]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/skbuff.h
vmci: propagate msghdr all way down to __qp_memcpy_to_queue()
[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;
a8f820aa 153struct iov_iter;
fd11a83d 154struct napi_struct;
1da177e4 155
5f79e0f9 156#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
157struct nf_conntrack {
158 atomic_t use;
1da177e4 159};
5f79e0f9 160#endif
1da177e4 161
34666d46 162#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 163struct nf_bridge_info {
bf1ac5ca
ED
164 atomic_t use;
165 unsigned int mask;
166 struct net_device *physindev;
167 struct net_device *physoutdev;
168 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
169};
170#endif
171
1da177e4
LT
172struct sk_buff_head {
173 /* These two members must be first. */
174 struct sk_buff *next;
175 struct sk_buff *prev;
176
177 __u32 qlen;
178 spinlock_t lock;
179};
180
181struct sk_buff;
182
9d4dde52
IC
183/* To allow 64K frame to be packed as single skb without frag_list we
184 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
185 * buffers which do not start on a page boundary.
186 *
187 * Since GRO uses frags we allocate at least 16 regardless of page
188 * size.
a715dea3 189 */
9d4dde52 190#if (65536/PAGE_SIZE + 1) < 16
eec00954 191#define MAX_SKB_FRAGS 16UL
a715dea3 192#else
9d4dde52 193#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 194#endif
1da177e4
LT
195
196typedef struct skb_frag_struct skb_frag_t;
197
198struct skb_frag_struct {
a8605c60
IC
199 struct {
200 struct page *p;
201 } page;
cb4dfe56 202#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
203 __u32 page_offset;
204 __u32 size;
cb4dfe56
ED
205#else
206 __u16 page_offset;
207 __u16 size;
208#endif
1da177e4
LT
209};
210
9e903e08
ED
211static inline unsigned int skb_frag_size(const skb_frag_t *frag)
212{
213 return frag->size;
214}
215
216static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
217{
218 frag->size = size;
219}
220
221static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
222{
223 frag->size += delta;
224}
225
226static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
227{
228 frag->size -= delta;
229}
230
ac45f602
PO
231#define HAVE_HW_TIME_STAMP
232
233/**
d3a21be8 234 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
235 * @hwtstamp: hardware time stamp transformed into duration
236 * since arbitrary point in time
ac45f602
PO
237 *
238 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 239 * skb->tstamp.
ac45f602
PO
240 *
241 * hwtstamps can only be compared against other hwtstamps from
242 * the same device.
243 *
244 * This structure is attached to packets as part of the
245 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
246 */
247struct skb_shared_hwtstamps {
248 ktime_t hwtstamp;
ac45f602
PO
249};
250
2244d07b
OH
251/* Definitions for tx_flags in struct skb_shared_info */
252enum {
253 /* generate hardware time stamp */
254 SKBTX_HW_TSTAMP = 1 << 0,
255
e7fd2885 256 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
257 SKBTX_SW_TSTAMP = 1 << 1,
258
259 /* device driver is going to provide hardware time stamp */
260 SKBTX_IN_PROGRESS = 1 << 2,
261
a6686f2f 262 /* device driver supports TX zero-copy buffers */
62b1a8ab 263 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
264
265 /* generate wifi status information (where possible) */
62b1a8ab 266 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
267
268 /* This indicates at least one fragment might be overwritten
269 * (as in vmsplice(), sendfile() ...)
270 * If we need to compute a TX checksum, we'll need to copy
271 * all frags to avoid possible bad checksum
272 */
273 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
274
275 /* generate software time stamp when entering packet scheduling */
276 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
277
278 /* generate software timestamp on peer data acknowledgment */
279 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
280};
281
e1c8a607
WB
282#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
283 SKBTX_SCHED_TSTAMP | \
284 SKBTX_ACK_TSTAMP)
f24b9be5
WB
285#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
286
a6686f2f
SM
287/*
288 * The callback notifies userspace to release buffers when skb DMA is done in
289 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
290 * The zerocopy_success argument is true if zero copy transmit occurred,
291 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
292 * The ctx field is used to track device context.
293 * The desc field is used to track userspace buffer index.
a6686f2f
SM
294 */
295struct ubuf_info {
e19d6763 296 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 297 void *ctx;
a6686f2f 298 unsigned long desc;
ac45f602
PO
299};
300
1da177e4
LT
301/* This data is invariant across clones and lives at
302 * the end of the header data, ie. at skb->end.
303 */
304struct skb_shared_info {
9f42f126
IC
305 unsigned char nr_frags;
306 __u8 tx_flags;
7967168c
HX
307 unsigned short gso_size;
308 /* Warning: this field is not always filled in (UFO)! */
309 unsigned short gso_segs;
310 unsigned short gso_type;
1da177e4 311 struct sk_buff *frag_list;
ac45f602 312 struct skb_shared_hwtstamps hwtstamps;
09c2d251 313 u32 tskey;
9f42f126 314 __be32 ip6_frag_id;
ec7d2f2c
ED
315
316 /*
317 * Warning : all fields before dataref are cleared in __alloc_skb()
318 */
319 atomic_t dataref;
320
69e3c75f
JB
321 /* Intermediate layers must ensure that destructor_arg
322 * remains valid until skb destructor */
323 void * destructor_arg;
a6686f2f 324
fed66381
ED
325 /* must be last field, see pskb_expand_head() */
326 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
327};
328
329/* We divide dataref into two halves. The higher 16 bits hold references
330 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
331 * the entire skb->data. A clone of a headerless skb holds the length of
332 * the header in skb->hdr_len.
1da177e4
LT
333 *
334 * All users must obey the rule that the skb->data reference count must be
335 * greater than or equal to the payload reference count.
336 *
337 * Holding a reference to the payload part means that the user does not
338 * care about modifications to the header part of skb->data.
339 */
340#define SKB_DATAREF_SHIFT 16
341#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
342
d179cd12
DM
343
344enum {
c8753d55
VS
345 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
346 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
347 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
348};
349
7967168c
HX
350enum {
351 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 352 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
353
354 /* This indicates the skb is from an untrusted source. */
355 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
356
357 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
358 SKB_GSO_TCP_ECN = 1 << 3,
359
360 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
361
362 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
363
364 SKB_GSO_GRE = 1 << 6,
73136267 365
4b28252c 366 SKB_GSO_GRE_CSUM = 1 << 7,
0d89d203 367
4b28252c 368 SKB_GSO_IPIP = 1 << 8,
cb32f511 369
4b28252c 370 SKB_GSO_SIT = 1 << 9,
61c1db7f 371
4b28252c 372 SKB_GSO_UDP_TUNNEL = 1 << 10,
0f4f4ffa
TH
373
374 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c 375
59b93b41 376 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
7967168c
HX
377};
378
2e07fa9c
ACM
379#if BITS_PER_LONG > 32
380#define NET_SKBUFF_DATA_USES_OFFSET 1
381#endif
382
383#ifdef NET_SKBUFF_DATA_USES_OFFSET
384typedef unsigned int sk_buff_data_t;
385#else
386typedef unsigned char *sk_buff_data_t;
387#endif
388
363ec392
ED
389/**
390 * struct skb_mstamp - multi resolution time stamps
391 * @stamp_us: timestamp in us resolution
392 * @stamp_jiffies: timestamp in jiffies
393 */
394struct skb_mstamp {
395 union {
396 u64 v64;
397 struct {
398 u32 stamp_us;
399 u32 stamp_jiffies;
400 };
401 };
402};
403
404/**
405 * skb_mstamp_get - get current timestamp
406 * @cl: place to store timestamps
407 */
408static inline void skb_mstamp_get(struct skb_mstamp *cl)
409{
410 u64 val = local_clock();
411
412 do_div(val, NSEC_PER_USEC);
413 cl->stamp_us = (u32)val;
414 cl->stamp_jiffies = (u32)jiffies;
415}
416
417/**
418 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
419 * @t1: pointer to newest sample
420 * @t0: pointer to oldest sample
421 */
422static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
423 const struct skb_mstamp *t0)
424{
425 s32 delta_us = t1->stamp_us - t0->stamp_us;
426 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
427
428 /* If delta_us is negative, this might be because interval is too big,
429 * or local_clock() drift is too big : fallback using jiffies.
430 */
431 if (delta_us <= 0 ||
432 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
433
434 delta_us = jiffies_to_usecs(delta_jiffies);
435
436 return delta_us;
437}
438
439
1da177e4
LT
440/**
441 * struct sk_buff - socket buffer
442 * @next: Next buffer in list
443 * @prev: Previous buffer in list
363ec392 444 * @tstamp: Time we arrived/left
56b17425 445 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 446 * @sk: Socket we are owned by
1da177e4 447 * @dev: Device we arrived on/are leaving by
d84e0bd7 448 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 449 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 450 * @sp: the security path, used for xfrm
1da177e4
LT
451 * @len: Length of actual data
452 * @data_len: Data length
453 * @mac_len: Length of link layer header
334a8132 454 * @hdr_len: writable header length of cloned skb
663ead3b
HX
455 * @csum: Checksum (must include start/offset pair)
456 * @csum_start: Offset from skb->head where checksumming should start
457 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 458 * @priority: Packet queueing priority
60ff7467 459 * @ignore_df: allow local fragmentation
1da177e4 460 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 461 * @ip_summed: Driver fed us an IP checksum
1da177e4 462 * @nohdr: Payload reference only, must not modify header
d84e0bd7 463 * @nfctinfo: Relationship of this skb to the connection
1da177e4 464 * @pkt_type: Packet class
c83c2486 465 * @fclone: skbuff clone status
c83c2486 466 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
467 * @peeked: this packet has been seen already, so stats have been
468 * done for it, don't do them again
ba9dda3a 469 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
470 * @protocol: Packet protocol from driver
471 * @destructor: Destruct function
472 * @nfct: Associated connection, if any
1da177e4 473 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 474 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
475 * @tc_index: Traffic control index
476 * @tc_verd: traffic control verdict
61b905da 477 * @hash: the packet hash
d84e0bd7 478 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 479 * @xmit_more: More SKBs are pending for this queue
553a5672 480 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 481 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 482 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 483 * ports.
a3b18ddb 484 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
485 * @wifi_acked_valid: wifi_acked was set
486 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 487 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
06021292 488 * @napi_id: id of the NAPI struct this skb came from
984bc16c 489 * @secmark: security marking
d84e0bd7
DB
490 * @mark: Generic packet mark
491 * @dropcount: total number of sk_receive_queue overflows
86a9bad3 492 * @vlan_proto: vlan encapsulation protocol
6aa895b0 493 * @vlan_tci: vlan tag control information
0d89d203 494 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
495 * @inner_transport_header: Inner transport layer header (encapsulation)
496 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 497 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
498 * @transport_header: Transport layer header
499 * @network_header: Network layer header
500 * @mac_header: Link layer header
501 * @tail: Tail pointer
502 * @end: End pointer
503 * @head: Head of buffer
504 * @data: Data head pointer
505 * @truesize: Buffer size
506 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
507 */
508
509struct sk_buff {
363ec392 510 union {
56b17425
ED
511 struct {
512 /* These two members must be first. */
513 struct sk_buff *next;
514 struct sk_buff *prev;
515
516 union {
517 ktime_t tstamp;
518 struct skb_mstamp skb_mstamp;
519 };
520 };
521 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 522 };
da3f5cf1 523 struct sock *sk;
1da177e4 524 struct net_device *dev;
1da177e4 525
1da177e4
LT
526 /*
527 * This is the control buffer. It is free to use for every
528 * layer. Please put your private variables there. If you
529 * want to keep them across layers you have to do a skb_clone()
530 * first. This is owned by whoever has the skb queued ATM.
