]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - include/linux/skbuff.h
tcp: do not fake tcp headers in tcp_send_rcvq()
[mirror_ubuntu-zesty-kernel.git] / include / linux / skbuff.h
CommitLineData
1da177e4
LT
1/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
1da177e4 17#include <linux/kernel.h>
fe55f6d5 18#include <linux/kmemcheck.h>
1da177e4
LT
19#include <linux/compiler.h>
20#include <linux/time.h>
187f1882 21#include <linux/bug.h>
1da177e4
LT
22#include <linux/cache.h>
23
60063497 24#include <linux/atomic.h>
1da177e4
LT
25#include <asm/types.h>
26#include <linux/spinlock.h>
1da177e4 27#include <linux/net.h>
3fc7e8a6 28#include <linux/textsearch.h>
1da177e4 29#include <net/checksum.h>
a80958f4 30#include <linux/rcupdate.h>
97fc2f08 31#include <linux/dmaengine.h>
b7aa0bf7 32#include <linux/hrtimer.h>
131ea667 33#include <linux/dma-mapping.h>
c8f44aff 34#include <linux/netdev_features.h>
363ec392 35#include <linux/sched.h>
5203cd28 36#include <net/flow_keys.h>
1da177e4 37
78ea85f1
DB
38/* A. Checksumming of received packets by device.
39 *
40 * CHECKSUM_NONE:
41 *
42 * Device failed to checksum this packet e.g. due to lack of capabilities.
43 * The packet contains full (though not verified) checksum in packet but
44 * not in skb->csum. Thus, skb->csum is undefined in this case.
45 *
46 * CHECKSUM_UNNECESSARY:
47 *
48 * The hardware you're dealing with doesn't calculate the full checksum
49 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
50 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
51 * if their checksums are okay. skb->csum is still undefined in this case
52 * though. It is a bad option, but, unfortunately, nowadays most vendors do
53 * this. Apparently with the secret goal to sell you new devices, when you
54 * will add new protocol to your host, f.e. IPv6 8)
55 *
56 * CHECKSUM_UNNECESSARY is applicable to following protocols:
57 * TCP: IPv6 and IPv4.
58 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
59 * zero UDP checksum for either IPv4 or IPv6, the networking stack
60 * may perform further validation in this case.
61 * GRE: only if the checksum is present in the header.
62 * SCTP: indicates the CRC in SCTP header has been validated.
63 *
64 * skb->csum_level indicates the number of consecutive checksums found in
65 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
66 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
67 * and a device is able to verify the checksums for UDP (possibly zero),
68 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
69 * two. If the device were only able to verify the UDP checksum and not
70 * GRE, either because it doesn't support GRE checksum of because GRE
71 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
72 * not considered in this case).
78ea85f1
DB
73 *
74 * CHECKSUM_COMPLETE:
75 *
76 * This is the most generic way. The device supplied checksum of the _whole_
77 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
78 * hardware doesn't need to parse L3/L4 headers to implement this.
79 *
80 * Note: Even if device supports only some protocols, but is able to produce
81 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
82 *
83 * CHECKSUM_PARTIAL:
84 *
85 * This is identical to the case for output below. This may occur on a packet
86 * received directly from another Linux OS, e.g., a virtualized Linux kernel
87 * on the same host. The packet can be treated in the same way as
88 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
89 * checksum must be filled in by the OS or the hardware.
90 *
91 * B. Checksumming on output.
92 *
93 * CHECKSUM_NONE:
94 *
95 * The skb was already checksummed by the protocol, or a checksum is not
96 * required.
97 *
98 * CHECKSUM_PARTIAL:
99 *
100 * The device is required to checksum the packet as seen by hard_start_xmit()
101 * from skb->csum_start up to the end, and to record/write the checksum at
102 * offset skb->csum_start + skb->csum_offset.
103 *
104 * The device must show its capabilities in dev->features, set up at device
105 * setup time, e.g. netdev_features.h:
106 *
107 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
108 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
109 * IPv4. Sigh. Vendors like this way for an unknown reason.
110 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
111 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
112 * NETIF_F_... - Well, you get the picture.
113 *
114 * CHECKSUM_UNNECESSARY:
115 *
116 * Normally, the device will do per protocol specific checksumming. Protocol
117 * implementations that do not want the NIC to perform the checksum
118 * calculation should use this flag in their outgoing skbs.
119 *
120 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
121 * offload. Correspondingly, the FCoE protocol driver
122 * stack should use CHECKSUM_UNNECESSARY.
123 *
124 * Any questions? No questions, good. --ANK
125 */
126
60476372 127/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
128#define CHECKSUM_NONE 0
129#define CHECKSUM_UNNECESSARY 1
130#define CHECKSUM_COMPLETE 2
131#define CHECKSUM_PARTIAL 3
1da177e4 132
77cffe23
TH
133/* Maximum value in skb->csum_level */
134#define SKB_MAX_CSUM_LEVEL 3
135
0bec8c88 136#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 137#define SKB_WITH_OVERHEAD(X) \
deea84b0 138 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
139#define SKB_MAX_ORDER(X, ORDER) \
140 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
141#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
142#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
143
87fb4b7b
ED
144/* return minimum truesize of one skb containing X bytes of data */
145#define SKB_TRUESIZE(X) ((X) + \
146 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
147 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
148
1da177e4 149struct net_device;
716ea3a7 150struct scatterlist;
9c55e01c 151struct pipe_inode_info;
1da177e4 152
5f79e0f9 153#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
154struct nf_conntrack {
155 atomic_t use;
1da177e4 156};
5f79e0f9 157#endif
1da177e4
LT
158
159#ifdef CONFIG_BRIDGE_NETFILTER
160struct nf_bridge_info {
bf1ac5ca
ED
161 atomic_t use;
162 unsigned int mask;
163 struct net_device *physindev;
164 struct net_device *physoutdev;
165 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
166};
167#endif
168
1da177e4
LT
169struct sk_buff_head {
170 /* These two members must be first. */
171 struct sk_buff *next;
172 struct sk_buff *prev;
173
174 __u32 qlen;
175 spinlock_t lock;
176};
177
178struct sk_buff;
179
9d4dde52
IC
180/* To allow 64K frame to be packed as single skb without frag_list we
181 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
182 * buffers which do not start on a page boundary.
183 *
184 * Since GRO uses frags we allocate at least 16 regardless of page
185 * size.
a715dea3 186 */
9d4dde52 187#if (65536/PAGE_SIZE + 1) < 16
eec00954 188#define MAX_SKB_FRAGS 16UL
a715dea3 189#else
9d4dde52 190#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 191#endif
1da177e4
LT
192
193typedef struct skb_frag_struct skb_frag_t;
194
195struct skb_frag_struct {
a8605c60
IC
196 struct {
197 struct page *p;
198 } page;
cb4dfe56 199#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
200 __u32 page_offset;
201 __u32 size;
cb4dfe56
ED
202#else
203 __u16 page_offset;
204 __u16 size;
205#endif
1da177e4
LT
206};
207
9e903e08
ED
208static inline unsigned int skb_frag_size(const skb_frag_t *frag)
209{
210 return frag->size;
211}
212
213static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
214{
215 frag->size = size;
216}
217
218static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
219{
220 frag->size += delta;
221}
222
223static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
224{
225 frag->size -= delta;
226}
227
ac45f602
PO
228#define HAVE_HW_TIME_STAMP
229
230/**
d3a21be8 231 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
232 * @hwtstamp: hardware time stamp transformed into duration
233 * since arbitrary point in time
ac45f602
PO
234 *
235 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 236 * skb->tstamp.
ac45f602
PO
237 *
238 * hwtstamps can only be compared against other hwtstamps from
239 * the same device.
240 *
241 * This structure is attached to packets as part of the
242 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
243 */
244struct skb_shared_hwtstamps {
245 ktime_t hwtstamp;
ac45f602
PO
246};
247
2244d07b
OH
248/* Definitions for tx_flags in struct skb_shared_info */
249enum {
250 /* generate hardware time stamp */
251 SKBTX_HW_TSTAMP = 1 << 0,
252
e7fd2885 253 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
254 SKBTX_SW_TSTAMP = 1 << 1,
255
256 /* device driver is going to provide hardware time stamp */
257 SKBTX_IN_PROGRESS = 1 << 2,
258
a6686f2f 259 /* device driver supports TX zero-copy buffers */
62b1a8ab 260 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
261
262 /* generate wifi status information (where possible) */
62b1a8ab 263 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
264
265 /* This indicates at least one fragment might be overwritten
266 * (as in vmsplice(), sendfile() ...)
267 * If we need to compute a TX checksum, we'll need to copy
268 * all frags to avoid possible bad checksum
269 */
270 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
271
272 /* generate software time stamp when entering packet scheduling */
273 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
274
275 /* generate software timestamp on peer data acknowledgment */
276 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
277};
278
e1c8a607
WB
279#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
280 SKBTX_SCHED_TSTAMP | \
281 SKBTX_ACK_TSTAMP)
f24b9be5
WB
282#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
283
a6686f2f
SM
284/*
285 * The callback notifies userspace to release buffers when skb DMA is done in
286 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
287 * The zerocopy_success argument is true if zero copy transmit occurred,
288 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
289 * The ctx field is used to track device context.
290 * The desc field is used to track userspace buffer index.
a6686f2f
SM
291 */
292struct ubuf_info {
e19d6763 293 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 294 void *ctx;
a6686f2f 295 unsigned long desc;
ac45f602
PO
296};
297
1da177e4
LT
298/* This data is invariant across clones and lives at
299 * the end of the header data, ie. at skb->end.
300 */
301struct skb_shared_info {
9f42f126
IC
302 unsigned char nr_frags;
303 __u8 tx_flags;
7967168c
HX
304 unsigned short gso_size;
305 /* Warning: this field is not always filled in (UFO)! */
306 unsigned short gso_segs;
307 unsigned short gso_type;
1da177e4 308 struct sk_buff *frag_list;
ac45f602 309 struct skb_shared_hwtstamps hwtstamps;
09c2d251 310 u32 tskey;
9f42f126 311 __be32 ip6_frag_id;
ec7d2f2c
ED
312
313 /*
314 * Warning : all fields before dataref are cleared in __alloc_skb()
315 */
316 atomic_t dataref;
317
69e3c75f
JB
318 /* Intermediate layers must ensure that destructor_arg
319 * remains valid until skb destructor */
320 void * destructor_arg;
a6686f2f 321
fed66381
ED
322 /* must be last field, see pskb_expand_head() */
323 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
324};
325
326/* We divide dataref into two halves. The higher 16 bits hold references
327 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
328 * the entire skb->data. A clone of a headerless skb holds the length of
329 * the header in skb->hdr_len.
1da177e4
LT
330 *
331 * All users must obey the rule that the skb->data reference count must be
332 * greater than or equal to the payload reference count.
333 *
334 * Holding a reference to the payload part means that the user does not
335 * care about modifications to the header part of skb->data.
