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1640
27a884dc
ACM
1641#ifdef NET_SKBUFF_DATA_USES_OFFSET
1642static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1643{
1644 return skb->head + skb->tail;
1645}
1646
1647static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1648{
1649 skb->tail = skb->data - skb->head;
1650}
1651
1652static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1653{
1654 skb_reset_tail_pointer(skb);
1655 skb->tail += offset;
1656}
7cc46190 1657
27a884dc
ACM
1658#else /* NET_SKBUFF_DATA_USES_OFFSET */
1659static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1660{
1661 return skb->tail;
1662}
1663
1664static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1665{
1666 skb->tail = skb->data;
1667}
1668
1669static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1670{
1671 skb->tail = skb->data + offset;
1672}
4305b541 1673
27a884dc
ACM
1674#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1675
1da177e4
LT
1676/*
1677 * Add data to an sk_buff
1678 */
0c7ddf36 1679unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1680unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1681static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1682{
27a884dc 1683 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1684 SKB_LINEAR_ASSERT(skb);
1685 skb->tail += len;
1686 skb->len += len;
1687 return tmp;
1688}
1689
7965bd4d 1690unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1691static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1692{
1693 skb->data -= len;
1694 skb->len += len;
1695 return skb->data;
1696}
1697
7965bd4d 1698unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1699static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1700{
1701 skb->len -= len;
1702 BUG_ON(skb->len < skb->data_len);
1703 return skb->data += len;
1704}
1705
47d29646
DM
1706static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1707{
1708 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1709}
1710
7965bd4d 1711unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1712
1713static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1714{
1715 if (len > skb_headlen(skb) &&
987c402a 1716 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1717 return NULL;
1718 skb->len -= len;
1719 return skb->data += len;
1720}
1721
1722static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1723{
1724 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1725}
1726
1727static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1728{
1729 if (likely(len <= skb_headlen(skb)))
1730 return 1;
1731 if (unlikely(len > skb->len))
1732 return 0;
987c402a 1733 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1734}
1735
1736/**
1737 * skb_headroom - bytes at buffer head
1738 * @skb: buffer to check
1739 *
1740 * Return the number of bytes of free space at the head of an &sk_buff.
1741 */
c2636b4d 1742static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1743{
1744 return skb->data - skb->head;
1745}
1746
1747/**
1748 * skb_tailroom - bytes at buffer end
1749 * @skb: buffer to check
1750 *
1751 * Return the number of bytes of free space at the tail of an sk_buff
1752 */
1753static inline int skb_tailroom(const struct sk_buff *skb)
1754{
4305b541 1755 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1756}
1757
a21d4572
ED
1758/**
1759 * skb_availroom - bytes at buffer end
1760 * @skb: buffer to check
1761 *
1762 * Return the number of bytes of free space at the tail of an sk_buff
1763 * allocated by sk_stream_alloc()
1764 */
1765static inline int skb_availroom(const struct sk_buff *skb)
1766{
16fad69c
ED
1767 if (skb_is_nonlinear(skb))
1768 return 0;
1769
1770 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1771}
1772
1da177e4
LT
1773/**
1774 * skb_reserve - adjust headroom
1775 * @skb: buffer to alter
1776 * @len: bytes to move
1777 *
1778 * Increase the headroom of an empty &sk_buff by reducing the tail
1779 * room. This is only allowed for an empty buffer.
1780 */
8243126c 1781static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1782{
1783 skb->data += len;
1784 skb->tail += len;
1785}
1786
8bce6d7d
TH
1787#define ENCAP_TYPE_ETHER 0
1788#define ENCAP_TYPE_IPPROTO 1
1789
1790static inline void skb_set_inner_protocol(struct sk_buff *skb,
1791 __be16 protocol)
1792{
1793 skb->inner_protocol = protocol;
1794 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1795}
1796
1797static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1798 __u8 ipproto)
1799{
1800 skb->inner_ipproto = ipproto;
1801 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1802}
1803
6a674e9c
JG
1804static inline void skb_reset_inner_headers(struct sk_buff *skb)
1805{
aefbd2b3 1806 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1807 skb->inner_network_header = skb->network_header;
1808 skb->inner_transport_header = skb->transport_header;
1809}
1810
0b5c9db1
JP
1811static inline void skb_reset_mac_len(struct sk_buff *skb)
1812{
1813 skb->mac_len = skb->network_header - skb->mac_header;
1814}
1815
6a674e9c
JG
1816static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1817 *skb)
1818{
1819 return skb->head + skb->inner_transport_header;
1820}
1821
1822static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1823{
1824 skb->inner_transport_header = skb->data - skb->head;
1825}
1826
1827static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1828 const int offset)
1829{
1830 skb_reset_inner_transport_header(skb);
1831 skb->inner_transport_header += offset;
1832}
1833
1834static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1835{
1836 return skb->head + skb->inner_network_header;
1837}
1838
1839static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1840{
1841 skb->inner_network_header = skb->data - skb->head;
1842}
1843
1844static inline void skb_set_inner_network_header(struct sk_buff *skb,
1845 const int offset)
1846{
1847 skb_reset_inner_network_header(skb);
1848 skb->inner_network_header += offset;
1849}
1850
aefbd2b3
PS
1851static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1852{
1853 return skb->head + skb->inner_mac_header;
1854}
1855
1856static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1857{
1858 skb->inner_mac_header = skb->data - skb->head;
1859}
1860
1861static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1862 const int offset)
1863{
1864 skb_reset_inner_mac_header(skb);
1865 skb->inner_mac_header += offset;
1866}
fda55eca
ED
1867static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1868{
35d04610 1869 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1870}
1871
9c70220b
ACM
1872static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1873{
2e07fa9c 1874 return skb->head + skb->transport_header;
9c70220b
ACM
1875}
1876
badff6d0
ACM
1877static inline void skb_reset_transport_header(struct sk_buff *skb)
1878{