531 */
da3f5cf1 532 char cb[48] __aligned(8);
1da177e4 533
7fee226a 534 unsigned long _skb_refdst;
b1937227 535 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
536#ifdef CONFIG_XFRM
537 struct sec_path *sp;
b1937227
ED
538#endif
539#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
540 struct nf_conntrack *nfct;
541#endif
85224844 542#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 543 struct nf_bridge_info *nf_bridge;
da3f5cf1 544#endif
1da177e4 545 unsigned int len,
334a8132
PM
546 data_len;
547 __u16 mac_len,
548 hdr_len;
b1937227
ED
549
550 /* Following fields are _not_ copied in __copy_skb_header()
551 * Note that queue_mapping is here mostly to fill a hole.
552 */
fe55f6d5 553 kmemcheck_bitfield_begin(flags1);
b1937227
ED
554 __u16 queue_mapping;
555 __u8 cloned:1,
6869c4d8 556 nohdr:1,
b84f4cc9 557 fclone:2,
a59322be 558 peeked:1,
b1937227
ED
559 head_frag:1,
560 xmit_more:1;
561 /* one bit hole */
fe55f6d5 562 kmemcheck_bitfield_end(flags1);
4031ae6e 563
b1937227
ED
564 /* fields enclosed in headers_start/headers_end are copied
565 * using a single memcpy() in __copy_skb_header()
566 */
ebcf34f3 567 /* private: */
b1937227 568 __u32 headers_start[0];
ebcf34f3 569 /* public: */
4031ae6e 570
233577a2
HFS
571/* if you move pkt_type around you also must adapt those constants */
572#ifdef __BIG_ENDIAN_BITFIELD
573#define PKT_TYPE_MAX (7 << 5)
574#else
575#define PKT_TYPE_MAX 7
1da177e4 576#endif
233577a2 577#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 578
233577a2 579 __u8 __pkt_type_offset[0];
b1937227 580 __u8 pkt_type:3;
c93bdd0e 581 __u8 pfmemalloc:1;
b1937227
ED
582 __u8 ignore_df:1;
583 __u8 nfctinfo:3;
584
585 __u8 nf_trace:1;
586 __u8 ip_summed:2;
3853b584 587 __u8 ooo_okay:1;
61b905da 588 __u8 l4_hash:1;
a3b18ddb 589 __u8 sw_hash:1;
6e3e939f
JB
590 __u8 wifi_acked_valid:1;
591 __u8 wifi_acked:1;
b1937227 592
3bdc0eba 593 __u8 no_fcs:1;
77cffe23 594 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 595 __u8 encapsulation:1;
7e2b10c1 596 __u8 encap_hdr_csum:1;
5d0c2b95 597 __u8 csum_valid:1;
7e3cead5 598 __u8 csum_complete_sw:1;
b1937227
ED
599 __u8 csum_level:2;
600 __u8 csum_bad:1;
fe55f6d5 601
b1937227
ED
602#ifdef CONFIG_IPV6_NDISC_NODETYPE
603 __u8 ndisc_nodetype:2;
604#endif
605 __u8 ipvs_property:1;
8bce6d7d 606 __u8 inner_protocol_type:1;
e585f236
TH
607 __u8 remcsum_offload:1;
608 /* 3 or 5 bit hole */
b1937227
ED
609
610#ifdef CONFIG_NET_SCHED
611 __u16 tc_index; /* traffic control index */
612#ifdef CONFIG_NET_CLS_ACT
613 __u16 tc_verd; /* traffic control verdict */
614#endif
615#endif
fe55f6d5 616
b1937227
ED
617 union {
618 __wsum csum;
619 struct {
620 __u16 csum_start;
621 __u16 csum_offset;
622 };
623 };
624 __u32 priority;
625 int skb_iif;
626 __u32 hash;
627 __be16 vlan_proto;
628 __u16 vlan_tci;
7bced397
DW
629#ifdef CONFIG_NET_RX_BUSY_POLL
630 unsigned int napi_id;
97fc2f08 631#endif
984bc16c
JM
632#ifdef CONFIG_NETWORK_SECMARK
633 __u32 secmark;
634#endif
3b885787
NH
635 union {
636 __u32 mark;
637 __u32 dropcount;
16fad69c 638 __u32 reserved_tailroom;
3b885787 639 };
1da177e4 640
8bce6d7d
TH
641 union {
642 __be16 inner_protocol;
643 __u8 inner_ipproto;
644 };
645
1a37e412
SH
646 __u16 inner_transport_header;
647 __u16 inner_network_header;
648 __u16 inner_mac_header;
b1937227
ED
649
650 __be16 protocol;
1a37e412
SH
651 __u16 transport_header;
652 __u16 network_header;
653 __u16 mac_header;
b1937227 654
ebcf34f3 655 /* private: */
b1937227 656 __u32 headers_end[0];
ebcf34f3 657 /* public: */
b1937227 658
1da177e4 659 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 660 sk_buff_data_t tail;
4305b541 661 sk_buff_data_t end;
1da177e4 662 unsigned char *head,
4305b541 663 *data;
27a884dc
ACM
664 unsigned int truesize;
665 atomic_t users;
1da177e4
LT
666};
667
668#ifdef __KERNEL__
669/*
670 * Handling routines are only of interest to the kernel
671 */
672#include <linux/slab.h>
673
1da177e4 674
c93bdd0e
MG
675#define SKB_ALLOC_FCLONE 0x01
676#define SKB_ALLOC_RX 0x02
fd11a83d 677#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
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
932bc4d7
JA
722/**
723 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
724 * @skb: buffer
725 * @dst: dst entry
726 *
727 * Sets skb dst, assuming a reference was not taken on dst.
728 * If dst entry is cached, we do not take reference and dst_release
729 * will be avoided by refdst_drop. If dst entry is not cached, we take
730 * reference, so that last dst_release can destroy the dst immediately.
731 */
732static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
733{
dbfc4fb7
HFS
734 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
735 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 736}
7fee226a
ED
737
738/**
25985edc 739 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
740 * @skb: buffer
741 */
742static inline bool skb_dst_is_noref(const struct sk_buff *skb)
743{
744 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
745}
746
511c3f92
ED
747static inline struct rtable *skb_rtable(const struct sk_buff *skb)
748{
adf30907 749 return (struct rtable *)skb_dst(skb);
511c3f92
ED
750}
751
7965bd4d
JP
752void kfree_skb(struct sk_buff *skb);
753void kfree_skb_list(struct sk_buff *segs);
754void skb_tx_error(struct sk_buff *skb);
755void consume_skb(struct sk_buff *skb);
756void __kfree_skb(struct sk_buff *skb);
d7e8883c 757extern struct kmem_cache *skbuff_head_cache;
bad43ca8 758
7965bd4d
JP
759void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
760bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
761 bool *fragstolen, int *delta_truesize);
bad43ca8 762
7965bd4d
JP
763struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
764 int node);
765struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 766static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 767 gfp_t priority)
d179cd12 768{
564824b0 769 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
770}
771
2e4e4410
ED
772struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
773 unsigned long data_len,
774 int max_page_order,
775 int *errcode,
776 gfp_t gfp_mask);
777
d0bf4a9e
ED
778/* Layout of fast clones : [skb1][skb2][fclone_ref] */
779struct sk_buff_fclones {
780 struct sk_buff skb1;
781
782 struct sk_buff skb2;
783
784 atomic_t fclone_ref;
785};
786
787/**
788 * skb_fclone_busy - check if fclone is busy
789 * @skb: buffer
790 *
791 * Returns true is skb is a fast clone, and its clone is not freed.
39bb5e62
ED
792 * Some drivers call skb_orphan() in their ndo_start_xmit(),
793 * so we also check that this didnt happen.
d0bf4a9e 794 */
39bb5e62
ED
795static inline bool skb_fclone_busy(const struct sock *sk,
796 const struct sk_buff *skb)
d0bf4a9e
ED
797{
798 const struct sk_buff_fclones *fclones;
799
800 fclones = container_of(skb, struct sk_buff_fclones, skb1);
801
802 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 803 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 804 fclones->skb2.sk == sk;
d0bf4a9e
ED
805}
806
d179cd12 807static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 808 gfp_t priority)
d179cd12 809{
c93bdd0e 810 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
811}
812
7965bd4d 813struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
814static inline struct sk_buff *alloc_skb_head(gfp_t priority)
815{
816 return __alloc_skb_head(priority, -1);
817}
818
7965bd4d
JP
819struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
820int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
821struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
822struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
823struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
824 gfp_t gfp_mask, bool fclone);
825static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
826 gfp_t gfp_mask)
827{
828 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
829}
7965bd4d
JP
830
831int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
832struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
833 unsigned int headroom);
834struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
835 int newtailroom, gfp_t priority);
25a91d8d
FD
836int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
837 int offset, int len);
7965bd4d
JP
838int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
839 int len);
840int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
841int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 842#define dev_kfree_skb(a) consume_skb(a)
1da177e4 843
7965bd4d
JP
844int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
845 int getfrag(void *from, char *to, int offset,
846 int len, int odd, struct sk_buff *skb),
847 void *from, int length);
e89e9cf5 848
d94d9fee 849struct skb_seq_state {
677e90ed
TG
850 __u32 lower_offset;
851 __u32 upper_offset;
852 __u32 frag_idx;
853 __u32 stepped_offset;
854 struct sk_buff *root_skb;
855 struct sk_buff *cur_skb;
856 __u8 *frag_data;
857};
858
7965bd4d
JP
859void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
860 unsigned int to, struct skb_seq_state *st);
861unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
862 struct skb_seq_state *st);
863void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 864
7965bd4d
JP
865unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
866 unsigned int to, struct ts_config *config,
867 struct ts_state *state);
3fc7e8a6 868
09323cc4
TH
869/*
870 * Packet hash types specify the type of hash in skb_set_hash.
871 *
872 * Hash types refer to the protocol layer addresses which are used to
873 * construct a packet's hash. The hashes are used to differentiate or identify
874 * flows of the protocol layer for the hash type. Hash types are either
875 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
876 *
877 * Properties of hashes:
878 *
879 * 1) Two packets in different flows have different hash values
880 * 2) Two packets in the same flow should have the same hash value
881 *
882 * A hash at a higher layer is considered to be more specific. A driver should
883 * set the most specific hash possible.
884 *
885 * A driver cannot indicate a more specific hash than the layer at which a hash
886 * was computed. For instance an L3 hash cannot be set as an L4 hash.
887 *
888 * A driver may indicate a hash level which is less specific than the
889 * actual layer the hash was computed on. For instance, a hash computed
890 * at L4 may be considered an L3 hash. This should only be done if the
891 * driver can't unambiguously determine that the HW computed the hash at
892 * the higher layer. Note that the "should" in the second property above
893 * permits this.
894 */
895enum pkt_hash_types {
896 PKT_HASH_TYPE_NONE, /* Undefined type */
897 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
898 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
899 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
900};
901
902static inline void
903skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
904{
61b905da 905 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 906 skb->sw_hash = 0;
61b905da 907 skb->hash = hash;
09323cc4
TH
908}
909
3958afa1
TH
910void __skb_get_hash(struct sk_buff *skb);
911static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 912{
a3b18ddb 913 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 914 __skb_get_hash(skb);
bfb564e7 915
61b905da 916 return skb->hash;
bfb564e7
KK
917}
918
57bdf7f4
TH
919static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
920{
61b905da 921 return skb->hash;
57bdf7f4
TH
922}
923
7539fadc
TH
924static inline void skb_clear_hash(struct sk_buff *skb)
925{
61b905da 926 skb->hash = 0;
a3b18ddb 927 skb->sw_hash = 0;
61b905da 928 skb->l4_hash = 0;
7539fadc
TH
929}
930
931static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
932{
61b905da 933 if (!skb->l4_hash)
7539fadc
TH
934 skb_clear_hash(skb);
935}
936
3df7a74e
TH
937static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
938{
61b905da 939 to->hash = from->hash;
a3b18ddb 940 to->sw_hash = from->sw_hash;
61b905da 941 to->l4_hash = from->l4_hash;
3df7a74e
TH
942};
943
4305b541
ACM
944#ifdef NET_SKBUFF_DATA_USES_OFFSET
945static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
946{
947 return skb->head + skb->end;
948}
ec47ea82
AD
949
950static inline unsigned int skb_end_offset(const struct sk_buff *skb)
951{
952 return skb->end;
953}
4305b541
ACM
954#else
955static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
956{
957 return skb->end;
958}
ec47ea82
AD
959
960static inline unsigned int skb_end_offset(const struct sk_buff *skb)
961{
962 return skb->end - skb->head;
963}
4305b541
ACM
964#endif
965
1da177e4 966/* Internal */
4305b541 967#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 968
ac45f602
PO
969static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
970{
971 return &skb_shinfo(skb)->hwtstamps;
972}
973
1da177e4
LT
974/**
975 * skb_queue_empty - check if a queue is empty
976 * @list: queue head
977 *
978 * Returns true if the queue is empty, false otherwise.