336 */
337#define SKB_DATAREF_SHIFT 16
338#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
339
d179cd12
DM
340
341enum {
342 SKB_FCLONE_UNAVAILABLE,
343 SKB_FCLONE_ORIG,
344 SKB_FCLONE_CLONE,
345};
346
7967168c
HX
347enum {
348 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 349 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
350
351 /* This indicates the skb is from an untrusted source. */
352 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
353
354 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
355 SKB_GSO_TCP_ECN = 1 << 3,
356
357 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
358
359 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
360
361 SKB_GSO_GRE = 1 << 6,
73136267 362
4b28252c 363 SKB_GSO_GRE_CSUM = 1 << 7,
0d89d203 364
4b28252c 365 SKB_GSO_IPIP = 1 << 8,
cb32f511 366
4b28252c 367 SKB_GSO_SIT = 1 << 9,
61c1db7f 368
4b28252c 369 SKB_GSO_UDP_TUNNEL = 1 << 10,
0f4f4ffa
TH
370
371 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c 372
4b28252c
TH
373 SKB_GSO_MPLS = 1 << 12,
374
7967168c
HX
375};
376
2e07fa9c
ACM
377#if BITS_PER_LONG > 32
378#define NET_SKBUFF_DATA_USES_OFFSET 1
379#endif
380
381#ifdef NET_SKBUFF_DATA_USES_OFFSET
382typedef unsigned int sk_buff_data_t;
383#else
384typedef unsigned char *sk_buff_data_t;
385#endif
386
363ec392
ED
387/**
388 * struct skb_mstamp - multi resolution time stamps
389 * @stamp_us: timestamp in us resolution
390 * @stamp_jiffies: timestamp in jiffies
391 */
392struct skb_mstamp {
393 union {
394 u64 v64;
395 struct {
396 u32 stamp_us;
397 u32 stamp_jiffies;
398 };
399 };
400};
401
402/**
403 * skb_mstamp_get - get current timestamp
404 * @cl: place to store timestamps
405 */
406static inline void skb_mstamp_get(struct skb_mstamp *cl)
407{
408 u64 val = local_clock();
409
410 do_div(val, NSEC_PER_USEC);
411 cl->stamp_us = (u32)val;
412 cl->stamp_jiffies = (u32)jiffies;
413}
414
415/**
416 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
417 * @t1: pointer to newest sample
418 * @t0: pointer to oldest sample
419 */
420static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
421 const struct skb_mstamp *t0)
422{
423 s32 delta_us = t1->stamp_us - t0->stamp_us;
424 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
425
426 /* If delta_us is negative, this might be because interval is too big,
427 * or local_clock() drift is too big : fallback using jiffies.
428 */
429 if (delta_us <= 0 ||
430 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
431
432 delta_us = jiffies_to_usecs(delta_jiffies);
433
434 return delta_us;
435}
436
437
1da177e4
LT
438/**
439 * struct sk_buff - socket buffer
440 * @next: Next buffer in list
441 * @prev: Previous buffer in list
363ec392 442 * @tstamp: Time we arrived/left
d84e0bd7 443 * @sk: Socket we are owned by
1da177e4 444 * @dev: Device we arrived on/are leaving by
d84e0bd7 445 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 446 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 447 * @sp: the security path, used for xfrm
1da177e4
LT
448 * @len: Length of actual data
449 * @data_len: Data length
450 * @mac_len: Length of link layer header
334a8132 451 * @hdr_len: writable header length of cloned skb
663ead3b
HX
452 * @csum: Checksum (must include start/offset pair)
453 * @csum_start: Offset from skb->head where checksumming should start
454 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 455 * @priority: Packet queueing priority
60ff7467 456 * @ignore_df: allow local fragmentation
1da177e4 457 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 458 * @ip_summed: Driver fed us an IP checksum
1da177e4 459 * @nohdr: Payload reference only, must not modify header
d84e0bd7 460 * @nfctinfo: Relationship of this skb to the connection
1da177e4 461 * @pkt_type: Packet class
c83c2486 462 * @fclone: skbuff clone status
c83c2486 463 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
464 * @peeked: this packet has been seen already, so stats have been
465 * done for it, don't do them again
ba9dda3a 466 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
467 * @protocol: Packet protocol from driver
468 * @destructor: Destruct function
469 * @nfct: Associated connection, if any
1da177e4 470 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 471 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
472 * @tc_index: Traffic control index
473 * @tc_verd: traffic control verdict
61b905da 474 * @hash: the packet hash
d84e0bd7 475 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 476 * @xmit_more: More SKBs are pending for this queue
553a5672 477 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 478 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 479 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 480 * ports.
a3b18ddb 481 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
482 * @wifi_acked_valid: wifi_acked was set
483 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 484 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
f4b8ea78
RD
485 * @dma_cookie: a cookie to one of several possible DMA operations
486 * done by skb DMA functions
06021292 487 * @napi_id: id of the NAPI struct this skb came from
984bc16c 488 * @secmark: security marking
d84e0bd7
DB
489 * @mark: Generic packet mark
490 * @dropcount: total number of sk_receive_queue overflows
86a9bad3 491 * @vlan_proto: vlan encapsulation protocol
6aa895b0 492 * @vlan_tci: vlan tag control information
0d89d203 493 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
494 * @inner_transport_header: Inner transport layer header (encapsulation)
495 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 496 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
497 * @transport_header: Transport layer header
498 * @network_header: Network layer header
499 * @mac_header: Link layer header
500 * @tail: Tail pointer
501 * @end: End pointer
502 * @head: Head of buffer
503 * @data: Data head pointer
504 * @truesize: Buffer size
505 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
506 */
507
508struct sk_buff {
509 /* These two members must be first. */
510 struct sk_buff *next;
511 struct sk_buff *prev;
512
363ec392
ED
513 union {
514 ktime_t tstamp;
515 struct skb_mstamp skb_mstamp;
516 };
da3f5cf1
FF
517
518 struct sock *sk;
1da177e4 519 struct net_device *dev;
1da177e4 520
1da177e4
LT
521 /*
522 * This is the control buffer. It is free to use for every
523 * layer. Please put your private variables there. If you
524 * want to keep them across layers you have to do a skb_clone()
525 * first. This is owned by whoever has the skb queued ATM.
526 */
da3f5cf1 527 char cb[48] __aligned(8);
1da177e4 528
7fee226a 529 unsigned long _skb_refdst;
da3f5cf1
FF
530#ifdef CONFIG_XFRM
531 struct sec_path *sp;
532#endif
1da177e4 533 unsigned int len,
334a8132
PM
534 data_len;
535 __u16 mac_len,
536 hdr_len;
ff1dcadb
AV
537 union {
538 __wsum csum;
663ead3b
HX
539 struct {
540 __u16 csum_start;
541 __u16 csum_offset;
542 };
ff1dcadb 543 };
1da177e4 544 __u32 priority;
fe55f6d5 545 kmemcheck_bitfield_begin(flags1);
60ff7467 546 __u8 ignore_df:1,
1cbb3380
TG
547 cloned:1,
548 ip_summed:2,
6869c4d8
HW
549 nohdr:1,
550 nfctinfo:3;
233577a2
HFS
551
552/* if you move pkt_type around you also must adapt those constants */
553#ifdef __BIG_ENDIAN_BITFIELD
554#define PKT_TYPE_MAX (7 << 5)
555#else
556#define PKT_TYPE_MAX 7
557#endif
558#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
559
560 __u8 __pkt_type_offset[0];
d179cd12 561 __u8 pkt_type:3,
b84f4cc9 562 fclone:2,
ba9dda3a 563 ipvs_property:1,
a59322be 564 peeked:1,
ba9dda3a 565 nf_trace:1;
fe55f6d5 566 kmemcheck_bitfield_end(flags1);
4ab408de 567 __be16 protocol;
1da177e4
LT
568
569 void (*destructor)(struct sk_buff *skb);
9fb9cbb1 570#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 571 struct nf_conntrack *nfct;
2fc72c7b 572#endif
1da177e4
LT
573#ifdef CONFIG_BRIDGE_NETFILTER
574 struct nf_bridge_info *nf_bridge;
575#endif
f25f4e44 576
8964be4a 577 int skb_iif;
4031ae6e 578
61b905da 579 __u32 hash;
4031ae6e 580
86a9bad3 581 __be16 vlan_proto;
4031ae6e
AD
582 __u16 vlan_tci;
583
1da177e4 584#ifdef CONFIG_NET_SCHED
b6b99eb5 585 __u16 tc_index; /* traffic control index */
1da177e4 586#ifdef CONFIG_NET_CLS_ACT
b6b99eb5 587 __u16 tc_verd; /* traffic control verdict */
1da177e4 588#endif
1da177e4 589#endif
fe55f6d5 590
0a14842f 591 __u16 queue_mapping;
fe55f6d5 592 kmemcheck_bitfield_begin(flags2);
0b725a2c 593 __u8 xmit_more:1;
de357cc0 594#ifdef CONFIG_IPV6_NDISC_NODETYPE
8a4eb573 595 __u8 ndisc_nodetype:2;
d0f09804 596#endif
c93bdd0e 597 __u8 pfmemalloc:1;
3853b584 598 __u8 ooo_okay:1;
61b905da 599 __u8 l4_hash:1;
a3b18ddb 600 __u8 sw_hash:1;
6e3e939f
JB
601 __u8 wifi_acked_valid:1;
602 __u8 wifi_acked:1;
3bdc0eba 603 __u8 no_fcs:1;
d3836f21 604 __u8 head_frag:1;
77cffe23 605 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 606 __u8 encapsulation:1;
7e2b10c1 607 __u8 encap_hdr_csum:1;
5d0c2b95 608 __u8 csum_valid:1;
7e3cead5 609 __u8 csum_complete_sw:1;
5c21403d 610 /* 1/3 bit hole (depending on ndisc_nodetype presence) */
fe55f6d5
VN
611 kmemcheck_bitfield_end(flags2);
612
e0d1095a 613#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
06021292
ET
614 union {
615 unsigned int napi_id;
616 dma_cookie_t dma_cookie;
617 };
97fc2f08 618#endif
984bc16c
JM
619#ifdef CONFIG_NETWORK_SECMARK
620 __u32 secmark;
621#endif
3b885787
NH
622 union {
623 __u32 mark;
624 __u32 dropcount;
16fad69c 625 __u32 reserved_tailroom;
3b885787 626 };
1da177e4 627
de20fe8e 628 kmemcheck_bitfield_begin(flags3);
77cffe23 629 __u8 csum_level:2;
5a212329
TH
630 __u8 csum_bad:1;
631 /* 13 bit hole */
de20fe8e
TH
632 kmemcheck_bitfield_end(flags3);
633
0d89d203 634 __be16 inner_protocol;
1a37e412
SH
635 __u16 inner_transport_header;
636 __u16 inner_network_header;
637 __u16 inner_mac_header;
638 __u16 transport_header;
639 __u16 network_header;
640 __u16 mac_header;
1da177e4 641 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 642 sk_buff_data_t tail;
4305b541 643 sk_buff_data_t end;
1da177e4 644 unsigned char *head,
4305b541 645 *data;
27a884dc
ACM
646 unsigned int truesize;
647 atomic_t users;
1da177e4
LT
648};
649
650#ifdef __KERNEL__
651/*
652 * Handling routines are only of interest to the kernel
653 */
654#include <linux/slab.h>
655
1da177e4 656
c93bdd0e
MG
657#define SKB_ALLOC_FCLONE 0x01
658#define SKB_ALLOC_RX 0x02
659
660/* Returns true if the skb was allocated from PFMEMALLOC reserves */
661static inline bool skb_pfmemalloc(const struct sk_buff *skb)
662{
663 return unlikely(skb->pfmemalloc);
664}
665
7fee226a
ED
666/*
667 * skb might have a dst pointer attached, refcounted or not.
668 * _skb_refdst low order bit is set if refcount was _not_ taken
669 */
670#define SKB_DST_NOREF 1UL
671#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
672
673/**
674 * skb_dst - returns skb dst_entry
675 * @skb: buffer
676 *
677 * Returns skb dst_entry, regardless of reference taken or not.
678 */
adf30907
ED
679static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
680{
7fee226a
ED
681 /* If refdst was not refcounted, check we still are in a
682 * rcu_read_lock section
683 */
684 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
685 !rcu_read_lock_held() &&
686 !rcu_read_lock_bh_held());
687 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
688}
689
7fee226a
ED
690/**
691 * skb_dst_set - sets skb dst
692 * @skb: buffer
693 * @dst: dst entry
694 *
695 * Sets skb dst, assuming a reference was taken on dst and should
696 * be released by skb_dst_drop()
697 */
adf30907
ED
698static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
699{
7fee226a
ED
700 skb->_skb_refdst = (unsigned long)dst;
701}
702
7965bd4d
JP
703void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
704 bool force);
932bc4d7
JA
705
706/**
707 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
708 * @skb: buffer
709 * @dst: dst entry
710 *
711 * Sets skb dst, assuming a reference was not taken on dst.
712 * If dst entry is cached, we do not take reference and dst_release
713 * will be avoided by refdst_drop. If dst entry is not cached, we take
714 * reference, so that last dst_release can destroy the dst immediately.
715 */
716static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
717{
718 __skb_dst_set_noref(skb, dst, false);
719}
720
721/**
722 * skb_dst_set_noref_force - sets skb dst, without taking reference
723 * @skb: buffer
724 * @dst: dst entry
725 *
726 * Sets skb dst, assuming a reference was not taken on dst.