2e07fa9c 1879 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1880}
1881
967b05f6
ACM
1882static inline void skb_set_transport_header(struct sk_buff *skb,
1883 const int offset)
1884{
2e07fa9c
ACM
1885 skb_reset_transport_header(skb);
1886 skb->transport_header += offset;
ea2ae17d
ACM
1887}
1888
d56f90a7
ACM
1889static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1890{
2e07fa9c 1891 return skb->head + skb->network_header;
d56f90a7
ACM
1892}
1893
c1d2bbe1
ACM
1894static inline void skb_reset_network_header(struct sk_buff *skb)
1895{
2e07fa9c 1896 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1897}
1898
c14d2450
ACM
1899static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1900{
2e07fa9c
ACM
1901 skb_reset_network_header(skb);
1902 skb->network_header += offset;
c14d2450
ACM
1903}
1904
2e07fa9c 1905static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1906{
2e07fa9c 1907 return skb->head + skb->mac_header;
bbe735e4
ACM
1908}
1909
2e07fa9c 1910static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1911{
35d04610 1912 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1913}
1914
1915static inline void skb_reset_mac_header(struct sk_buff *skb)
1916{
1917 skb->mac_header = skb->data - skb->head;
1918}
1919
1920static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1921{
1922 skb_reset_mac_header(skb);
1923 skb->mac_header += offset;
1924}
1925
0e3da5bb
TT
1926static inline void skb_pop_mac_header(struct sk_buff *skb)
1927{
1928 skb->mac_header = skb->network_header;
1929}
1930
fbbdb8f0
YX
1931static inline void skb_probe_transport_header(struct sk_buff *skb,
1932 const int offset_hint)
1933{
1934 struct flow_keys keys;
1935
1936 if (skb_transport_header_was_set(skb))
1937 return;
1938 else if (skb_flow_dissect(skb, &keys))
1939 skb_set_transport_header(skb, keys.thoff);
1940 else
1941 skb_set_transport_header(skb, offset_hint);
1942}
1943
03606895
ED
1944static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1945{
1946 if (skb_mac_header_was_set(skb)) {
1947 const unsigned char *old_mac = skb_mac_header(skb);
1948
1949 skb_set_mac_header(skb, -skb->mac_len);
1950 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1951 }
1952}
1953
04fb451e
MM
1954static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1955{
1956 return skb->csum_start - skb_headroom(skb);
1957}
1958
2e07fa9c
ACM
1959static inline int skb_transport_offset(const struct sk_buff *skb)
1960{
1961 return skb_transport_header(skb) - skb->data;
1962}
1963
1964static inline u32 skb_network_header_len(const struct sk_buff *skb)
1965{
1966 return skb->transport_header - skb->network_header;
1967}
1968
6a674e9c
JG
1969static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1970{
1971 return skb->inner_transport_header - skb->inner_network_header;
1972}
1973
2e07fa9c
ACM
1974static inline int skb_network_offset(const struct sk_buff *skb)
1975{
1976 return skb_network_header(skb) - skb->data;
1977}
48d49d0c 1978
6a674e9c
JG
1979static inline int skb_inner_network_offset(const struct sk_buff *skb)
1980{
1981 return skb_inner_network_header(skb) - skb->data;
1982}
1983
f9599ce1
CG
1984static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1985{
1986 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1987}
1988
1da177e4
LT
1989/*
1990 * CPUs often take a performance hit when accessing unaligned memory
1991 * locations. The actual performance hit varies, it can be small if the
1992 * hardware handles it or large if we have to take an exception and fix it
1993 * in software.
1994 *
1995 * Since an ethernet header is 14 bytes network drivers often end up with
1996 * the IP header at an unaligned offset. The IP header can be aligned by
1997 * shifting the start of the packet by 2 bytes. Drivers should do this
1998 * with:
1999 *
8660c124 2000 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2001 *
2002 * The downside to this alignment of the IP header is that the DMA is now
2003 * unaligned. On some architectures the cost of an unaligned DMA is high
2004 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2005 *
1da177e4
LT
2006 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2007 * to be overridden.
2008 */
2009#ifndef NET_IP_ALIGN
2010#define NET_IP_ALIGN 2
2011#endif
2012
025be81e
AB
2013/*
2014 * The networking layer reserves some headroom in skb data (via
2015 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2016 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2017 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2018 *
2019 * Unfortunately this headroom changes the DMA alignment of the resulting
2020 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2021 * on some architectures. An architecture can override this value,
2022 * perhaps setting it to a cacheline in size (since that will maintain
2023 * cacheline alignment of the DMA). It must be a power of 2.
2024 *
d6301d3d 2025 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2026 * headroom, you should not reduce this.
5933dd2f
ED
2027 *
2028 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2029 * to reduce average number of cache lines per packet.
2030 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2031 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2032 */
2033#ifndef NET_SKB_PAD
5933dd2f 2034#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2035#endif
2036
7965bd4d 2037int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2038
2039static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2040{
c4264f27 2041 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2042 WARN_ON(1);
2043 return;
2044 }
27a884dc
ACM
2045 skb->len = len;
2046 skb_set_tail_pointer(skb, len);
1da177e4
LT
2047}
2048
7965bd4d 2049void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2050
2051static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2052{
3cc0e873
HX
2053 if (skb->data_len)
2054 return ___pskb_trim(skb, len);
2055 __skb_trim(skb, len);
2056 return 0;
1da177e4
LT
2057}
2058
2059static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2060{
2061 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2062}
2063
e9fa4f7b
HX
2064/**
2065 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2066 * @skb: buffer to alter
2067 * @len: new length
2068 *
2069 * This is identical to pskb_trim except that the caller knows that
2070 * the skb is not cloned so we should never get an error due to out-
2071 * of-memory.
2072 */
2073static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2074{
2075 int err = pskb_trim(skb, len);
2076 BUG_ON(err);
2077}
2078
1da177e4
LT
2079/**
2080 * skb_orphan - orphan a buffer
2081 * @skb: buffer to orphan
2082 *
2083 * If a buffer currently has an owner then we call the owner's
2084 * destructor function and make the @skb unowned. The buffer continues
2085 * to exist but is no longer charged to its former owner.
