]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - include/linux/skbuff.h
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
[mirror_ubuntu-zesty-kernel.git] / include / linux / skbuff.h
CommitLineData
1da177e4
LT
1/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
1da177e4 17#include <linux/kernel.h>
fe55f6d5 18#include <linux/kmemcheck.h>
1da177e4
LT
19#include <linux/compiler.h>
20#include <linux/time.h>
187f1882 21#include <linux/bug.h>
1da177e4 22#include <linux/cache.h>
56b17425 23#include <linux/rbtree.h>
51f3d02b 24#include <linux/socket.h>
1da177e4 25
60063497 26#include <linux/atomic.h>
1da177e4
LT
27#include <asm/types.h>
28#include <linux/spinlock.h>
1da177e4 29#include <linux/net.h>
3fc7e8a6 30#include <linux/textsearch.h>
1da177e4 31#include <net/checksum.h>
a80958f4 32#include <linux/rcupdate.h>
b7aa0bf7 33#include <linux/hrtimer.h>
131ea667 34#include <linux/dma-mapping.h>
c8f44aff 35#include <linux/netdev_features.h>
363ec392 36#include <linux/sched.h>
5203cd28 37#include <net/flow_keys.h>
1da177e4 38
78ea85f1
DB
39/* A. Checksumming of received packets by device.
40 *
41 * CHECKSUM_NONE:
42 *
43 * Device failed to checksum this packet e.g. due to lack of capabilities.
44 * The packet contains full (though not verified) checksum in packet but
45 * not in skb->csum. Thus, skb->csum is undefined in this case.
46 *
47 * CHECKSUM_UNNECESSARY:
48 *
49 * The hardware you're dealing with doesn't calculate the full checksum
50 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
51 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
52 * if their checksums are okay. skb->csum is still undefined in this case
53 * though. It is a bad option, but, unfortunately, nowadays most vendors do
54 * this. Apparently with the secret goal to sell you new devices, when you
55 * will add new protocol to your host, f.e. IPv6 8)
56 *
57 * CHECKSUM_UNNECESSARY is applicable to following protocols:
58 * TCP: IPv6 and IPv4.
59 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
60 * zero UDP checksum for either IPv4 or IPv6, the networking stack
61 * may perform further validation in this case.
62 * GRE: only if the checksum is present in the header.
63 * SCTP: indicates the CRC in SCTP header has been validated.
64 *
65 * skb->csum_level indicates the number of consecutive checksums found in
66 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
67 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
68 * and a device is able to verify the checksums for UDP (possibly zero),
69 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
70 * two. If the device were only able to verify the UDP checksum and not
71 * GRE, either because it doesn't support GRE checksum of because GRE
72 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
73 * not considered in this case).
78ea85f1
DB
74 *
75 * CHECKSUM_COMPLETE:
76 *
77 * This is the most generic way. The device supplied checksum of the _whole_
78 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
79 * hardware doesn't need to parse L3/L4 headers to implement this.
80 *
81 * Note: Even if device supports only some protocols, but is able to produce
82 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
83 *
84 * CHECKSUM_PARTIAL:
85 *
86 * This is identical to the case for output below. This may occur on a packet
87 * received directly from another Linux OS, e.g., a virtualized Linux kernel
88 * on the same host. The packet can be treated in the same way as
89 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
90 * checksum must be filled in by the OS or the hardware.
91 *
92 * B. Checksumming on output.
93 *
94 * CHECKSUM_NONE:
95 *
96 * The skb was already checksummed by the protocol, or a checksum is not
97 * required.
98 *
99 * CHECKSUM_PARTIAL:
100 *
101 * The device is required to checksum the packet as seen by hard_start_xmit()
102 * from skb->csum_start up to the end, and to record/write the checksum at
103 * offset skb->csum_start + skb->csum_offset.
104 *
105 * The device must show its capabilities in dev->features, set up at device
106 * setup time, e.g. netdev_features.h:
107 *
108 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
109 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
110 * IPv4. Sigh. Vendors like this way for an unknown reason.
111 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
112 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
113 * NETIF_F_... - Well, you get the picture.
114 *
115 * CHECKSUM_UNNECESSARY:
116 *
117 * Normally, the device will do per protocol specific checksumming. Protocol
118 * implementations that do not want the NIC to perform the checksum
119 * calculation should use this flag in their outgoing skbs.
120 *
121 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
122 * offload. Correspondingly, the FCoE protocol driver
123 * stack should use CHECKSUM_UNNECESSARY.
124 *
125 * Any questions? No questions, good. --ANK
126 */
127
60476372 128/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
129#define CHECKSUM_NONE 0
130#define CHECKSUM_UNNECESSARY 1
131#define CHECKSUM_COMPLETE 2
132#define CHECKSUM_PARTIAL 3
1da177e4 133
77cffe23
TH
134/* Maximum value in skb->csum_level */
135#define SKB_MAX_CSUM_LEVEL 3
136
0bec8c88 137#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 138#define SKB_WITH_OVERHEAD(X) \
deea84b0 139 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
140#define SKB_MAX_ORDER(X, ORDER) \
141 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
142#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
143#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
144
87fb4b7b
ED
145/* return minimum truesize of one skb containing X bytes of data */
146#define SKB_TRUESIZE(X) ((X) + \
147 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
148 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
149
1da177e4 150struct net_device;
716ea3a7 151struct scatterlist;
9c55e01c 152struct pipe_inode_info;
1da177e4 153
5f79e0f9 154#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
155struct nf_conntrack {
156 atomic_t use;
1da177e4 157};
5f79e0f9 158#endif
1da177e4 159
34666d46 160#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 161struct nf_bridge_info {
bf1ac5ca
ED
162 atomic_t use;
163 unsigned int mask;
164 struct net_device *physindev;
165 struct net_device *physoutdev;
166 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
167};
168#endif
169
1da177e4
LT
170struct sk_buff_head {
171 /* These two members must be first. */
172 struct sk_buff *next;
173 struct sk_buff *prev;
174
175 __u32 qlen;
176 spinlock_t lock;
177};
178
179struct sk_buff;
180
9d4dde52
IC
181/* To allow 64K frame to be packed as single skb without frag_list we
182 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
183 * buffers which do not start on a page boundary.
184 *
185 * Since GRO uses frags we allocate at least 16 regardless of page
186 * size.
a715dea3 187 */
9d4dde52 188#if (65536/PAGE_SIZE + 1) < 16
eec00954 189#define MAX_SKB_FRAGS 16UL
a715dea3 190#else
9d4dde52 191#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 192#endif
1da177e4
LT
193
194typedef struct skb_frag_struct skb_frag_t;
195
196struct skb_frag_struct {
a8605c60
IC
197 struct {
198 struct page *p;
199 } page;
cb4dfe56 200#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
201 __u32 page_offset;
202 __u32 size;
cb4dfe56
ED
203#else
204 __u16 page_offset;
205 __u16 size;
206#endif
1da177e4
LT
207};
208
9e903e08
ED
209static inline unsigned int skb_frag_size(const skb_frag_t *frag)
210{
211 return frag->size;
212}
213
214static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
215{
216 frag->size = size;
217}
218
219static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
220{
221 frag->size += delta;
222}
223
224static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
225{
226 frag->size -= delta;
227}
228
ac45f602
PO
229#define HAVE_HW_TIME_STAMP
230
231/**
d3a21be8 232 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
233 * @hwtstamp: hardware time stamp transformed into duration
234 * since arbitrary point in time
ac45f602
PO
235 *
236 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 237 * skb->tstamp.
ac45f602
PO
238 *
239 * hwtstamps can only be compared against other hwtstamps from
240 * the same device.
241 *
242 * This structure is attached to packets as part of the
243 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
244 */
245struct skb_shared_hwtstamps {
246 ktime_t hwtstamp;
ac45f602
PO
247};
248
2244d07b
OH
249/* Definitions for tx_flags in struct skb_shared_info */
250enum {
251 /* generate hardware time stamp */
252 SKBTX_HW_TSTAMP = 1 << 0,
253
e7fd2885 254 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
255 SKBTX_SW_TSTAMP = 1 << 1,
256
257 /* device driver is going to provide hardware time stamp */
258 SKBTX_IN_PROGRESS = 1 << 2,
259
a6686f2f 260 /* device driver supports TX zero-copy buffers */
62b1a8ab 261 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
262
263 /* generate wifi status information (where possible) */
62b1a8ab 264 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
265
266 /* This indicates at least one fragment might be overwritten
267 * (as in vmsplice(), sendfile() ...)
268 * If we need to compute a TX checksum, we'll need to copy
269 * all frags to avoid possible bad checksum
270 */
271 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
272
273 /* generate software time stamp when entering packet scheduling */
274 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
275
276 /* generate software timestamp on peer data acknowledgment */
277 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
278};
279
e1c8a607
WB
280#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
281 SKBTX_SCHED_TSTAMP | \
282 SKBTX_ACK_TSTAMP)
f24b9be5
WB
283#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
284
a6686f2f
SM
285/*
286 * The callback notifies userspace to release buffers when skb DMA is done in
287 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
288 * The zerocopy_success argument is true if zero copy transmit occurred,
289 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
290 * The ctx field is used to track device context.
291 * The desc field is used to track userspace buffer index.
a6686f2f
SM
292 */
293struct ubuf_info {
e19d6763 294 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 295 void *ctx;
a6686f2f 296 unsigned long desc;
ac45f602
PO
297};
298
1da177e4
LT
299/* This data is invariant across clones and lives at
300 * the end of the header data, ie. at skb->end.
301 */
302struct skb_shared_info {
9f42f126
IC
303 unsigned char nr_frags;
304 __u8 tx_flags;
7967168c
HX
305 unsigned short gso_size;
306 /* Warning: this field is not always filled in (UFO)! */
307 unsigned short gso_segs;
308 unsigned short gso_type;
1da177e4 309 struct sk_buff *frag_list;
ac45f602 310 struct skb_shared_hwtstamps hwtstamps;
09c2d251 311 u32 tskey;
9f42f126 312 __be32 ip6_frag_id;
ec7d2f2c
ED
313
314 /*
315 * Warning : all fields before dataref are cleared in __alloc_skb()
316 */
317 atomic_t dataref;
318
69e3c75f
JB
319 /* Intermediate layers must ensure that destructor_arg
320 * remains valid until skb destructor */
321 void * destructor_arg;
a6686f2f 322
fed66381
ED
323 /* must be last field, see pskb_expand_head() */
324 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
325};
326
327/* We divide dataref into two halves. The higher 16 bits hold references
328 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
329 * the entire skb->data. A clone of a headerless skb holds the length of
330 * the header in skb->hdr_len.
1da177e4
LT
331 *
332 * All users must obey the rule that the skb->data reference count must be
333 * greater than or equal to the payload reference count.
334 *
335 * Holding a reference to the payload part means that the user does not
336 * care about modifications to the header part of skb->data.
337 */
338#define SKB_DATAREF_SHIFT 16
339#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
340
d179cd12
DM
341
342enum {
c8753d55
VS
343 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
344 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
345 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
346 SKB_FCLONE_FREE, /* this companion fclone skb is available */
d179cd12
DM
347};
348
7967168c
HX
349enum {
350 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 351 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
352
353 /* This indicates the skb is from an untrusted source. */
354 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
355
356 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
357 SKB_GSO_TCP_ECN = 1 << 3,
358
359 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
360
361 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
362
363 SKB_GSO_GRE = 1 << 6,
73136267 364
4b28252c 365 SKB_GSO_GRE_CSUM = 1 << 7,
0d89d203 366
4b28252c 367 SKB_GSO_IPIP = 1 << 8,
cb32f511 368
4b28252c 369 SKB_GSO_SIT = 1 << 9,
61c1db7f 370
4b28252c 371 SKB_GSO_UDP_TUNNEL = 1 << 10,
0f4f4ffa
TH
372
373 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c 374
59b93b41 375 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
7967168c
HX
376};
377
2e07fa9c
ACM
378#if BITS_PER_LONG > 32
379#define NET_SKBUFF_DATA_USES_OFFSET 1
380#endif
381
382#ifdef NET_SKBUFF_DATA_USES_OFFSET
383typedef unsigned int sk_buff_data_t;
384#else
385typedef unsigned char *sk_buff_data_t;
386#endif
387
363ec392
ED
388/**
389 * struct skb_mstamp - multi resolution time stamps
390 * @stamp_us: timestamp in us resolution
391 * @stamp_jiffies: timestamp in jiffies
392 */
393struct skb_mstamp {
394 union {
395 u64 v64;
396 struct {
397 u32 stamp_us;
398 u32 stamp_jiffies;
399 };
400 };
401};
402
403/**
404 * skb_mstamp_get - get current timestamp
405 * @cl: place to store timestamps
406 */
407static inline void skb_mstamp_get(struct skb_mstamp *cl)
408{
409 u64 val = local_clock();
410
411 do_div(val, NSEC_PER_USEC);
412 cl->stamp_us = (u32)val;
413 cl->stamp_jiffies = (u32)jiffies;
414}
415
416/**
417 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
418 * @t1: pointer to newest sample
419 * @t0: pointer to oldest sample
420 */
421static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
422 const struct skb_mstamp *t0)
423{
424 s32 delta_us = t1->stamp_us - t0->stamp_us;
425 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
426
427 /* If delta_us is negative, this might be because interval is too big,
428 * or local_clock() drift is too big : fallback using jiffies.