979 */
980static inline int skb_queue_empty(const struct sk_buff_head *list)
981{
fd44b93c 982 return list->next == (const struct sk_buff *) list;
1da177e4
LT
983}
984
fc7ebb21
DM
985/**
986 * skb_queue_is_last - check if skb is the last entry in the queue
987 * @list: queue head
988 * @skb: buffer
989 *
990 * Returns true if @skb is the last buffer on the list.
991 */
992static inline bool skb_queue_is_last(const struct sk_buff_head *list,
993 const struct sk_buff *skb)
994{
fd44b93c 995 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
996}
997
832d11c5
IJ
998/**
999 * skb_queue_is_first - check if skb is the first entry in the queue
1000 * @list: queue head
1001 * @skb: buffer
1002 *
1003 * Returns true if @skb is the first buffer on the list.
1004 */
1005static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1006 const struct sk_buff *skb)
1007{
fd44b93c 1008 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1009}
1010
249c8b42
DM
1011/**
1012 * skb_queue_next - return the next packet in the queue
1013 * @list: queue head
1014 * @skb: current buffer
1015 *
1016 * Return the next packet in @list after @skb. It is only valid to
1017 * call this if skb_queue_is_last() evaluates to false.
1018 */
1019static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1020 const struct sk_buff *skb)
1021{
1022 /* This BUG_ON may seem severe, but if we just return then we
1023 * are going to dereference garbage.
1024 */
1025 BUG_ON(skb_queue_is_last(list, skb));
1026 return skb->next;
1027}
1028
832d11c5
IJ
1029/**
1030 * skb_queue_prev - return the prev packet in the queue
1031 * @list: queue head
1032 * @skb: current buffer
1033 *
1034 * Return the prev packet in @list before @skb. It is only valid to
1035 * call this if skb_queue_is_first() evaluates to false.
1036 */
1037static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1038 const struct sk_buff *skb)
1039{
1040 /* This BUG_ON may seem severe, but if we just return then we
1041 * are going to dereference garbage.
1042 */
1043 BUG_ON(skb_queue_is_first(list, skb));
1044 return skb->prev;
1045}
1046
1da177e4
LT
1047/**
1048 * skb_get - reference buffer
1049 * @skb: buffer to reference
1050 *
1051 * Makes another reference to a socket buffer and returns a pointer
1052 * to the buffer.
1053 */
1054static inline struct sk_buff *skb_get(struct sk_buff *skb)
1055{
1056 atomic_inc(&skb->users);
1057 return skb;
1058}
1059
1060/*
1061 * If users == 1, we are the only owner and are can avoid redundant
1062 * atomic change.
1063 */
1064
1da177e4
LT
1065/**
1066 * skb_cloned - is the buffer a clone
1067 * @skb: buffer to check
1068 *
1069 * Returns true if the buffer was generated with skb_clone() and is
1070 * one of multiple shared copies of the buffer. Cloned buffers are
1071 * shared data so must not be written to under normal circumstances.
1072 */
1073static inline int skb_cloned(const struct sk_buff *skb)
1074{
1075 return skb->cloned &&
1076 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1077}
1078
14bbd6a5
PS
1079static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1080{
1081 might_sleep_if(pri & __GFP_WAIT);
1082
1083 if (skb_cloned(skb))
1084 return pskb_expand_head(skb, 0, 0, pri);
1085
1086 return 0;
1087}
1088
1da177e4
LT
1089/**
1090 * skb_header_cloned - is the header a clone
1091 * @skb: buffer to check
1092 *
1093 * Returns true if modifying the header part of the buffer requires
1094 * the data to be copied.
1095 */
1096static inline int skb_header_cloned(const struct sk_buff *skb)
1097{
1098 int dataref;
1099
1100 if (!skb->cloned)
1101 return 0;
1102
1103 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1104 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1105 return dataref != 1;
1106}
1107
1108/**
1109 * skb_header_release - release reference to header
1110 * @skb: buffer to operate on
1111 *
1112 * Drop a reference to the header part of the buffer. This is done
1113 * by acquiring a payload reference. You must not read from the header
1114 * part of skb->data after this.
f4a775d1 1115 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1116 */
1117static inline void skb_header_release(struct sk_buff *skb)
1118{
1119 BUG_ON(skb->nohdr);
1120 skb->nohdr = 1;
1121 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1122}
1123
f4a775d1
ED
1124/**
1125 * __skb_header_release - release reference to header
1126 * @skb: buffer to operate on
1127 *
1128 * Variant of skb_header_release() assuming skb is private to caller.
1129 * We can avoid one atomic operation.
1130 */
1131static inline void __skb_header_release(struct sk_buff *skb)
1132{
1133 skb->nohdr = 1;
1134 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1135}
1136
1137
1da177e4
LT
1138/**
1139 * skb_shared - is the buffer shared
1140 * @skb: buffer to check
1141 *
1142 * Returns true if more than one person has a reference to this
1143 * buffer.
1144 */
1145static inline int skb_shared(const struct sk_buff *skb)
1146{
1147 return atomic_read(&skb->users) != 1;
1148}
1149
1150/**
1151 * skb_share_check - check if buffer is shared and if so clone it
1152 * @skb: buffer to check
1153 * @pri: priority for memory allocation
1154 *
1155 * If the buffer is shared the buffer is cloned and the old copy
1156 * drops a reference. A new clone with a single reference is returned.
1157 * If the buffer is not shared the original buffer is returned. When
1158 * being called from interrupt status or with spinlocks held pri must
1159 * be GFP_ATOMIC.
1160 *
1161 * NULL is returned on a memory allocation failure.
1162 */
47061bc4 1163static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1164{
1165 might_sleep_if(pri & __GFP_WAIT);
1166 if (skb_shared(skb)) {
1167 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1168
1169 if (likely(nskb))
1170 consume_skb(skb);
1171 else
1172 kfree_skb(skb);
1da177e4
LT
1173 skb = nskb;
1174 }
1175 return skb;
1176}
1177
1178/*
1179 * Copy shared buffers into a new sk_buff. We effectively do COW on
1180 * packets to handle cases where we have a local reader and forward
1181 * and a couple of other messy ones. The normal one is tcpdumping
1182 * a packet thats being forwarded.
1183 */
1184
1185/**
1186 * skb_unshare - make a copy of a shared buffer
1187 * @skb: buffer to check
1188 * @pri: priority for memory allocation
1189 *
1190 * If the socket buffer is a clone then this function creates a new
1191 * copy of the data, drops a reference count on the old copy and returns
1192 * the new copy with the reference count at 1. If the buffer is not a clone
1193 * the original buffer is returned. When called with a spinlock held or
1194 * from interrupt state @pri must be %GFP_ATOMIC
1195 *
1196 * %NULL is returned on a memory allocation failure.
1197 */
e2bf521d 1198static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1199 gfp_t pri)
1da177e4
LT
1200{
1201 might_sleep_if(pri & __GFP_WAIT);
1202 if (skb_cloned(skb)) {
1203 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1204
1205 /* Free our shared copy */
1206 if (likely(nskb))
1207 consume_skb(skb);
1208 else
1209 kfree_skb(skb);
1da177e4
LT
1210 skb = nskb;
1211 }
1212 return skb;
1213}
1214
1215/**
1a5778aa 1216 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1217 * @list_: list to peek at
1218 *
1219 * Peek an &sk_buff. Unlike most other operations you _MUST_
1220 * be careful with this one. A peek leaves the buffer on the
1221 * list and someone else may run off with it. You must hold
1222 * the appropriate locks or have a private queue to do this.
1223 *
1224 * Returns %NULL for an empty list or a pointer to the head element.
1225 * The reference count is not incremented and the reference is therefore
1226 * volatile. Use with caution.
1227 */
05bdd2f1 1228static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1229{
18d07000
ED
1230 struct sk_buff *skb = list_->next;
1231
1232 if (skb == (struct sk_buff *)list_)
1233 skb = NULL;
1234 return skb;
1da177e4
LT
1235}
1236
da5ef6e5
PE
1237/**
1238 * skb_peek_next - peek skb following the given one from a queue
1239 * @skb: skb to start from
1240 * @list_: list to peek at
1241 *
1242 * Returns %NULL when the end of the list is met or a pointer to the
1243 * next element. The reference count is not incremented and the
1244 * reference is therefore volatile. Use with caution.
1245 */
1246static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1247 const struct sk_buff_head *list_)
1248{
1249 struct sk_buff *next = skb->next;
18d07000 1250
da5ef6e5
PE
1251 if (next == (struct sk_buff *)list_)
1252 next = NULL;
1253 return next;
1254}
1255
1da177e4 1256/**
1a5778aa 1257 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1258 * @list_: list to peek at
1259 *
1260 * Peek an &sk_buff. Unlike most other operations you _MUST_
1261 * be careful with this one. A peek leaves the buffer on the
1262 * list and someone else may run off with it. You must hold
1263 * the appropriate locks or have a private queue to do this.
1264 *
1265 * Returns %NULL for an empty list or a pointer to the tail element.
1266 * The reference count is not incremented and the reference is therefore
1267 * volatile. Use with caution.
1268 */
05bdd2f1 1269static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1270{
18d07000
ED
1271 struct sk_buff *skb = list_->prev;
1272
1273 if (skb == (struct sk_buff *)list_)
1274 skb = NULL;
1275 return skb;
1276
1da177e4
LT
1277}
1278
1279/**
1280 * skb_queue_len - get queue length
1281 * @list_: list to measure
1282 *
1283 * Return the length of an &sk_buff queue.
1284 */
1285static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1286{
1287 return list_->qlen;
1288}
1289
67fed459
DM
1290/**
1291 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1292 * @list: queue to initialize
1293 *
1294 * This initializes only the list and queue length aspects of
1295 * an sk_buff_head object. This allows to initialize the list
1296 * aspects of an sk_buff_head without reinitializing things like
1297 * the spinlock. It can also be used for on-stack sk_buff_head
1298 * objects where the spinlock is known to not be used.
1299 */
1300static inline void __skb_queue_head_init(struct sk_buff_head *list)
1301{
1302 list->prev = list->next = (struct sk_buff *)list;
1303 list->qlen = 0;
1304}
1305
76f10ad0
AV
1306/*
1307 * This function creates a split out lock class for each invocation;
1308 * this is needed for now since a whole lot of users of the skb-queue
1309 * infrastructure in drivers have different locking usage (in hardirq)
1310 * than the networking core (in softirq only). In the long run either the
1311 * network layer or drivers should need annotation to consolidate the
1312 * main types of usage into 3 classes.
1313 */
1da177e4
LT
1314static inline void skb_queue_head_init(struct sk_buff_head *list)
1315{
1316 spin_lock_init(&list->lock);
67fed459 1317 __skb_queue_head_init(list);
1da177e4
LT
1318}
1319
c2ecba71
PE
1320static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1321 struct lock_class_key *class)
1322{
1323 skb_queue_head_init(list);
1324 lockdep_set_class(&list->lock, class);
1325}
1326
1da177e4 1327/*
bf299275 1328 * Insert an sk_buff on a list.