2086 */
2087static inline void skb_orphan(struct sk_buff *skb)
2088{
c34a7612 2089 if (skb->destructor) {
1da177e4 2090 skb->destructor(skb);
c34a7612
ED
2091 skb->destructor = NULL;
2092 skb->sk = NULL;
376c7311
ED
2093 } else {
2094 BUG_ON(skb->sk);
c34a7612 2095 }
1da177e4
LT
2096}
2097
a353e0ce
MT
2098/**
2099 * skb_orphan_frags - orphan the frags contained in a buffer
2100 * @skb: buffer to orphan frags from
2101 * @gfp_mask: allocation mask for replacement pages
2102 *
2103 * For each frag in the SKB which needs a destructor (i.e. has an
2104 * owner) create a copy of that frag and release the original
2105 * page by calling the destructor.
2106 */
2107static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2108{
2109 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2110 return 0;
2111 return skb_copy_ubufs(skb, gfp_mask);
2112}
2113
1da177e4
LT
2114/**
2115 * __skb_queue_purge - empty a list
2116 * @list: list to empty
2117 *
2118 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2119 * the list and one reference dropped. This function does not take the
2120 * list lock and the caller must hold the relevant locks to use it.
2121 */
7965bd4d 2122void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2123static inline void __skb_queue_purge(struct sk_buff_head *list)
2124{
2125 struct sk_buff *skb;
2126 while ((skb = __skb_dequeue(list)) != NULL)
2127 kfree_skb(skb);
2128}
2129
e5e67305
AD
2130#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2131#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2132#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2133
7965bd4d 2134void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2135
7965bd4d
JP
2136struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2137 gfp_t gfp_mask);
8af27456
CH
2138
2139/**
2140 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2141 * @dev: network device to receive on
2142 * @length: length to allocate
2143 *
2144 * Allocate a new &sk_buff and assign it a usage count of one. The
2145 * buffer has unspecified headroom built in. Users should allocate
2146 * the headroom they think they need without accounting for the
2147 * built in space. The built in space is used for optimisations.
2148 *
2149 * %NULL is returned if there is no free memory. Although this function
2150 * allocates memory it can be called from an interrupt.
2151 */
2152static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2153 unsigned int length)
8af27456
CH
2154{
2155 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2156}
2157
6f532612
ED
2158/* legacy helper around __netdev_alloc_skb() */
2159static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2160 gfp_t gfp_mask)
2161{
2162 return __netdev_alloc_skb(NULL, length, gfp_mask);
2163}
2164
2165/* legacy helper around netdev_alloc_skb() */
2166static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2167{
2168 return netdev_alloc_skb(NULL, length);
2169}
2170
2171
4915a0de
ED
2172static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2173 unsigned int length, gfp_t gfp)
61321bbd 2174{
4915a0de 2175 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2176
2177 if (NET_IP_ALIGN && skb)
2178 skb_reserve(skb, NET_IP_ALIGN);
2179 return skb;
2180}
2181
4915a0de
ED
2182static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2183 unsigned int length)
2184{
2185 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2186}
2187
bc6fc9fa
FF
2188/**
2189 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
2190 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2191 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2192 * @order: size of the allocation
2193 *
2194 * Allocate a new page.
2195 *
2196 * %NULL is returned if there is no free memory.
2197*/
2198static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2199 struct sk_buff *skb,
2200 unsigned int order)
2201{
2202 struct page *page;
2203
2204 gfp_mask |= __GFP_COLD;
2205
2206 if (!(gfp_mask & __GFP_NOMEMALLOC))
2207 gfp_mask |= __GFP_MEMALLOC;
2208
2209 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2210 if (skb && page && page->pfmemalloc)
2211 skb->pfmemalloc = true;
2212
2213 return page;
2214}
2215
2216/**
2217 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2218 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2219 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2220 *
2221 * Allocate a new page.
2222 *
2223 * %NULL is returned if there is no free memory.
2224 */
2225static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2226 struct sk_buff *skb)
2227{
2228 return __skb_alloc_pages(gfp_mask, skb, 0);
2229}
2230
2231/**
2232 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2233 * @page: The page that was allocated from skb_alloc_page
2234 * @skb: The skb that may need pfmemalloc set
2235 */
2236static inline void skb_propagate_pfmemalloc(struct page *page,
2237 struct sk_buff *skb)
2238{
2239 if (page && page->pfmemalloc)
2240 skb->pfmemalloc = true;
2241}
2242
131ea667 2243/**
e227867f 2244 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2245 * @frag: the paged fragment
2246 *
2247 * Returns the &struct page associated with @frag.
2248 */
2249static inline struct page *skb_frag_page(const skb_frag_t *frag)
2250{
a8605c60 2251 return frag->page.p;
131ea667
IC
2252}
2253
2254/**
2255 * __skb_frag_ref - take an addition reference on a paged fragment.
2256 * @frag: the paged fragment
2257 *
2258 * Takes an additional reference on the paged fragment @frag.
2259 */
2260static inline void __skb_frag_ref(skb_frag_t *frag)
2261{
2262 get_page(skb_frag_page(frag));
2263}
2264
2265/**
2266 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2267 * @skb: the buffer
2268 * @f: the fragment offset.
2269 *
2270 * Takes an additional reference on the @f'th paged fragment of @skb.
2271 */
2272static inline void skb_frag_ref(struct sk_buff *skb, int f)
2273{
2274 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2275}
2276
2277/**
2278 * __skb_frag_unref - release a reference on a paged fragment.
2279 * @frag: the paged fragment
2280 *
2281 * Releases a reference on the paged fragment @frag.
2282 */
2283static inline void __skb_frag_unref(skb_frag_t *frag)
2284{
2285 put_page(skb_frag_page(frag));
2286}
2287
2288/**
2289 * skb_frag_unref - release a reference on a paged fragment of an skb.
2290 * @skb: the buffer
2291 * @f: the fragment offset
2292 *
2293 * Releases a reference on the @f'th paged fragment of @skb.
2294 */
2295static inline void skb_frag_unref(struct sk_buff *skb, int f)
2296{
2297 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2298}
2299
2300/**
2301 * skb_frag_address - gets the address of the data contained in a paged fragment
2302 * @frag: the paged fragment buffer
2303 *
2304 * Returns the address of the data within @frag. The page must already
2305 * be mapped.
