]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - include/linux/skbuff.h
Merge branch 'for-davem-2' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
[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;
a8f820aa 153struct iov_iter;
1da177e4 154
5f79e0f9 155#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
156struct nf_conntrack {
157 atomic_t use;
1da177e4 158};
5f79e0f9 159#endif
1da177e4 160
34666d46 161#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 162struct nf_bridge_info {
bf1ac5ca
ED
163 atomic_t use;
164 unsigned int mask;
165 struct net_device *physindev;
166 struct net_device *physoutdev;
167 unsigned long data[32 / sizeof(unsigned long)];
1da177e4
LT
168};
169#endif
170
1da177e4
LT
171struct sk_buff_head {
172 /* These two members must be first. */
173 struct sk_buff *next;
174 struct sk_buff *prev;
175
176 __u32 qlen;
177 spinlock_t lock;
178};
179
180struct sk_buff;
181
9d4dde52
IC
182/* To allow 64K frame to be packed as single skb without frag_list we
183 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
184 * buffers which do not start on a page boundary.
185 *
186 * Since GRO uses frags we allocate at least 16 regardless of page
187 * size.
a715dea3 188 */
9d4dde52 189#if (65536/PAGE_SIZE + 1) < 16
eec00954 190#define MAX_SKB_FRAGS 16UL
a715dea3 191#else
9d4dde52 192#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 193#endif
1da177e4
LT
194
195typedef struct skb_frag_struct skb_frag_t;
196
197struct skb_frag_struct {
a8605c60
IC
198 struct {
199 struct page *p;
200 } page;
cb4dfe56 201#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
202 __u32 page_offset;
203 __u32 size;
cb4dfe56
ED
204#else
205 __u16 page_offset;
206 __u16 size;
207#endif
1da177e4
LT
208};
209
9e903e08
ED
210static inline unsigned int skb_frag_size(const skb_frag_t *frag)
211{
212 return frag->size;
213}
214
215static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
216{
217 frag->size = size;
218}
219
220static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
221{
222 frag->size += delta;
223}
224
225static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
226{
227 frag->size -= delta;
228}
229
ac45f602
PO
230#define HAVE_HW_TIME_STAMP
231
232/**
d3a21be8 233 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
234 * @hwtstamp: hardware time stamp transformed into duration
235 * since arbitrary point in time
ac45f602
PO
236 *
237 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 238 * skb->tstamp.
ac45f602
PO
239 *
240 * hwtstamps can only be compared against other hwtstamps from
241 * the same device.
242 *
243 * This structure is attached to packets as part of the
244 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
245 */
246struct skb_shared_hwtstamps {
247 ktime_t hwtstamp;
ac45f602
PO
248};
249
2244d07b
OH
250/* Definitions for tx_flags in struct skb_shared_info */
251enum {
252 /* generate hardware time stamp */
253 SKBTX_HW_TSTAMP = 1 << 0,
254
e7fd2885 255 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
256 SKBTX_SW_TSTAMP = 1 << 1,
257
258 /* device driver is going to provide hardware time stamp */
259 SKBTX_IN_PROGRESS = 1 << 2,
260
a6686f2f 261 /* device driver supports TX zero-copy buffers */
62b1a8ab 262 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
263
264 /* generate wifi status information (where possible) */
62b1a8ab 265 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
266
267 /* This indicates at least one fragment might be overwritten
268 * (as in vmsplice(), sendfile() ...)
269 * If we need to compute a TX checksum, we'll need to copy
270 * all frags to avoid possible bad checksum
271 */
272 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
273
274 /* generate software time stamp when entering packet scheduling */
275 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
276
277 /* generate software timestamp on peer data acknowledgment */
278 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
279};
280
e1c8a607
WB
281#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
282 SKBTX_SCHED_TSTAMP | \
283 SKBTX_ACK_TSTAMP)
f24b9be5
WB
284#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
285
a6686f2f
SM
286/*
287 * The callback notifies userspace to release buffers when skb DMA is done in
288 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
289 * The zerocopy_success argument is true if zero copy transmit occurred,
290 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
291 * The ctx field is used to track device context.
292 * The desc field is used to track userspace buffer index.
a6686f2f
SM
293 */
294struct ubuf_info {
e19d6763 295 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 296 void *ctx;
a6686f2f 297 unsigned long desc;
ac45f602
PO
298};
299
1da177e4
LT
300/* This data is invariant across clones and lives at
301 * the end of the header data, ie. at skb->end.