1da177e4
LT
1329 *
1330 * The "__skb_xxxx()" functions are the non-atomic ones that
1331 * can only be called with interrupts disabled.
1332 */
7965bd4d
JP
1333void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1334 struct sk_buff_head *list);
bf299275
GR
1335static inline void __skb_insert(struct sk_buff *newsk,
1336 struct sk_buff *prev, struct sk_buff *next,
1337 struct sk_buff_head *list)
1338{
1339 newsk->next = next;
1340 newsk->prev = prev;
1341 next->prev = prev->next = newsk;
1342 list->qlen++;
1343}
1da177e4 1344
67fed459
DM
1345static inline void __skb_queue_splice(const struct sk_buff_head *list,
1346 struct sk_buff *prev,
1347 struct sk_buff *next)
1348{
1349 struct sk_buff *first = list->next;
1350 struct sk_buff *last = list->prev;
1351
1352 first->prev = prev;
1353 prev->next = first;
1354
1355 last->next = next;
1356 next->prev = last;
1357}
1358
1359/**
1360 * skb_queue_splice - join two skb lists, this is designed for stacks
1361 * @list: the new list to add
1362 * @head: the place to add it in the first list
1363 */
1364static inline void skb_queue_splice(const struct sk_buff_head *list,
1365 struct sk_buff_head *head)
1366{
1367 if (!skb_queue_empty(list)) {
1368 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1369 head->qlen += list->qlen;
67fed459
DM
1370 }
1371}
1372
1373/**
d9619496 1374 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1375 * @list: the new list to add
1376 * @head: the place to add it in the first list
1377 *
1378 * The list at @list is reinitialised
1379 */
1380static inline void skb_queue_splice_init(struct sk_buff_head *list,
1381 struct sk_buff_head *head)
1382{
1383 if (!skb_queue_empty(list)) {
1384 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1385 head->qlen += list->qlen;
67fed459
DM
1386 __skb_queue_head_init(list);
1387 }
1388}
1389
1390/**
1391 * skb_queue_splice_tail - join two skb lists, each list being a queue
1392 * @list: the new list to add
1393 * @head: the place to add it in the first list
1394 */
1395static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1396 struct sk_buff_head *head)
1397{
1398 if (!skb_queue_empty(list)) {
1399 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1400 head->qlen += list->qlen;
67fed459
DM
1401 }
1402}
1403
1404/**
d9619496 1405 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1406 * @list: the new list to add
1407 * @head: the place to add it in the first list
1408 *
1409 * Each of the lists is a queue.
1410 * The list at @list is reinitialised
1411 */
1412static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1413 struct sk_buff_head *head)
1414{
1415 if (!skb_queue_empty(list)) {
1416 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1417 head->qlen += list->qlen;
67fed459
DM
1418 __skb_queue_head_init(list);
1419 }
1420}
1421
1da177e4 1422/**
300ce174 1423 * __skb_queue_after - queue a buffer at the list head
1da177e4 1424 * @list: list to use
300ce174 1425 * @prev: place after this buffer
1da177e4
LT
1426 * @newsk: buffer to queue
1427 *
300ce174 1428 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1429 * and you must therefore hold required locks before calling it.
1430 *
1431 * A buffer cannot be placed on two lists at the same time.
1432 */
300ce174
SH
1433static inline void __skb_queue_after(struct sk_buff_head *list,
1434 struct sk_buff *prev,
1435 struct sk_buff *newsk)
1da177e4 1436{
bf299275 1437 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1438}
1439
7965bd4d
JP
1440void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1441 struct sk_buff_head *list);
7de6c033 1442
f5572855
GR
1443static inline void __skb_queue_before(struct sk_buff_head *list,
1444 struct sk_buff *next,
1445 struct sk_buff *newsk)
1446{
1447 __skb_insert(newsk, next->prev, next, list);
1448}
1449
300ce174
SH
1450/**
1451 * __skb_queue_head - queue a buffer at the list head
1452 * @list: list to use
1453 * @newsk: buffer to queue
1454 *
1455 * Queue a buffer at the start of a list. This function takes no locks
1456 * and you must therefore hold required locks before calling it.
1457 *
1458 * A buffer cannot be placed on two lists at the same time.
1459 */
7965bd4d 1460void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1461static inline void __skb_queue_head(struct sk_buff_head *list,
1462 struct sk_buff *newsk)
1463{
1464 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1465}
1466
1da177e4
LT
1467/**
1468 * __skb_queue_tail - queue a buffer at the list tail
1469 * @list: list to use
1470 * @newsk: buffer to queue
1471 *
1472 * Queue a buffer at the end of a list. This function takes no locks
1473 * and you must therefore hold required locks before calling it.
1474 *
1475 * A buffer cannot be placed on two lists at the same time.
1476 */
7965bd4d 1477void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1478static inline void __skb_queue_tail(struct sk_buff_head *list,
1479 struct sk_buff *newsk)
1480{
f5572855 1481 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1482}
1483
1da177e4
LT
1484/*
1485 * remove sk_buff from list. _Must_ be called atomically, and with
1486 * the list known..
1487 */
7965bd4d 1488void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1489static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1490{
1491 struct sk_buff *next, *prev;
1492
1493 list->qlen--;
1494 next = skb->next;
1495 prev = skb->prev;
1496 skb->next = skb->prev = NULL;
1da177e4
LT
1497 next->prev = prev;
1498 prev->next = next;
1499}
1500
f525c06d
GR
1501/**
1502 * __skb_dequeue - remove from the head of the queue
1503 * @list: list to dequeue from
1504 *
1505 * Remove the head of the list. This function does not take any locks
1506 * so must be used with appropriate locks held only. The head item is
1507 * returned or %NULL if the list is empty.
1508 */
7965bd4d 1509struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1510static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1511{
1512 struct sk_buff *skb = skb_peek(list);
1513 if (skb)
1514 __skb_unlink(skb, list);
1515 return skb;
1516}
1da177e4
LT
1517
1518/**
1519 * __skb_dequeue_tail - remove from the tail of the queue
1520 * @list: list to dequeue from
1521 *
1522 * Remove the tail of the list. This function does not take any locks
1523 * so must be used with appropriate locks held only. The tail item is
1524 * returned or %NULL if the list is empty.
1525 */
7965bd4d 1526struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1527static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1528{
1529 struct sk_buff *skb = skb_peek_tail(list);
1530 if (skb)
1531 __skb_unlink(skb, list);
1532 return skb;
1533}
1534
1535
bdcc0924 1536static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1537{
1538 return skb->data_len;
1539}
1540
1541static inline unsigned int skb_headlen(const struct sk_buff *skb)
1542{
1543 return skb->len - skb->data_len;
1544}
1545
1546static inline int skb_pagelen(const struct sk_buff *skb)
1547{
1548 int i, len = 0;
1549
1550 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1551 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1552 return len + skb_headlen(skb);
1553}
1554
131ea667
IC
1555/**
1556 * __skb_fill_page_desc - initialise a paged fragment in an skb
1557 * @skb: buffer containing fragment to be initialised
1558 * @i: paged fragment index to initialise
1559 * @page: the page to use for this fragment
1560 * @off: the offset to the data with @page
1561 * @size: the length of the data
1562 *
1563 * Initialises the @i'th fragment of @skb to point to &size bytes at
1564 * offset @off within @page.
1565 *
1566 * Does not take any additional reference on the fragment.
1567 */
1568static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1569 struct page *page, int off, int size)
1da177e4
LT
1570{
1571 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1572
c48a11c7
MG
1573 /*
1574 * Propagate page->pfmemalloc to the skb if we can. The problem is
1575 * that not all callers have unique ownership of the page. If
1576 * pfmemalloc is set, we check the mapping as a mapping implies
1577 * page->index is set (index and pfmemalloc share space).
1578 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1579 * do not lose pfmemalloc information as the pages would not be
1580 * allocated using __GFP_MEMALLOC.
1581 */
a8605c60 1582 frag->page.p = page;
1da177e4 1583 frag->page_offset = off;
9e903e08 1584 skb_frag_size_set(frag, size);
cca7af38
PE
1585
1586 page = compound_head(page);
1587 if (page->pfmemalloc && !page->mapping)
1588 skb->pfmemalloc = true;
131ea667
IC
1589}
1590
1591/**
1592 * skb_fill_page_desc - initialise a paged fragment in an skb
1593 * @skb: buffer containing fragment to be initialised
1594 * @i: paged fragment index to initialise
1595 * @page: the page to use for this fragment
1596 * @off: the offset to the data with @page
1597 * @size: the length of the data
1598 *
1599 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1600 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1601 * addition updates @skb such that @i is the last fragment.
1602 *
1603 * Does not take any additional reference on the fragment.
1604 */
1605static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1606 struct page *page, int off, int size)
1607{
1608 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1609 skb_shinfo(skb)->nr_frags = i + 1;
1610}
1611
7965bd4d
JP
1612void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1613 int size, unsigned int truesize);
654bed16 1614
f8e617e1
JW
1615void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1616 unsigned int truesize);
1617
1da177e4 1618#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1619#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1620#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1621
27a884dc
ACM
1622#ifdef NET_SKBUFF_DATA_USES_OFFSET
1623static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1624{
1625 return skb->head + skb->tail;
1626}
1627
1628static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1629{
1630 skb->tail = skb->data - skb->head;
1631}
1632
1633static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1634{
1635 skb_reset_tail_pointer(skb);
1636 skb->tail += offset;
1637}
7cc46190 1638
27a884dc
ACM
1639#else /* NET_SKBUFF_DATA_USES_OFFSET */
1640static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1641{
1642 return skb->tail;
1643}
1644
1645static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1646{
1647 skb->tail = skb->data;
1648}
1649
1650static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1651{
1652 skb->tail = skb->data + offset;
1653}
4305b541 1654
27a884dc
ACM
1655#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1656
1da177e4
LT
1657/*
1658 * Add data to an sk_buff
1659 */
0c7ddf36 1660unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1661unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1662static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1663{
27a884dc 1664 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1665 SKB_LINEAR_ASSERT(skb);
1666 skb->tail += len;
1667 skb->len += len;
1668 return tmp;
1669}
1670
7965bd4d 1671unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1672static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1673{
1674 skb->data -= len;
1675 skb->len += len;
1676 return skb->data;
1677}
1678
7965bd4d 1679unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1680static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1681{
1682 skb->len -= len;
1683 BUG_ON(skb->len < skb->data_len);
1684 return skb->data += len;
1685}
1686
47d29646
DM
1687static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1688{
1689 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1690}
1691
7965bd4d 1692unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1693
1694static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1695{
1696 if (len > skb_headlen(skb) &&
987c402a 1697 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1698 return NULL;
1699 skb->len -= len;
1700 return skb->data += len;
1701}
1702
1703static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1704{
1705 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1706}
1707
1708static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1709{
1710 if (likely(len <= skb_headlen(skb)))
1711 return 1;
1712 if (unlikely(len > skb->len))
1713 return 0;
987c402a 1714 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1715}
1716
1717/**
1718 * skb_headroom - bytes at buffer head
1719 * @skb: buffer to check
1720 *
1721 * Return the number of bytes of free space at the head of an &sk_buff.
1722 */
c2636b4d 1723static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1724{
1725 return skb->data - skb->head;
1726}
1727
1728/**
1729 * skb_tailroom - bytes at buffer end
1730 * @skb: buffer to check
1731 *
1732 * Return the number of bytes of free space at the tail of an sk_buff
1733 */
1734static inline int skb_tailroom(const struct sk_buff *skb)
1735{
4305b541 1736 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1737}
1738
a21d4572
ED
1739/**
1740 * skb_availroom - bytes at buffer end
1741 * @skb: buffer to check
1742 *
1743 * Return the number of bytes of free space at the tail of an sk_buff
1744 * allocated by sk_stream_alloc()
1745 */
1746static inline int skb_availroom(const struct sk_buff *skb)
1747{
16fad69c
ED
1748 if (skb_is_nonlinear(skb))
1749 return 0;
1750
1751 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1752}
1753
1da177e4
LT
1754/**
1755 * skb_reserve - adjust headroom
1756 * @skb: buffer to alter
1757 * @len: bytes to move
1758 *
1759 * Increase the headroom of an empty &sk_buff by reducing the tail
1760 * room. This is only allowed for an empty buffer.