2306 */
2307static inline void *skb_frag_address(const skb_frag_t *frag)
2308{
2309 return page_address(skb_frag_page(frag)) + frag->page_offset;
2310}
2311
2312/**
2313 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2314 * @frag: the paged fragment buffer
2315 *
2316 * Returns the address of the data within @frag. Checks that the page
2317 * is mapped and returns %NULL otherwise.
2318 */
2319static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2320{
2321 void *ptr = page_address(skb_frag_page(frag));
2322 if (unlikely(!ptr))
2323 return NULL;
2324
2325 return ptr + frag->page_offset;
2326}
2327
2328/**
2329 * __skb_frag_set_page - sets the page contained in a paged fragment
2330 * @frag: the paged fragment
2331 * @page: the page to set
2332 *
2333 * Sets the fragment @frag to contain @page.
2334 */
2335static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2336{
a8605c60 2337 frag->page.p = page;
131ea667
IC
2338}
2339
2340/**
2341 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2342 * @skb: the buffer
2343 * @f: the fragment offset
2344 * @page: the page to set
2345 *
2346 * Sets the @f'th fragment of @skb to contain @page.
2347 */
2348static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2349 struct page *page)
2350{
2351 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2352}
2353
400dfd3a
ED
2354bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2355
131ea667
IC
2356/**
2357 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2358 * @dev: the device to map the fragment to
131ea667
IC
2359 * @frag: the paged fragment to map
2360 * @offset: the offset within the fragment (starting at the
2361 * fragment's own offset)
2362 * @size: the number of bytes to map
f83347df 2363 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2364 *
2365 * Maps the page associated with @frag to @device.
2366 */
2367static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2368 const skb_frag_t *frag,
2369 size_t offset, size_t size,
2370 enum dma_data_direction dir)
2371{
2372 return dma_map_page(dev, skb_frag_page(frag),
2373 frag->page_offset + offset, size, dir);
2374}
2375
117632e6
ED
2376static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2377 gfp_t gfp_mask)
2378{
2379 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2380}
2381
bad93e9d
OP
2382
2383static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2384 gfp_t gfp_mask)
2385{
2386 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2387}
2388
2389
334a8132
PM
2390/**
2391 * skb_clone_writable - is the header of a clone writable
2392 * @skb: buffer to check
2393 * @len: length up to which to write
2394 *
2395 * Returns true if modifying the header part of the cloned buffer
2396 * does not requires the data to be copied.
2397 */
05bdd2f1 2398static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2399{
2400 return !skb_header_cloned(skb) &&
2401 skb_headroom(skb) + len <= skb->hdr_len;
2402}
2403
d9cc2048
HX
2404static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2405 int cloned)
2406{
2407 int delta = 0;
2408
d9cc2048
HX
2409 if (headroom > skb_headroom(skb))
2410 delta = headroom - skb_headroom(skb);
2411
2412 if (delta || cloned)
2413 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2414 GFP_ATOMIC);
2415 return 0;
2416}
2417
1da177e4
LT
2418/**
2419 * skb_cow - copy header of skb when it is required
2420 * @skb: buffer to cow
2421 * @headroom: needed headroom
2422 *
2423 * If the skb passed lacks sufficient headroom or its data part
2424 * is shared, data is reallocated. If reallocation fails, an error
2425 * is returned and original skb is not changed.
2426 *
2427 * The result is skb with writable area skb->head...skb->tail
2428 * and at least @headroom of space at head.
2429 */
2430static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2431{
d9cc2048
HX
2432 return __skb_cow(skb, headroom, skb_cloned(skb));
2433}
1da177e4 2434
d9cc2048
HX
2435/**
2436 * skb_cow_head - skb_cow but only making the head writable
2437 * @skb: buffer to cow
2438 * @headroom: needed headroom
2439 *
2440 * This function is identical to skb_cow except that we replace the
2441 * skb_cloned check by skb_header_cloned. It should be used when
2442 * you only need to push on some header and do not need to modify
2443 * the data.
2444 */
2445static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2446{
2447 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2448}
2449
2450/**
2451 * skb_padto - pad an skbuff up to a minimal size
2452 * @skb: buffer to pad
2453 * @len: minimal length
2454 *
2455 * Pads up a buffer to ensure the trailing bytes exist and are
2456 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2457 * is untouched. Otherwise it is extended. Returns zero on
2458 * success. The skb is freed on error.
1da177e4
LT
2459 */
2460
5b057c6b 2461static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2462{
2463 unsigned int size = skb->len;
2464 if (likely(size >= len))
5b057c6b 2465 return 0;
987c402a 2466 return skb_pad(skb, len - size);
1da177e4
LT
2467}
2468
2469static inline int skb_add_data(struct sk_buff *skb,
2470 char __user *from, int copy)
2471{
2472 const int off = skb->len;
2473
2474 if (skb->ip_summed == CHECKSUM_NONE) {
2475 int err = 0;
5084205f 2476 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2477 copy, 0, &err);
2478 if (!err) {
2479 skb->csum = csum_block_add(skb->csum, csum, off);
2480 return 0;
2481 }
2482 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2483 return 0;
2484
2485 __skb_trim(skb, off);
2486 return -EFAULT;
2487}
2488
38ba0a65
ED
2489static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2490 const struct page *page, int off)
1da177e4
LT
2491{
2492 if (i) {
9e903e08 2493 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2494
ea2ab693 2495 return page == skb_frag_page(frag) &&
9e903e08 2496 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2497 }
38ba0a65 2498 return false;
1da177e4
LT
2499}
2500
364c6bad
HX
2501static inline int __skb_linearize(struct sk_buff *skb)
2502{
2503 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2504}
2505
1da177e4
LT
2506/**
2507 * skb_linearize - convert paged skb to linear one
2508 * @skb: buffer to linarize
1da177e4
LT
2509 *
2510 * If there is no free memory -ENOMEM is returned, otherwise zero
2511 * is returned and the old skb data released.