429 */
430 if (delta_us <= 0 ||
431 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
432
433 delta_us = jiffies_to_usecs(delta_jiffies);
434
435 return delta_us;
436}
437
438
1da177e4
LT
439/**
440 * struct sk_buff - socket buffer
441 * @next: Next buffer in list
442 * @prev: Previous buffer in list
363ec392 443 * @tstamp: Time we arrived/left
56b17425 444 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 445 * @sk: Socket we are owned by
1da177e4 446 * @dev: Device we arrived on/are leaving by
d84e0bd7 447 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 448 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 449 * @sp: the security path, used for xfrm
1da177e4
LT
450 * @len: Length of actual data
451 * @data_len: Data length
452 * @mac_len: Length of link layer header
334a8132 453 * @hdr_len: writable header length of cloned skb
663ead3b
HX
454 * @csum: Checksum (must include start/offset pair)
455 * @csum_start: Offset from skb->head where checksumming should start
456 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 457 * @priority: Packet queueing priority
60ff7467 458 * @ignore_df: allow local fragmentation
1da177e4 459 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 460 * @ip_summed: Driver fed us an IP checksum
1da177e4 461 * @nohdr: Payload reference only, must not modify header
d84e0bd7 462 * @nfctinfo: Relationship of this skb to the connection
1da177e4 463 * @pkt_type: Packet class
c83c2486 464 * @fclone: skbuff clone status
c83c2486 465 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
466 * @peeked: this packet has been seen already, so stats have been
467 * done for it, don't do them again
ba9dda3a 468 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
469 * @protocol: Packet protocol from driver
470 * @destructor: Destruct function
471 * @nfct: Associated connection, if any
1da177e4 472 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 473 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
474 * @tc_index: Traffic control index
475 * @tc_verd: traffic control verdict
61b905da 476 * @hash: the packet hash
d84e0bd7 477 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 478 * @xmit_more: More SKBs are pending for this queue
553a5672 479 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 480 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 481 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 482 * ports.
a3b18ddb 483 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
484 * @wifi_acked_valid: wifi_acked was set
485 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 486 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
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 {
363ec392 509 union {
56b17425
ED
510 struct {
511 /* These two members must be first. */
512 struct sk_buff *next;
513 struct sk_buff *prev;
514
515 union {
516 ktime_t tstamp;
517 struct skb_mstamp skb_mstamp;
518 };
519 };
520 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 521 };
da3f5cf1 522 struct sock *sk;
1da177e4 523 struct net_device *dev;
1da177e4 524
1da177e4
LT
525 /*
526 * This is the control buffer. It is free to use for every
527 * layer. Please put your private variables there. If you
528 * want to keep them across layers you have to do a skb_clone()
529 * first. This is owned by whoever has the skb queued ATM.
530 */
da3f5cf1 531 char cb[48] __aligned(8);
1da177e4 532
7fee226a 533 unsigned long _skb_refdst;
b1937227 534 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
535#ifdef CONFIG_XFRM
536 struct sec_path *sp;
b1937227
ED
537#endif
538#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
539 struct nf_conntrack *nfct;
540#endif
85224844 541#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 542 struct nf_bridge_info *nf_bridge;
da3f5cf1 543#endif
1da177e4 544 unsigned int len,
334a8132
PM
545 data_len;
546 __u16 mac_len,
547 hdr_len;
b1937227
ED
548
549 /* Following fields are _not_ copied in __copy_skb_header()
550 * Note that queue_mapping is here mostly to fill a hole.
551 */
fe55f6d5 552 kmemcheck_bitfield_begin(flags1);
b1937227
ED
553 __u16 queue_mapping;
554 __u8 cloned:1,
6869c4d8 555 nohdr:1,
b84f4cc9 556 fclone:2,
a59322be 557 peeked:1,
b1937227
ED
558 head_frag:1,
559 xmit_more:1;
560 /* one bit hole */
fe55f6d5 561 kmemcheck_bitfield_end(flags1);
4031ae6e 562
b1937227
ED
563 /* fields enclosed in headers_start/headers_end are copied
564 * using a single memcpy() in __copy_skb_header()
565 */
ebcf34f3 566 /* private: */
b1937227 567 __u32 headers_start[0];
ebcf34f3 568 /* public: */
4031ae6e 569
233577a2
HFS
570/* if you move pkt_type around you also must adapt those constants */
571#ifdef __BIG_ENDIAN_BITFIELD
572#define PKT_TYPE_MAX (7 << 5)
573#else
574#define PKT_TYPE_MAX 7
1da177e4 575#endif
233577a2 576#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 577
233577a2 578 __u8 __pkt_type_offset[0];
b1937227 579 __u8 pkt_type:3;
c93bdd0e 580 __u8 pfmemalloc:1;
b1937227
ED
581 __u8 ignore_df:1;
582 __u8 nfctinfo:3;
583
584 __u8 nf_trace:1;
585 __u8 ip_summed:2;
3853b584 586 __u8 ooo_okay:1;
61b905da 587 __u8 l4_hash:1;
a3b18ddb 588 __u8 sw_hash:1;
6e3e939f
JB
589 __u8 wifi_acked_valid:1;
590 __u8 wifi_acked:1;
b1937227 591
3bdc0eba 592 __u8 no_fcs:1;
77cffe23 593 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 594 __u8 encapsulation:1;
7e2b10c1 595 __u8 encap_hdr_csum:1;
5d0c2b95 596 __u8 csum_valid:1;
7e3cead5 597 __u8 csum_complete_sw:1;
b1937227
ED
598 __u8 csum_level:2;
599 __u8 csum_bad:1;
fe55f6d5 600
b1937227
ED
601#ifdef CONFIG_IPV6_NDISC_NODETYPE
602 __u8 ndisc_nodetype:2;
603#endif
604 __u8 ipvs_property:1;
8bce6d7d 605 __u8 inner_protocol_type:1;
e585f236
TH
606 __u8 remcsum_offload:1;
607 /* 3 or 5 bit hole */
b1937227
ED
608
609#ifdef CONFIG_NET_SCHED
610 __u16 tc_index; /* traffic control index */
611#ifdef CONFIG_NET_CLS_ACT
612 __u16 tc_verd; /* traffic control verdict */
613#endif
614#endif
fe55f6d5 615
b1937227
ED
616 union {
617 __wsum csum;
618 struct {
619 __u16 csum_start;
620 __u16 csum_offset;
621 };
622 };
623 __u32 priority;
624 int skb_iif;
625 __u32 hash;
626 __be16 vlan_proto;
627 __u16 vlan_tci;
7bced397
DW
628#ifdef CONFIG_NET_RX_BUSY_POLL
629 unsigned int napi_id;
97fc2f08 630#endif
984bc16c
JM
631#ifdef CONFIG_NETWORK_SECMARK
632 __u32 secmark;
633#endif
3b885787
NH
634 union {
635 __u32 mark;
636 __u32 dropcount;
16fad69c 637 __u32 reserved_tailroom;
3b885787 638 };
1da177e4 639
8bce6d7d
TH
640 union {
641 __be16 inner_protocol;
642 __u8 inner_ipproto;
643 };
644
1a37e412
SH
645 __u16 inner_transport_header;
646 __u16 inner_network_header;
647 __u16 inner_mac_header;
b1937227
ED
648
649 __be16 protocol;
1a37e412
SH
650 __u16 transport_header;
651 __u16 network_header;
652 __u16 mac_header;
b1937227 653
ebcf34f3 654 /* private: */
b1937227 655 __u32 headers_end[0];
ebcf34f3 656 /* public: */
b1937227 657
1da177e4 658 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 659 sk_buff_data_t tail;
4305b541 660 sk_buff_data_t end;
1da177e4 661 unsigned char *head,
4305b541 662 *data;
27a884dc
ACM
663 unsigned int truesize;
664 atomic_t users;
1da177e4
LT
665};
666
667#ifdef __KERNEL__
668/*
669 * Handling routines are only of interest to the kernel
670 */
671#include <linux/slab.h>
672
1da177e4 673
c93bdd0e
MG
674#define SKB_ALLOC_FCLONE 0x01
675#define SKB_ALLOC_RX 0x02
676
677/* Returns true if the skb was allocated from PFMEMALLOC reserves */
678static inline bool skb_pfmemalloc(const struct sk_buff *skb)
679{
680 return unlikely(skb->pfmemalloc);
681}
682
7fee226a
ED
683/*
684 * skb might have a dst pointer attached, refcounted or not.
685 * _skb_refdst low order bit is set if refcount was _not_ taken
686 */
687#define SKB_DST_NOREF 1UL
688#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
689
690/**
691 * skb_dst - returns skb dst_entry
692 * @skb: buffer
693 *
694 * Returns skb dst_entry, regardless of reference taken or not.
695 */
adf30907
ED
696static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
697{
7fee226a
ED
698 /* If refdst was not refcounted, check we still are in a
699 * rcu_read_lock section
700 */
701 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
702 !rcu_read_lock_held() &&
703 !rcu_read_lock_bh_held());
704 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
705}
706
7fee226a
ED
707/**
708 * skb_dst_set - sets skb dst
709 * @skb: buffer
710 * @dst: dst entry
711 *
712 * Sets skb dst, assuming a reference was taken on dst and should
713 * be released by skb_dst_drop()
714 */
adf30907
ED
715static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
716{
7fee226a
ED
717 skb->_skb_refdst = (unsigned long)dst;
718}
719
7965bd4d
JP
720void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
721 bool force);
932bc4d7
JA
722
723/**
724 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
725 * @skb: buffer
726 * @dst: dst entry
727 *
728 * Sets skb dst, assuming a reference was not taken on dst.
729 * If dst entry is cached, we do not take reference and dst_release
730 * will be avoided by refdst_drop. If dst entry is not cached, we take
731 * reference, so that last dst_release can destroy the dst immediately.
732 */
733static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
734{
735 __skb_dst_set_noref(skb, dst, false);
736}
737
738/**
739 * skb_dst_set_noref_force - sets skb dst, without taking reference
740 * @skb: buffer
741 * @dst: dst entry
742 *
743 * Sets skb dst, assuming a reference was not taken on dst.
744 * No reference is taken and no dst_release will be called. While for
745 * cached dsts deferred reclaim is a basic feature, for entries that are
746 * not cached it is caller's job to guarantee that last dst_release for
747 * provided dst happens when nobody uses it, eg. after a RCU grace period.
748 */
749static inline void skb_dst_set_noref_force(struct sk_buff *skb,
750 struct dst_entry *dst)
751{
752 __skb_dst_set_noref(skb, dst, true);
753}
7fee226a
ED
754
755/**
25985edc 756 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
757 * @skb: buffer
758 */
759static inline bool skb_dst_is_noref(const struct sk_buff *skb)
760{
761 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
762}
763
511c3f92
ED
764static inline struct rtable *skb_rtable(const struct sk_buff *skb)
765{
adf30907 766 return (struct rtable *)skb_dst(skb);
511c3f92
ED
767}
768
7965bd4d
JP
769void kfree_skb(struct sk_buff *skb);
770void kfree_skb_list(struct sk_buff *segs);
771void skb_tx_error(struct sk_buff *skb);
772void consume_skb(struct sk_buff *skb);
773void __kfree_skb(struct sk_buff *skb);
d7e8883c 774extern struct kmem_cache *skbuff_head_cache;
bad43ca8 775
7965bd4d
JP
776void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
777bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
778 bool *fragstolen, int *delta_truesize);
bad43ca8 779
7965bd4d
JP
780struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
781 int node);
782struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 783static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 784 gfp_t priority)
d179cd12 785{
564824b0 786 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
787}
788
2e4e4410
ED
789struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
790 unsigned long data_len,
791 int max_page_order,
792 int *errcode,
793 gfp_t gfp_mask);
794
d0bf4a9e
ED
795/* Layout of fast clones : [skb1][skb2][fclone_ref] */
796struct sk_buff_fclones {
797 struct sk_buff skb1;
798
799 struct sk_buff skb2;
800
801 atomic_t fclone_ref;
802};
803
804/**
805 * skb_fclone_busy - check if fclone is busy
806 * @skb: buffer
807 *
808 * Returns true is skb is a fast clone, and its clone is not freed.