727 * No reference is taken and no dst_release will be called. While for
728 * cached dsts deferred reclaim is a basic feature, for entries that are
729 * not cached it is caller's job to guarantee that last dst_release for
730 * provided dst happens when nobody uses it, eg. after a RCU grace period.
731 */
732static inline void skb_dst_set_noref_force(struct sk_buff *skb,
733 struct dst_entry *dst)
734{
735 __skb_dst_set_noref(skb, dst, true);
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
772static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 773 gfp_t priority)
d179cd12 774{
c93bdd0e 775 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
776}
777
7965bd4d 778struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
779static inline struct sk_buff *alloc_skb_head(gfp_t priority)
780{
781 return __alloc_skb_head(priority, -1);
782}
783
7965bd4d
JP
784struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
785int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
786struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
787struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
788struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
789 gfp_t gfp_mask, bool fclone);
790static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
791 gfp_t gfp_mask)
792{
793 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
794}
7965bd4d
JP
795
796int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
797struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
798 unsigned int headroom);
799struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
800 int newtailroom, gfp_t priority);
25a91d8d
FD
801int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
802 int offset, int len);
7965bd4d
JP
803int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
804 int len);
805int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
806int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 807#define dev_kfree_skb(a) consume_skb(a)
1da177e4 808
7965bd4d
JP
809int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
810 int getfrag(void *from, char *to, int offset,
811 int len, int odd, struct sk_buff *skb),
812 void *from, int length);
e89e9cf5 813
d94d9fee 814struct skb_seq_state {
677e90ed
TG
815 __u32 lower_offset;
816 __u32 upper_offset;
817 __u32 frag_idx;
818 __u32 stepped_offset;
819 struct sk_buff *root_skb;
820 struct sk_buff *cur_skb;
821 __u8 *frag_data;
822};
823
7965bd4d
JP
824void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
825 unsigned int to, struct skb_seq_state *st);
826unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
827 struct skb_seq_state *st);
828void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 829
7965bd4d
JP
830unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
831 unsigned int to, struct ts_config *config,
832 struct ts_state *state);
3fc7e8a6 833
09323cc4
TH
834/*
835 * Packet hash types specify the type of hash in skb_set_hash.
836 *
837 * Hash types refer to the protocol layer addresses which are used to
838 * construct a packet's hash. The hashes are used to differentiate or identify
839 * flows of the protocol layer for the hash type. Hash types are either
840 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
841 *
842 * Properties of hashes:
843 *
844 * 1) Two packets in different flows have different hash values
845 * 2) Two packets in the same flow should have the same hash value
846 *
847 * A hash at a higher layer is considered to be more specific. A driver should
848 * set the most specific hash possible.
849 *
850 * A driver cannot indicate a more specific hash than the layer at which a hash
851 * was computed. For instance an L3 hash cannot be set as an L4 hash.
852 *
853 * A driver may indicate a hash level which is less specific than the
854 * actual layer the hash was computed on. For instance, a hash computed
855 * at L4 may be considered an L3 hash. This should only be done if the
856 * driver can't unambiguously determine that the HW computed the hash at
857 * the higher layer. Note that the "should" in the second property above
858 * permits this.
859 */
860enum pkt_hash_types {
861 PKT_HASH_TYPE_NONE, /* Undefined type */
862 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
863 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
864 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
865};
866
867static inline void
868skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
869{
61b905da 870 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 871 skb->sw_hash = 0;
61b905da 872 skb->hash = hash;
09323cc4
TH
873}
874
3958afa1
TH
875void __skb_get_hash(struct sk_buff *skb);
876static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 877{
a3b18ddb 878 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 879 __skb_get_hash(skb);
bfb564e7 880
61b905da 881 return skb->hash;
bfb564e7
KK
882}
883
57bdf7f4
TH
884static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
885{
61b905da 886 return skb->hash;
57bdf7f4
TH
887}
888
7539fadc
TH
889static inline void skb_clear_hash(struct sk_buff *skb)
890{
61b905da 891 skb->hash = 0;
a3b18ddb 892 skb->sw_hash = 0;
61b905da 893 skb->l4_hash = 0;
7539fadc
TH
894}
895
896static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
897{
61b905da 898 if (!skb->l4_hash)
7539fadc
TH
899 skb_clear_hash(skb);
900}
901
3df7a74e
TH
902static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
903{
61b905da 904 to->hash = from->hash;
a3b18ddb 905 to->sw_hash = from->sw_hash;
61b905da 906 to->l4_hash = from->l4_hash;
3df7a74e
TH
907};
908
4305b541
ACM
909#ifdef NET_SKBUFF_DATA_USES_OFFSET
910static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
911{
912 return skb->head + skb->end;
913}
ec47ea82
AD
914
915static inline unsigned int skb_end_offset(const struct sk_buff *skb)
916{
917 return skb->end;
918}
4305b541
ACM
919#else
920static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
921{
922 return skb->end;
923}
ec47ea82
AD
924
925static inline unsigned int skb_end_offset(const struct sk_buff *skb)
926{
927 return skb->end - skb->head;
928}
4305b541
ACM
929#endif
930
1da177e4 931/* Internal */
4305b541 932#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 933
ac45f602
PO
934static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
935{
936 return &skb_shinfo(skb)->hwtstamps;
937}
938
1da177e4
LT
939/**
940 * skb_queue_empty - check if a queue is empty
941 * @list: queue head
942 *
943 * Returns true if the queue is empty, false otherwise.
944 */
945static inline int skb_queue_empty(const struct sk_buff_head *list)
946{
fd44b93c 947 return list->next == (const struct sk_buff *) list;
1da177e4
LT
948}
949
fc7ebb21
DM
950/**
951 * skb_queue_is_last - check if skb is the last entry in the queue
952 * @list: queue head
953 * @skb: buffer
954 *
955 * Returns true if @skb is the last buffer on the list.
956 */
957static inline bool skb_queue_is_last(const struct sk_buff_head *list,
958 const struct sk_buff *skb)
959{
fd44b93c 960 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
961}
962
832d11c5
IJ
963/**
964 * skb_queue_is_first - check if skb is the first entry in the queue
965 * @list: queue head
966 * @skb: buffer
967 *
968 * Returns true if @skb is the first buffer on the list.
969 */
970static inline bool skb_queue_is_first(const struct sk_buff_head *list,
971 const struct sk_buff *skb)
972{
fd44b93c 973 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
974}
975
249c8b42
DM
976/**
977 * skb_queue_next - return the next packet in the queue
978 * @list: queue head
979 * @skb: current buffer
980 *
981 * Return the next packet in @list after @skb. It is only valid to
982 * call this if skb_queue_is_last() evaluates to false.
983 */
984static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
985 const struct sk_buff *skb)
986{
987 /* This BUG_ON may seem severe, but if we just return then we
988 * are going to dereference garbage.
989 */
990 BUG_ON(skb_queue_is_last(list, skb));
991 return skb->next;
992}
993
832d11c5
IJ
994/**
995 * skb_queue_prev - return the prev packet in the queue
996 * @list: queue head
997 * @skb: current buffer
998 *
999 * Return the prev packet in @list before @skb. It is only valid to
1000 * call this if skb_queue_is_first() evaluates to false.
1001 */
1002static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1003 const struct sk_buff *skb)
1004{
1005 /* This BUG_ON may seem severe, but if we just return then we
1006 * are going to dereference garbage.
1007 */
1008 BUG_ON(skb_queue_is_first(list, skb));
1009 return skb->prev;
1010}
1011
1da177e4
LT
1012/**
1013 * skb_get - reference buffer
1014 * @skb: buffer to reference
1015 *
1016 * Makes another reference to a socket buffer and returns a pointer
1017 * to the buffer.
1018 */
1019static inline struct sk_buff *skb_get(struct sk_buff *skb)
1020{
1021 atomic_inc(&skb->users);
1022 return skb;
1023}
1024
1025/*
1026 * If users == 1, we are the only owner and are can avoid redundant
1027 * atomic change.
1028 */
1029
1da177e4
LT
1030/**
1031 * skb_cloned - is the buffer a clone
1032 * @skb: buffer to check
1033 *
1034 * Returns true if the buffer was generated with skb_clone() and is
1035 * one of multiple shared copies of the buffer. Cloned buffers are
1036 * shared data so must not be written to under normal circumstances.
1037 */
1038static inline int skb_cloned(const struct sk_buff *skb)
1039{
1040 return skb->cloned &&
1041 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1042}
1043
14bbd6a5
PS
1044static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1045{
1046 might_sleep_if(pri & __GFP_WAIT);
1047
1048 if (skb_cloned(skb))
1049 return pskb_expand_head(skb, 0, 0, pri);
1050
1051 return 0;
1052}
1053
1da177e4
LT
1054/**
1055 * skb_header_cloned - is the header a clone
1056 * @skb: buffer to check
1057 *
1058 * Returns true if modifying the header part of the buffer requires
1059 * the data to be copied.
1060 */
1061static inline int skb_header_cloned(const struct sk_buff *skb)
1062{
1063 int dataref;
1064
1065 if (!skb->cloned)
1066 return 0;
1067
1068 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1069 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1070 return dataref != 1;
1071}
1072
1073/**
1074 * skb_header_release - release reference to header
1075 * @skb: buffer to operate on
1076 *
1077 * Drop a reference to the header part of the buffer. This is done
1078 * by acquiring a payload reference. You must not read from the header
1079 * part of skb->data after this.
1080 */
1081static inline void skb_header_release(struct sk_buff *skb)
1082{
1083 BUG_ON(skb->nohdr);
1084 skb->nohdr = 1;
1085 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1086}
1087
1088/**
1089 * skb_shared - is the buffer shared
1090 * @skb: buffer to check
1091 *
1092 * Returns true if more than one person has a reference to this
1093 * buffer.
1094 */
1095static inline int skb_shared(const struct sk_buff *skb)
1096{
1097 return atomic_read(&skb->users) != 1;
1098}
1099
1100/**
1101 * skb_share_check - check if buffer is shared and if so clone it
1102 * @skb: buffer to check
1103 * @pri: priority for memory allocation
1104 *
1105 * If the buffer is shared the buffer is cloned and the old copy
1106 * drops a reference. A new clone with a single reference is returned.
1107 * If the buffer is not shared the original buffer is returned. When
1108 * being called from interrupt status or with spinlocks held pri must
1109 * be GFP_ATOMIC.
1110 *
1111 * NULL is returned on a memory allocation failure.
1112 */
47061bc4 1113static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1114{
1115 might_sleep_if(pri & __GFP_WAIT);
1116 if (skb_shared(skb)) {
1117 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1118
1119 if (likely(nskb))
1120 consume_skb(skb);
1121 else
1122 kfree_skb(skb);
1da177e4
LT
1123 skb = nskb;
1124 }
1125 return skb;
1126}
1127
1128/*
1129 * Copy shared buffers into a new sk_buff. We effectively do COW on
1130 * packets to handle cases where we have a local reader and forward
1131 * and a couple of other messy ones. The normal one is tcpdumping
1132 * a packet thats being forwarded.
1133 */
1134
1135/**
1136 * skb_unshare - make a copy of a shared buffer
1137 * @skb: buffer to check
1138 * @pri: priority for memory allocation
1139 *
1140 * If the socket buffer is a clone then this function creates a new
1141 * copy of the data, drops a reference count on the old copy and returns
1142 * the new copy with the reference count at 1. If the buffer is not a clone
1143 * the original buffer is returned. When called with a spinlock held or
1144 * from interrupt state @pri must be %GFP_ATOMIC
1145 *
1146 * %NULL is returned on a memory allocation failure.
1147 */
e2bf521d 1148static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1149 gfp_t pri)
1da177e4
LT
1150{
1151 might_sleep_if(pri & __GFP_WAIT);
1152 if (skb_cloned(skb)) {
1153 struct sk_buff *nskb = skb_copy(skb, pri);
1154 kfree_skb(skb); /* Free our shared copy */
1155 skb = nskb;
1156 }
1157 return skb;
1158}
1159
1160/**
1a5778aa 1161 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1162 * @list_: list to peek at
1163 *
1164 * Peek an &sk_buff. Unlike most other operations you _MUST_
1165 * be careful with this one. A peek leaves the buffer on the
1166 * list and someone else may run off with it. You must hold
1167 * the appropriate locks or have a private queue to do this.