1761 */
8243126c 1762static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1763{
1764 skb->data += len;
1765 skb->tail += len;
1766}
1767
8bce6d7d
TH
1768#define ENCAP_TYPE_ETHER 0
1769#define ENCAP_TYPE_IPPROTO 1
1770
1771static inline void skb_set_inner_protocol(struct sk_buff *skb,
1772 __be16 protocol)
1773{
1774 skb->inner_protocol = protocol;
1775 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1776}
1777
1778static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1779 __u8 ipproto)
1780{
1781 skb->inner_ipproto = ipproto;
1782 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1783}
1784
6a674e9c
JG
1785static inline void skb_reset_inner_headers(struct sk_buff *skb)
1786{
aefbd2b3 1787 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1788 skb->inner_network_header = skb->network_header;
1789 skb->inner_transport_header = skb->transport_header;
1790}
1791
0b5c9db1
JP
1792static inline void skb_reset_mac_len(struct sk_buff *skb)
1793{
1794 skb->mac_len = skb->network_header - skb->mac_header;
1795}
1796
6a674e9c
JG
1797static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1798 *skb)
1799{
1800 return skb->head + skb->inner_transport_header;
1801}
1802
1803static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1804{
1805 skb->inner_transport_header = skb->data - skb->head;
1806}
1807
1808static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1809 const int offset)
1810{
1811 skb_reset_inner_transport_header(skb);
1812 skb->inner_transport_header += offset;
1813}
1814
1815static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1816{
1817 return skb->head + skb->inner_network_header;
1818}
1819
1820static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1821{
1822 skb->inner_network_header = skb->data - skb->head;
1823}
1824
1825static inline void skb_set_inner_network_header(struct sk_buff *skb,
1826 const int offset)
1827{
1828 skb_reset_inner_network_header(skb);
1829 skb->inner_network_header += offset;
1830}
1831
aefbd2b3
PS
1832static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1833{
1834 return skb->head + skb->inner_mac_header;
1835}
1836
1837static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1838{
1839 skb->inner_mac_header = skb->data - skb->head;
1840}
1841
1842static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1843 const int offset)
1844{
1845 skb_reset_inner_mac_header(skb);
1846 skb->inner_mac_header += offset;
1847}
fda55eca
ED
1848static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1849{
35d04610 1850 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1851}
1852
9c70220b
ACM
1853static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1854{
2e07fa9c 1855 return skb->head + skb->transport_header;
9c70220b
ACM
1856}
1857
badff6d0
ACM
1858static inline void skb_reset_transport_header(struct sk_buff *skb)
1859{
2e07fa9c 1860 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1861}
1862
967b05f6
ACM
1863static inline void skb_set_transport_header(struct sk_buff *skb,
1864 const int offset)
1865{
2e07fa9c
ACM
1866 skb_reset_transport_header(skb);
1867 skb->transport_header += offset;
ea2ae17d
ACM
1868}
1869
d56f90a7
ACM
1870static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1871{
2e07fa9c 1872 return skb->head + skb->network_header;
d56f90a7
ACM
1873}
1874
c1d2bbe1
ACM
1875static inline void skb_reset_network_header(struct sk_buff *skb)
1876{
2e07fa9c 1877 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1878}
1879
c14d2450
ACM
1880static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1881{
2e07fa9c
ACM
1882 skb_reset_network_header(skb);
1883 skb->network_header += offset;
c14d2450
ACM
1884}
1885
2e07fa9c 1886static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1887{
2e07fa9c 1888 return skb->head + skb->mac_header;
bbe735e4
ACM
1889}
1890
2e07fa9c 1891static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1892{
35d04610 1893 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1894}
1895
1896static inline void skb_reset_mac_header(struct sk_buff *skb)
1897{
1898 skb->mac_header = skb->data - skb->head;
1899}
1900
1901static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1902{
1903 skb_reset_mac_header(skb);
1904 skb->mac_header += offset;
1905}
1906
0e3da5bb
TT
1907static inline void skb_pop_mac_header(struct sk_buff *skb)
1908{
1909 skb->mac_header = skb->network_header;
1910}
1911
fbbdb8f0
YX
1912static inline void skb_probe_transport_header(struct sk_buff *skb,
1913 const int offset_hint)
1914{
1915 struct flow_keys keys;
1916
1917 if (skb_transport_header_was_set(skb))
1918 return;
1919 else if (skb_flow_dissect(skb, &keys))
1920 skb_set_transport_header(skb, keys.thoff);
1921 else
1922 skb_set_transport_header(skb, offset_hint);
1923}
1924
03606895
ED
1925static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1926{
1927 if (skb_mac_header_was_set(skb)) {
1928 const unsigned char *old_mac = skb_mac_header(skb);
1929
1930 skb_set_mac_header(skb, -skb->mac_len);
1931 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1932 }
1933}
1934
04fb451e
MM
1935static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1936{
1937 return skb->csum_start - skb_headroom(skb);
1938}
1939
2e07fa9c
ACM
1940static inline int skb_transport_offset(const struct sk_buff *skb)
1941{
1942 return skb_transport_header(skb) - skb->data;
1943}
1944
1945static inline u32 skb_network_header_len(const struct sk_buff *skb)
1946{
1947 return skb->transport_header - skb->network_header;
1948}
1949
6a674e9c
JG
1950static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1951{
1952 return skb->inner_transport_header - skb->inner_network_header;
1953}
1954
2e07fa9c
ACM
1955static inline int skb_network_offset(const struct sk_buff *skb)
1956{
1957 return skb_network_header(skb) - skb->data;
1958}
48d49d0c 1959
6a674e9c
JG
1960static inline int skb_inner_network_offset(const struct sk_buff *skb)
1961{
1962 return skb_inner_network_header(skb) - skb->data;
1963}
1964
f9599ce1
CG
1965static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1966{
1967 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1968}
1969
1da177e4
LT
1970/*
1971 * CPUs often take a performance hit when accessing unaligned memory
1972 * locations. The actual performance hit varies, it can be small if the
1973 * hardware handles it or large if we have to take an exception and fix it
1974 * in software.
1975 *
1976 * Since an ethernet header is 14 bytes network drivers often end up with
1977 * the IP header at an unaligned offset. The IP header can be aligned by
1978 * shifting the start of the packet by 2 bytes. Drivers should do this
1979 * with:
1980 *
8660c124 1981 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1982 *
1983 * The downside to this alignment of the IP header is that the DMA is now
1984 * unaligned. On some architectures the cost of an unaligned DMA is high
1985 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1986 *
1da177e4
LT
1987 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1988 * to be overridden.
1989 */
1990#ifndef NET_IP_ALIGN
1991#define NET_IP_ALIGN 2
1992#endif
1993
025be81e
AB
1994/*
1995 * The networking layer reserves some headroom in skb data (via
1996 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1997 * the header has to grow. In the default case, if the header has to grow
d6301d3d 1998 * 32 bytes or less we avoid the reallocation.
025be81e
AB
1999 *
2000 * Unfortunately this headroom changes the DMA alignment of the resulting
2001 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2002 * on some architectures. An architecture can override this value,
2003 * perhaps setting it to a cacheline in size (since that will maintain
2004 * cacheline alignment of the DMA). It must be a power of 2.
2005 *
d6301d3d 2006 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2007 * headroom, you should not reduce this.
5933dd2f
ED
2008 *
2009 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2010 * to reduce average number of cache lines per packet.
2011 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2012 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2013 */
2014#ifndef NET_SKB_PAD
5933dd2f 2015#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2016#endif
2017
7965bd4d 2018int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2019
2020static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2021{
c4264f27 2022 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2023 WARN_ON(1);
2024 return;
2025 }
27a884dc
ACM
2026 skb->len = len;
2027 skb_set_tail_pointer(skb, len);
1da177e4
LT
2028}
2029
7965bd4d 2030void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2031
2032static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2033{
3cc0e873
HX
2034 if (skb->data_len)
2035 return ___pskb_trim(skb, len);
2036 __skb_trim(skb, len);
2037 return 0;
1da177e4
LT
2038}
2039
2040static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2041{
2042 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2043}
2044
e9fa4f7b
HX
2045/**
2046 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2047 * @skb: buffer to alter
2048 * @len: new length
2049 *
2050 * This is identical to pskb_trim except that the caller knows that
2051 * the skb is not cloned so we should never get an error due to out-
2052 * of-memory.
2053 */
2054static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2055{
2056 int err = pskb_trim(skb, len);
2057 BUG_ON(err);
2058}
2059
1da177e4
LT
2060/**
2061 * skb_orphan - orphan a buffer
2062 * @skb: buffer to orphan
2063 *
2064 * If a buffer currently has an owner then we call the owner's
2065 * destructor function and make the @skb unowned. The buffer continues
2066 * to exist but is no longer charged to its former owner.
2067 */
2068static inline void skb_orphan(struct sk_buff *skb)
2069{
c34a7612 2070 if (skb->destructor) {
1da177e4 2071 skb->destructor(skb);
c34a7612
ED
2072 skb->destructor = NULL;
2073 skb->sk = NULL;
376c7311
ED
2074 } else {
2075 BUG_ON(skb->sk);
c34a7612 2076 }
1da177e4
LT
2077}
2078
a353e0ce
MT
2079/**
2080 * skb_orphan_frags - orphan the frags contained in a buffer
2081 * @skb: buffer to orphan frags from
2082 * @gfp_mask: allocation mask for replacement pages
2083 *
2084 * For each frag in the SKB which needs a destructor (i.e. has an
2085 * owner) create a copy of that frag and release the original
2086 * page by calling the destructor.
2087 */
2088static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2089{
2090 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2091 return 0;
2092 return skb_copy_ubufs(skb, gfp_mask);
2093}
2094
1da177e4
LT
2095/**
2096 * __skb_queue_purge - empty a list
2097 * @list: list to empty
2098 *
2099 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2100 * the list and one reference dropped. This function does not take the
2101 * list lock and the caller must hold the relevant locks to use it.
2102 */
7965bd4d 2103void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2104static inline void __skb_queue_purge(struct sk_buff_head *list)
2105{
2106 struct sk_buff *skb;
2107 while ((skb = __skb_dequeue(list)) != NULL)
2108 kfree_skb(skb);
2109}
2110
e5e67305
AD
2111#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2112#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2113#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2114
7965bd4d 2115void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2116
7965bd4d
JP
2117struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2118 gfp_t gfp_mask);
8af27456
CH
2119
2120/**
2121 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2122 * @dev: network device to receive on
2123 * @length: length to allocate
2124 *
2125 * Allocate a new &sk_buff and assign it a usage count of one. The
2126 * buffer has unspecified headroom built in. Users should allocate
2127 * the headroom they think they need without accounting for the
2128 * built in space. The built in space is used for optimisations.
2129 *
2130 * %NULL is returned if there is no free memory. Although this function
2131 * allocates memory it can be called from an interrupt.