39bb5e62
ED
809 * Some drivers call skb_orphan() in their ndo_start_xmit(),
810 * so we also check that this didnt happen.
d0bf4a9e 811 */
39bb5e62
ED
812static inline bool skb_fclone_busy(const struct sock *sk,
813 const struct sk_buff *skb)
d0bf4a9e
ED
814{
815 const struct sk_buff_fclones *fclones;
816
817 fclones = container_of(skb, struct sk_buff_fclones, skb1);
818
819 return skb->fclone == SKB_FCLONE_ORIG &&
39bb5e62
ED
820 fclones->skb2.fclone == SKB_FCLONE_CLONE &&
821 fclones->skb2.sk == sk;
d0bf4a9e
ED
822}
823
d179cd12 824static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 825 gfp_t priority)
d179cd12 826{
c93bdd0e 827 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
828}
829
7965bd4d 830struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
831static inline struct sk_buff *alloc_skb_head(gfp_t priority)
832{
833 return __alloc_skb_head(priority, -1);
834}
835
7965bd4d
JP
836struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
837int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
838struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
839struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
840struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
841 gfp_t gfp_mask, bool fclone);
842static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
843 gfp_t gfp_mask)
844{
845 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
846}
7965bd4d
JP
847
848int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
849struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
850 unsigned int headroom);
851struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
852 int newtailroom, gfp_t priority);
25a91d8d
FD
853int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
854 int offset, int len);
7965bd4d
JP
855int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
856 int len);
857int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
858int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 859#define dev_kfree_skb(a) consume_skb(a)
1da177e4 860
7965bd4d
JP
861int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
862 int getfrag(void *from, char *to, int offset,
863 int len, int odd, struct sk_buff *skb),
864 void *from, int length);
e89e9cf5 865
d94d9fee 866struct skb_seq_state {
677e90ed
TG
867 __u32 lower_offset;
868 __u32 upper_offset;
869 __u32 frag_idx;
870 __u32 stepped_offset;
871 struct sk_buff *root_skb;
872 struct sk_buff *cur_skb;
873 __u8 *frag_data;
874};
875
7965bd4d
JP
876void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
877 unsigned int to, struct skb_seq_state *st);
878unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
879 struct skb_seq_state *st);
880void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 881
7965bd4d
JP
882unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
883 unsigned int to, struct ts_config *config,
884 struct ts_state *state);
3fc7e8a6 885
09323cc4
TH
886/*
887 * Packet hash types specify the type of hash in skb_set_hash.
888 *
889 * Hash types refer to the protocol layer addresses which are used to
890 * construct a packet's hash. The hashes are used to differentiate or identify
891 * flows of the protocol layer for the hash type. Hash types are either
892 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
893 *
894 * Properties of hashes:
895 *
896 * 1) Two packets in different flows have different hash values
897 * 2) Two packets in the same flow should have the same hash value
898 *
899 * A hash at a higher layer is considered to be more specific. A driver should
900 * set the most specific hash possible.
901 *
902 * A driver cannot indicate a more specific hash than the layer at which a hash
903 * was computed. For instance an L3 hash cannot be set as an L4 hash.
904 *
905 * A driver may indicate a hash level which is less specific than the
906 * actual layer the hash was computed on. For instance, a hash computed
907 * at L4 may be considered an L3 hash. This should only be done if the
908 * driver can't unambiguously determine that the HW computed the hash at
909 * the higher layer. Note that the "should" in the second property above
910 * permits this.
911 */
912enum pkt_hash_types {
913 PKT_HASH_TYPE_NONE, /* Undefined type */
914 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
915 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
916 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
917};
918
919static inline void
920skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
921{
61b905da 922 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 923 skb->sw_hash = 0;
61b905da 924 skb->hash = hash;
09323cc4
TH
925}
926
3958afa1
TH
927void __skb_get_hash(struct sk_buff *skb);
928static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 929{
a3b18ddb 930 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 931 __skb_get_hash(skb);
bfb564e7 932
61b905da 933 return skb->hash;
bfb564e7
KK
934}
935
57bdf7f4
TH
936static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
937{
61b905da 938 return skb->hash;
57bdf7f4
TH
939}
940
7539fadc
TH
941static inline void skb_clear_hash(struct sk_buff *skb)
942{
61b905da 943 skb->hash = 0;
a3b18ddb 944 skb->sw_hash = 0;
61b905da 945 skb->l4_hash = 0;
7539fadc
TH
946}
947
948static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
949{
61b905da 950 if (!skb->l4_hash)
7539fadc
TH
951 skb_clear_hash(skb);
952}
953
3df7a74e
TH
954static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
955{
61b905da 956 to->hash = from->hash;
a3b18ddb 957 to->sw_hash = from->sw_hash;
61b905da 958 to->l4_hash = from->l4_hash;
3df7a74e
TH
959};
960
4305b541
ACM
961#ifdef NET_SKBUFF_DATA_USES_OFFSET
962static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
963{
964 return skb->head + skb->end;
965}
ec47ea82
AD
966
967static inline unsigned int skb_end_offset(const struct sk_buff *skb)
968{
969 return skb->end;
970}
4305b541
ACM
971#else
972static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
973{
974 return skb->end;
975}
ec47ea82
AD
976
977static inline unsigned int skb_end_offset(const struct sk_buff *skb)
978{
979 return skb->end - skb->head;
980}
4305b541
ACM
981#endif
982
1da177e4 983/* Internal */
4305b541 984#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 985
ac45f602
PO
986static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
987{
988 return &skb_shinfo(skb)->hwtstamps;
989}
990
1da177e4
LT
991/**
992 * skb_queue_empty - check if a queue is empty
993 * @list: queue head
994 *
995 * Returns true if the queue is empty, false otherwise.
996 */
997static inline int skb_queue_empty(const struct sk_buff_head *list)
998{
fd44b93c 999 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1000}
1001
fc7ebb21
DM
1002/**
1003 * skb_queue_is_last - check if skb is the last entry in the queue
1004 * @list: queue head
1005 * @skb: buffer
1006 *
1007 * Returns true if @skb is the last buffer on the list.
1008 */
1009static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1010 const struct sk_buff *skb)
1011{
fd44b93c 1012 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1013}
1014
832d11c5
IJ
1015/**
1016 * skb_queue_is_first - check if skb is the first entry in the queue
1017 * @list: queue head
1018 * @skb: buffer
1019 *
1020 * Returns true if @skb is the first buffer on the list.
1021 */
1022static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1023 const struct sk_buff *skb)
1024{
fd44b93c 1025 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1026}
1027
249c8b42
DM
1028/**
1029 * skb_queue_next - return the next packet in the queue
1030 * @list: queue head
1031 * @skb: current buffer
1032 *
1033 * Return the next packet in @list after @skb. It is only valid to
1034 * call this if skb_queue_is_last() evaluates to false.
1035 */
1036static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1037 const struct sk_buff *skb)
1038{
1039 /* This BUG_ON may seem severe, but if we just return then we
1040 * are going to dereference garbage.
1041 */
1042 BUG_ON(skb_queue_is_last(list, skb));
1043 return skb->next;
1044}
1045
832d11c5
IJ
1046/**
1047 * skb_queue_prev - return the prev packet in the queue
1048 * @list: queue head
1049 * @skb: current buffer
1050 *
1051 * Return the prev packet in @list before @skb. It is only valid to
1052 * call this if skb_queue_is_first() evaluates to false.
1053 */
1054static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1055 const struct sk_buff *skb)
1056{
1057 /* This BUG_ON may seem severe, but if we just return then we
1058 * are going to dereference garbage.
1059 */
1060 BUG_ON(skb_queue_is_first(list, skb));
1061 return skb->prev;
1062}
1063
1da177e4
LT
1064/**
1065 * skb_get - reference buffer
1066 * @skb: buffer to reference
1067 *
1068 * Makes another reference to a socket buffer and returns a pointer
1069 * to the buffer.
1070 */
1071static inline struct sk_buff *skb_get(struct sk_buff *skb)
1072{
1073 atomic_inc(&skb->users);
1074 return skb;
1075}
1076
1077/*
1078 * If users == 1, we are the only owner and are can avoid redundant
1079 * atomic change.
1080 */
1081
1da177e4
LT
1082/**
1083 * skb_cloned - is the buffer a clone
1084 * @skb: buffer to check
1085 *
1086 * Returns true if the buffer was generated with skb_clone() and is
1087 * one of multiple shared copies of the buffer. Cloned buffers are
1088 * shared data so must not be written to under normal circumstances.
1089 */
1090static inline int skb_cloned(const struct sk_buff *skb)
1091{
1092 return skb->cloned &&
1093 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1094}
1095
14bbd6a5
PS
1096static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1097{
1098 might_sleep_if(pri & __GFP_WAIT);
1099
1100 if (skb_cloned(skb))
1101 return pskb_expand_head(skb, 0, 0, pri);
1102
1103 return 0;
1104}
1105
1da177e4
LT
1106/**
1107 * skb_header_cloned - is the header a clone
1108 * @skb: buffer to check
1109 *
1110 * Returns true if modifying the header part of the buffer requires
1111 * the data to be copied.
1112 */
1113static inline int skb_header_cloned(const struct sk_buff *skb)
1114{
1115 int dataref;
1116
1117 if (!skb->cloned)
1118 return 0;
1119
1120 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1121 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1122 return dataref != 1;
1123}
1124
1125/**
1126 * skb_header_release - release reference to header
1127 * @skb: buffer to operate on
1128 *
1129 * Drop a reference to the header part of the buffer. This is done
1130 * by acquiring a payload reference. You must not read from the header
1131 * part of skb->data after this.
f4a775d1 1132 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1133 */
1134static inline void skb_header_release(struct sk_buff *skb)
1135{
1136 BUG_ON(skb->nohdr);
1137 skb->nohdr = 1;
1138 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1139}
1140
f4a775d1
ED
1141/**
1142 * __skb_header_release - release reference to header
1143 * @skb: buffer to operate on
1144 *
1145 * Variant of skb_header_release() assuming skb is private to caller.
1146 * We can avoid one atomic operation.
1147 */
1148static inline void __skb_header_release(struct sk_buff *skb)
1149{
1150 skb->nohdr = 1;
1151 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1152}
1153
1154
1da177e4
LT
1155/**
1156 * skb_shared - is the buffer shared
1157 * @skb: buffer to check
1158 *
1159 * Returns true if more than one person has a reference to this
1160 * buffer.
1161 */
1162static inline int skb_shared(const struct sk_buff *skb)
1163{
1164 return atomic_read(&skb->users) != 1;
1165}
1166
1167/**
1168 * skb_share_check - check if buffer is shared and if so clone it
1169 * @skb: buffer to check
1170 * @pri: priority for memory allocation
1171 *
1172 * If the buffer is shared the buffer is cloned and the old copy
1173 * drops a reference. A new clone with a single reference is returned.
1174 * If the buffer is not shared the original buffer is returned. When
1175 * being called from interrupt status or with spinlocks held pri must
1176 * be GFP_ATOMIC.
1177 *
1178 * NULL is returned on a memory allocation failure.
1179 */
47061bc4 1180static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1181{
1182 might_sleep_if(pri & __GFP_WAIT);
1183 if (skb_shared(skb)) {
1184 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1185
1186 if (likely(nskb))
1187 consume_skb(skb);
1188 else
1189 kfree_skb(skb);
1da177e4
LT
1190 skb = nskb;
1191 }
1192 return skb;
1193}
1194
1195/*
1196 * Copy shared buffers into a new sk_buff. We effectively do COW on
1197 * packets to handle cases where we have a local reader and forward
1198 * and a couple of other messy ones. The normal one is tcpdumping
1199 * a packet thats being forwarded.