1168 *
1169 * Returns %NULL for an empty list or a pointer to the head element.
1170 * The reference count is not incremented and the reference is therefore
1171 * volatile. Use with caution.
1172 */
05bdd2f1 1173static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1174{
18d07000
ED
1175 struct sk_buff *skb = list_->next;
1176
1177 if (skb == (struct sk_buff *)list_)
1178 skb = NULL;
1179 return skb;
1da177e4
LT
1180}
1181
da5ef6e5
PE
1182/**
1183 * skb_peek_next - peek skb following the given one from a queue
1184 * @skb: skb to start from
1185 * @list_: list to peek at
1186 *
1187 * Returns %NULL when the end of the list is met or a pointer to the
1188 * next element. The reference count is not incremented and the
1189 * reference is therefore volatile. Use with caution.
1190 */
1191static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1192 const struct sk_buff_head *list_)
1193{
1194 struct sk_buff *next = skb->next;
18d07000 1195
da5ef6e5
PE
1196 if (next == (struct sk_buff *)list_)
1197 next = NULL;
1198 return next;
1199}
1200
1da177e4 1201/**
1a5778aa 1202 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1203 * @list_: list to peek at
1204 *
1205 * Peek an &sk_buff. Unlike most other operations you _MUST_
1206 * be careful with this one. A peek leaves the buffer on the
1207 * list and someone else may run off with it. You must hold
1208 * the appropriate locks or have a private queue to do this.
1209 *
1210 * Returns %NULL for an empty list or a pointer to the tail element.
1211 * The reference count is not incremented and the reference is therefore
1212 * volatile. Use with caution.
1213 */
05bdd2f1 1214static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1215{
18d07000
ED
1216 struct sk_buff *skb = list_->prev;
1217
1218 if (skb == (struct sk_buff *)list_)
1219 skb = NULL;
1220 return skb;
1221
1da177e4
LT
1222}
1223
1224/**
1225 * skb_queue_len - get queue length
1226 * @list_: list to measure
1227 *
1228 * Return the length of an &sk_buff queue.
1229 */
1230static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1231{
1232 return list_->qlen;
1233}
1234
67fed459
DM
1235/**
1236 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1237 * @list: queue to initialize
1238 *
1239 * This initializes only the list and queue length aspects of
1240 * an sk_buff_head object. This allows to initialize the list
1241 * aspects of an sk_buff_head without reinitializing things like
1242 * the spinlock. It can also be used for on-stack sk_buff_head
1243 * objects where the spinlock is known to not be used.
1244 */
1245static inline void __skb_queue_head_init(struct sk_buff_head *list)
1246{
1247 list->prev = list->next = (struct sk_buff *)list;
1248 list->qlen = 0;
1249}
1250
76f10ad0
AV
1251/*
1252 * This function creates a split out lock class for each invocation;
1253 * this is needed for now since a whole lot of users of the skb-queue
1254 * infrastructure in drivers have different locking usage (in hardirq)
1255 * than the networking core (in softirq only). In the long run either the
1256 * network layer or drivers should need annotation to consolidate the
1257 * main types of usage into 3 classes.
1258 */
1da177e4
LT
1259static inline void skb_queue_head_init(struct sk_buff_head *list)
1260{
1261 spin_lock_init(&list->lock);
67fed459 1262 __skb_queue_head_init(list);
1da177e4
LT
1263}
1264
c2ecba71
PE
1265static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1266 struct lock_class_key *class)
1267{
1268 skb_queue_head_init(list);
1269 lockdep_set_class(&list->lock, class);
1270}
1271
1da177e4 1272/*
bf299275 1273 * Insert an sk_buff on a list.
1da177e4
LT
1274 *
1275 * The "__skb_xxxx()" functions are the non-atomic ones that
1276 * can only be called with interrupts disabled.
1277 */
7965bd4d
JP
1278void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1279 struct sk_buff_head *list);
bf299275
GR
1280static inline void __skb_insert(struct sk_buff *newsk,
1281 struct sk_buff *prev, struct sk_buff *next,
1282 struct sk_buff_head *list)
1283{
1284 newsk->next = next;
1285 newsk->prev = prev;
1286 next->prev = prev->next = newsk;
1287 list->qlen++;
1288}
1da177e4 1289
67fed459
DM
1290static inline void __skb_queue_splice(const struct sk_buff_head *list,
1291 struct sk_buff *prev,
1292 struct sk_buff *next)
1293{
1294 struct sk_buff *first = list->next;
1295 struct sk_buff *last = list->prev;
1296
1297 first->prev = prev;
1298 prev->next = first;
1299
1300 last->next = next;
1301 next->prev = last;
1302}
1303
1304/**
1305 * skb_queue_splice - join two skb lists, this is designed for stacks
1306 * @list: the new list to add
1307 * @head: the place to add it in the first list
1308 */
1309static inline void skb_queue_splice(const struct sk_buff_head *list,
1310 struct sk_buff_head *head)
1311{
1312 if (!skb_queue_empty(list)) {
1313 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1314 head->qlen += list->qlen;
67fed459
DM
1315 }
1316}
1317
1318/**
d9619496 1319 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1320 * @list: the new list to add
1321 * @head: the place to add it in the first list
1322 *
1323 * The list at @list is reinitialised
1324 */
1325static inline void skb_queue_splice_init(struct sk_buff_head *list,
1326 struct sk_buff_head *head)
1327{
1328 if (!skb_queue_empty(list)) {
1329 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1330 head->qlen += list->qlen;
67fed459
DM
1331 __skb_queue_head_init(list);
1332 }
1333}
1334
1335/**
1336 * skb_queue_splice_tail - join two skb lists, each list being a queue
1337 * @list: the new list to add
1338 * @head: the place to add it in the first list
1339 */
1340static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1341 struct sk_buff_head *head)
1342{
1343 if (!skb_queue_empty(list)) {
1344 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1345 head->qlen += list->qlen;
67fed459
DM
1346 }
1347}
1348
1349/**
d9619496 1350 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1351 * @list: the new list to add
1352 * @head: the place to add it in the first list
1353 *
1354 * Each of the lists is a queue.
1355 * The list at @list is reinitialised
1356 */
1357static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1358 struct sk_buff_head *head)
1359{
1360 if (!skb_queue_empty(list)) {
1361 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1362 head->qlen += list->qlen;
67fed459
DM
1363 __skb_queue_head_init(list);
1364 }
1365}
1366
1da177e4 1367/**
300ce174 1368 * __skb_queue_after - queue a buffer at the list head
1da177e4 1369 * @list: list to use
300ce174 1370 * @prev: place after this buffer
1da177e4
LT
1371 * @newsk: buffer to queue
1372 *
300ce174 1373 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1374 * and you must therefore hold required locks before calling it.
1375 *
1376 * A buffer cannot be placed on two lists at the same time.
1377 */
300ce174
SH
1378static inline void __skb_queue_after(struct sk_buff_head *list,
1379 struct sk_buff *prev,
1380 struct sk_buff *newsk)
1da177e4 1381{
bf299275 1382 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1383}
1384
7965bd4d
JP
1385void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1386 struct sk_buff_head *list);
7de6c033 1387
f5572855
GR
1388static inline void __skb_queue_before(struct sk_buff_head *list,
1389 struct sk_buff *next,
1390 struct sk_buff *newsk)
1391{
1392 __skb_insert(newsk, next->prev, next, list);
1393}
1394
300ce174
SH
1395/**
1396 * __skb_queue_head - queue a buffer at the list head
1397 * @list: list to use
1398 * @newsk: buffer to queue
1399 *
1400 * Queue a buffer at the start of a list. This function takes no locks
1401 * and you must therefore hold required locks before calling it.
1402 *
1403 * A buffer cannot be placed on two lists at the same time.
1404 */
7965bd4d 1405void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1406static inline void __skb_queue_head(struct sk_buff_head *list,
1407 struct sk_buff *newsk)
1408{
1409 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1410}
1411
1da177e4
LT
1412/**
1413 * __skb_queue_tail - queue a buffer at the list tail
1414 * @list: list to use
1415 * @newsk: buffer to queue
1416 *
1417 * Queue a buffer at the end of a list. This function takes no locks
1418 * and you must therefore hold required locks before calling it.
1419 *
1420 * A buffer cannot be placed on two lists at the same time.
1421 */
7965bd4d 1422void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1423static inline void __skb_queue_tail(struct sk_buff_head *list,
1424 struct sk_buff *newsk)
1425{
f5572855 1426 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1427}
1428
1da177e4
LT
1429/*
1430 * remove sk_buff from list. _Must_ be called atomically, and with
1431 * the list known..
1432 */
7965bd4d 1433void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1434static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1435{
1436 struct sk_buff *next, *prev;
1437
1438 list->qlen--;
1439 next = skb->next;
1440 prev = skb->prev;
1441 skb->next = skb->prev = NULL;
1da177e4
LT
1442 next->prev = prev;
1443 prev->next = next;
1444}
1445
f525c06d
GR
1446/**
1447 * __skb_dequeue - remove from the head of the queue
1448 * @list: list to dequeue from
1449 *
1450 * Remove the head of the list. This function does not take any locks
1451 * so must be used with appropriate locks held only. The head item is
1452 * returned or %NULL if the list is empty.
1453 */
7965bd4d 1454struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1455static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1456{
1457 struct sk_buff *skb = skb_peek(list);
1458 if (skb)
1459 __skb_unlink(skb, list);
1460 return skb;
1461}
1da177e4
LT
1462
1463/**
1464 * __skb_dequeue_tail - remove from the tail of the queue
1465 * @list: list to dequeue from
1466 *
1467 * Remove the tail of the list. This function does not take any locks
1468 * so must be used with appropriate locks held only. The tail item is
1469 * returned or %NULL if the list is empty.
1470 */
7965bd4d 1471struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1472static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1473{
1474 struct sk_buff *skb = skb_peek_tail(list);
1475 if (skb)
1476 __skb_unlink(skb, list);
1477 return skb;
1478}
1479
1480
bdcc0924 1481static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1482{
1483 return skb->data_len;
1484}
1485
1486static inline unsigned int skb_headlen(const struct sk_buff *skb)
1487{
1488 return skb->len - skb->data_len;
1489}
1490
1491static inline int skb_pagelen(const struct sk_buff *skb)
1492{
1493 int i, len = 0;
1494
1495 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1496 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1497 return len + skb_headlen(skb);
1498}
1499
131ea667
IC
1500/**
1501 * __skb_fill_page_desc - initialise a paged fragment in an skb
1502 * @skb: buffer containing fragment to be initialised
1503 * @i: paged fragment index to initialise
1504 * @page: the page to use for this fragment
1505 * @off: the offset to the data with @page
1506 * @size: the length of the data
1507 *
1508 * Initialises the @i'th fragment of @skb to point to &size bytes at
1509 * offset @off within @page.
1510 *
1511 * Does not take any additional reference on the fragment.
1512 */
1513static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1514 struct page *page, int off, int size)
1da177e4
LT
1515{
1516 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1517
c48a11c7
MG
1518 /*
1519 * Propagate page->pfmemalloc to the skb if we can. The problem is
1520 * that not all callers have unique ownership of the page. If
1521 * pfmemalloc is set, we check the mapping as a mapping implies
1522 * page->index is set (index and pfmemalloc share space).
1523 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1524 * do not lose pfmemalloc information as the pages would not be
1525 * allocated using __GFP_MEMALLOC.
1526 */
a8605c60 1527 frag->page.p = page;
1da177e4 1528 frag->page_offset = off;
9e903e08 1529 skb_frag_size_set(frag, size);
cca7af38
PE
1530
1531 page = compound_head(page);
1532 if (page->pfmemalloc && !page->mapping)
1533 skb->pfmemalloc = true;
131ea667
IC
1534}
1535
1536/**
1537 * skb_fill_page_desc - initialise a paged fragment in an skb
1538 * @skb: buffer containing fragment to be initialised
1539 * @i: paged fragment index to initialise
1540 * @page: the page to use for this fragment
1541 * @off: the offset to the data with @page
1542 * @size: the length of the data
1543 *
1544 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1545 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1546 * addition updates @skb such that @i is the last fragment.