2132 */
2133static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2134 unsigned int length)
8af27456
CH
2135{
2136 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2137}
2138
6f532612
ED
2139/* legacy helper around __netdev_alloc_skb() */
2140static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2141 gfp_t gfp_mask)
2142{
2143 return __netdev_alloc_skb(NULL, length, gfp_mask);
2144}
2145
2146/* legacy helper around netdev_alloc_skb() */
2147static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2148{
2149 return netdev_alloc_skb(NULL, length);
2150}
2151
2152
4915a0de
ED
2153static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2154 unsigned int length, gfp_t gfp)
61321bbd 2155{
4915a0de 2156 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2157
2158 if (NET_IP_ALIGN && skb)
2159 skb_reserve(skb, NET_IP_ALIGN);
2160 return skb;
2161}
2162
4915a0de
ED
2163static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2164 unsigned int length)
2165{
2166 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2167}
2168
ffde7328 2169void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2170struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2171 unsigned int length, gfp_t gfp_mask);
2172static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2173 unsigned int length)
2174{
2175 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2176}
ffde7328 2177
71dfda58
AD
2178/**
2179 * __dev_alloc_pages - allocate page for network Rx
2180 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2181 * @order: size of the allocation
2182 *
2183 * Allocate a new page.
2184 *
2185 * %NULL is returned if there is no free memory.
2186*/
2187static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2188 unsigned int order)
2189{
2190 /* This piece of code contains several assumptions.
2191 * 1. This is for device Rx, therefor a cold page is preferred.
2192 * 2. The expectation is the user wants a compound page.
2193 * 3. If requesting a order 0 page it will not be compound
2194 * due to the check to see if order has a value in prep_new_page
2195 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2196 * code in gfp_to_alloc_flags that should be enforcing this.
2197 */
2198 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2199
2200 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2201}
2202
2203static inline struct page *dev_alloc_pages(unsigned int order)
2204{
2205 return __dev_alloc_pages(GFP_ATOMIC, order);
2206}
2207
2208/**
2209 * __dev_alloc_page - allocate a page for network Rx
2210 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2211 *
2212 * Allocate a new page.
2213 *
2214 * %NULL is returned if there is no free memory.
2215 */
2216static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2217{
2218 return __dev_alloc_pages(gfp_mask, 0);
2219}
2220
2221static inline struct page *dev_alloc_page(void)
2222{
2223 return __dev_alloc_page(GFP_ATOMIC);
2224}
2225
0614002b
MG
2226/**
2227 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2228 * @page: The page that was allocated from skb_alloc_page
2229 * @skb: The skb that may need pfmemalloc set
2230 */
2231static inline void skb_propagate_pfmemalloc(struct page *page,
2232 struct sk_buff *skb)
2233{
2234 if (page && page->pfmemalloc)
2235 skb->pfmemalloc = true;
2236}
2237
131ea667 2238/**
e227867f 2239 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2240 * @frag: the paged fragment
2241 *
2242 * Returns the &struct page associated with @frag.
2243 */
2244static inline struct page *skb_frag_page(const skb_frag_t *frag)
2245{
a8605c60 2246 return frag->page.p;
131ea667
IC
2247}
2248
2249/**
2250 * __skb_frag_ref - take an addition reference on a paged fragment.
2251 * @frag: the paged fragment
2252 *
2253 * Takes an additional reference on the paged fragment @frag.
2254 */
2255static inline void __skb_frag_ref(skb_frag_t *frag)
2256{
2257 get_page(skb_frag_page(frag));
2258}
2259
2260/**
2261 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2262 * @skb: the buffer
2263 * @f: the fragment offset.
2264 *
2265 * Takes an additional reference on the @f'th paged fragment of @skb.
2266 */
2267static inline void skb_frag_ref(struct sk_buff *skb, int f)
2268{
2269 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2270}
2271
2272/**
2273 * __skb_frag_unref - release a reference on a paged fragment.
2274 * @frag: the paged fragment
2275 *
2276 * Releases a reference on the paged fragment @frag.
2277 */
2278static inline void __skb_frag_unref(skb_frag_t *frag)
2279{
2280 put_page(skb_frag_page(frag));
2281}
2282
2283/**
2284 * skb_frag_unref - release a reference on a paged fragment of an skb.
2285 * @skb: the buffer
2286 * @f: the fragment offset
2287 *
2288 * Releases a reference on the @f'th paged fragment of @skb.
2289 */
2290static inline void skb_frag_unref(struct sk_buff *skb, int f)
2291{
2292 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2293}
2294
2295/**
2296 * skb_frag_address - gets the address of the data contained in a paged fragment
2297 * @frag: the paged fragment buffer
2298 *
2299 * Returns the address of the data within @frag. The page must already
2300 * be mapped.
2301 */
2302static inline void *skb_frag_address(const skb_frag_t *frag)
2303{
2304 return page_address(skb_frag_page(frag)) + frag->page_offset;
2305}
2306
2307/**
2308 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2309 * @frag: the paged fragment buffer
2310 *
2311 * Returns the address of the data within @frag. Checks that the page
2312 * is mapped and returns %NULL otherwise.
2313 */
2314static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2315{
2316 void *ptr = page_address(skb_frag_page(frag));
2317 if (unlikely(!ptr))
2318 return NULL;
2319
2320 return ptr + frag->page_offset;
2321}
2322
2323/**
2324 * __skb_frag_set_page - sets the page contained in a paged fragment
2325 * @frag: the paged fragment
2326 * @page: the page to set
2327 *
2328 * Sets the fragment @frag to contain @page.
2329 */
2330static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2331{
a8605c60 2332 frag->page.p = page;
131ea667
IC
2333}
2334
2335/**
2336 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2337 * @skb: the buffer
2338 * @f: the fragment offset
2339 * @page: the page to set
2340 *
2341 * Sets the @f'th fragment of @skb to contain @page.
2342 */
2343static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2344 struct page *page)
2345{
2346 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2347}
2348
400dfd3a
ED
2349bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2350
131ea667
IC
2351/**
2352 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2353 * @dev: the device to map the fragment to
131ea667
IC
2354 * @frag: the paged fragment to map
2355 * @offset: the offset within the fragment (starting at the
2356 * fragment's own offset)
2357 * @size: the number of bytes to map
f83347df 2358 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2359 *
2360 * Maps the page associated with @frag to @device.
2361 */
2362static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2363 const skb_frag_t *frag,
2364 size_t offset, size_t size,
2365 enum dma_data_direction dir)
2366{
2367 return dma_map_page(dev, skb_frag_page(frag),
2368 frag->page_offset + offset, size, dir);
2369}
2370
117632e6
ED
2371static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2372 gfp_t gfp_mask)
2373{
2374 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2375}
2376
bad93e9d
OP
2377
2378static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2379 gfp_t gfp_mask)
2380{
2381 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2382}
2383
2384
334a8132
PM
2385/**
2386 * skb_clone_writable - is the header of a clone writable
2387 * @skb: buffer to check
2388 * @len: length up to which to write
2389 *
2390 * Returns true if modifying the header part of the cloned buffer
2391 * does not requires the data to be copied.
2392 */
05bdd2f1 2393static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2394{
2395 return !skb_header_cloned(skb) &&
2396 skb_headroom(skb) + len <= skb->hdr_len;
2397}
2398
d9cc2048
HX
2399static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2400 int cloned)
2401{
2402 int delta = 0;
2403
d9cc2048
HX
2404 if (headroom > skb_headroom(skb))
2405 delta = headroom - skb_headroom(skb);
2406
2407 if (delta || cloned)
2408 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2409 GFP_ATOMIC);
2410 return 0;
2411}
2412
1da177e4
LT
2413/**
2414 * skb_cow - copy header of skb when it is required
2415 * @skb: buffer to cow
2416 * @headroom: needed headroom
2417 *
2418 * If the skb passed lacks sufficient headroom or its data part
2419 * is shared, data is reallocated. If reallocation fails, an error
2420 * is returned and original skb is not changed.
2421 *
2422 * The result is skb with writable area skb->head...skb->tail
2423 * and at least @headroom of space at head.
2424 */
2425static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2426{
d9cc2048
HX
2427 return __skb_cow(skb, headroom, skb_cloned(skb));
2428}
1da177e4 2429
d9cc2048
HX
2430/**
2431 * skb_cow_head - skb_cow but only making the head writable
2432 * @skb: buffer to cow
2433 * @headroom: needed headroom
2434 *
2435 * This function is identical to skb_cow except that we replace the
2436 * skb_cloned check by skb_header_cloned. It should be used when
2437 * you only need to push on some header and do not need to modify
2438 * the data.
2439 */
2440static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2441{
2442 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2443}
2444
2445/**
2446 * skb_padto - pad an skbuff up to a minimal size
2447 * @skb: buffer to pad
2448 * @len: minimal length
2449 *
2450 * Pads up a buffer to ensure the trailing bytes exist and are
2451 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2452 * is untouched. Otherwise it is extended. Returns zero on
2453 * success. The skb is freed on error.
1da177e4 2454 */
5b057c6b 2455static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2456{
2457 unsigned int size = skb->len;
2458 if (likely(size >= len))
5b057c6b 2459 return 0;
987c402a 2460 return skb_pad(skb, len - size);
1da177e4
LT
2461}
2462
9c0c1124
AD
2463/**
2464 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2465 * @skb: buffer to pad
2466 * @len: minimal length
2467 *
2468 * Pads up a buffer to ensure the trailing bytes exist and are
2469 * blanked. If the buffer already contains sufficient data it
2470 * is untouched. Otherwise it is extended. Returns zero on
2471 * success. The skb is freed on error.
2472 */
2473static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2474{
2475 unsigned int size = skb->len;
2476
2477 if (unlikely(size < len)) {
2478 len -= size;
2479 if (skb_pad(skb, len))
2480 return -ENOMEM;
2481 __skb_put(skb, len);
2482 }
2483 return 0;
2484}
2485
1da177e4
LT
2486static inline int skb_add_data(struct sk_buff *skb,
2487 char __user *from, int copy)
2488{
2489 const int off = skb->len;
2490
2491 if (skb->ip_summed == CHECKSUM_NONE) {
2492 int err = 0;
5084205f 2493 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2494 copy, 0, &err);
2495 if (!err) {
2496 skb->csum = csum_block_add(skb->csum, csum, off);
2497 return 0;
2498 }
2499 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2500 return 0;
2501
2502 __skb_trim(skb, off);
2503 return -EFAULT;
2504}
2505
38ba0a65
ED
2506static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2507 const struct page *page, int off)
1da177e4
LT
2508{
2509 if (i) {
9e903e08 2510 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2511
ea2ab693 2512 return page == skb_frag_page(frag) &&
9e903e08 2513 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2514 }
38ba0a65 2515 return false;
1da177e4
LT
2516}
2517
364c6bad
HX
2518static inline int __skb_linearize(struct sk_buff *skb)
2519{
2520 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2521}
2522
1da177e4
LT
2523/**
2524 * skb_linearize - convert paged skb to linear one
2525 * @skb: buffer to linarize
1da177e4
LT
2526 *
2527 * If there is no free memory -ENOMEM is returned, otherwise zero
2528 * is returned and the old skb data released.
2529 */
364c6bad
HX
2530static inline int skb_linearize(struct sk_buff *skb)
2531{
2532 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2533}
2534
cef401de
ED
2535/**
2536 * skb_has_shared_frag - can any frag be overwritten
2537 * @skb: buffer to test
2538 *
2539 * Return true if the skb has at least one frag that might be modified
2540 * by an external entity (as in vmsplice()/sendfile())
2541 */
2542static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2543{
c9af6db4
PS
2544 return skb_is_nonlinear(skb) &&
2545 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2546}
2547
364c6bad
HX
2548/**
2549 * skb_linearize_cow - make sure skb is linear and writable
2550 * @skb: buffer to process
2551 *
2552 * If there is no free memory -ENOMEM is returned, otherwise zero
2553 * is returned and the old skb data released.
2554 */
2555static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2556{
364c6bad
HX
2557 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2558 __skb_linearize(skb) : 0;
1da177e4
LT
2559}
2560
2561/**
2562 * skb_postpull_rcsum - update checksum for received skb after pull
2563 * @skb: buffer to update
2564 * @start: start of data before pull
2565 * @len: length of data pulled
2566 *
2567 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2568 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2569 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2570 */
2571
2572static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2573 const void *start, unsigned int len)
1da177e4 2574{
84fa7933 2575 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2576 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2577}
2578
cbb042f9
HX
2579unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2580
7ce5a27f
DM
2581/**
2582 * pskb_trim_rcsum - trim received skb and update checksum
2583 * @skb: buffer to trim
2584 * @len: new length
2585 *
2586 * This is exactly the same as pskb_trim except that it ensures the
2587 * checksum of received packets are still valid after the operation.