1200 */
1201
1202/**
1203 * skb_unshare - make a copy of a shared buffer
1204 * @skb: buffer to check
1205 * @pri: priority for memory allocation
1206 *
1207 * If the socket buffer is a clone then this function creates a new
1208 * copy of the data, drops a reference count on the old copy and returns
1209 * the new copy with the reference count at 1. If the buffer is not a clone
1210 * the original buffer is returned. When called with a spinlock held or
1211 * from interrupt state @pri must be %GFP_ATOMIC
1212 *
1213 * %NULL is returned on a memory allocation failure.
1214 */
e2bf521d 1215static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1216 gfp_t pri)
1da177e4
LT
1217{
1218 might_sleep_if(pri & __GFP_WAIT);
1219 if (skb_cloned(skb)) {
1220 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1221
1222 /* Free our shared copy */
1223 if (likely(nskb))
1224 consume_skb(skb);
1225 else
1226 kfree_skb(skb);
1da177e4
LT
1227 skb = nskb;
1228 }
1229 return skb;
1230}
1231
1232/**
1a5778aa 1233 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1234 * @list_: list to peek at
1235 *
1236 * Peek an &sk_buff. Unlike most other operations you _MUST_
1237 * be careful with this one. A peek leaves the buffer on the
1238 * list and someone else may run off with it. You must hold
1239 * the appropriate locks or have a private queue to do this.
1240 *
1241 * Returns %NULL for an empty list or a pointer to the head element.
1242 * The reference count is not incremented and the reference is therefore
1243 * volatile. Use with caution.
1244 */
05bdd2f1 1245static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1246{
18d07000
ED
1247 struct sk_buff *skb = list_->next;
1248
1249 if (skb == (struct sk_buff *)list_)
1250 skb = NULL;
1251 return skb;
1da177e4
LT
1252}
1253
da5ef6e5
PE
1254/**
1255 * skb_peek_next - peek skb following the given one from a queue
1256 * @skb: skb to start from
1257 * @list_: list to peek at
1258 *
1259 * Returns %NULL when the end of the list is met or a pointer to the
1260 * next element. The reference count is not incremented and the
1261 * reference is therefore volatile. Use with caution.
1262 */
1263static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1264 const struct sk_buff_head *list_)
1265{
1266 struct sk_buff *next = skb->next;
18d07000 1267
da5ef6e5
PE
1268 if (next == (struct sk_buff *)list_)
1269 next = NULL;
1270 return next;
1271}
1272
1da177e4 1273/**
1a5778aa 1274 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1275 * @list_: list to peek at
1276 *
1277 * Peek an &sk_buff. Unlike most other operations you _MUST_
1278 * be careful with this one. A peek leaves the buffer on the
1279 * list and someone else may run off with it. You must hold
1280 * the appropriate locks or have a private queue to do this.
1281 *
1282 * Returns %NULL for an empty list or a pointer to the tail element.
1283 * The reference count is not incremented and the reference is therefore
1284 * volatile. Use with caution.
1285 */
05bdd2f1 1286static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1287{
18d07000
ED
1288 struct sk_buff *skb = list_->prev;
1289
1290 if (skb == (struct sk_buff *)list_)
1291 skb = NULL;
1292 return skb;
1293
1da177e4
LT
1294}
1295
1296/**
1297 * skb_queue_len - get queue length
1298 * @list_: list to measure
1299 *
1300 * Return the length of an &sk_buff queue.
1301 */
1302static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1303{
1304 return list_->qlen;
1305}
1306
67fed459
DM
1307/**
1308 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1309 * @list: queue to initialize
1310 *
1311 * This initializes only the list and queue length aspects of
1312 * an sk_buff_head object. This allows to initialize the list
1313 * aspects of an sk_buff_head without reinitializing things like
1314 * the spinlock. It can also be used for on-stack sk_buff_head
1315 * objects where the spinlock is known to not be used.
1316 */
1317static inline void __skb_queue_head_init(struct sk_buff_head *list)
1318{
1319 list->prev = list->next = (struct sk_buff *)list;
1320 list->qlen = 0;
1321}
1322
76f10ad0
AV
1323/*
1324 * This function creates a split out lock class for each invocation;
1325 * this is needed for now since a whole lot of users of the skb-queue
1326 * infrastructure in drivers have different locking usage (in hardirq)
1327 * than the networking core (in softirq only). In the long run either the
1328 * network layer or drivers should need annotation to consolidate the
1329 * main types of usage into 3 classes.
1330 */
1da177e4
LT
1331static inline void skb_queue_head_init(struct sk_buff_head *list)
1332{
1333 spin_lock_init(&list->lock);
67fed459 1334 __skb_queue_head_init(list);
1da177e4
LT
1335}
1336
c2ecba71
PE
1337static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1338 struct lock_class_key *class)
1339{
1340 skb_queue_head_init(list);
1341 lockdep_set_class(&list->lock, class);
1342}
1343
1da177e4 1344/*
bf299275 1345 * Insert an sk_buff on a list.
1da177e4
LT
1346 *
1347 * The "__skb_xxxx()" functions are the non-atomic ones that
1348 * can only be called with interrupts disabled.
1349 */
7965bd4d
JP
1350void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1351 struct sk_buff_head *list);
bf299275
GR
1352static inline void __skb_insert(struct sk_buff *newsk,
1353 struct sk_buff *prev, struct sk_buff *next,
1354 struct sk_buff_head *list)
1355{
1356 newsk->next = next;
1357 newsk->prev = prev;
1358 next->prev = prev->next = newsk;
1359 list->qlen++;
1360}
1da177e4 1361
67fed459
DM
1362static inline void __skb_queue_splice(const struct sk_buff_head *list,
1363 struct sk_buff *prev,
1364 struct sk_buff *next)
1365{
1366 struct sk_buff *first = list->next;
1367 struct sk_buff *last = list->prev;
1368
1369 first->prev = prev;
1370 prev->next = first;
1371
1372 last->next = next;
1373 next->prev = last;
1374}
1375
1376/**
1377 * skb_queue_splice - join two skb lists, this is designed for stacks
1378 * @list: the new list to add
1379 * @head: the place to add it in the first list
1380 */
1381static inline void skb_queue_splice(const struct sk_buff_head *list,
1382 struct sk_buff_head *head)
1383{
1384 if (!skb_queue_empty(list)) {
1385 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1386 head->qlen += list->qlen;
67fed459
DM
1387 }
1388}
1389
1390/**
d9619496 1391 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1392 * @list: the new list to add
1393 * @head: the place to add it in the first list
1394 *
1395 * The list at @list is reinitialised
1396 */
1397static inline void skb_queue_splice_init(struct sk_buff_head *list,
1398 struct sk_buff_head *head)
1399{
1400 if (!skb_queue_empty(list)) {
1401 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1402 head->qlen += list->qlen;
67fed459
DM
1403 __skb_queue_head_init(list);
1404 }
1405}
1406
1407/**
1408 * skb_queue_splice_tail - join two skb lists, each list being a queue
1409 * @list: the new list to add
1410 * @head: the place to add it in the first list
1411 */
1412static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1413 struct sk_buff_head *head)
1414{
1415 if (!skb_queue_empty(list)) {
1416 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1417 head->qlen += list->qlen;
67fed459
DM
1418 }
1419}
1420
1421/**
d9619496 1422 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1423 * @list: the new list to add
1424 * @head: the place to add it in the first list
1425 *
1426 * Each of the lists is a queue.
1427 * The list at @list is reinitialised
1428 */
1429static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1430 struct sk_buff_head *head)
1431{
1432 if (!skb_queue_empty(list)) {
1433 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1434 head->qlen += list->qlen;
67fed459
DM
1435 __skb_queue_head_init(list);
1436 }
1437}
1438
1da177e4 1439/**
300ce174 1440 * __skb_queue_after - queue a buffer at the list head
1da177e4 1441 * @list: list to use
300ce174 1442 * @prev: place after this buffer
1da177e4
LT
1443 * @newsk: buffer to queue
1444 *
300ce174 1445 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1446 * and you must therefore hold required locks before calling it.
1447 *
1448 * A buffer cannot be placed on two lists at the same time.
1449 */
300ce174
SH
1450static inline void __skb_queue_after(struct sk_buff_head *list,
1451 struct sk_buff *prev,
1452 struct sk_buff *newsk)
1da177e4 1453{
bf299275 1454 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1455}
1456
7965bd4d
JP
1457void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1458 struct sk_buff_head *list);
7de6c033 1459
f5572855
GR
1460static inline void __skb_queue_before(struct sk_buff_head *list,
1461 struct sk_buff *next,
1462 struct sk_buff *newsk)
1463{
1464 __skb_insert(newsk, next->prev, next, list);
1465}
1466
300ce174
SH
1467/**
1468 * __skb_queue_head - queue a buffer at the list head
1469 * @list: list to use
1470 * @newsk: buffer to queue
1471 *
1472 * Queue a buffer at the start of a list. This function takes no locks
1473 * and you must therefore hold required locks before calling it.
1474 *
1475 * A buffer cannot be placed on two lists at the same time.
1476 */
7965bd4d 1477void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1478static inline void __skb_queue_head(struct sk_buff_head *list,
1479 struct sk_buff *newsk)
1480{
1481 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1482}
1483
1da177e4
LT
1484/**
1485 * __skb_queue_tail - queue a buffer at the list tail
1486 * @list: list to use
1487 * @newsk: buffer to queue
1488 *
1489 * Queue a buffer at the end of a list. This function takes no locks
1490 * and you must therefore hold required locks before calling it.
1491 *
1492 * A buffer cannot be placed on two lists at the same time.
1493 */
7965bd4d 1494void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1495static inline void __skb_queue_tail(struct sk_buff_head *list,
1496 struct sk_buff *newsk)
1497{
f5572855 1498 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1499}
1500
1da177e4
LT
1501/*
1502 * remove sk_buff from list. _Must_ be called atomically, and with
1503 * the list known..
1504 */
7965bd4d 1505void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1506static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1507{
1508 struct sk_buff *next, *prev;
1509
1510 list->qlen--;
1511 next = skb->next;
1512 prev = skb->prev;
1513 skb->next = skb->prev = NULL;
1da177e4
LT
1514 next->prev = prev;
1515 prev->next = next;
1516}
1517
f525c06d
GR
1518/**
1519 * __skb_dequeue - remove from the head of the queue
1520 * @list: list to dequeue from
1521 *
1522 * Remove the head of the list. This function does not take any locks
1523 * so must be used with appropriate locks held only. The head item is
1524 * returned or %NULL if the list is empty.
1525 */
7965bd4d 1526struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1527static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1528{
1529 struct sk_buff *skb = skb_peek(list);
1530 if (skb)
1531 __skb_unlink(skb, list);
1532 return skb;
1533}
1da177e4
LT
1534
1535/**
1536 * __skb_dequeue_tail - remove from the tail of the queue
1537 * @list: list to dequeue from
1538 *
1539 * Remove the tail of the list. This function does not take any locks
1540 * so must be used with appropriate locks held only. The tail item is
1541 * returned or %NULL if the list is empty.
1542 */
7965bd4d 1543struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1544static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1545{
1546 struct sk_buff *skb = skb_peek_tail(list);
1547 if (skb)
1548 __skb_unlink(skb, list);
1549 return skb;
1550}
1551
1552
bdcc0924 1553static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1554{
1555 return skb->data_len;
1556}
1557
1558static inline unsigned int skb_headlen(const struct sk_buff *skb)
1559{
1560 return skb->len - skb->data_len;
1561}
1562
1563static inline int skb_pagelen(const struct sk_buff *skb)
1564{
1565 int i, len = 0;
1566
1567 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1568 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1569 return len + skb_headlen(skb);
1570}
1571
131ea667
IC
1572/**
1573 * __skb_fill_page_desc - initialise a paged fragment in an skb
1574 * @skb: buffer containing fragment to be initialised
1575 * @i: paged fragment index to initialise
1576 * @page: the page to use for this fragment
1577 * @off: the offset to the data with @page
1578 * @size: the length of the data
1579 *
1580 * Initialises the @i'th fragment of @skb to point to &size bytes at
1581 * offset @off within @page.
1582 *
1583 * Does not take any additional reference on the fragment.
1584 */
1585static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1586 struct page *page, int off, int size)
1da177e4
LT
1587{
1588 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1589
c48a11c7
MG
1590 /*
1591 * Propagate page->pfmemalloc to the skb if we can. The problem is
1592 * that not all callers have unique ownership of the page. If
1593 * pfmemalloc is set, we check the mapping as a mapping implies
1594 * page->index is set (index and pfmemalloc share space).