2512 */
364c6bad
HX
2513static inline int skb_linearize(struct sk_buff *skb)
2514{
2515 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2516}
2517
cef401de
ED
2518/**
2519 * skb_has_shared_frag - can any frag be overwritten
2520 * @skb: buffer to test
2521 *
2522 * Return true if the skb has at least one frag that might be modified
2523 * by an external entity (as in vmsplice()/sendfile())
2524 */
2525static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2526{
c9af6db4
PS
2527 return skb_is_nonlinear(skb) &&
2528 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2529}
2530
364c6bad
HX
2531/**
2532 * skb_linearize_cow - make sure skb is linear and writable
2533 * @skb: buffer to process
2534 *
2535 * If there is no free memory -ENOMEM is returned, otherwise zero
2536 * is returned and the old skb data released.
2537 */
2538static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2539{
364c6bad
HX
2540 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2541 __skb_linearize(skb) : 0;
1da177e4
LT
2542}
2543
2544/**
2545 * skb_postpull_rcsum - update checksum for received skb after pull
2546 * @skb: buffer to update
2547 * @start: start of data before pull
2548 * @len: length of data pulled
2549 *
2550 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2551 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2552 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2553 */
2554
2555static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2556 const void *start, unsigned int len)
1da177e4 2557{
84fa7933 2558 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2559 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2560}
2561
cbb042f9
HX
2562unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2563
7ce5a27f
DM
2564/**
2565 * pskb_trim_rcsum - trim received skb and update checksum
2566 * @skb: buffer to trim
2567 * @len: new length
2568 *
2569 * This is exactly the same as pskb_trim except that it ensures the
2570 * checksum of received packets are still valid after the operation.
2571 */
2572
2573static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2574{
2575 if (likely(len >= skb->len))
2576 return 0;
2577 if (skb->ip_summed == CHECKSUM_COMPLETE)
2578 skb->ip_summed = CHECKSUM_NONE;
2579 return __pskb_trim(skb, len);
2580}
2581
1da177e4
LT
2582#define skb_queue_walk(queue, skb) \
2583 for (skb = (queue)->next; \
a1e4891f 2584 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2585 skb = skb->next)
2586
46f8914e
JC
2587#define skb_queue_walk_safe(queue, skb, tmp) \
2588 for (skb = (queue)->next, tmp = skb->next; \
2589 skb != (struct sk_buff *)(queue); \
2590 skb = tmp, tmp = skb->next)
2591
1164f52a 2592#define skb_queue_walk_from(queue, skb) \
a1e4891f 2593 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2594 skb = skb->next)
2595
2596#define skb_queue_walk_from_safe(queue, skb, tmp) \
2597 for (tmp = skb->next; \
2598 skb != (struct sk_buff *)(queue); \
2599 skb = tmp, tmp = skb->next)
2600
300ce174
SH
2601#define skb_queue_reverse_walk(queue, skb) \
2602 for (skb = (queue)->prev; \
a1e4891f 2603 skb != (struct sk_buff *)(queue); \
300ce174
SH
2604 skb = skb->prev)
2605
686a2955
DM
2606#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2607 for (skb = (queue)->prev, tmp = skb->prev; \
2608 skb != (struct sk_buff *)(queue); \
2609 skb = tmp, tmp = skb->prev)
2610
2611#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2612 for (tmp = skb->prev; \
2613 skb != (struct sk_buff *)(queue); \
2614 skb = tmp, tmp = skb->prev)
1da177e4 2615
21dc3301 2616static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2617{
2618 return skb_shinfo(skb)->frag_list != NULL;
2619}
2620
2621static inline void skb_frag_list_init(struct sk_buff *skb)
2622{
2623 skb_shinfo(skb)->frag_list = NULL;
2624}
2625
2626static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2627{
2628 frag->next = skb_shinfo(skb)->frag_list;
2629 skb_shinfo(skb)->frag_list = frag;
2630}
2631
2632#define skb_walk_frags(skb, iter) \
2633 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2634
7965bd4d
JP
2635struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2636 int *peeked, int *off, int *err);
2637struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2638 int *err);
2639unsigned int datagram_poll(struct file *file, struct socket *sock,
2640 struct poll_table_struct *wait);
2641int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2642 struct iovec *to, int size);
51f3d02b
DM
2643static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2644 struct msghdr *msg, int size)
2645{
2646 return skb_copy_datagram_iovec(from, offset, msg->msg_iov, size);
2647}
7965bd4d
JP
2648int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2649 struct iovec *iov);
2650int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2651 const struct iovec *from, int from_offset,
2652 int len);
2653int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2654 int offset, size_t count);
2655int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2656 const struct iovec *to, int to_offset,
2657 int size);
2658void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2659void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2660int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2661int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2662int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2663__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2664 int len, __wsum csum);
2665int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2666 struct pipe_inode_info *pipe, unsigned int len,
2667 unsigned int flags);
2668void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2669unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2670int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2671 int len, int hlen);
7965bd4d
JP
2672void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2673int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2674void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2675unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2676struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2677struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
20380731 2678
2817a336
DB
2679struct skb_checksum_ops {
2680 __wsum (*update)(const void *mem, int len, __wsum wsum);
2681 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2682};
2683
2684__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2685 __wsum csum, const struct skb_checksum_ops *ops);
2686__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2687 __wsum csum);
2688
690e36e7
DM
2689static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2690 int len, void *data, int hlen, void *buffer)
1da177e4 2691{
55820ee2 2692 if (hlen - offset >= len)
690e36e7 2693 return data + offset;
1da177e4 2694
690e36e7
DM
2695 if (!skb ||
2696 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2697 return NULL;
2698
2699 return buffer;
2700}
2701
690e36e7
DM
2702static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2703 int len, void *buffer)
2704{
2705 return __skb_header_pointer(skb, offset, len, skb->data,
2706 skb_headlen(skb), buffer);
2707}
2708
4262e5cc
DB
2709/**
2710 * skb_needs_linearize - check if we need to linearize a given skb
2711 * depending on the given device features.
2712 * @skb: socket buffer to check
2713 * @features: net device features
2714 *
2715 * Returns true if either:
2716 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2717 * 2. skb is fragmented and the device does not support SG.