1547 *
1548 * Does not take any additional reference on the fragment.
1549 */
1550static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1551 struct page *page, int off, int size)
1552{
1553 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1554 skb_shinfo(skb)->nr_frags = i + 1;
1555}
1556
7965bd4d
JP
1557void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1558 int size, unsigned int truesize);
654bed16 1559
f8e617e1
JW
1560void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1561 unsigned int truesize);
1562
1da177e4 1563#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1564#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1565#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1566
27a884dc
ACM
1567#ifdef NET_SKBUFF_DATA_USES_OFFSET
1568static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1569{
1570 return skb->head + skb->tail;
1571}
1572
1573static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1574{
1575 skb->tail = skb->data - skb->head;
1576}
1577
1578static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1579{
1580 skb_reset_tail_pointer(skb);
1581 skb->tail += offset;
1582}
7cc46190 1583
27a884dc
ACM
1584#else /* NET_SKBUFF_DATA_USES_OFFSET */
1585static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1586{
1587 return skb->tail;
1588}
1589
1590static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1591{
1592 skb->tail = skb->data;
1593}
1594
1595static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1596{
1597 skb->tail = skb->data + offset;
1598}
4305b541 1599
27a884dc
ACM
1600#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1601
1da177e4
LT
1602/*
1603 * Add data to an sk_buff
1604 */
0c7ddf36 1605unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1606unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1607static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1608{
27a884dc 1609 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1610 SKB_LINEAR_ASSERT(skb);
1611 skb->tail += len;
1612 skb->len += len;
1613 return tmp;
1614}
1615
7965bd4d 1616unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1617static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1618{
1619 skb->data -= len;
1620 skb->len += len;
1621 return skb->data;
1622}
1623
7965bd4d 1624unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1625static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1626{
1627 skb->len -= len;
1628 BUG_ON(skb->len < skb->data_len);
1629 return skb->data += len;
1630}
1631
47d29646
DM
1632static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1633{
1634 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1635}
1636
7965bd4d 1637unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1638
1639static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1640{
1641 if (len > skb_headlen(skb) &&
987c402a 1642 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1643 return NULL;
1644 skb->len -= len;
1645 return skb->data += len;
1646}
1647
1648static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1649{
1650 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1651}
1652
1653static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1654{
1655 if (likely(len <= skb_headlen(skb)))
1656 return 1;
1657 if (unlikely(len > skb->len))
1658 return 0;
987c402a 1659 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1660}
1661
1662/**
1663 * skb_headroom - bytes at buffer head
1664 * @skb: buffer to check
1665 *
1666 * Return the number of bytes of free space at the head of an &sk_buff.
1667 */
c2636b4d 1668static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1669{
1670 return skb->data - skb->head;
1671}
1672
1673/**
1674 * skb_tailroom - bytes at buffer end
1675 * @skb: buffer to check
1676 *
1677 * Return the number of bytes of free space at the tail of an sk_buff
1678 */
1679static inline int skb_tailroom(const struct sk_buff *skb)
1680{
4305b541 1681 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1682}
1683
a21d4572
ED
1684/**
1685 * skb_availroom - bytes at buffer end
1686 * @skb: buffer to check
1687 *
1688 * Return the number of bytes of free space at the tail of an sk_buff
1689 * allocated by sk_stream_alloc()
1690 */
1691static inline int skb_availroom(const struct sk_buff *skb)
1692{
16fad69c
ED
1693 if (skb_is_nonlinear(skb))
1694 return 0;
1695
1696 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1697}
1698
1da177e4
LT
1699/**
1700 * skb_reserve - adjust headroom
1701 * @skb: buffer to alter
1702 * @len: bytes to move
1703 *
1704 * Increase the headroom of an empty &sk_buff by reducing the tail
1705 * room. This is only allowed for an empty buffer.
1706 */
8243126c 1707static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1708{
1709 skb->data += len;
1710 skb->tail += len;
1711}
1712
6a674e9c
JG
1713static inline void skb_reset_inner_headers(struct sk_buff *skb)
1714{
aefbd2b3 1715 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1716 skb->inner_network_header = skb->network_header;
1717 skb->inner_transport_header = skb->transport_header;
1718}
1719
0b5c9db1
JP
1720static inline void skb_reset_mac_len(struct sk_buff *skb)
1721{
1722 skb->mac_len = skb->network_header - skb->mac_header;
1723}
1724
6a674e9c
JG
1725static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1726 *skb)
1727{
1728 return skb->head + skb->inner_transport_header;
1729}
1730
1731static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1732{
1733 skb->inner_transport_header = skb->data - skb->head;
1734}
1735
1736static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1737 const int offset)
1738{
1739 skb_reset_inner_transport_header(skb);
1740 skb->inner_transport_header += offset;
1741}
1742
1743static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1744{
1745 return skb->head + skb->inner_network_header;
1746}
1747
1748static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1749{
1750 skb->inner_network_header = skb->data - skb->head;
1751}
1752
1753static inline void skb_set_inner_network_header(struct sk_buff *skb,
1754 const int offset)
1755{
1756 skb_reset_inner_network_header(skb);
1757 skb->inner_network_header += offset;
1758}
1759
aefbd2b3
PS
1760static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1761{
1762 return skb->head + skb->inner_mac_header;
1763}
1764
1765static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1766{
1767 skb->inner_mac_header = skb->data - skb->head;
1768}
1769
1770static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1771 const int offset)
1772{
1773 skb_reset_inner_mac_header(skb);
1774 skb->inner_mac_header += offset;
1775}
fda55eca
ED
1776static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1777{
35d04610 1778 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1779}
1780
9c70220b
ACM
1781static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1782{
2e07fa9c 1783 return skb->head + skb->transport_header;
9c70220b
ACM
1784}
1785
badff6d0
ACM
1786static inline void skb_reset_transport_header(struct sk_buff *skb)
1787{
2e07fa9c 1788 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1789}
1790
967b05f6
ACM
1791static inline void skb_set_transport_header(struct sk_buff *skb,
1792 const int offset)
1793{
2e07fa9c
ACM
1794 skb_reset_transport_header(skb);
1795 skb->transport_header += offset;
ea2ae17d
ACM
1796}
1797
d56f90a7
ACM
1798static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1799{
2e07fa9c 1800 return skb->head + skb->network_header;
d56f90a7
ACM
1801}
1802
c1d2bbe1
ACM
1803static inline void skb_reset_network_header(struct sk_buff *skb)
1804{
2e07fa9c 1805 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1806}
1807
c14d2450
ACM
1808static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1809{
2e07fa9c
ACM
1810 skb_reset_network_header(skb);
1811 skb->network_header += offset;
c14d2450
ACM
1812}
1813
2e07fa9c 1814static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1815{
2e07fa9c 1816 return skb->head + skb->mac_header;
bbe735e4
ACM
1817}
1818
2e07fa9c 1819static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1820{
35d04610 1821 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1822}
1823
1824static inline void skb_reset_mac_header(struct sk_buff *skb)
1825{
1826 skb->mac_header = skb->data - skb->head;
1827}
1828
1829static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1830{
1831 skb_reset_mac_header(skb);
1832 skb->mac_header += offset;
1833}
1834
0e3da5bb
TT
1835static inline void skb_pop_mac_header(struct sk_buff *skb)
1836{
1837 skb->mac_header = skb->network_header;
1838}
1839
fbbdb8f0
YX
1840static inline void skb_probe_transport_header(struct sk_buff *skb,
1841 const int offset_hint)
1842{
1843 struct flow_keys keys;
1844
1845 if (skb_transport_header_was_set(skb))
1846 return;
1847 else if (skb_flow_dissect(skb, &keys))
1848 skb_set_transport_header(skb, keys.thoff);
1849 else
1850 skb_set_transport_header(skb, offset_hint);
1851}
1852
03606895
ED
1853static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1854{
1855 if (skb_mac_header_was_set(skb)) {
1856 const unsigned char *old_mac = skb_mac_header(skb);
1857
1858 skb_set_mac_header(skb, -skb->mac_len);
1859 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1860 }
1861}
1862
04fb451e
MM
1863static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1864{
1865 return skb->csum_start - skb_headroom(skb);
1866}
1867
2e07fa9c
ACM
1868static inline int skb_transport_offset(const struct sk_buff *skb)
1869{
1870 return skb_transport_header(skb) - skb->data;
1871}
1872
1873static inline u32 skb_network_header_len(const struct sk_buff *skb)
1874{
1875 return skb->transport_header - skb->network_header;
1876}
1877
6a674e9c
JG
1878static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1879{
1880 return skb->inner_transport_header - skb->inner_network_header;
1881}
1882
2e07fa9c
ACM
1883static inline int skb_network_offset(const struct sk_buff *skb)
1884{
1885 return skb_network_header(skb) - skb->data;
1886}
48d49d0c 1887
6a674e9c
JG
1888static inline int skb_inner_network_offset(const struct sk_buff *skb)
1889{
1890 return skb_inner_network_header(skb) - skb->data;
1891}
1892
f9599ce1
CG
1893static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1894{
1895 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1896}
1897
1da177e4
LT
1898/*
1899 * CPUs often take a performance hit when accessing unaligned memory
1900 * locations. The actual performance hit varies, it can be small if the
1901 * hardware handles it or large if we have to take an exception and fix it
1902 * in software.
1903 *
1904 * Since an ethernet header is 14 bytes network drivers often end up with
1905 * the IP header at an unaligned offset. The IP header can be aligned by
1906 * shifting the start of the packet by 2 bytes. Drivers should do this
1907 * with:
1908 *
8660c124 1909 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1910 *
1911 * The downside to this alignment of the IP header is that the DMA is now
1912 * unaligned. On some architectures the cost of an unaligned DMA is high
1913 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1914 *
1da177e4
LT
1915 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1916 * to be overridden.
1917 */
1918#ifndef NET_IP_ALIGN
1919#define NET_IP_ALIGN 2
1920#endif
1921
025be81e
AB
1922/*
1923 * The networking layer reserves some headroom in skb data (via
1924 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1925 * the header has to grow. In the default case, if the header has to grow
d6301d3d 1926 * 32 bytes or less we avoid the reallocation.
025be81e
AB
1927 *
1928 * Unfortunately this headroom changes the DMA alignment of the resulting
1929 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1930 * on some architectures. An architecture can override this value,
1931 * perhaps setting it to a cacheline in size (since that will maintain
1932 * cacheline alignment of the DMA). It must be a power of 2.
1933 *
d6301d3d 1934 * Various parts of the networking layer expect at least 32 bytes of
025be81e 1935 * headroom, you should not reduce this.
5933dd2f
ED
1936 *
1937 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1938 * to reduce average number of cache lines per packet.
1939 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 1940 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
1941 */
1942#ifndef NET_SKB_PAD
5933dd2f 1943#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
1944#endif
1945
7965bd4d 1946int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1947
1948static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1949{
c4264f27 1950 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
1951 WARN_ON(1);
1952 return;
1953 }
27a884dc
ACM
1954 skb->len = len;
1955 skb_set_tail_pointer(skb, len);
1da177e4
LT
1956}
1957
7965bd4d 1958void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1959
1960static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1961{
3cc0e873
HX
1962 if (skb->data_len)
1963 return ___pskb_trim(skb, len);
1964 __skb_trim(skb, len);
1965 return 0;
1da177e4
LT
1966}
1967
1968static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1969{
1970 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1971}
1972
e9fa4f7b
HX
1973/**
1974 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1975 * @skb: buffer to alter
1976 * @len: new length
1977 *
1978 * This is identical to pskb_trim except that the caller knows that
1979 * the skb is not cloned so we should never get an error due to out-
1980 * of-memory.
1981 */
1982static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1983{
1984 int err = pskb_trim(skb, len);
1985 BUG_ON(err);
1986}
1987
1da177e4
LT
1988/**
1989 * skb_orphan - orphan a buffer
1990 * @skb: buffer to orphan
1991 *
1992 * If a buffer currently has an owner then we call the owner's
1993 * destructor function and make the @skb unowned. The buffer continues
1994 * to exist but is no longer charged to its former owner.