1595 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1596 * do not lose pfmemalloc information as the pages would not be
1597 * allocated using __GFP_MEMALLOC.
1598 */
a8605c60 1599 frag->page.p = page;
1da177e4 1600 frag->page_offset = off;
9e903e08 1601 skb_frag_size_set(frag, size);
cca7af38
PE
1602
1603 page = compound_head(page);
1604 if (page->pfmemalloc && !page->mapping)
1605 skb->pfmemalloc = true;
131ea667
IC
1606}
1607
1608/**
1609 * skb_fill_page_desc - initialise a paged fragment in an skb
1610 * @skb: buffer containing fragment to be initialised
1611 * @i: paged fragment index to initialise
1612 * @page: the page to use for this fragment
1613 * @off: the offset to the data with @page
1614 * @size: the length of the data
1615 *
1616 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1617 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1618 * addition updates @skb such that @i is the last fragment.
1619 *
1620 * Does not take any additional reference on the fragment.
1621 */
1622static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1623 struct page *page, int off, int size)
1624{
1625 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1626 skb_shinfo(skb)->nr_frags = i + 1;
1627}
1628
7965bd4d
JP
1629void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1630 int size, unsigned int truesize);
654bed16 1631
f8e617e1
JW
1632void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1633 unsigned int truesize);
1634
1da177e4 1635#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1636#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1637#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1638
27a884dc
ACM
1639#ifdef NET_SKBUFF_DATA_USES_OFFSET
1640static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1641{
1642 return skb->head + skb->tail;
1643}
1644
1645static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1646{
1647 skb->tail = skb->data - skb->head;
1648}
1649
1650static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1651{
1652 skb_reset_tail_pointer(skb);
1653 skb->tail += offset;
1654}
7cc46190 1655
27a884dc
ACM
1656#else /* NET_SKBUFF_DATA_USES_OFFSET */
1657static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1658{
1659 return skb->tail;
1660}
1661
1662static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1663{
1664 skb->tail = skb->data;
1665}
1666
1667static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1668{
1669 skb->tail = skb->data + offset;
1670}
4305b541 1671
27a884dc
ACM
1672#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1673
1da177e4
LT
1674/*
1675 * Add data to an sk_buff
1676 */
0c7ddf36 1677unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1678unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1679static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1680{
27a884dc 1681 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1682 SKB_LINEAR_ASSERT(skb);
1683 skb->tail += len;
1684 skb->len += len;
1685 return tmp;
1686}
1687
7965bd4d 1688unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1689static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1690{
1691 skb->data -= len;
1692 skb->len += len;
1693 return skb->data;
1694}
1695
7965bd4d 1696unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1697static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1698{
1699 skb->len -= len;
1700 BUG_ON(skb->len < skb->data_len);
1701 return skb->data += len;
1702}
1703
47d29646
DM
1704static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1705{
1706 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1707}
1708
7965bd4d 1709unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1710
1711static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1712{
1713 if (len > skb_headlen(skb) &&
987c402a 1714 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1715 return NULL;
1716 skb->len -= len;
1717 return skb->data += len;
1718}
1719
1720static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1721{
1722 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1723}
1724
1725static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1726{
1727 if (likely(len <= skb_headlen(skb)))
1728 return 1;
1729 if (unlikely(len > skb->len))
1730 return 0;
987c402a 1731 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1732}
1733
1734/**
1735 * skb_headroom - bytes at buffer head
1736 * @skb: buffer to check
1737 *
1738 * Return the number of bytes of free space at the head of an &sk_buff.
1739 */
c2636b4d 1740static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1741{
1742 return skb->data - skb->head;
1743}
1744
1745/**
1746 * skb_tailroom - bytes at buffer end
1747 * @skb: buffer to check
1748 *
1749 * Return the number of bytes of free space at the tail of an sk_buff
1750 */
1751static inline int skb_tailroom(const struct sk_buff *skb)
1752{
4305b541 1753 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1754}
1755
a21d4572
ED
1756/**
1757 * skb_availroom - bytes at buffer end
1758 * @skb: buffer to check
1759 *
1760 * Return the number of bytes of free space at the tail of an sk_buff
1761 * allocated by sk_stream_alloc()
1762 */
1763static inline int skb_availroom(const struct sk_buff *skb)
1764{
16fad69c
ED
1765 if (skb_is_nonlinear(skb))
1766 return 0;
1767
1768 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1769}
1770
1da177e4
LT
1771/**
1772 * skb_reserve - adjust headroom
1773 * @skb: buffer to alter
1774 * @len: bytes to move
1775 *
1776 * Increase the headroom of an empty &sk_buff by reducing the tail
1777 * room. This is only allowed for an empty buffer.
1778 */
8243126c 1779static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1780{
1781 skb->data += len;
1782 skb->tail += len;
1783}
1784
8bce6d7d
TH
1785#define ENCAP_TYPE_ETHER 0
1786#define ENCAP_TYPE_IPPROTO 1
1787
1788static inline void skb_set_inner_protocol(struct sk_buff *skb,
1789 __be16 protocol)
1790{
1791 skb->inner_protocol = protocol;
1792 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1793}
1794
1795static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1796 __u8 ipproto)
1797{
1798 skb->inner_ipproto = ipproto;
1799 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1800}
1801
6a674e9c
JG
1802static inline void skb_reset_inner_headers(struct sk_buff *skb)
1803{
aefbd2b3 1804 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1805 skb->inner_network_header = skb->network_header;
1806 skb->inner_transport_header = skb->transport_header;
1807}
1808
0b5c9db1
JP
1809static inline void skb_reset_mac_len(struct sk_buff *skb)
1810{
1811 skb->mac_len = skb->network_header - skb->mac_header;
1812}
1813
6a674e9c
JG
1814static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1815 *skb)
1816{
1817 return skb->head + skb->inner_transport_header;
1818}
1819
1820static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1821{
1822 skb->inner_transport_header = skb->data - skb->head;
1823}
1824
1825static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1826 const int offset)
1827{
1828 skb_reset_inner_transport_header(skb);
1829 skb->inner_transport_header += offset;
1830}
1831
1832static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1833{
1834 return skb->head + skb->inner_network_header;
1835}
1836
1837static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1838{
1839 skb->inner_network_header = skb->data - skb->head;
1840}
1841
1842static inline void skb_set_inner_network_header(struct sk_buff *skb,
1843 const int offset)
1844{
1845 skb_reset_inner_network_header(skb);
1846 skb->inner_network_header += offset;
1847}
1848
aefbd2b3
PS
1849static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1850{
1851 return skb->head + skb->inner_mac_header;
1852}
1853
1854static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1855{
1856 skb->inner_mac_header = skb->data - skb->head;
1857}
1858
1859static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1860 const int offset)
1861{
1862 skb_reset_inner_mac_header(skb);
1863 skb->inner_mac_header += offset;
1864}
fda55eca
ED
1865static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1866{
35d04610 1867 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1868}
1869
9c70220b
ACM
1870static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1871{
2e07fa9c 1872 return skb->head + skb->transport_header;
9c70220b
ACM
1873}
1874
badff6d0
ACM
1875static inline void skb_reset_transport_header(struct sk_buff *skb)
1876{
2e07fa9c 1877 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1878}
1879
967b05f6
ACM
1880static inline void skb_set_transport_header(struct sk_buff *skb,
1881 const int offset)
1882{
2e07fa9c
ACM
1883 skb_reset_transport_header(skb);
1884 skb->transport_header += offset;
ea2ae17d
ACM
1885}
1886
d56f90a7
ACM
1887static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1888{
2e07fa9c 1889 return skb->head + skb->network_header;
d56f90a7
ACM
1890}
1891
c1d2bbe1
ACM
1892static inline void skb_reset_network_header(struct sk_buff *skb)
1893{
2e07fa9c 1894 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1895}
1896
c14d2450
ACM
1897static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1898{
2e07fa9c
ACM
1899 skb_reset_network_header(skb);
1900 skb->network_header += offset;
c14d2450
ACM
1901}
1902
2e07fa9c 1903static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1904{
2e07fa9c 1905 return skb->head + skb->mac_header;
bbe735e4
ACM
1906}
1907
2e07fa9c 1908static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1909{
35d04610 1910 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1911}
1912
1913static inline void skb_reset_mac_header(struct sk_buff *skb)
1914{
1915 skb->mac_header = skb->data - skb->head;
1916}
1917
1918static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1919{
1920 skb_reset_mac_header(skb);
1921 skb->mac_header += offset;
1922}
1923
0e3da5bb
TT
1924static inline void skb_pop_mac_header(struct sk_buff *skb)
1925{
1926 skb->mac_header = skb->network_header;
1927}
1928
fbbdb8f0
YX
1929static inline void skb_probe_transport_header(struct sk_buff *skb,
1930 const int offset_hint)
1931{
1932 struct flow_keys keys;
1933
1934 if (skb_transport_header_was_set(skb))
1935 return;
1936 else if (skb_flow_dissect(skb, &keys))
1937 skb_set_transport_header(skb, keys.thoff);
1938 else
1939 skb_set_transport_header(skb, offset_hint);
1940}
1941
03606895
ED
1942static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1943{
1944 if (skb_mac_header_was_set(skb)) {
1945 const unsigned char *old_mac = skb_mac_header(skb);
1946
1947 skb_set_mac_header(skb, -skb->mac_len);
1948 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1949 }
1950}
1951
04fb451e
MM
1952static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1953{
1954 return skb->csum_start - skb_headroom(skb);
1955}
1956
2e07fa9c
ACM
1957static inline int skb_transport_offset(const struct sk_buff *skb)
1958{
1959 return skb_transport_header(skb) - skb->data;
1960}
1961
1962static inline u32 skb_network_header_len(const struct sk_buff *skb)
1963{
1964 return skb->transport_header - skb->network_header;
1965}
1966
6a674e9c
JG
1967static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1968{
1969 return skb->inner_transport_header - skb->inner_network_header;
1970}
1971
2e07fa9c
ACM
1972static inline int skb_network_offset(const struct sk_buff *skb)
1973{
1974 return skb_network_header(skb) - skb->data;
1975}
48d49d0c 1976
6a674e9c
JG
1977static inline int skb_inner_network_offset(const struct sk_buff *skb)
1978{
1979 return skb_inner_network_header(skb) - skb->data;
1980}
1981
f9599ce1
CG
1982static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1983{
1984 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1985}
1986
1da177e4
LT
1987/*
1988 * CPUs often take a performance hit when accessing unaligned memory
1989 * locations. The actual performance hit varies, it can be small if the
1990 * hardware handles it or large if we have to take an exception and fix it
1991 * in software.
1992 *
1993 * Since an ethernet header is 14 bytes network drivers often end up with
1994 * the IP header at an unaligned offset. The IP header can be aligned by
1995 * shifting the start of the packet by 2 bytes. Drivers should do this
1996 * with:
1997 *
8660c124 1998 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1999 *
2000 * The downside to this alignment of the IP header is that the DMA is now
2001 * unaligned. On some architectures the cost of an unaligned DMA is high
2002 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2003 *
1da177e4
LT
2004 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2005 * to be overridden.
2006 */
2007#ifndef NET_IP_ALIGN
2008#define NET_IP_ALIGN 2
2009#endif
2010
025be81e
AB
2011/*
2012 * The networking layer reserves some headroom in skb data (via
2013 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2014 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2015 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2016 *
2017 * Unfortunately this headroom changes the DMA alignment of the resulting
2018 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2019 * on some architectures. An architecture can override this value,
2020 * perhaps setting it to a cacheline in size (since that will maintain
2021 * cacheline alignment of the DMA). It must be a power of 2.
2022 *
d6301d3d 2023 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2024 * headroom, you should not reduce this.
5933dd2f
ED
2025 *
2026 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2027 * to reduce average number of cache lines per packet.
2028 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2029 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2030 */
2031#ifndef NET_SKB_PAD
5933dd2f 2032#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2033#endif
2034
7965bd4d 2035int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2036
2037static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2038{
c4264f27 2039 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2040 WARN_ON(1);
2041 return;
2042 }
27a884dc
ACM
2043 skb->len = len;
2044 skb_set_tail_pointer(skb, len);
1da177e4
LT
2045}
2046
7965bd4d 2047void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2048
2049static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2050{
3cc0e873
HX
2051 if (skb->data_len)
2052 return ___pskb_trim(skb, len);
2053 __skb_trim(skb, len);
2054 return 0;
1da177e4
LT
2055}
2056
2057static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2058{
2059 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2060}
2061
e9fa4f7b
HX
2062/**
2063 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2064 * @skb: buffer to alter
2065 * @len: new length
2066 *
2067 * This is identical to pskb_trim except that the caller knows that
2068 * the skb is not cloned so we should never get an error due to out-
2069 * of-memory.