2588 */
2589
2590static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2591{
2592 if (likely(len >= skb->len))
2593 return 0;
2594 if (skb->ip_summed == CHECKSUM_COMPLETE)
2595 skb->ip_summed = CHECKSUM_NONE;
2596 return __pskb_trim(skb, len);
2597}
2598
1da177e4
LT
2599#define skb_queue_walk(queue, skb) \
2600 for (skb = (queue)->next; \
a1e4891f 2601 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2602 skb = skb->next)
2603
46f8914e
JC
2604#define skb_queue_walk_safe(queue, skb, tmp) \
2605 for (skb = (queue)->next, tmp = skb->next; \
2606 skb != (struct sk_buff *)(queue); \
2607 skb = tmp, tmp = skb->next)
2608
1164f52a 2609#define skb_queue_walk_from(queue, skb) \
a1e4891f 2610 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2611 skb = skb->next)
2612
2613#define skb_queue_walk_from_safe(queue, skb, tmp) \
2614 for (tmp = skb->next; \
2615 skb != (struct sk_buff *)(queue); \
2616 skb = tmp, tmp = skb->next)
2617
300ce174
SH
2618#define skb_queue_reverse_walk(queue, skb) \
2619 for (skb = (queue)->prev; \
a1e4891f 2620 skb != (struct sk_buff *)(queue); \
300ce174
SH
2621 skb = skb->prev)
2622
686a2955
DM
2623#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2624 for (skb = (queue)->prev, tmp = skb->prev; \
2625 skb != (struct sk_buff *)(queue); \
2626 skb = tmp, tmp = skb->prev)
2627
2628#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2629 for (tmp = skb->prev; \
2630 skb != (struct sk_buff *)(queue); \
2631 skb = tmp, tmp = skb->prev)
1da177e4 2632
21dc3301 2633static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2634{
2635 return skb_shinfo(skb)->frag_list != NULL;
2636}
2637
2638static inline void skb_frag_list_init(struct sk_buff *skb)
2639{
2640 skb_shinfo(skb)->frag_list = NULL;
2641}
2642
2643static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2644{
2645 frag->next = skb_shinfo(skb)->frag_list;
2646 skb_shinfo(skb)->frag_list = frag;
2647}
2648
2649#define skb_walk_frags(skb, iter) \
2650 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2651
7965bd4d
JP
2652struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2653 int *peeked, int *off, int *err);
2654struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2655 int *err);
2656unsigned int datagram_poll(struct file *file, struct socket *sock,
2657 struct poll_table_struct *wait);
c0371da6
AV
2658int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2659 struct iov_iter *to, int size);
51f3d02b
DM
2660static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2661 struct msghdr *msg, int size)
2662{
e5a4b0bb 2663 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 2664}
e5a4b0bb
AV
2665int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2666 struct msghdr *msg);
3a654f97
AV
2667int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2668 struct iov_iter *from, int len);
3a654f97 2669int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d
JP
2670void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2671void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2672int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2673int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2674int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2675__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2676 int len, __wsum csum);
2677int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2678 struct pipe_inode_info *pipe, unsigned int len,
2679 unsigned int flags);
2680void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2681unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2682int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2683 int len, int hlen);
7965bd4d
JP
2684void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2685int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2686void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2687unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2688struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2689struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 2690int skb_ensure_writable(struct sk_buff *skb, int write_len);
93515d53
JP
2691int skb_vlan_pop(struct sk_buff *skb);
2692int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
20380731 2693
6ce8e9ce
AV
2694static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2695{
c0371da6
AV
2696 /* XXX: stripping const */
2697 return memcpy_fromiovec(data, (struct iovec *)msg->msg_iter.iov, len);
6ce8e9ce
AV
2698}
2699
7eab8d9e
AV
2700static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2701{
e5a4b0bb 2702 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
2703}
2704
2817a336
DB
2705struct skb_checksum_ops {
2706 __wsum (*update)(const void *mem, int len, __wsum wsum);
2707 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2708};
2709
2710__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2711 __wsum csum, const struct skb_checksum_ops *ops);
2712__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2713 __wsum csum);
2714
690e36e7
DM
2715static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2716 int len, void *data, int hlen, void *buffer)
1da177e4 2717{
55820ee2 2718 if (hlen - offset >= len)
690e36e7 2719 return data + offset;
1da177e4 2720
690e36e7
DM
2721 if (!skb ||
2722 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2723 return NULL;
2724
2725 return buffer;
2726}
2727
690e36e7
DM
2728static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2729 int len, void *buffer)
2730{
2731 return __skb_header_pointer(skb, offset, len, skb->data,
2732 skb_headlen(skb), buffer);
2733}
2734
4262e5cc
DB
2735/**
2736 * skb_needs_linearize - check if we need to linearize a given skb
2737 * depending on the given device features.
2738 * @skb: socket buffer to check
2739 * @features: net device features
2740 *
2741 * Returns true if either:
2742 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2743 * 2. skb is fragmented and the device does not support SG.
2744 */
2745static inline bool skb_needs_linearize(struct sk_buff *skb,
2746 netdev_features_t features)
2747{
2748 return skb_is_nonlinear(skb) &&
2749 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2750 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2751}
2752
d626f62b
ACM
2753static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2754 void *to,
2755 const unsigned int len)
2756{
2757 memcpy(to, skb->data, len);
2758}
2759
2760static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2761 const int offset, void *to,
2762 const unsigned int len)
2763{
2764 memcpy(to, skb->data + offset, len);
2765}
2766
27d7ff46
ACM
2767static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2768 const void *from,
2769 const unsigned int len)
2770{
2771 memcpy(skb->data, from, len);
2772}
2773
2774static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2775 const int offset,
2776 const void *from,
2777 const unsigned int len)
2778{
2779 memcpy(skb->data + offset, from, len);
2780}
2781
7965bd4d 2782void skb_init(void);
1da177e4 2783
ac45f602
PO
2784static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2785{
2786 return skb->tstamp;
2787}
2788
a61bbcf2
PM
2789/**
2790 * skb_get_timestamp - get timestamp from a skb
2791 * @skb: skb to get stamp from
2792 * @stamp: pointer to struct timeval to store stamp in
2793 *
2794 * Timestamps are stored in the skb as offsets to a base timestamp.
2795 * This function converts the offset back to a struct timeval and stores
2796 * it in stamp.
2797 */
ac45f602
PO
2798static inline void skb_get_timestamp(const struct sk_buff *skb,
2799 struct timeval *stamp)
a61bbcf2 2800{
b7aa0bf7 2801 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2802}
2803
ac45f602
PO
2804static inline void skb_get_timestampns(const struct sk_buff *skb,
2805 struct timespec *stamp)
2806{
2807 *stamp = ktime_to_timespec(skb->tstamp);
2808}
2809
b7aa0bf7 2810static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2811{
b7aa0bf7 2812 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2813}
2814
164891aa
SH
2815static inline ktime_t net_timedelta(ktime_t t)
2816{
2817 return ktime_sub(ktime_get_real(), t);
2818}
2819
b9ce204f
IJ
2820static inline ktime_t net_invalid_timestamp(void)
2821{
2822 return ktime_set(0, 0);
2823}
a61bbcf2 2824
62bccb8c
AD
2825struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2826
c1f19b51
RC
2827#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2828
7965bd4d
JP
2829void skb_clone_tx_timestamp(struct sk_buff *skb);
2830bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2831
2832#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2833
2834static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2835{
2836}
2837
2838static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2839{
2840 return false;
2841}
2842
2843#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2844
2845/**
2846 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2847 *
da92b194
RC
2848 * PHY drivers may accept clones of transmitted packets for
2849 * timestamping via their phy_driver.txtstamp method. These drivers
2850 * must call this function to return the skb back to the stack, with
2851 * or without a timestamp.
2852 *
c1f19b51 2853 * @skb: clone of the the original outgoing packet
da92b194 2854 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2855 *
2856 */
2857void skb_complete_tx_timestamp(struct sk_buff *skb,
2858 struct skb_shared_hwtstamps *hwtstamps);
2859
e7fd2885
WB
2860void __skb_tstamp_tx(struct sk_buff *orig_skb,
2861 struct skb_shared_hwtstamps *hwtstamps,
2862 struct sock *sk, int tstype);
2863
ac45f602
PO
2864/**
2865 * skb_tstamp_tx - queue clone of skb with send time stamps
2866 * @orig_skb: the original outgoing packet
2867 * @hwtstamps: hardware time stamps, may be NULL if not available
2868 *
2869 * If the skb has a socket associated, then this function clones the
2870 * skb (thus sharing the actual data and optional structures), stores
2871 * the optional hardware time stamping information (if non NULL) or
2872 * generates a software time stamp (otherwise), then queues the clone
2873 * to the error queue of the socket. Errors are silently ignored.
2874 */
7965bd4d
JP
2875void skb_tstamp_tx(struct sk_buff *orig_skb,
2876 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2877
4507a715
RC
2878static inline void sw_tx_timestamp(struct sk_buff *skb)
2879{
2244d07b
OH
2880 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2881 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2882 skb_tstamp_tx(skb, NULL);
2883}
2884
2885/**
2886 * skb_tx_timestamp() - Driver hook for transmit timestamping
2887 *
2888 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2889 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2890 *
73409f3b
DM
2891 * Specifically, one should make absolutely sure that this function is
2892 * called before TX completion of this packet can trigger. Otherwise
2893 * the packet could potentially already be freed.
2894 *
4507a715
RC
2895 * @skb: A socket buffer.
2896 */
2897static inline void skb_tx_timestamp(struct sk_buff *skb)
2898{
c1f19b51 2899 skb_clone_tx_timestamp(skb);
4507a715
RC
2900 sw_tx_timestamp(skb);
2901}
2902
6e3e939f
JB
2903/**
2904 * skb_complete_wifi_ack - deliver skb with wifi status
2905 *
2906 * @skb: the original outgoing packet
2907 * @acked: ack status
2908 *
2909 */
2910void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2911
7965bd4d
JP
2912__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2913__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2914
60476372
HX
2915static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2916{
5d0c2b95 2917 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
60476372
HX
2918}
2919
fb286bb2
HX
2920/**
2921 * skb_checksum_complete - Calculate checksum of an entire packet
2922 * @skb: packet to process
2923 *
2924 * This function calculates the checksum over the entire packet plus
2925 * the value of skb->csum. The latter can be used to supply the
2926 * checksum of a pseudo header as used by TCP/UDP. It returns the
2927 * checksum.
2928 *
2929 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2930 * this function can be used to verify that checksum on received
2931 * packets. In that case the function should return zero if the
2932 * checksum is correct. In particular, this function will return zero
2933 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2934 * hardware has already verified the correctness of the checksum.
2935 */
4381ca3c 2936static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2937{
60476372
HX
2938 return skb_csum_unnecessary(skb) ?
2939 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2940}
2941
77cffe23
TH
2942static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2943{
2944 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2945 if (skb->csum_level == 0)
2946 skb->ip_summed = CHECKSUM_NONE;
2947 else
2948 skb->csum_level--;
2949 }
2950}
2951
2952static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2953{
2954 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2955 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2956 skb->csum_level++;
2957 } else if (skb->ip_summed == CHECKSUM_NONE) {
2958 skb->ip_summed = CHECKSUM_UNNECESSARY;
2959 skb->csum_level = 0;
2960 }
2961}
2962
5a212329
TH
2963static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2964{
2965 /* Mark current checksum as bad (typically called from GRO
2966 * path). In the case that ip_summed is CHECKSUM_NONE
2967 * this must be the first checksum encountered in the packet.