2718 */
2719static inline bool skb_needs_linearize(struct sk_buff *skb,
2720 netdev_features_t features)
2721{
2722 return skb_is_nonlinear(skb) &&
2723 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2724 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2725}
2726
d626f62b
ACM
2727static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2728 void *to,
2729 const unsigned int len)
2730{
2731 memcpy(to, skb->data, len);
2732}
2733
2734static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2735 const int offset, void *to,
2736 const unsigned int len)
2737{
2738 memcpy(to, skb->data + offset, len);
2739}
2740
27d7ff46
ACM
2741static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2742 const void *from,
2743 const unsigned int len)
2744{
2745 memcpy(skb->data, from, len);
2746}
2747
2748static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2749 const int offset,
2750 const void *from,
2751 const unsigned int len)
2752{
2753 memcpy(skb->data + offset, from, len);
2754}
2755
7965bd4d 2756void skb_init(void);
1da177e4 2757
ac45f602
PO
2758static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2759{
2760 return skb->tstamp;
2761}
2762
a61bbcf2
PM
2763/**
2764 * skb_get_timestamp - get timestamp from a skb
2765 * @skb: skb to get stamp from
2766 * @stamp: pointer to struct timeval to store stamp in
2767 *
2768 * Timestamps are stored in the skb as offsets to a base timestamp.
2769 * This function converts the offset back to a struct timeval and stores
2770 * it in stamp.
2771 */
ac45f602
PO
2772static inline void skb_get_timestamp(const struct sk_buff *skb,
2773 struct timeval *stamp)
a61bbcf2 2774{
b7aa0bf7 2775 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2776}
2777
ac45f602
PO
2778static inline void skb_get_timestampns(const struct sk_buff *skb,
2779 struct timespec *stamp)
2780{
2781 *stamp = ktime_to_timespec(skb->tstamp);
2782}
2783
b7aa0bf7 2784static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2785{
b7aa0bf7 2786 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2787}
2788
164891aa
SH
2789static inline ktime_t net_timedelta(ktime_t t)
2790{
2791 return ktime_sub(ktime_get_real(), t);
2792}
2793
b9ce204f
IJ
2794static inline ktime_t net_invalid_timestamp(void)
2795{
2796 return ktime_set(0, 0);
2797}
a61bbcf2 2798
62bccb8c
AD
2799struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2800
c1f19b51
RC
2801#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2802
7965bd4d
JP
2803void skb_clone_tx_timestamp(struct sk_buff *skb);
2804bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2805
2806#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2807
2808static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2809{
2810}
2811
2812static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2813{
2814 return false;
2815}
2816
2817#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2818
2819/**
2820 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2821 *
da92b194
RC
2822 * PHY drivers may accept clones of transmitted packets for
2823 * timestamping via their phy_driver.txtstamp method. These drivers
2824 * must call this function to return the skb back to the stack, with
2825 * or without a timestamp.
2826 *
c1f19b51 2827 * @skb: clone of the the original outgoing packet
da92b194 2828 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2829 *
2830 */
2831void skb_complete_tx_timestamp(struct sk_buff *skb,
2832 struct skb_shared_hwtstamps *hwtstamps);
2833
e7fd2885
WB
2834void __skb_tstamp_tx(struct sk_buff *orig_skb,
2835 struct skb_shared_hwtstamps *hwtstamps,
2836 struct sock *sk, int tstype);
2837
ac45f602
PO
2838/**
2839 * skb_tstamp_tx - queue clone of skb with send time stamps
2840 * @orig_skb: the original outgoing packet
2841 * @hwtstamps: hardware time stamps, may be NULL if not available
2842 *
2843 * If the skb has a socket associated, then this function clones the
2844 * skb (thus sharing the actual data and optional structures), stores
2845 * the optional hardware time stamping information (if non NULL) or
2846 * generates a software time stamp (otherwise), then queues the clone
2847 * to the error queue of the socket. Errors are silently ignored.
2848 */
7965bd4d
JP
2849void skb_tstamp_tx(struct sk_buff *orig_skb,
2850 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2851
4507a715
RC
2852static inline void sw_tx_timestamp(struct sk_buff *skb)
2853{
2244d07b
OH
2854 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2855 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2856 skb_tstamp_tx(skb, NULL);
2857}
2858
2859/**
2860 * skb_tx_timestamp() - Driver hook for transmit timestamping
2861 *
2862 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2863 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2864 *
73409f3b
DM
2865 * Specifically, one should make absolutely sure that this function is
2866 * called before TX completion of this packet can trigger. Otherwise
2867 * the packet could potentially already be freed.
2868 *
4507a715
RC
2869 * @skb: A socket buffer.
2870 */
2871static inline void skb_tx_timestamp(struct sk_buff *skb)
2872{
c1f19b51 2873 skb_clone_tx_timestamp(skb);
4507a715
RC
2874 sw_tx_timestamp(skb);
2875}
2876
6e3e939f
JB
2877/**
2878 * skb_complete_wifi_ack - deliver skb with wifi status
2879 *
2880 * @skb: the original outgoing packet
2881 * @acked: ack status
2882 *
2883 */
2884void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2885
7965bd4d
JP
2886__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2887__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2888
60476372
HX
2889static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2890{
5d0c2b95 2891 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
60476372
HX
2892}
2893
fb286bb2
HX
2894/**
2895 * skb_checksum_complete - Calculate checksum of an entire packet
2896 * @skb: packet to process
2897 *
2898 * This function calculates the checksum over the entire packet plus
2899 * the value of skb->csum. The latter can be used to supply the
2900 * checksum of a pseudo header as used by TCP/UDP. It returns the
2901 * checksum.
2902 *
2903 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2904 * this function can be used to verify that checksum on received
2905 * packets. In that case the function should return zero if the
2906 * checksum is correct. In particular, this function will return zero
2907 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2908 * hardware has already verified the correctness of the checksum.
2909 */
4381ca3c 2910static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2911{
60476372
HX
2912 return skb_csum_unnecessary(skb) ?
2913 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2914}
2915
77cffe23
TH
2916static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2917{
2918 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2919 if (skb->csum_level == 0)
2920 skb->ip_summed = CHECKSUM_NONE;
2921 else
2922 skb->csum_level--;
2923 }
2924}
2925
2926static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2927{
2928 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2929 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2930 skb->csum_level++;
2931 } else if (skb->ip_summed == CHECKSUM_NONE) {
2932 skb->ip_summed = CHECKSUM_UNNECESSARY;
2933 skb->csum_level = 0;
2934 }
2935}
2936
5a212329
TH
2937static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2938{
2939 /* Mark current checksum as bad (typically called from GRO
2940 * path). In the case that ip_summed is CHECKSUM_NONE
2941 * this must be the first checksum encountered in the packet.