2070 */
2071static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2072{
2073 int err = pskb_trim(skb, len);
2074 BUG_ON(err);
2075}
2076
1da177e4
LT
2077/**
2078 * skb_orphan - orphan a buffer
2079 * @skb: buffer to orphan
2080 *
2081 * If a buffer currently has an owner then we call the owner's
2082 * destructor function and make the @skb unowned. The buffer continues
2083 * to exist but is no longer charged to its former owner.
2084 */
2085static inline void skb_orphan(struct sk_buff *skb)
2086{
c34a7612 2087 if (skb->destructor) {
1da177e4 2088 skb->destructor(skb);
c34a7612
ED
2089 skb->destructor = NULL;
2090 skb->sk = NULL;
376c7311
ED
2091 } else {
2092 BUG_ON(skb->sk);
c34a7612 2093 }
1da177e4
LT
2094}
2095
a353e0ce
MT
2096/**
2097 * skb_orphan_frags - orphan the frags contained in a buffer
2098 * @skb: buffer to orphan frags from
2099 * @gfp_mask: allocation mask for replacement pages
2100 *
2101 * For each frag in the SKB which needs a destructor (i.e. has an
2102 * owner) create a copy of that frag and release the original
2103 * page by calling the destructor.
2104 */
2105static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2106{
2107 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2108 return 0;
2109 return skb_copy_ubufs(skb, gfp_mask);
2110}
2111
1da177e4
LT
2112/**
2113 * __skb_queue_purge - empty a list
2114 * @list: list to empty
2115 *
2116 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2117 * the list and one reference dropped. This function does not take the
2118 * list lock and the caller must hold the relevant locks to use it.
2119 */
7965bd4d 2120void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2121static inline void __skb_queue_purge(struct sk_buff_head *list)
2122{
2123 struct sk_buff *skb;
2124 while ((skb = __skb_dequeue(list)) != NULL)
2125 kfree_skb(skb);
2126}
2127
e5e67305
AD
2128#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2129#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2130#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2131
7965bd4d 2132void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2133
7965bd4d
JP
2134struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2135 gfp_t gfp_mask);
8af27456
CH
2136
2137/**
2138 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2139 * @dev: network device to receive on
2140 * @length: length to allocate
2141 *
2142 * Allocate a new &sk_buff and assign it a usage count of one. The
2143 * buffer has unspecified headroom built in. Users should allocate
2144 * the headroom they think they need without accounting for the
2145 * built in space. The built in space is used for optimisations.
2146 *
2147 * %NULL is returned if there is no free memory. Although this function
2148 * allocates memory it can be called from an interrupt.
2149 */
2150static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2151 unsigned int length)
8af27456
CH
2152{
2153 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2154}
2155
6f532612
ED
2156/* legacy helper around __netdev_alloc_skb() */
2157static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2158 gfp_t gfp_mask)
2159{
2160 return __netdev_alloc_skb(NULL, length, gfp_mask);
2161}
2162
2163/* legacy helper around netdev_alloc_skb() */
2164static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2165{
2166 return netdev_alloc_skb(NULL, length);
2167}
2168
2169
4915a0de
ED
2170static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2171 unsigned int length, gfp_t gfp)
61321bbd 2172{
4915a0de 2173 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2174
2175 if (NET_IP_ALIGN && skb)
2176 skb_reserve(skb, NET_IP_ALIGN);
2177 return skb;
2178}
2179
4915a0de
ED
2180static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2181 unsigned int length)
2182{
2183 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2184}
2185
bc6fc9fa
FF
2186/**
2187 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
0614002b
MG
2188 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2189 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2190 * @order: size of the allocation
2191 *
2192 * Allocate a new page.
2193 *
2194 * %NULL is returned if there is no free memory.
2195*/
2196static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2197 struct sk_buff *skb,
2198 unsigned int order)
2199{
2200 struct page *page;
2201
2202 gfp_mask |= __GFP_COLD;
2203
2204 if (!(gfp_mask & __GFP_NOMEMALLOC))
2205 gfp_mask |= __GFP_MEMALLOC;
2206
2207 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2208 if (skb && page && page->pfmemalloc)
2209 skb->pfmemalloc = true;
2210
2211 return page;
2212}
2213
2214/**
2215 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2216 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2217 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2218 *
2219 * Allocate a new page.
2220 *
2221 * %NULL is returned if there is no free memory.
2222 */
2223static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2224 struct sk_buff *skb)
2225{
2226 return __skb_alloc_pages(gfp_mask, skb, 0);
2227}
2228
2229/**
2230 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2231 * @page: The page that was allocated from skb_alloc_page
2232 * @skb: The skb that may need pfmemalloc set
2233 */
2234static inline void skb_propagate_pfmemalloc(struct page *page,
2235 struct sk_buff *skb)
2236{
2237 if (page && page->pfmemalloc)
2238 skb->pfmemalloc = true;
2239}
2240
131ea667 2241/**
e227867f 2242 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2243 * @frag: the paged fragment
2244 *
2245 * Returns the &struct page associated with @frag.
2246 */
2247static inline struct page *skb_frag_page(const skb_frag_t *frag)
2248{
a8605c60 2249 return frag->page.p;
131ea667
IC
2250}
2251
2252/**
2253 * __skb_frag_ref - take an addition reference on a paged fragment.
2254 * @frag: the paged fragment
2255 *
2256 * Takes an additional reference on the paged fragment @frag.
2257 */
2258static inline void __skb_frag_ref(skb_frag_t *frag)
2259{
2260 get_page(skb_frag_page(frag));
2261}
2262
2263/**
2264 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2265 * @skb: the buffer
2266 * @f: the fragment offset.
2267 *
2268 * Takes an additional reference on the @f'th paged fragment of @skb.
2269 */
2270static inline void skb_frag_ref(struct sk_buff *skb, int f)
2271{
2272 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2273}
2274
2275/**
2276 * __skb_frag_unref - release a reference on a paged fragment.
2277 * @frag: the paged fragment
2278 *
2279 * Releases a reference on the paged fragment @frag.
2280 */
2281static inline void __skb_frag_unref(skb_frag_t *frag)
2282{
2283 put_page(skb_frag_page(frag));
2284}
2285
2286/**
2287 * skb_frag_unref - release a reference on a paged fragment of an skb.
2288 * @skb: the buffer
2289 * @f: the fragment offset
2290 *
2291 * Releases a reference on the @f'th paged fragment of @skb.
2292 */
2293static inline void skb_frag_unref(struct sk_buff *skb, int f)
2294{
2295 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2296}
2297
2298/**
2299 * skb_frag_address - gets the address of the data contained in a paged fragment
2300 * @frag: the paged fragment buffer
2301 *
2302 * Returns the address of the data within @frag. The page must already
2303 * be mapped.
2304 */
2305static inline void *skb_frag_address(const skb_frag_t *frag)
2306{
2307 return page_address(skb_frag_page(frag)) + frag->page_offset;
2308}
2309
2310/**
2311 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2312 * @frag: the paged fragment buffer
2313 *
2314 * Returns the address of the data within @frag. Checks that the page
2315 * is mapped and returns %NULL otherwise.
2316 */
2317static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2318{
2319 void *ptr = page_address(skb_frag_page(frag));
2320 if (unlikely(!ptr))
2321 return NULL;
2322
2323 return ptr + frag->page_offset;
2324}
2325
2326/**
2327 * __skb_frag_set_page - sets the page contained in a paged fragment
2328 * @frag: the paged fragment
2329 * @page: the page to set
2330 *
2331 * Sets the fragment @frag to contain @page.
2332 */
2333static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2334{
a8605c60 2335 frag->page.p = page;
131ea667
IC
2336}
2337
2338/**
2339 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2340 * @skb: the buffer
2341 * @f: the fragment offset
2342 * @page: the page to set
2343 *
2344 * Sets the @f'th fragment of @skb to contain @page.
2345 */
2346static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2347 struct page *page)
2348{
2349 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2350}
2351
400dfd3a
ED
2352bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2353
131ea667
IC
2354/**
2355 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2356 * @dev: the device to map the fragment to
131ea667
IC
2357 * @frag: the paged fragment to map
2358 * @offset: the offset within the fragment (starting at the
2359 * fragment's own offset)
2360 * @size: the number of bytes to map
f83347df 2361 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2362 *
2363 * Maps the page associated with @frag to @device.
2364 */
2365static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2366 const skb_frag_t *frag,
2367 size_t offset, size_t size,
2368 enum dma_data_direction dir)
2369{
2370 return dma_map_page(dev, skb_frag_page(frag),
2371 frag->page_offset + offset, size, dir);
2372}
2373
117632e6
ED
2374static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2375 gfp_t gfp_mask)
2376{
2377 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2378}
2379
bad93e9d
OP
2380
2381static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2382 gfp_t gfp_mask)
2383{
2384 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2385}
2386
2387
334a8132
PM
2388/**
2389 * skb_clone_writable - is the header of a clone writable
2390 * @skb: buffer to check
2391 * @len: length up to which to write
2392 *
2393 * Returns true if modifying the header part of the cloned buffer
2394 * does not requires the data to be copied.
2395 */
05bdd2f1 2396static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2397{
2398 return !skb_header_cloned(skb) &&
2399 skb_headroom(skb) + len <= skb->hdr_len;
2400}
2401
d9cc2048
HX
2402static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2403 int cloned)
2404{
2405 int delta = 0;
2406
d9cc2048
HX
2407 if (headroom > skb_headroom(skb))
2408 delta = headroom - skb_headroom(skb);
2409
2410 if (delta || cloned)
2411 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2412 GFP_ATOMIC);
2413 return 0;
2414}
2415
1da177e4
LT
2416/**
2417 * skb_cow - copy header of skb when it is required
2418 * @skb: buffer to cow
2419 * @headroom: needed headroom
2420 *
2421 * If the skb passed lacks sufficient headroom or its data part
2422 * is shared, data is reallocated. If reallocation fails, an error
2423 * is returned and original skb is not changed.
2424 *
2425 * The result is skb with writable area skb->head...skb->tail
2426 * and at least @headroom of space at head.
2427 */
2428static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2429{
d9cc2048
HX
2430 return __skb_cow(skb, headroom, skb_cloned(skb));
2431}
1da177e4 2432
d9cc2048
HX
2433/**
2434 * skb_cow_head - skb_cow but only making the head writable
2435 * @skb: buffer to cow
2436 * @headroom: needed headroom
2437 *
2438 * This function is identical to skb_cow except that we replace the
2439 * skb_cloned check by skb_header_cloned. It should be used when
2440 * you only need to push on some header and do not need to modify
2441 * the data.
2442 */
2443static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2444{
2445 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2446}
2447
2448/**
2449 * skb_padto - pad an skbuff up to a minimal size
2450 * @skb: buffer to pad
2451 * @len: minimal length
2452 *
2453 * Pads up a buffer to ensure the trailing bytes exist and are
2454 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2455 * is untouched. Otherwise it is extended. Returns zero on
2456 * success. The skb is freed on error.
1da177e4
LT
2457 */
2458
5b057c6b 2459static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2460{
2461 unsigned int size = skb->len;
2462 if (likely(size >= len))
5b057c6b 2463 return 0;
987c402a 2464 return skb_pad(skb, len - size);
1da177e4
LT
2465}
2466
2467static inline int skb_add_data(struct sk_buff *skb,
2468 char __user *from, int copy)
2469{
2470 const int off = skb->len;
2471
2472 if (skb->ip_summed == CHECKSUM_NONE) {
2473 int err = 0;
5084205f 2474 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2475 copy, 0, &err);
2476 if (!err) {
2477 skb->csum = csum_block_add(skb->csum, csum, off);
2478 return 0;
2479 }
2480 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2481 return 0;
2482
2483 __skb_trim(skb, off);
2484 return -EFAULT;
2485}
2486
38ba0a65
ED
2487static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2488 const struct page *page, int off)
1da177e4
LT
2489{
2490 if (i) {
9e903e08 2491 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2492
ea2ab693 2493 return page == skb_frag_page(frag) &&
9e903e08 2494 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2495 }
38ba0a65 2496 return false;
1da177e4
LT
2497}
2498
364c6bad
HX
2499static inline int __skb_linearize(struct sk_buff *skb)
2500{
2501 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2502}
2503
1da177e4
LT
2504/**
2505 * skb_linearize - convert paged skb to linear one
2506 * @skb: buffer to linarize
1da177e4
LT
2507 *
2508 * If there is no free memory -ENOMEM is returned, otherwise zero
2509 * is returned and the old skb data released.