2968 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2969 * checksum after the last one validated. For UDP, a zero
2970 * checksum can not be marked as bad.
2971 */
2972
2973 if (skb->ip_summed == CHECKSUM_NONE ||
2974 skb->ip_summed == CHECKSUM_UNNECESSARY)
2975 skb->csum_bad = 1;
2976}
2977
76ba0aae
TH
2978/* Check if we need to perform checksum complete validation.
2979 *
2980 * Returns true if checksum complete is needed, false otherwise
2981 * (either checksum is unnecessary or zero checksum is allowed).
2982 */
2983static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2984 bool zero_okay,
2985 __sum16 check)
2986{
5d0c2b95
TH
2987 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2988 skb->csum_valid = 1;
77cffe23 2989 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
2990 return false;
2991 }
2992
2993 return true;
2994}
2995
2996/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2997 * in checksum_init.
2998 */
2999#define CHECKSUM_BREAK 76
3000
3001/* Validate (init) checksum based on checksum complete.
3002 *
3003 * Return values:
3004 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3005 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3006 * checksum is stored in skb->csum for use in __skb_checksum_complete
3007 * non-zero: value of invalid checksum
3008 *
3009 */
3010static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3011 bool complete,
3012 __wsum psum)
3013{
3014 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3015 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3016 skb->csum_valid = 1;
76ba0aae
TH
3017 return 0;
3018 }
5a212329
TH
3019 } else if (skb->csum_bad) {
3020 /* ip_summed == CHECKSUM_NONE in this case */
3021 return 1;
76ba0aae
TH
3022 }
3023
3024 skb->csum = psum;
3025
5d0c2b95
TH
3026 if (complete || skb->len <= CHECKSUM_BREAK) {
3027 __sum16 csum;
3028
3029 csum = __skb_checksum_complete(skb);
3030 skb->csum_valid = !csum;
3031 return csum;
3032 }
76ba0aae
TH
3033
3034 return 0;
3035}
3036
3037static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3038{
3039 return 0;
3040}
3041
3042/* Perform checksum validate (init). Note that this is a macro since we only
3043 * want to calculate the pseudo header which is an input function if necessary.
3044 * First we try to validate without any computation (checksum unnecessary) and
3045 * then calculate based on checksum complete calling the function to compute
3046 * pseudo header.
3047 *
3048 * Return values:
3049 * 0: checksum is validated or try to in skb_checksum_complete
3050 * non-zero: value of invalid checksum
3051 */
3052#define __skb_checksum_validate(skb, proto, complete, \
3053 zero_okay, check, compute_pseudo) \
3054({ \
3055 __sum16 __ret = 0; \
5d0c2b95 3056 skb->csum_valid = 0; \
76ba0aae
TH
3057 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3058 __ret = __skb_checksum_validate_complete(skb, \
3059 complete, compute_pseudo(skb, proto)); \
3060 __ret; \
3061})
3062
3063#define skb_checksum_init(skb, proto, compute_pseudo) \
3064 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3065
3066#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3067 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3068
3069#define skb_checksum_validate(skb, proto, compute_pseudo) \
3070 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3071
3072#define skb_checksum_validate_zero_check(skb, proto, check, \
3073 compute_pseudo) \
3074 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
3075
3076#define skb_checksum_simple_validate(skb) \
3077 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3078
d96535a1
TH
3079static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3080{
3081 return (skb->ip_summed == CHECKSUM_NONE &&
3082 skb->csum_valid && !skb->csum_bad);
3083}
3084
3085static inline void __skb_checksum_convert(struct sk_buff *skb,
3086 __sum16 check, __wsum pseudo)
3087{
3088 skb->csum = ~pseudo;
3089 skb->ip_summed = CHECKSUM_COMPLETE;
3090}
3091
3092#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3093do { \
3094 if (__skb_checksum_convert_check(skb)) \
3095 __skb_checksum_convert(skb, check, \
3096 compute_pseudo(skb, proto)); \
3097} while (0)
3098
5f79e0f9 3099#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3100void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3101static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3102{
3103 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3104 nf_conntrack_destroy(nfct);
1da177e4
LT
3105}
3106static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3107{
3108 if (nfct)
3109 atomic_inc(&nfct->use);
3110}
2fc72c7b 3111#endif
34666d46 3112#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3113static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3114{
3115 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3116 kfree(nf_bridge);
3117}
3118static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3119{
3120 if (nf_bridge)
3121 atomic_inc(&nf_bridge->use);
3122}
3123#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3124static inline void nf_reset(struct sk_buff *skb)
3125{
5f79e0f9 3126#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3127 nf_conntrack_put(skb->nfct);
3128 skb->nfct = NULL;
2fc72c7b 3129#endif
34666d46 3130#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3131 nf_bridge_put(skb->nf_bridge);
3132 skb->nf_bridge = NULL;
3133#endif
3134}
3135
124dff01
PM
3136static inline void nf_reset_trace(struct sk_buff *skb)
3137{
478b360a 3138#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3139 skb->nf_trace = 0;
3140#endif
a193a4ab
PM
3141}
3142
edda553c 3143/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3144static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3145 bool copy)
edda553c 3146{
5f79e0f9 3147#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3148 dst->nfct = src->nfct;
3149 nf_conntrack_get(src->nfct);
b1937227
ED
3150 if (copy)
3151 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3152#endif
34666d46 3153#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3154 dst->nf_bridge = src->nf_bridge;
3155 nf_bridge_get(src->nf_bridge);
3156#endif
478b360a 3157#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3158 if (copy)
3159 dst->nf_trace = src->nf_trace;
478b360a 3160#endif
edda553c
YK
3161}
3162
e7ac05f3
YK
3163static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3164{
e7ac05f3 3165#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3166 nf_conntrack_put(dst->nfct);
2fc72c7b 3167#endif
34666d46 3168#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3169 nf_bridge_put(dst->nf_bridge);
3170#endif
b1937227 3171 __nf_copy(dst, src, true);
e7ac05f3
YK
3172}
3173
984bc16c
JM
3174#ifdef CONFIG_NETWORK_SECMARK
3175static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3176{
3177 to->secmark = from->secmark;
3178}
3179
3180static inline void skb_init_secmark(struct sk_buff *skb)
3181{
3182 skb->secmark = 0;
3183}
3184#else
3185static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3186{ }
3187
3188static inline void skb_init_secmark(struct sk_buff *skb)
3189{ }
3190#endif
3191
574f7194
EB
3192static inline bool skb_irq_freeable(const struct sk_buff *skb)
3193{
3194 return !skb->destructor &&
3195#if IS_ENABLED(CONFIG_XFRM)
3196 !skb->sp &&
3197#endif
3198#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3199 !skb->nfct &&
3200#endif
3201 !skb->_skb_refdst &&
3202 !skb_has_frag_list(skb);
3203}
3204
f25f4e44
PWJ
3205static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3206{
f25f4e44 3207 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3208}
3209
9247744e 3210static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3211{
4e3ab47a 3212 return skb->queue_mapping;
4e3ab47a
PE
3213}
3214
f25f4e44
PWJ
3215static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3216{
f25f4e44 3217 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3218}
3219
d5a9e24a
DM
3220static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3221{
3222 skb->queue_mapping = rx_queue + 1;
3223}
3224
9247744e 3225static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3226{
3227 return skb->queue_mapping - 1;
3228}
3229
9247744e 3230static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3231{
a02cec21 3232 return skb->queue_mapping != 0;
d5a9e24a
DM
3233}
3234
0e001614 3235u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
7965bd4d 3236 unsigned int num_tx_queues);
9247744e 3237
def8b4fa
AD
3238static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3239{
0b3d8e08 3240#ifdef CONFIG_XFRM
def8b4fa 3241 return skb->sp;
def8b4fa 3242#else
def8b4fa 3243 return NULL;
def8b4fa 3244#endif
0b3d8e08 3245}
def8b4fa 3246
68c33163
PS
3247/* Keeps track of mac header offset relative to skb->head.
3248 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3249 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3250 * tunnel skb it points to outer mac header.
3251 * Keeps track of level of encapsulation of network headers.
3252 */
68c33163 3253struct skb_gso_cb {
3347c960
ED
3254 int mac_offset;
3255 int encap_level;
7e2b10c1 3256 __u16 csum_start;
68c33163
PS
3257};
3258#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3259
3260static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3261{
3262 return (skb_mac_header(inner_skb) - inner_skb->head) -
3263 SKB_GSO_CB(inner_skb)->mac_offset;
3264}
3265
1e2bd517
PS
3266static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3267{
3268 int new_headroom, headroom;
3269 int ret;
3270
3271 headroom = skb_headroom(skb);
3272 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3273 if (ret)
3274 return ret;
3275
3276 new_headroom = skb_headroom(skb);
3277 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3278 return 0;
3279}
3280
7e2b10c1
TH
3281/* Compute the checksum for a gso segment. First compute the checksum value
3282 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3283 * then add in skb->csum (checksum from csum_start to end of packet).
3284 * skb->csum and csum_start are then updated to reflect the checksum of the
3285 * resultant packet starting from the transport header-- the resultant checksum
3286 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3287 * header.
3288 */
3289static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3290{
3291 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3292 skb_transport_offset(skb);
3293 __u16 csum;
3294
3295 csum = csum_fold(csum_partial(skb_transport_header(skb),
3296 plen, skb->csum));
3297 skb->csum = res;
3298 SKB_GSO_CB(skb)->csum_start -= plen;
3299
3300 return csum;
3301}
3302
bdcc0924 3303static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3304{
3305 return skb_shinfo(skb)->gso_size;
3306}
3307
36a8f39e 3308/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3309static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3310{
3311 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3312}
3313
7965bd4d 3314void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3315
3316static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3317{
3318 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3319 * wanted then gso_type will be set. */
05bdd2f1
ED
3320 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3321
b78462eb
AD
3322 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3323 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3324 __skb_warn_lro_forwarding(skb);
3325 return true;
3326 }
3327 return false;
3328}
3329
35fc92a9
HX
3330static inline void skb_forward_csum(struct sk_buff *skb)
3331{
3332 /* Unfortunately we don't support this one. Any brave souls? */
3333 if (skb->ip_summed == CHECKSUM_COMPLETE)
3334 skb->ip_summed = CHECKSUM_NONE;
3335}
3336
bc8acf2c
ED
3337/**
3338 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3339 * @skb: skb to check
3340 *
3341 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3342 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3343 * use this helper, to document places where we make this assertion.
3344 */
05bdd2f1 3345static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3346{
3347#ifdef DEBUG
3348 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3349#endif
3350}
3351
f35d9d8a 3352bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3353
ed1f50c3
PD
3354int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3355
56193d1b
AD
3356u32 skb_get_poff(const struct sk_buff *skb);
3357u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3358 const struct flow_keys *keys, int hlen);
f77668dc 3359
3a7c1ee4
AD
3360/**
3361 * skb_head_is_locked - Determine if the skb->head is locked down
3362 * @skb: skb to check
3363 *
3364 * The head on skbs build around a head frag can be removed if they are
3365 * not cloned. This function returns true if the skb head is locked down
3366 * due to either being allocated via kmalloc, or by being a clone with
3367 * multiple references to the head.
3368 */
3369static inline bool skb_head_is_locked(const struct sk_buff *skb)
3370{
3371 return !skb->head_frag || skb_cloned(skb);
3372}
fe6cc55f
FW
3373
3374/**
3375 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3376 *
3377 * @skb: GSO skb
3378 *
3379 * skb_gso_network_seglen is used to determine the real size of the
3380 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3381 *
3382 * The MAC/L2 header is not accounted for.
3383 */
3384static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3385{
3386 unsigned int hdr_len = skb_transport_header(skb) -
3387 skb_network_header(skb);
3388 return hdr_len + skb_gso_transport_seglen(skb);
3389}
1da177e4
LT
3390#endif /* __KERNEL__ */
3391#endif /* _LINUX_SKBUFF_H */