2942 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2943 * checksum after the last one validated. For UDP, a zero
2944 * checksum can not be marked as bad.
2945 */
2946
2947 if (skb->ip_summed == CHECKSUM_NONE ||
2948 skb->ip_summed == CHECKSUM_UNNECESSARY)
2949 skb->csum_bad = 1;
2950}
2951
76ba0aae
TH
2952/* Check if we need to perform checksum complete validation.
2953 *
2954 * Returns true if checksum complete is needed, false otherwise
2955 * (either checksum is unnecessary or zero checksum is allowed).
2956 */
2957static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2958 bool zero_okay,
2959 __sum16 check)
2960{
5d0c2b95
TH
2961 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2962 skb->csum_valid = 1;
77cffe23 2963 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
2964 return false;
2965 }
2966
2967 return true;
2968}
2969
2970/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2971 * in checksum_init.
2972 */
2973#define CHECKSUM_BREAK 76
2974
2975/* Validate (init) checksum based on checksum complete.
2976 *
2977 * Return values:
2978 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2979 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2980 * checksum is stored in skb->csum for use in __skb_checksum_complete
2981 * non-zero: value of invalid checksum
2982 *
2983 */
2984static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2985 bool complete,
2986 __wsum psum)
2987{
2988 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2989 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 2990 skb->csum_valid = 1;
76ba0aae
TH
2991 return 0;
2992 }
5a212329
TH
2993 } else if (skb->csum_bad) {
2994 /* ip_summed == CHECKSUM_NONE in this case */
2995 return 1;
76ba0aae
TH
2996 }
2997
2998 skb->csum = psum;
2999
5d0c2b95
TH
3000 if (complete || skb->len <= CHECKSUM_BREAK) {
3001 __sum16 csum;
3002
3003 csum = __skb_checksum_complete(skb);
3004 skb->csum_valid = !csum;
3005 return csum;
3006 }
76ba0aae
TH
3007
3008 return 0;
3009}
3010
3011static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3012{
3013 return 0;
3014}
3015
3016/* Perform checksum validate (init). Note that this is a macro since we only
3017 * want to calculate the pseudo header which is an input function if necessary.
3018 * First we try to validate without any computation (checksum unnecessary) and
3019 * then calculate based on checksum complete calling the function to compute
3020 * pseudo header.
3021 *
3022 * Return values:
3023 * 0: checksum is validated or try to in skb_checksum_complete
3024 * non-zero: value of invalid checksum
3025 */
3026#define __skb_checksum_validate(skb, proto, complete, \
3027 zero_okay, check, compute_pseudo) \
3028({ \
3029 __sum16 __ret = 0; \
5d0c2b95 3030 skb->csum_valid = 0; \
76ba0aae
TH
3031 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3032 __ret = __skb_checksum_validate_complete(skb, \
3033 complete, compute_pseudo(skb, proto)); \
3034 __ret; \
3035})
3036
3037#define skb_checksum_init(skb, proto, compute_pseudo) \
3038 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3039
3040#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3041 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3042
3043#define skb_checksum_validate(skb, proto, compute_pseudo) \
3044 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3045
3046#define skb_checksum_validate_zero_check(skb, proto, check, \
3047 compute_pseudo) \
3048 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
3049
3050#define skb_checksum_simple_validate(skb) \
3051 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3052
d96535a1
TH
3053static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3054{
3055 return (skb->ip_summed == CHECKSUM_NONE &&
3056 skb->csum_valid && !skb->csum_bad);
3057}
3058
3059static inline void __skb_checksum_convert(struct sk_buff *skb,
3060 __sum16 check, __wsum pseudo)
3061{
3062 skb->csum = ~pseudo;
3063 skb->ip_summed = CHECKSUM_COMPLETE;
3064}
3065
3066#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3067do { \
3068 if (__skb_checksum_convert_check(skb)) \
3069 __skb_checksum_convert(skb, check, \
3070 compute_pseudo(skb, proto)); \
3071} while (0)
3072
5f79e0f9 3073#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3074void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3075static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3076{
3077 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3078 nf_conntrack_destroy(nfct);
1da177e4
LT
3079}
3080static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3081{
3082 if (nfct)
3083 atomic_inc(&nfct->use);
3084}
2fc72c7b 3085#endif
34666d46 3086#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3087static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3088{
3089 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3090 kfree(nf_bridge);
3091}
3092static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3093{
3094 if (nf_bridge)
3095 atomic_inc(&nf_bridge->use);
3096}
3097#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3098static inline void nf_reset(struct sk_buff *skb)
3099{
5f79e0f9 3100#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3101 nf_conntrack_put(skb->nfct);
3102 skb->nfct = NULL;
2fc72c7b 3103#endif
34666d46 3104#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3105 nf_bridge_put(skb->nf_bridge);
3106 skb->nf_bridge = NULL;
3107#endif
3108}
3109
124dff01
PM
3110static inline void nf_reset_trace(struct sk_buff *skb)
3111{
478b360a 3112#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3113 skb->nf_trace = 0;
3114#endif
a193a4ab
PM
3115}
3116
edda553c 3117/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3118static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3119 bool copy)
edda553c 3120{
5f79e0f9 3121#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3122 dst->nfct = src->nfct;
3123 nf_conntrack_get(src->nfct);
b1937227
ED
3124 if (copy)
3125 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3126#endif
34666d46 3127#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3128 dst->nf_bridge = src->nf_bridge;
3129 nf_bridge_get(src->nf_bridge);
3130#endif
478b360a 3131#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3132 if (copy)
3133 dst->nf_trace = src->nf_trace;
478b360a 3134#endif
edda553c
YK
3135}
3136
e7ac05f3
YK
3137static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3138{
e7ac05f3 3139#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3140 nf_conntrack_put(dst->nfct);
2fc72c7b 3141#endif
34666d46 3142#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3143 nf_bridge_put(dst->nf_bridge);
3144#endif
b1937227 3145 __nf_copy(dst, src, true);
e7ac05f3
YK
3146}
3147
984bc16c
JM
3148#ifdef CONFIG_NETWORK_SECMARK
3149static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3150{
3151 to->secmark = from->secmark;