2510 */
364c6bad
HX
2511static inline int skb_linearize(struct sk_buff *skb)
2512{
2513 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2514}
2515
cef401de
ED
2516/**
2517 * skb_has_shared_frag - can any frag be overwritten
2518 * @skb: buffer to test
2519 *
2520 * Return true if the skb has at least one frag that might be modified
2521 * by an external entity (as in vmsplice()/sendfile())
2522 */
2523static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2524{
c9af6db4
PS
2525 return skb_is_nonlinear(skb) &&
2526 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2527}
2528
364c6bad
HX
2529/**
2530 * skb_linearize_cow - make sure skb is linear and writable
2531 * @skb: buffer to process
2532 *
2533 * If there is no free memory -ENOMEM is returned, otherwise zero
2534 * is returned and the old skb data released.
2535 */
2536static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2537{
364c6bad
HX
2538 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2539 __skb_linearize(skb) : 0;
1da177e4
LT
2540}
2541
2542/**
2543 * skb_postpull_rcsum - update checksum for received skb after pull
2544 * @skb: buffer to update
2545 * @start: start of data before pull
2546 * @len: length of data pulled
2547 *
2548 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2549 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2550 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2551 */
2552
2553static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2554 const void *start, unsigned int len)
1da177e4 2555{
84fa7933 2556 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2557 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2558}
2559
cbb042f9
HX
2560unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2561
7ce5a27f
DM
2562/**
2563 * pskb_trim_rcsum - trim received skb and update checksum
2564 * @skb: buffer to trim
2565 * @len: new length
2566 *
2567 * This is exactly the same as pskb_trim except that it ensures the
2568 * checksum of received packets are still valid after the operation.
2569 */
2570
2571static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2572{
2573 if (likely(len >= skb->len))
2574 return 0;
2575 if (skb->ip_summed == CHECKSUM_COMPLETE)
2576 skb->ip_summed = CHECKSUM_NONE;
2577 return __pskb_trim(skb, len);
2578}
2579
1da177e4
LT
2580#define skb_queue_walk(queue, skb) \
2581 for (skb = (queue)->next; \
a1e4891f 2582 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2583 skb = skb->next)
2584
46f8914e
JC
2585#define skb_queue_walk_safe(queue, skb, tmp) \
2586 for (skb = (queue)->next, tmp = skb->next; \
2587 skb != (struct sk_buff *)(queue); \
2588 skb = tmp, tmp = skb->next)
2589
1164f52a 2590#define skb_queue_walk_from(queue, skb) \
a1e4891f 2591 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2592 skb = skb->next)
2593
2594#define skb_queue_walk_from_safe(queue, skb, tmp) \
2595 for (tmp = skb->next; \
2596 skb != (struct sk_buff *)(queue); \
2597 skb = tmp, tmp = skb->next)
2598
300ce174
SH
2599#define skb_queue_reverse_walk(queue, skb) \
2600 for (skb = (queue)->prev; \
a1e4891f 2601 skb != (struct sk_buff *)(queue); \
300ce174
SH
2602 skb = skb->prev)
2603
686a2955
DM
2604#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2605 for (skb = (queue)->prev, tmp = skb->prev; \
2606 skb != (struct sk_buff *)(queue); \
2607 skb = tmp, tmp = skb->prev)
2608
2609#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2610 for (tmp = skb->prev; \
2611 skb != (struct sk_buff *)(queue); \
2612 skb = tmp, tmp = skb->prev)
1da177e4 2613
21dc3301 2614static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2615{
2616 return skb_shinfo(skb)->frag_list != NULL;
2617}
2618
2619static inline void skb_frag_list_init(struct sk_buff *skb)
2620{
2621 skb_shinfo(skb)->frag_list = NULL;
2622}
2623
2624static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2625{
2626 frag->next = skb_shinfo(skb)->frag_list;
2627 skb_shinfo(skb)->frag_list = frag;
2628}
2629
2630#define skb_walk_frags(skb, iter) \
2631 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2632
7965bd4d
JP
2633struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2634 int *peeked, int *off, int *err);
2635struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2636 int *err);
2637unsigned int datagram_poll(struct file *file, struct socket *sock,
2638 struct poll_table_struct *wait);
2639int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2640 struct iovec *to, int size);
51f3d02b
DM
2641static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2642 struct msghdr *msg, int size)
2643{
2644 return skb_copy_datagram_iovec(from, offset, msg->msg_iov, size);
2645}
7965bd4d
JP
2646int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2647 struct iovec *iov);
2648int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2649 const struct iovec *from, int from_offset,
2650 int len);
2651int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2652 int offset, size_t count);
2653int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2654 const struct iovec *to, int to_offset,
2655 int size);
2656void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2657void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2658int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2659int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2660int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2661__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2662 int len, __wsum csum);
2663int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2664 struct pipe_inode_info *pipe, unsigned int len,
2665 unsigned int flags);
2666void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2667unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2668int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2669 int len, int hlen);
7965bd4d
JP
2670void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2671int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2672void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2673unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2674struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2675struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
20380731 2676
2817a336
DB
2677struct skb_checksum_ops {
2678 __wsum (*update)(const void *mem, int len, __wsum wsum);
2679 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2680};
2681
2682__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2683 __wsum csum, const struct skb_checksum_ops *ops);
2684__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2685 __wsum csum);
2686
690e36e7
DM
2687static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2688 int len, void *data, int hlen, void *buffer)
1da177e4 2689{
55820ee2 2690 if (hlen - offset >= len)
690e36e7 2691 return data + offset;
1da177e4 2692
690e36e7
DM
2693 if (!skb ||
2694 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2695 return NULL;
2696
2697 return buffer;
2698}
2699
690e36e7
DM
2700static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2701 int len, void *buffer)
2702{
2703 return __skb_header_pointer(skb, offset, len, skb->data,
2704 skb_headlen(skb), buffer);
2705}
2706
4262e5cc
DB
2707/**
2708 * skb_needs_linearize - check if we need to linearize a given skb
2709 * depending on the given device features.
2710 * @skb: socket buffer to check
2711 * @features: net device features
2712 *
2713 * Returns true if either:
2714 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2715 * 2. skb is fragmented and the device does not support SG.
2716 */
2717static inline bool skb_needs_linearize(struct sk_buff *skb,
2718 netdev_features_t features)
2719{
2720 return skb_is_nonlinear(skb) &&
2721 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2722 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2723}
2724
d626f62b
ACM
2725static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2726 void *to,
2727 const unsigned int len)
2728{
2729 memcpy(to, skb->data, len);
2730}
2731
2732static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2733 const int offset, void *to,
2734 const unsigned int len)
2735{
2736 memcpy(to, skb->data + offset, len);
2737}
2738
27d7ff46
ACM
2739static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2740 const void *from,
2741 const unsigned int len)
2742{
2743 memcpy(skb->data, from, len);
2744}
2745
2746static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2747 const int offset,
2748 const void *from,
2749 const unsigned int len)
2750{
2751 memcpy(skb->data + offset, from, len);
2752}
2753
7965bd4d 2754void skb_init(void);
1da177e4 2755
ac45f602
PO
2756static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2757{
2758 return skb->tstamp;
2759}
2760
a61bbcf2
PM
2761/**
2762 * skb_get_timestamp - get timestamp from a skb
2763 * @skb: skb to get stamp from
2764 * @stamp: pointer to struct timeval to store stamp in
2765 *
2766 * Timestamps are stored in the skb as offsets to a base timestamp.
2767 * This function converts the offset back to a struct timeval and stores
2768 * it in stamp.
2769 */
ac45f602
PO
2770static inline void skb_get_timestamp(const struct sk_buff *skb,
2771 struct timeval *stamp)
a61bbcf2 2772{
b7aa0bf7 2773 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2774}
2775
ac45f602
PO
2776static inline void skb_get_timestampns(const struct sk_buff *skb,
2777 struct timespec *stamp)
2778{
2779 *stamp = ktime_to_timespec(skb->tstamp);
2780}
2781
b7aa0bf7 2782static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2783{
b7aa0bf7 2784 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2785}
2786
164891aa
SH
2787static inline ktime_t net_timedelta(ktime_t t)
2788{
2789 return ktime_sub(ktime_get_real(), t);
2790}
2791
b9ce204f
IJ
2792static inline ktime_t net_invalid_timestamp(void)
2793{
2794 return ktime_set(0, 0);
2795}
a61bbcf2 2796
62bccb8c
AD
2797struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2798
c1f19b51
RC
2799#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2800
7965bd4d
JP
2801void skb_clone_tx_timestamp(struct sk_buff *skb);
2802bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2803
2804#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2805
2806static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2807{
2808}
2809
2810static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2811{
2812 return false;
2813}
2814
2815#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2816
2817/**
2818 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2819 *
da92b194
RC
2820 * PHY drivers may accept clones of transmitted packets for
2821 * timestamping via their phy_driver.txtstamp method. These drivers
2822 * must call this function to return the skb back to the stack, with
2823 * or without a timestamp.
2824 *
c1f19b51 2825 * @skb: clone of the the original outgoing packet
da92b194 2826 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2827 *
2828 */
2829void skb_complete_tx_timestamp(struct sk_buff *skb,
2830 struct skb_shared_hwtstamps *hwtstamps);
2831
e7fd2885
WB
2832void __skb_tstamp_tx(struct sk_buff *orig_skb,
2833 struct skb_shared_hwtstamps *hwtstamps,
2834 struct sock *sk, int tstype);
2835
ac45f602
PO
2836/**
2837 * skb_tstamp_tx - queue clone of skb with send time stamps
2838 * @orig_skb: the original outgoing packet
2839 * @hwtstamps: hardware time stamps, may be NULL if not available
2840 *
2841 * If the skb has a socket associated, then this function clones the
2842 * skb (thus sharing the actual data and optional structures), stores
2843 * the optional hardware time stamping information (if non NULL) or
2844 * generates a software time stamp (otherwise), then queues the clone
2845 * to the error queue of the socket. Errors are silently ignored.
2846 */
7965bd4d
JP
2847void skb_tstamp_tx(struct sk_buff *orig_skb,
2848 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2849
4507a715
RC
2850static inline void sw_tx_timestamp(struct sk_buff *skb)
2851{
2244d07b
OH
2852 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2853 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2854 skb_tstamp_tx(skb, NULL);
2855}
2856
2857/**
2858 * skb_tx_timestamp() - Driver hook for transmit timestamping
2859 *
2860 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2861 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2862 *
73409f3b
DM
2863 * Specifically, one should make absolutely sure that this function is
2864 * called before TX completion of this packet can trigger. Otherwise
2865 * the packet could potentially already be freed.
2866 *
4507a715
RC
2867 * @skb: A socket buffer.
2868 */
2869static inline void skb_tx_timestamp(struct sk_buff *skb)
2870{
c1f19b51 2871 skb_clone_tx_timestamp(skb);
4507a715
RC
2872 sw_tx_timestamp(skb);
2873}
2874
6e3e939f
JB
2875/**
2876 * skb_complete_wifi_ack - deliver skb with wifi status
2877 *
2878 * @skb: the original outgoing packet
2879 * @acked: ack status
2880 *
2881 */
2882void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2883
7965bd4d
JP
2884__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2885__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2886
60476372
HX
2887static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2888{
5d0c2b95 2889 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
60476372
HX
2890}
2891
fb286bb2
HX
2892/**
2893 * skb_checksum_complete - Calculate checksum of an entire packet
2894 * @skb: packet to process
2895 *
2896 * This function calculates the checksum over the entire packet plus
2897 * the value of skb->csum. The latter can be used to supply the
2898 * checksum of a pseudo header as used by TCP/UDP. It returns the
2899 * checksum.
2900 *
2901 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2902 * this function can be used to verify that checksum on received
2903 * packets. In that case the function should return zero if the
2904 * checksum is correct. In particular, this function will return zero
2905 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2906 * hardware has already verified the correctness of the checksum.
2907 */
4381ca3c 2908static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2909{
60476372
HX
2910 return skb_csum_unnecessary(skb) ?