1995 */
1996static inline void skb_orphan(struct sk_buff *skb)
1997{
c34a7612 1998 if (skb->destructor) {
1da177e4 1999 skb->destructor(skb);
c34a7612
ED
2000 skb->destructor = NULL;
2001 skb->sk = NULL;
376c7311
ED
2002 } else {
2003 BUG_ON(skb->sk);
c34a7612 2004 }
1da177e4
LT
2005}
2006
a353e0ce
MT
2007/**
2008 * skb_orphan_frags - orphan the frags contained in a buffer
2009 * @skb: buffer to orphan frags from
2010 * @gfp_mask: allocation mask for replacement pages
2011 *
2012 * For each frag in the SKB which needs a destructor (i.e. has an
2013 * owner) create a copy of that frag and release the original
2014 * page by calling the destructor.
2015 */
2016static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2017{
2018 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2019 return 0;
2020 return skb_copy_ubufs(skb, gfp_mask);
2021}
2022
1da177e4
LT
2023/**
2024 * __skb_queue_purge - empty a list
2025 * @list: list to empty
2026 *
2027 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2028 * the list and one reference dropped. This function does not take the
2029 * list lock and the caller must hold the relevant locks to use it.
2030 */
7965bd4d 2031void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2032static inline void __skb_queue_purge(struct sk_buff_head *list)
2033{
2034 struct sk_buff *skb;
2035 while ((skb = __skb_dequeue(list)) != NULL)
2036 kfree_skb(skb);
2037}
2038
e5e67305
AD
2039#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2040#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2041#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2042
7965bd4d 2043void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2044
7965bd4d
JP
2045struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2046 gfp_t gfp_mask);
8af27456
CH
2047
2048/**
2049 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2050 * @dev: network device to receive on
2051 * @length: length to allocate
2052 *
2053 * Allocate a new &sk_buff and assign it a usage count of one. The
2054 * buffer has unspecified headroom built in. Users should allocate
2055 * the headroom they think they need without accounting for the
2056 * built in space. The built in space is used for optimisations.
2057 *
2058 * %NULL is returned if there is no free memory. Although this function
2059 * allocates memory it can be called from an interrupt.
2060 */
2061static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2062 unsigned int length)
8af27456
CH
2063{
2064 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2065}
2066
6f532612
ED
2067/* legacy helper around __netdev_alloc_skb() */
2068static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2069 gfp_t gfp_mask)
2070{
2071 return __netdev_alloc_skb(NULL, length, gfp_mask);
2072}
2073
2074/* legacy helper around netdev_alloc_skb() */
2075static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2076{
2077 return netdev_alloc_skb(NULL, length);
2078}
2079
2080
4915a0de
ED
2081static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2082 unsigned int length, gfp_t gfp)
61321bbd 2083{
4915a0de 2084 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2085
2086 if (NET_IP_ALIGN && skb)
2087 skb_reserve(skb, NET_IP_ALIGN);
2088 return skb;
2089}
2090
4915a0de
ED
2091static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2092 unsigned int length)
2093{
2094 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2095}
2096
bc6fc9fa
FF
2097/**
2098 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
2099 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2100 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2101 * @order: size of the allocation
2102 *
2103 * Allocate a new page.
2104 *
2105 * %NULL is returned if there is no free memory.
2106*/
2107static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2108 struct sk_buff *skb,
2109 unsigned int order)
2110{
2111 struct page *page;
2112
2113 gfp_mask |= __GFP_COLD;
2114
2115 if (!(gfp_mask & __GFP_NOMEMALLOC))
2116 gfp_mask |= __GFP_MEMALLOC;
2117
2118 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2119 if (skb && page && page->pfmemalloc)
2120 skb->pfmemalloc = true;
2121
2122 return page;
2123}
2124
2125/**
2126 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2127 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2128 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2129 *
2130 * Allocate a new page.
2131 *
2132 * %NULL is returned if there is no free memory.
2133 */
2134static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2135 struct sk_buff *skb)
2136{
2137 return __skb_alloc_pages(gfp_mask, skb, 0);
2138}
2139
2140/**
2141 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2142 * @page: The page that was allocated from skb_alloc_page
2143 * @skb: The skb that may need pfmemalloc set
2144 */
2145static inline void skb_propagate_pfmemalloc(struct page *page,
2146 struct sk_buff *skb)
2147{
2148 if (page && page->pfmemalloc)
2149 skb->pfmemalloc = true;
2150}
2151
131ea667 2152/**
e227867f 2153 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2154 * @frag: the paged fragment
2155 *
2156 * Returns the &struct page associated with @frag.
2157 */
2158static inline struct page *skb_frag_page(const skb_frag_t *frag)
2159{
a8605c60 2160 return frag->page.p;
131ea667
IC
2161}
2162
2163/**
2164 * __skb_frag_ref - take an addition reference on a paged fragment.
2165 * @frag: the paged fragment
2166 *
2167 * Takes an additional reference on the paged fragment @frag.
2168 */
2169static inline void __skb_frag_ref(skb_frag_t *frag)
2170{
2171 get_page(skb_frag_page(frag));
2172}
2173
2174/**
2175 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2176 * @skb: the buffer
2177 * @f: the fragment offset.
2178 *
2179 * Takes an additional reference on the @f'th paged fragment of @skb.
2180 */
2181static inline void skb_frag_ref(struct sk_buff *skb, int f)
2182{
2183 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2184}
2185
2186/**
2187 * __skb_frag_unref - release a reference on a paged fragment.
2188 * @frag: the paged fragment
2189 *
2190 * Releases a reference on the paged fragment @frag.
2191 */
2192static inline void __skb_frag_unref(skb_frag_t *frag)
2193{
2194 put_page(skb_frag_page(frag));
2195}
2196
2197/**
2198 * skb_frag_unref - release a reference on a paged fragment of an skb.
2199 * @skb: the buffer
2200 * @f: the fragment offset
2201 *
2202 * Releases a reference on the @f'th paged fragment of @skb.
2203 */
2204static inline void skb_frag_unref(struct sk_buff *skb, int f)
2205{
2206 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2207}
2208
2209/**
2210 * skb_frag_address - gets the address of the data contained in a paged fragment
2211 * @frag: the paged fragment buffer
2212 *
2213 * Returns the address of the data within @frag. The page must already
2214 * be mapped.
2215 */
2216static inline void *skb_frag_address(const skb_frag_t *frag)
2217{
2218 return page_address(skb_frag_page(frag)) + frag->page_offset;
2219}
2220
2221/**
2222 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2223 * @frag: the paged fragment buffer
2224 *
2225 * Returns the address of the data within @frag. Checks that the page
2226 * is mapped and returns %NULL otherwise.
2227 */
2228static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2229{
2230 void *ptr = page_address(skb_frag_page(frag));
2231 if (unlikely(!ptr))
2232 return NULL;
2233
2234 return ptr + frag->page_offset;
2235}
2236
2237/**
2238 * __skb_frag_set_page - sets the page contained in a paged fragment
2239 * @frag: the paged fragment
2240 * @page: the page to set
2241 *
2242 * Sets the fragment @frag to contain @page.
2243 */
2244static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2245{
a8605c60 2246 frag->page.p = page;
131ea667
IC
2247}
2248
2249/**
2250 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2251 * @skb: the buffer
2252 * @f: the fragment offset
2253 * @page: the page to set
2254 *
2255 * Sets the @f'th fragment of @skb to contain @page.
2256 */
2257static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2258 struct page *page)
2259{
2260 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2261}
2262
400dfd3a
ED
2263bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2264
131ea667
IC
2265/**
2266 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2267 * @dev: the device to map the fragment to
131ea667
IC
2268 * @frag: the paged fragment to map
2269 * @offset: the offset within the fragment (starting at the
2270 * fragment's own offset)
2271 * @size: the number of bytes to map
f83347df 2272 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2273 *
2274 * Maps the page associated with @frag to @device.
2275 */
2276static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2277 const skb_frag_t *frag,
2278 size_t offset, size_t size,
2279 enum dma_data_direction dir)
2280{
2281 return dma_map_page(dev, skb_frag_page(frag),
2282 frag->page_offset + offset, size, dir);
2283}
2284
117632e6
ED
2285static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2286 gfp_t gfp_mask)
2287{
2288 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2289}
2290
bad93e9d
OP
2291
2292static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2293 gfp_t gfp_mask)
2294{
2295 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2296}
2297
2298
334a8132
PM
2299/**
2300 * skb_clone_writable - is the header of a clone writable
2301 * @skb: buffer to check
2302 * @len: length up to which to write
2303 *
2304 * Returns true if modifying the header part of the cloned buffer
2305 * does not requires the data to be copied.
2306 */
05bdd2f1 2307static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2308{
2309 return !skb_header_cloned(skb) &&
2310 skb_headroom(skb) + len <= skb->hdr_len;
2311}
2312
d9cc2048
HX
2313static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2314 int cloned)
2315{
2316 int delta = 0;
2317
d9cc2048
HX
2318 if (headroom > skb_headroom(skb))
2319 delta = headroom - skb_headroom(skb);
2320
2321 if (delta || cloned)
2322 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2323 GFP_ATOMIC);
2324 return 0;
2325}
2326
1da177e4
LT
2327/**
2328 * skb_cow - copy header of skb when it is required
2329 * @skb: buffer to cow
2330 * @headroom: needed headroom
2331 *
2332 * If the skb passed lacks sufficient headroom or its data part
2333 * is shared, data is reallocated. If reallocation fails, an error
2334 * is returned and original skb is not changed.
2335 *
2336 * The result is skb with writable area skb->head...skb->tail
2337 * and at least @headroom of space at head.
2338 */
2339static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2340{
d9cc2048
HX
2341 return __skb_cow(skb, headroom, skb_cloned(skb));
2342}
1da177e4 2343
d9cc2048
HX
2344/**
2345 * skb_cow_head - skb_cow but only making the head writable
2346 * @skb: buffer to cow
2347 * @headroom: needed headroom
2348 *
2349 * This function is identical to skb_cow except that we replace the
2350 * skb_cloned check by skb_header_cloned. It should be used when
2351 * you only need to push on some header and do not need to modify
2352 * the data.
2353 */
2354static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2355{
2356 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2357}
2358
2359/**
2360 * skb_padto - pad an skbuff up to a minimal size
2361 * @skb: buffer to pad
2362 * @len: minimal length
2363 *
2364 * Pads up a buffer to ensure the trailing bytes exist and are
2365 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2366 * is untouched. Otherwise it is extended. Returns zero on
2367 * success. The skb is freed on error.
1da177e4
LT
2368 */
2369
5b057c6b 2370static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2371{
2372 unsigned int size = skb->len;
2373 if (likely(size >= len))
5b057c6b 2374 return 0;
987c402a 2375 return skb_pad(skb, len - size);
1da177e4
LT
2376}
2377
2378static inline int skb_add_data(struct sk_buff *skb,
2379 char __user *from, int copy)
2380{
2381 const int off = skb->len;
2382
2383 if (skb->ip_summed == CHECKSUM_NONE) {
2384 int err = 0;
5084205f 2385 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2386 copy, 0, &err);
2387 if (!err) {
2388 skb->csum = csum_block_add(skb->csum, csum, off);
2389 return 0;
2390 }
2391 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2392 return 0;
2393
2394 __skb_trim(skb, off);
2395 return -EFAULT;
2396}
2397
38ba0a65
ED
2398static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2399 const struct page *page, int off)
1da177e4
LT
2400{
2401 if (i) {
9e903e08 2402 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2403
ea2ab693 2404 return page == skb_frag_page(frag) &&
9e903e08 2405 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2406 }
38ba0a65 2407 return false;
1da177e4
LT
2408}
2409
364c6bad
HX
2410static inline int __skb_linearize(struct sk_buff *skb)
2411{
2412 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2413}
2414
1da177e4
LT
2415/**
2416 * skb_linearize - convert paged skb to linear one
2417 * @skb: buffer to linarize
1da177e4
LT
2418 *
2419 * If there is no free memory -ENOMEM is returned, otherwise zero
2420 * is returned and the old skb data released.