3152}
3153
3154static inline void skb_init_secmark(struct sk_buff *skb)
3155{
3156 skb->secmark = 0;
3157}
3158#else
3159static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3160{ }
3161
3162static inline void skb_init_secmark(struct sk_buff *skb)
3163{ }
3164#endif
3165
574f7194
EB
3166static inline bool skb_irq_freeable(const struct sk_buff *skb)
3167{
3168 return !skb->destructor &&
3169#if IS_ENABLED(CONFIG_XFRM)
3170 !skb->sp &&
3171#endif
3172#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3173 !skb->nfct &&
3174#endif
3175 !skb->_skb_refdst &&
3176 !skb_has_frag_list(skb);
3177}
3178
f25f4e44
PWJ
3179static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3180{
f25f4e44 3181 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3182}
3183
9247744e 3184static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3185{
4e3ab47a 3186 return skb->queue_mapping;
4e3ab47a
PE
3187}
3188
f25f4e44
PWJ
3189static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3190{
f25f4e44 3191 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3192}
3193
d5a9e24a
DM
3194static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3195{
3196 skb->queue_mapping = rx_queue + 1;
3197}
3198
9247744e 3199static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3200{
3201 return skb->queue_mapping - 1;
3202}
3203
9247744e 3204static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3205{
a02cec21 3206 return skb->queue_mapping != 0;
d5a9e24a
DM
3207}
3208
0e001614 3209u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
7965bd4d 3210 unsigned int num_tx_queues);
9247744e 3211
def8b4fa
AD
3212static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3213{
0b3d8e08 3214#ifdef CONFIG_XFRM
def8b4fa 3215 return skb->sp;
def8b4fa 3216#else
def8b4fa 3217 return NULL;
def8b4fa 3218#endif
0b3d8e08 3219}
def8b4fa 3220
68c33163
PS
3221/* Keeps track of mac header offset relative to skb->head.
3222 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3223 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3224 * tunnel skb it points to outer mac header.
3225 * Keeps track of level of encapsulation of network headers.
3226 */
68c33163 3227struct skb_gso_cb {
3347c960
ED
3228 int mac_offset;
3229 int encap_level;
7e2b10c1 3230 __u16 csum_start;
68c33163
PS
3231};
3232#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3233
3234static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3235{
3236 return (skb_mac_header(inner_skb) - inner_skb->head) -
3237 SKB_GSO_CB(inner_skb)->mac_offset;
3238}
3239
1e2bd517
PS
3240static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3241{
3242 int new_headroom, headroom;
3243 int ret;
3244
3245 headroom = skb_headroom(skb);
3246 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3247 if (ret)
3248 return ret;
3249
3250 new_headroom = skb_headroom(skb);
3251 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3252 return 0;
3253}
3254
7e2b10c1
TH
3255/* Compute the checksum for a gso segment. First compute the checksum value
3256 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3257 * then add in skb->csum (checksum from csum_start to end of packet).
3258 * skb->csum and csum_start are then updated to reflect the checksum of the
3259 * resultant packet starting from the transport header-- the resultant checksum
3260 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3261 * header.
3262 */
3263static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3264{
3265 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3266 skb_transport_offset(skb);
3267 __u16 csum;
3268
3269 csum = csum_fold(csum_partial(skb_transport_header(skb),
3270 plen, skb->csum));
3271 skb->csum = res;
3272 SKB_GSO_CB(skb)->csum_start -= plen;
3273
3274 return csum;
3275}
3276
bdcc0924 3277static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3278{
3279 return skb_shinfo(skb)->gso_size;
3280}
3281
36a8f39e 3282/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3283static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3284{
3285 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3286}
3287
7965bd4d 3288void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3289
3290static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3291{
3292 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3293 * wanted then gso_type will be set. */
05bdd2f1
ED
3294 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3295
b78462eb
AD
3296 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3297 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3298 __skb_warn_lro_forwarding(skb);
3299 return true;
3300 }
3301 return false;
3302}
3303
35fc92a9
HX
3304static inline void skb_forward_csum(struct sk_buff *skb)
3305{
3306 /* Unfortunately we don't support this one. Any brave souls? */
3307 if (skb->ip_summed == CHECKSUM_COMPLETE)
3308 skb->ip_summed = CHECKSUM_NONE;
3309}
3310
bc8acf2c
ED
3311/**
3312 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3313 * @skb: skb to check
3314 *
3315 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3316 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3317 * use this helper, to document places where we make this assertion.
3318 */
05bdd2f1 3319static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3320{
3321#ifdef DEBUG
3322 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3323#endif
3324}
3325
f35d9d8a 3326bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3327
ed1f50c3
PD
3328int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3329
56193d1b
AD
3330u32 skb_get_poff(const struct sk_buff *skb);
3331u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3332 const struct flow_keys *keys, int hlen);
f77668dc 3333
3a7c1ee4
AD
3334/**
3335 * skb_head_is_locked - Determine if the skb->head is locked down
3336 * @skb: skb to check
3337 *
3338 * The head on skbs build around a head frag can be removed if they are
3339 * not cloned. This function returns true if the skb head is locked down
3340 * due to either being allocated via kmalloc, or by being a clone with
3341 * multiple references to the head.
3342 */
3343static inline bool skb_head_is_locked(const struct sk_buff *skb)
3344{
3345 return !skb->head_frag || skb_cloned(skb);
3346}
fe6cc55f
FW
3347
3348/**
3349 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3350 *
3351 * @skb: GSO skb
3352 *
3353 * skb_gso_network_seglen is used to determine the real size of the
3354 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3355 *
3356 * The MAC/L2 header is not accounted for.
3357 */
3358static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3359{
3360 unsigned int hdr_len = skb_transport_header(skb) -
3361 skb_network_header(skb);
3362 return hdr_len + skb_gso_transport_seglen(skb);
3363}
1da177e4
LT
3364#endif /* __KERNEL__ */
3365#endif /* _LINUX_SKBUFF_H */