2911 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2912}
2913
77cffe23
TH
2914static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2915{
2916 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2917 if (skb->csum_level == 0)
2918 skb->ip_summed = CHECKSUM_NONE;
2919 else
2920 skb->csum_level--;
2921 }
2922}
2923
2924static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2925{
2926 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2927 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2928 skb->csum_level++;
2929 } else if (skb->ip_summed == CHECKSUM_NONE) {
2930 skb->ip_summed = CHECKSUM_UNNECESSARY;
2931 skb->csum_level = 0;
2932 }
2933}
2934
5a212329
TH
2935static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2936{
2937 /* Mark current checksum as bad (typically called from GRO
2938 * path). In the case that ip_summed is CHECKSUM_NONE
2939 * this must be the first checksum encountered in the packet.
2940 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2941 * checksum after the last one validated. For UDP, a zero
2942 * checksum can not be marked as bad.
2943 */
2944
2945 if (skb->ip_summed == CHECKSUM_NONE ||
2946 skb->ip_summed == CHECKSUM_UNNECESSARY)
2947 skb->csum_bad = 1;
2948}
2949
76ba0aae
TH
2950/* Check if we need to perform checksum complete validation.
2951 *
2952 * Returns true if checksum complete is needed, false otherwise
2953 * (either checksum is unnecessary or zero checksum is allowed).
2954 */
2955static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2956 bool zero_okay,
2957 __sum16 check)
2958{
5d0c2b95
TH
2959 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2960 skb->csum_valid = 1;
77cffe23 2961 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
2962 return false;
2963 }
2964
2965 return true;
2966}
2967
2968/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2969 * in checksum_init.
2970 */
2971#define CHECKSUM_BREAK 76
2972
2973/* Validate (init) checksum based on checksum complete.
2974 *
2975 * Return values:
2976 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2977 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2978 * checksum is stored in skb->csum for use in __skb_checksum_complete
2979 * non-zero: value of invalid checksum
2980 *
2981 */
2982static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2983 bool complete,
2984 __wsum psum)
2985{
2986 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2987 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 2988 skb->csum_valid = 1;
76ba0aae
TH
2989 return 0;
2990 }
5a212329
TH
2991 } else if (skb->csum_bad) {
2992 /* ip_summed == CHECKSUM_NONE in this case */
2993 return 1;
76ba0aae
TH
2994 }
2995
2996 skb->csum = psum;
2997
5d0c2b95
TH
2998 if (complete || skb->len <= CHECKSUM_BREAK) {
2999 __sum16 csum;
3000
3001 csum = __skb_checksum_complete(skb);
3002 skb->csum_valid = !csum;
3003 return csum;
3004 }
76ba0aae
TH
3005
3006 return 0;
3007}
3008
3009static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3010{
3011 return 0;
3012}
3013
3014/* Perform checksum validate (init). Note that this is a macro since we only
3015 * want to calculate the pseudo header which is an input function if necessary.
3016 * First we try to validate without any computation (checksum unnecessary) and
3017 * then calculate based on checksum complete calling the function to compute
3018 * pseudo header.
3019 *
3020 * Return values:
3021 * 0: checksum is validated or try to in skb_checksum_complete
3022 * non-zero: value of invalid checksum
3023 */
3024#define __skb_checksum_validate(skb, proto, complete, \
3025 zero_okay, check, compute_pseudo) \
3026({ \
3027 __sum16 __ret = 0; \
5d0c2b95 3028 skb->csum_valid = 0; \
76ba0aae
TH
3029 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3030 __ret = __skb_checksum_validate_complete(skb, \
3031 complete, compute_pseudo(skb, proto)); \
3032 __ret; \
3033})
3034
3035#define skb_checksum_init(skb, proto, compute_pseudo) \
3036 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3037
3038#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3039 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3040
3041#define skb_checksum_validate(skb, proto, compute_pseudo) \
3042 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3043
3044#define skb_checksum_validate_zero_check(skb, proto, check, \
3045 compute_pseudo) \
3046 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
3047
3048#define skb_checksum_simple_validate(skb) \
3049 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3050
d96535a1
TH
3051static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3052{
3053 return (skb->ip_summed == CHECKSUM_NONE &&
3054 skb->csum_valid && !skb->csum_bad);
3055}
3056
3057static inline void __skb_checksum_convert(struct sk_buff *skb,
3058 __sum16 check, __wsum pseudo)
3059{
3060 skb->csum = ~pseudo;
3061 skb->ip_summed = CHECKSUM_COMPLETE;
3062}
3063
3064#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3065do { \
3066 if (__skb_checksum_convert_check(skb)) \
3067 __skb_checksum_convert(skb, check, \
3068 compute_pseudo(skb, proto)); \
3069} while (0)
3070
5f79e0f9 3071#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3072void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3073static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3074{
3075 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3076 nf_conntrack_destroy(nfct);
1da177e4
LT
3077}
3078static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3079{
3080 if (nfct)
3081 atomic_inc(&nfct->use);
3082}
2fc72c7b 3083#endif
34666d46 3084#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3085static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3086{
3087 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3088 kfree(nf_bridge);
3089}
3090static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3091{
3092 if (nf_bridge)
3093 atomic_inc(&nf_bridge->use);
3094}
3095#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3096static inline void nf_reset(struct sk_buff *skb)
3097{
5f79e0f9 3098#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3099 nf_conntrack_put(skb->nfct);
3100 skb->nfct = NULL;
2fc72c7b 3101#endif
34666d46 3102#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3103 nf_bridge_put(skb->nf_bridge);
3104 skb->nf_bridge = NULL;
3105#endif
3106}
3107
124dff01
PM
3108static inline void nf_reset_trace(struct sk_buff *skb)
3109{
478b360a 3110#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3111 skb->nf_trace = 0;
3112#endif
a193a4ab
PM
3113}
3114
edda553c 3115/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3116static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3117 bool copy)
edda553c 3118{
5f79e0f9 3119#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3120 dst->nfct = src->nfct;
3121 nf_conntrack_get(src->nfct);
b1937227
ED
3122 if (copy)
3123 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3124#endif
34666d46 3125#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3126 dst->nf_bridge = src->nf_bridge;
3127 nf_bridge_get(src->nf_bridge);
3128#endif
478b360a 3129#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3130 if (copy)
3131 dst->nf_trace = src->nf_trace;
478b360a 3132#endif
edda553c
YK
3133}
3134
e7ac05f3
YK
3135static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3136{
e7ac05f3 3137#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3138 nf_conntrack_put(dst->nfct);
2fc72c7b 3139#endif
34666d46 3140#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3141 nf_bridge_put(dst->nf_bridge);
3142#endif
b1937227 3143 __nf_copy(dst, src, true);
e7ac05f3
YK
3144}
3145
984bc16c
JM
3146#ifdef CONFIG_NETWORK_SECMARK
3147static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3148{
3149 to->secmark = from->secmark;
3150}
3151
3152static inline void skb_init_secmark(struct sk_buff *skb)
3153{
3154 skb->secmark = 0;
3155}
3156#else
3157static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3158{ }
3159
3160static inline void skb_init_secmark(struct sk_buff *skb)
3161{ }
3162#endif
3163
574f7194
EB
3164static inline bool skb_irq_freeable(const struct sk_buff *skb)
3165{
3166 return !skb->destructor &&
3167#if IS_ENABLED(CONFIG_XFRM)
3168 !skb->sp &&
3169#endif
3170#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3171 !skb->nfct &&
3172#endif
3173 !skb->_skb_refdst &&
3174 !skb_has_frag_list(skb);
3175}
3176
f25f4e44
PWJ
3177static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3178{
f25f4e44 3179 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3180}
3181
9247744e 3182static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3183{
4e3ab47a 3184 return skb->queue_mapping;
4e3ab47a
PE
3185}
3186
f25f4e44
PWJ
3187static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3188{
f25f4e44 3189 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3190}
3191
d5a9e24a
DM
3192static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3193{
3194 skb->queue_mapping = rx_queue + 1;
3195}
3196
9247744e 3197static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3198{
3199 return skb->queue_mapping - 1;
3200}
3201
9247744e 3202static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3203{
a02cec21 3204 return skb->queue_mapping != 0;
d5a9e24a
DM
3205}
3206
0e001614 3207u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
7965bd4d 3208 unsigned int num_tx_queues);
9247744e 3209
def8b4fa
AD
3210static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3211{
0b3d8e08 3212#ifdef CONFIG_XFRM
def8b4fa 3213 return skb->sp;
def8b4fa 3214#else
def8b4fa 3215 return NULL;
def8b4fa 3216#endif
0b3d8e08 3217}
def8b4fa 3218
68c33163
PS
3219/* Keeps track of mac header offset relative to skb->head.
3220 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3221 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3222 * tunnel skb it points to outer mac header.
3223 * Keeps track of level of encapsulation of network headers.
3224 */
68c33163 3225struct skb_gso_cb {
3347c960
ED
3226 int mac_offset;
3227 int encap_level;
7e2b10c1 3228 __u16 csum_start;
68c33163
PS
3229};
3230#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3231
3232static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3233{
3234 return (skb_mac_header(inner_skb) - inner_skb->head) -
3235 SKB_GSO_CB(inner_skb)->mac_offset;
3236}
3237
1e2bd517
PS
3238static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3239{
3240 int new_headroom, headroom;
3241 int ret;
3242
3243 headroom = skb_headroom(skb);
3244 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3245 if (ret)
3246 return ret;
3247
3248 new_headroom = skb_headroom(skb);
3249 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3250 return 0;
3251}
3252
7e2b10c1
TH
3253/* Compute the checksum for a gso segment. First compute the checksum value
3254 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3255 * then add in skb->csum (checksum from csum_start to end of packet).
3256 * skb->csum and csum_start are then updated to reflect the checksum of the
3257 * resultant packet starting from the transport header-- the resultant checksum
3258 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3259 * header.
3260 */
3261static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3262{
3263 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3264 skb_transport_offset(skb);
3265 __u16 csum;
3266
3267 csum = csum_fold(csum_partial(skb_transport_header(skb),
3268 plen, skb->csum));
3269 skb->csum = res;
3270 SKB_GSO_CB(skb)->csum_start -= plen;
3271
3272 return csum;
3273}
3274
bdcc0924 3275static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3276{
3277 return skb_shinfo(skb)->gso_size;
3278}
3279
36a8f39e 3280/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3281static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3282{
3283 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3284}
3285
7965bd4d 3286void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3287
3288static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3289{
3290 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3291 * wanted then gso_type will be set. */
05bdd2f1
ED
3292 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3293
b78462eb
AD
3294 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3295 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3296 __skb_warn_lro_forwarding(skb);
3297 return true;
3298 }
3299 return false;
3300}
3301
35fc92a9
HX
3302static inline void skb_forward_csum(struct sk_buff *skb)
3303{
3304 /* Unfortunately we don't support this one. Any brave souls? */
3305 if (skb->ip_summed == CHECKSUM_COMPLETE)
3306 skb->ip_summed = CHECKSUM_NONE;
3307}
3308
bc8acf2c
ED
3309/**
3310 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3311 * @skb: skb to check
3312 *
3313 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3314 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3315 * use this helper, to document places where we make this assertion.
3316 */
05bdd2f1 3317static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3318{
3319#ifdef DEBUG
3320 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3321#endif
3322}
3323
f35d9d8a 3324bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3325
ed1f50c3
PD
3326int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3327
56193d1b
AD
3328u32 skb_get_poff(const struct sk_buff *skb);
3329u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3330 const struct flow_keys *keys, int hlen);
f77668dc 3331
3a7c1ee4
AD
3332/**
3333 * skb_head_is_locked - Determine if the skb->head is locked down
3334 * @skb: skb to check
3335 *
3336 * The head on skbs build around a head frag can be removed if they are
3337 * not cloned. This function returns true if the skb head is locked down
3338 * due to either being allocated via kmalloc, or by being a clone with
3339 * multiple references to the head.
3340 */
3341static inline bool skb_head_is_locked(const struct sk_buff *skb)
3342{
3343 return !skb->head_frag || skb_cloned(skb);
3344}
fe6cc55f
FW
3345
3346/**
3347 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3348 *
3349 * @skb: GSO skb
3350 *
3351 * skb_gso_network_seglen is used to determine the real size of the
3352 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3353 *
3354 * The MAC/L2 header is not accounted for.
3355 */
3356static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3357{
3358 unsigned int hdr_len = skb_transport_header(skb) -
3359 skb_network_header(skb);
3360 return hdr_len + skb_gso_transport_seglen(skb);
3361}
1da177e4
LT
3362#endif /* __KERNEL__ */
3363#endif /* _LINUX_SKBUFF_H */