2421 */
364c6bad
HX
2422static inline int skb_linearize(struct sk_buff *skb)
2423{
2424 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2425}
2426
cef401de
ED
2427/**
2428 * skb_has_shared_frag - can any frag be overwritten
2429 * @skb: buffer to test
2430 *
2431 * Return true if the skb has at least one frag that might be modified
2432 * by an external entity (as in vmsplice()/sendfile())
2433 */
2434static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2435{
c9af6db4
PS
2436 return skb_is_nonlinear(skb) &&
2437 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2438}
2439
364c6bad
HX
2440/**
2441 * skb_linearize_cow - make sure skb is linear and writable
2442 * @skb: buffer to process
2443 *
2444 * If there is no free memory -ENOMEM is returned, otherwise zero
2445 * is returned and the old skb data released.
2446 */
2447static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2448{
364c6bad
HX
2449 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2450 __skb_linearize(skb) : 0;
1da177e4
LT
2451}
2452
2453/**
2454 * skb_postpull_rcsum - update checksum for received skb after pull
2455 * @skb: buffer to update
2456 * @start: start of data before pull
2457 * @len: length of data pulled
2458 *
2459 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2460 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2461 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2462 */
2463
2464static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2465 const void *start, unsigned int len)
1da177e4 2466{
84fa7933 2467 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2468 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2469}
2470
cbb042f9
HX
2471unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2472
7ce5a27f
DM
2473/**
2474 * pskb_trim_rcsum - trim received skb and update checksum
2475 * @skb: buffer to trim
2476 * @len: new length
2477 *
2478 * This is exactly the same as pskb_trim except that it ensures the
2479 * checksum of received packets are still valid after the operation.
2480 */
2481
2482static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2483{
2484 if (likely(len >= skb->len))
2485 return 0;
2486 if (skb->ip_summed == CHECKSUM_COMPLETE)
2487 skb->ip_summed = CHECKSUM_NONE;
2488 return __pskb_trim(skb, len);
2489}
2490
1da177e4
LT
2491#define skb_queue_walk(queue, skb) \
2492 for (skb = (queue)->next; \
a1e4891f 2493 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2494 skb = skb->next)
2495
46f8914e
JC
2496#define skb_queue_walk_safe(queue, skb, tmp) \
2497 for (skb = (queue)->next, tmp = skb->next; \
2498 skb != (struct sk_buff *)(queue); \
2499 skb = tmp, tmp = skb->next)
2500
1164f52a 2501#define skb_queue_walk_from(queue, skb) \
a1e4891f 2502 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2503 skb = skb->next)
2504
2505#define skb_queue_walk_from_safe(queue, skb, tmp) \
2506 for (tmp = skb->next; \
2507 skb != (struct sk_buff *)(queue); \
2508 skb = tmp, tmp = skb->next)
2509
300ce174
SH
2510#define skb_queue_reverse_walk(queue, skb) \
2511 for (skb = (queue)->prev; \
a1e4891f 2512 skb != (struct sk_buff *)(queue); \
300ce174
SH
2513 skb = skb->prev)
2514
686a2955
DM
2515#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2516 for (skb = (queue)->prev, tmp = skb->prev; \
2517 skb != (struct sk_buff *)(queue); \
2518 skb = tmp, tmp = skb->prev)
2519
2520#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2521 for (tmp = skb->prev; \
2522 skb != (struct sk_buff *)(queue); \
2523 skb = tmp, tmp = skb->prev)
1da177e4 2524
21dc3301 2525static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2526{
2527 return skb_shinfo(skb)->frag_list != NULL;
2528}
2529
2530static inline void skb_frag_list_init(struct sk_buff *skb)
2531{
2532 skb_shinfo(skb)->frag_list = NULL;
2533}
2534
2535static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2536{
2537 frag->next = skb_shinfo(skb)->frag_list;
2538 skb_shinfo(skb)->frag_list = frag;
2539}
2540
2541#define skb_walk_frags(skb, iter) \
2542 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2543
7965bd4d
JP
2544struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2545 int *peeked, int *off, int *err);
2546struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2547 int *err);
2548unsigned int datagram_poll(struct file *file, struct socket *sock,
2549 struct poll_table_struct *wait);
2550int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2551 struct iovec *to, int size);
2552int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2553 struct iovec *iov);
2554int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2555 const struct iovec *from, int from_offset,
2556 int len);
2557int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2558 int offset, size_t count);
2559int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2560 const struct iovec *to, int to_offset,
2561 int size);
2562void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2563void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2564int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2565int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2566int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2567__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2568 int len, __wsum csum);
2569int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2570 struct pipe_inode_info *pipe, unsigned int len,
2571 unsigned int flags);
2572void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2573unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2574int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2575 int len, int hlen);
7965bd4d
JP
2576void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2577int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2578void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2579unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2580struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2581struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
20380731 2582
2817a336
DB
2583struct skb_checksum_ops {
2584 __wsum (*update)(const void *mem, int len, __wsum wsum);
2585 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2586};
2587
2588__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2589 __wsum csum, const struct skb_checksum_ops *ops);
2590__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2591 __wsum csum);
2592
690e36e7
DM
2593static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2594 int len, void *data, int hlen, void *buffer)
1da177e4 2595{
55820ee2 2596 if (hlen - offset >= len)
690e36e7 2597 return data + offset;
1da177e4 2598
690e36e7
DM
2599 if (!skb ||
2600 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2601 return NULL;
2602
2603 return buffer;
2604}
2605
690e36e7
DM
2606static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2607 int len, void *buffer)
2608{
2609 return __skb_header_pointer(skb, offset, len, skb->data,
2610 skb_headlen(skb), buffer);
2611}
2612
4262e5cc
DB
2613/**
2614 * skb_needs_linearize - check if we need to linearize a given skb
2615 * depending on the given device features.
2616 * @skb: socket buffer to check
2617 * @features: net device features
2618 *
2619 * Returns true if either:
2620 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2621 * 2. skb is fragmented and the device does not support SG.
2622 */
2623static inline bool skb_needs_linearize(struct sk_buff *skb,
2624 netdev_features_t features)
2625{
2626 return skb_is_nonlinear(skb) &&
2627 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2628 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2629}
2630
d626f62b
ACM
2631static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2632 void *to,
2633 const unsigned int len)
2634{
2635 memcpy(to, skb->data, len);
2636}
2637
2638static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2639 const int offset, void *to,
2640 const unsigned int len)
2641{
2642 memcpy(to, skb->data + offset, len);
2643}
2644
27d7ff46
ACM
2645static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2646 const void *from,
2647 const unsigned int len)
2648{
2649 memcpy(skb->data, from, len);
2650}
2651
2652static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2653 const int offset,
2654 const void *from,
2655 const unsigned int len)
2656{
2657 memcpy(skb->data + offset, from, len);
2658}
2659
7965bd4d 2660void skb_init(void);
1da177e4 2661
ac45f602
PO
2662static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2663{
2664 return skb->tstamp;
2665}
2666
a61bbcf2
PM
2667/**
2668 * skb_get_timestamp - get timestamp from a skb
2669 * @skb: skb to get stamp from
2670 * @stamp: pointer to struct timeval to store stamp in
2671 *
2672 * Timestamps are stored in the skb as offsets to a base timestamp.
2673 * This function converts the offset back to a struct timeval and stores
2674 * it in stamp.
2675 */
ac45f602
PO
2676static inline void skb_get_timestamp(const struct sk_buff *skb,
2677 struct timeval *stamp)
a61bbcf2 2678{
b7aa0bf7 2679 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2680}
2681
ac45f602
PO
2682static inline void skb_get_timestampns(const struct sk_buff *skb,
2683 struct timespec *stamp)
2684{
2685 *stamp = ktime_to_timespec(skb->tstamp);
2686}
2687
b7aa0bf7 2688static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2689{
b7aa0bf7 2690 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2691}
2692
164891aa
SH
2693static inline ktime_t net_timedelta(ktime_t t)
2694{
2695 return ktime_sub(ktime_get_real(), t);
2696}
2697
b9ce204f
IJ
2698static inline ktime_t net_invalid_timestamp(void)
2699{
2700 return ktime_set(0, 0);
2701}
a61bbcf2 2702
62bccb8c
AD
2703struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2704
c1f19b51
RC
2705#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2706
7965bd4d
JP
2707void skb_clone_tx_timestamp(struct sk_buff *skb);
2708bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2709
2710#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2711
2712static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2713{
2714}
2715
2716static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2717{
2718 return false;
2719}
2720
2721#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2722
2723/**
2724 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2725 *
da92b194
RC
2726 * PHY drivers may accept clones of transmitted packets for
2727 * timestamping via their phy_driver.txtstamp method. These drivers
2728 * must call this function to return the skb back to the stack, with
2729 * or without a timestamp.
2730 *
c1f19b51 2731 * @skb: clone of the the original outgoing packet
da92b194 2732 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2733 *
2734 */
2735void skb_complete_tx_timestamp(struct sk_buff *skb,
2736 struct skb_shared_hwtstamps *hwtstamps);
2737
e7fd2885
WB
2738void __skb_tstamp_tx(struct sk_buff *orig_skb,
2739 struct skb_shared_hwtstamps *hwtstamps,
2740 struct sock *sk, int tstype);
2741
ac45f602
PO
2742/**
2743 * skb_tstamp_tx - queue clone of skb with send time stamps
2744 * @orig_skb: the original outgoing packet
2745 * @hwtstamps: hardware time stamps, may be NULL if not available
2746 *
2747 * If the skb has a socket associated, then this function clones the
2748 * skb (thus sharing the actual data and optional structures), stores
2749 * the optional hardware time stamping information (if non NULL) or
2750 * generates a software time stamp (otherwise), then queues the clone
2751 * to the error queue of the socket. Errors are silently ignored.
2752 */
7965bd4d
JP
2753void skb_tstamp_tx(struct sk_buff *orig_skb,
2754 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2755
4507a715
RC
2756static inline void sw_tx_timestamp(struct sk_buff *skb)
2757{
2244d07b
OH
2758 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2759 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2760 skb_tstamp_tx(skb, NULL);
2761}
2762
2763/**
2764 * skb_tx_timestamp() - Driver hook for transmit timestamping
2765 *
2766 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2767 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2768 *
73409f3b
DM
2769 * Specifically, one should make absolutely sure that this function is
2770 * called before TX completion of this packet can trigger. Otherwise
2771 * the packet could potentially already be freed.
2772 *
4507a715
RC
2773 * @skb: A socket buffer.
2774 */
2775static inline void skb_tx_timestamp(struct sk_buff *skb)
2776{
c1f19b51 2777 skb_clone_tx_timestamp(skb);
4507a715
RC
2778 sw_tx_timestamp(skb);
2779}
2780
6e3e939f
JB
2781/**
2782 * skb_complete_wifi_ack - deliver skb with wifi status
2783 *
2784 * @skb: the original outgoing packet
2785 * @acked: ack status
2786 *
2787 */
2788void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2789
7965bd4d
JP
2790__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2791__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2792
60476372
HX
2793static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2794{
5d0c2b95 2795 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
60476372
HX
2796}
2797
fb286bb2
HX
2798/**
2799 * skb_checksum_complete - Calculate checksum of an entire packet
2800 * @skb: packet to process
2801 *
2802 * This function calculates the checksum over the entire packet plus
2803 * the value of skb->csum. The latter can be used to supply the
2804 * checksum of a pseudo header as used by TCP/UDP. It returns the
2805 * checksum.
2806 *
2807 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2808 * this function can be used to verify that checksum on received
2809 * packets. In that case the function should return zero if the
2810 * checksum is correct. In particular, this function will return zero
2811 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2812 * hardware has already verified the correctness of the checksum.
2813 */
4381ca3c 2814static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2815{
60476372
HX
2816 return skb_csum_unnecessary(skb) ?
2817 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2818}
2819
77cffe23
TH
2820static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2821{
2822 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2823 if (skb->csum_level == 0)
2824 skb->ip_summed = CHECKSUM_NONE;
2825 else
2826 skb->csum_level--;
2827 }
2828}
2829
2830static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2831{
2832 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2833 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2834 skb->csum_level++;
2835 } else if (skb->ip_summed == CHECKSUM_NONE) {
2836 skb->ip_summed = CHECKSUM_UNNECESSARY;
2837 skb->csum_level = 0;
2838 }
2839}
2840
5a212329
TH
2841static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2842{
2843 /* Mark current checksum as bad (typically called from GRO
2844 * path). In the case that ip_summed is CHECKSUM_NONE
2845 * this must be the first checksum encountered in the packet.
2846 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2847 * checksum after the last one validated. For UDP, a zero
2848 * checksum can not be marked as bad.
2849 */
2850
2851 if (skb->ip_summed == CHECKSUM_NONE ||
2852 skb->ip_summed == CHECKSUM_UNNECESSARY)
2853 skb->csum_bad = 1;
2854}
2855
76ba0aae
TH
2856/* Check if we need to perform checksum complete validation.
2857 *
2858 * Returns true if checksum complete is needed, false otherwise
2859 * (either checksum is unnecessary or zero checksum is allowed).
2860 */
2861static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2862 bool zero_okay,
2863 __sum16 check)
2864{
5d0c2b95
TH
2865 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2866 skb->csum_valid = 1;
77cffe23 2867 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
2868 return false;
2869 }
2870
2871 return true;
2872}
2873
2874/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2875 * in checksum_init.
2876 */
2877#define CHECKSUM_BREAK 76
2878
2879/* Validate (init) checksum based on checksum complete.
2880 *
2881 * Return values:
2882 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2883 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2884 * checksum is stored in skb->csum for use in __skb_checksum_complete
2885 * non-zero: value of invalid checksum
2886 *
2887 */
2888static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2889 bool complete,
2890 __wsum psum)
2891{
2892 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2893 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 2894 skb->csum_valid = 1;
76ba0aae
TH
2895 return 0;
2896 }
5a212329
TH
2897 } else if (skb->csum_bad) {
2898 /* ip_summed == CHECKSUM_NONE in this case */
2899 return 1;
76ba0aae
TH
2900 }
2901
2902 skb->csum = psum;
2903
5d0c2b95
TH
2904 if (complete || skb->len <= CHECKSUM_BREAK) {
2905 __sum16 csum;
2906
2907 csum = __skb_checksum_complete(skb);
2908 skb->csum_valid = !csum;
2909 return csum;
2910 }
76ba0aae
TH
2911
2912 return 0;
2913}
2914
2915static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2916{
2917 return 0;
2918}
2919
2920/* Perform checksum validate (init). Note that this is a macro since we only
2921 * want to calculate the pseudo header which is an input function if necessary.
2922 * First we try to validate without any computation (checksum unnecessary) and
2923 * then calculate based on checksum complete calling the function to compute
2924 * pseudo header.
2925 *
2926 * Return values:
2927 * 0: checksum is validated or try to in skb_checksum_complete
2928 * non-zero: value of invalid checksum
2929 */
2930#define __skb_checksum_validate(skb, proto, complete, \
2931 zero_okay, check, compute_pseudo) \
2932({ \
2933 __sum16 __ret = 0; \
5d0c2b95 2934 skb->csum_valid = 0; \
76ba0aae
TH
2935 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2936 __ret = __skb_checksum_validate_complete(skb, \
2937 complete, compute_pseudo(skb, proto)); \
2938 __ret; \
2939})
2940
2941#define skb_checksum_init(skb, proto, compute_pseudo) \
2942 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2943
2944#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2945 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2946
2947#define skb_checksum_validate(skb, proto, compute_pseudo) \
2948 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2949
2950#define skb_checksum_validate_zero_check(skb, proto, check, \
2951 compute_pseudo) \
2952 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2953
2954#define skb_checksum_simple_validate(skb) \
2955 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2956
d96535a1
TH
2957static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
2958{
2959 return (skb->ip_summed == CHECKSUM_NONE &&
2960 skb->csum_valid && !skb->csum_bad);
2961}
2962
2963static inline void __skb_checksum_convert(struct sk_buff *skb,
2964 __sum16 check, __wsum pseudo)
2965{
2966 skb->csum = ~pseudo;
2967 skb->ip_summed = CHECKSUM_COMPLETE;
2968}
2969
2970#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
2971do { \
2972 if (__skb_checksum_convert_check(skb)) \
2973 __skb_checksum_convert(skb, check, \
2974 compute_pseudo(skb, proto)); \
2975} while (0)
2976
5f79e0f9 2977#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 2978void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
2979static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2980{
2981 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 2982 nf_conntrack_destroy(nfct);
1da177e4
LT
2983}
2984static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2985{
2986 if (nfct)
2987 atomic_inc(&nfct->use);
2988}
2fc72c7b 2989#endif
1da177e4
LT
2990#ifdef CONFIG_BRIDGE_NETFILTER
2991static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2992{
2993 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2994 kfree(nf_bridge);
2995}
2996static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2997{
2998 if (nf_bridge)
2999 atomic_inc(&nf_bridge->use);
3000}
3001#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3002static inline void nf_reset(struct sk_buff *skb)
3003{
5f79e0f9 3004#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3005 nf_conntrack_put(skb->nfct);
3006 skb->nfct = NULL;
2fc72c7b 3007#endif
a193a4ab
PM
3008#ifdef CONFIG_BRIDGE_NETFILTER
3009 nf_bridge_put(skb->nf_bridge);
3010 skb->nf_bridge = NULL;
3011#endif
3012}
3013
124dff01
PM
3014static inline void nf_reset_trace(struct sk_buff *skb)
3015{
478b360a 3016#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3017 skb->nf_trace = 0;
3018#endif
a193a4ab
PM
3019}
3020
edda553c
YK
3021/* Note: This doesn't put any conntrack and bridge info in dst. */
3022static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3023{
5f79e0f9 3024#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3025 dst->nfct = src->nfct;
3026 nf_conntrack_get(src->nfct);
3027 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3028#endif
edda553c
YK
3029#ifdef CONFIG_BRIDGE_NETFILTER
3030 dst->nf_bridge = src->nf_bridge;
3031 nf_bridge_get(src->nf_bridge);
3032#endif
478b360a
FW
3033#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3034 dst->nf_trace = src->nf_trace;
3035#endif
edda553c
YK
3036}
3037
e7ac05f3
YK
3038static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3039{
e7ac05f3 3040#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3041 nf_conntrack_put(dst->nfct);
2fc72c7b 3042#endif
e7ac05f3
YK
3043#ifdef CONFIG_BRIDGE_NETFILTER
3044 nf_bridge_put(dst->nf_bridge);
3045#endif
3046 __nf_copy(dst, src);
3047}
3048
984bc16c
JM
3049#ifdef CONFIG_NETWORK_SECMARK
3050static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3051{
3052 to->secmark = from->secmark;
3053}
3054
3055static inline void skb_init_secmark(struct sk_buff *skb)
3056{
3057 skb->secmark = 0;
3058}
3059#else
3060static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3061{ }
3062
3063static inline void skb_init_secmark(struct sk_buff *skb)
3064{ }
3065#endif
3066
574f7194
EB
3067static inline bool skb_irq_freeable(const struct sk_buff *skb)
3068{
3069 return !skb->destructor &&
3070#if IS_ENABLED(CONFIG_XFRM)
3071 !skb->sp &&
3072#endif
3073#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3074 !skb->nfct &&
3075#endif
3076 !skb->_skb_refdst &&
3077 !skb_has_frag_list(skb);
3078}
3079
f25f4e44
PWJ
3080static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3081{
f25f4e44 3082 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3083}
3084
9247744e 3085static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3086{
4e3ab47a 3087 return skb->queue_mapping;
4e3ab47a
PE
3088}
3089
f25f4e44
PWJ
3090static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3091{
f25f4e44 3092 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3093}
3094
d5a9e24a
DM
3095static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3096{
3097 skb->queue_mapping = rx_queue + 1;
3098}
3099
9247744e 3100static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3101{
3102 return skb->queue_mapping - 1;
3103}
3104
9247744e 3105static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3106{
a02cec21 3107 return skb->queue_mapping != 0;
d5a9e24a
DM
3108}
3109
0e001614 3110u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
7965bd4d 3111 unsigned int num_tx_queues);
9247744e 3112
def8b4fa
AD
3113static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3114{
0b3d8e08 3115#ifdef CONFIG_XFRM
def8b4fa 3116 return skb->sp;
def8b4fa 3117#else
def8b4fa 3118 return NULL;
def8b4fa 3119#endif
0b3d8e08 3120}
def8b4fa 3121
68c33163
PS
3122/* Keeps track of mac header offset relative to skb->head.
3123 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3124 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3125 * tunnel skb it points to outer mac header.
3126 * Keeps track of level of encapsulation of network headers.
3127 */
68c33163 3128struct skb_gso_cb {
3347c960
ED
3129 int mac_offset;
3130 int encap_level;
7e2b10c1 3131 __u16 csum_start;
68c33163
PS
3132};
3133#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3134
3135static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3136{
3137 return (skb_mac_header(inner_skb) - inner_skb->head) -
3138 SKB_GSO_CB(inner_skb)->mac_offset;
3139}
3140
1e2bd517
PS
3141static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3142{
3143 int new_headroom, headroom;
3144 int ret;
3145
3146 headroom = skb_headroom(skb);
3147 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3148 if (ret)
3149 return ret;
3150
3151 new_headroom = skb_headroom(skb);
3152 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3153 return 0;
3154}
3155
7e2b10c1
TH
3156/* Compute the checksum for a gso segment. First compute the checksum value
3157 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3158 * then add in skb->csum (checksum from csum_start to end of packet).
3159 * skb->csum and csum_start are then updated to reflect the checksum of the
3160 * resultant packet starting from the transport header-- the resultant checksum
3161 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3162 * header.
3163 */
3164static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3165{
3166 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3167 skb_transport_offset(skb);
3168 __u16 csum;
3169
3170 csum = csum_fold(csum_partial(skb_transport_header(skb),
3171 plen, skb->csum));
3172 skb->csum = res;
3173 SKB_GSO_CB(skb)->csum_start -= plen;
3174
3175 return csum;
3176}
3177
bdcc0924 3178static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3179{
3180 return skb_shinfo(skb)->gso_size;
3181}
3182
36a8f39e 3183/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3184static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3185{
3186 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3187}
3188
7965bd4d 3189void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3190
3191static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3192{
3193 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3194 * wanted then gso_type will be set. */
05bdd2f1
ED
3195 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3196
b78462eb
AD
3197 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3198 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3199 __skb_warn_lro_forwarding(skb);
3200 return true;
3201 }
3202 return false;
3203}
3204
35fc92a9
HX
3205static inline void skb_forward_csum(struct sk_buff *skb)
3206{
3207 /* Unfortunately we don't support this one. Any brave souls? */
3208 if (skb->ip_summed == CHECKSUM_COMPLETE)
3209 skb->ip_summed = CHECKSUM_NONE;
3210}
3211
bc8acf2c
ED
3212/**
3213 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3214 * @skb: skb to check
3215 *
3216 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3217 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3218 * use this helper, to document places where we make this assertion.
3219 */
05bdd2f1 3220static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3221{
3222#ifdef DEBUG
3223 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3224#endif
3225}
3226
f35d9d8a 3227bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3228
ed1f50c3
PD
3229int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3230
56193d1b
AD
3231u32 skb_get_poff(const struct sk_buff *skb);
3232u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3233 const struct flow_keys *keys, int hlen);
f77668dc 3234
3a7c1ee4
AD
3235/**
3236 * skb_head_is_locked - Determine if the skb->head is locked down
3237 * @skb: skb to check
3238 *
3239 * The head on skbs build around a head frag can be removed if they are
3240 * not cloned. This function returns true if the skb head is locked down
3241 * due to either being allocated via kmalloc, or by being a clone with
3242 * multiple references to the head.
3243 */
3244static inline bool skb_head_is_locked(const struct sk_buff *skb)
3245{
3246 return !skb->head_frag || skb_cloned(skb);
3247}
fe6cc55f
FW
3248
3249/**
3250 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3251 *
3252 * @skb: GSO skb
3253 *
3254 * skb_gso_network_seglen is used to determine the real size of the
3255 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3256 *
3257 * The MAC/L2 header is not accounted for.
3258 */
3259static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3260{
3261 unsigned int hdr_len = skb_transport_header(skb) -
3262 skb_network_header(skb);
3263 return hdr_len + skb_gso_transport_seglen(skb);
3264}
1da177e4
LT
3265#endif /* __KERNEL__ */
3266#endif /* _LINUX_SKBUFF_H */