302 */
303struct skb_shared_info {
9f42f126
IC
304 unsigned char nr_frags;
305 __u8 tx_flags;
7967168c
HX
306 unsigned short gso_size;
307 /* Warning: this field is not always filled in (UFO)! */
308 unsigned short gso_segs;
309 unsigned short gso_type;
1da177e4 310 struct sk_buff *frag_list;
ac45f602 311 struct skb_shared_hwtstamps hwtstamps;
09c2d251 312 u32 tskey;
9f42f126 313 __be32 ip6_frag_id;
ec7d2f2c
ED
314
315 /*
316 * Warning : all fields before dataref are cleared in __alloc_skb()
317 */
318 atomic_t dataref;
319
69e3c75f
JB
320 /* Intermediate layers must ensure that destructor_arg
321 * remains valid until skb destructor */
322 void * destructor_arg;
a6686f2f 323
fed66381
ED
324 /* must be last field, see pskb_expand_head() */
325 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
326};
327
328/* We divide dataref into two halves. The higher 16 bits hold references
329 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
330 * the entire skb->data. A clone of a headerless skb holds the length of
331 * the header in skb->hdr_len.
1da177e4
LT
332 *
333 * All users must obey the rule that the skb->data reference count must be
334 * greater than or equal to the payload reference count.
335 *
336 * Holding a reference to the payload part means that the user does not
337 * care about modifications to the header part of skb->data.
338 */
339#define SKB_DATAREF_SHIFT 16
340#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
341
d179cd12
DM
342
343enum {
c8753d55
VS
344 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
345 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
346 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
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
932bc4d7
JA
720/**
721 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
722 * @skb: buffer
723 * @dst: dst entry
724 *
725 * Sets skb dst, assuming a reference was not taken on dst.
726 * If dst entry is cached, we do not take reference and dst_release
727 * will be avoided by refdst_drop. If dst entry is not cached, we take
728 * reference, so that last dst_release can destroy the dst immediately.
729 */
730static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
731{
dbfc4fb7
HFS
732 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
733 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 734}
7fee226a
ED
735
736/**
25985edc 737 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
738 * @skb: buffer
739 */
740static inline bool skb_dst_is_noref(const struct sk_buff *skb)
741{
742 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
743}
744
511c3f92
ED
745static inline struct rtable *skb_rtable(const struct sk_buff *skb)
746{
adf30907 747 return (struct rtable *)skb_dst(skb);
511c3f92
ED
748}
749
7965bd4d
JP
750void kfree_skb(struct sk_buff *skb);
751void kfree_skb_list(struct sk_buff *segs);
752void skb_tx_error(struct sk_buff *skb);
753void consume_skb(struct sk_buff *skb);
754void __kfree_skb(struct sk_buff *skb);
d7e8883c 755extern struct kmem_cache *skbuff_head_cache;
bad43ca8 756
7965bd4d
JP
757void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
758bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
759 bool *fragstolen, int *delta_truesize);
bad43ca8 760
7965bd4d
JP
761struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
762 int node);
763struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 764static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 765 gfp_t priority)
d179cd12 766{
564824b0 767 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
768}
769
2e4e4410
ED
770struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
771 unsigned long data_len,
772 int max_page_order,
773 int *errcode,
774 gfp_t gfp_mask);
775
d0bf4a9e
ED
776/* Layout of fast clones : [skb1][skb2][fclone_ref] */
777struct sk_buff_fclones {
778 struct sk_buff skb1;
779
780 struct sk_buff skb2;
781
782 atomic_t fclone_ref;
783};
784
785/**
786 * skb_fclone_busy - check if fclone is busy
787 * @skb: buffer
788 *
789 * Returns true is skb is a fast clone, and its clone is not freed.
39bb5e62
ED
790 * Some drivers call skb_orphan() in their ndo_start_xmit(),
791 * so we also check that this didnt happen.
d0bf4a9e 792 */
39bb5e62
ED
793static inline bool skb_fclone_busy(const struct sock *sk,
794 const struct sk_buff *skb)
d0bf4a9e
ED
795{
796 const struct sk_buff_fclones *fclones;
797
798 fclones = container_of(skb, struct sk_buff_fclones, skb1);
799
800 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 801 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 802 fclones->skb2.sk == sk;
d0bf4a9e
ED
803}
804
d179cd12 805static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 806 gfp_t priority)
d179cd12 807{
c93bdd0e 808 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
809}
810
7965bd4d 811struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
812static inline struct sk_buff *alloc_skb_head(gfp_t priority)
813{
814 return __alloc_skb_head(priority, -1);
815}
816
7965bd4d
JP
817struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
818int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
819struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
820struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
821struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
822 gfp_t gfp_mask, bool fclone);
823static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
824 gfp_t gfp_mask)
825{
826 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
827}
7965bd4d
JP
828
829int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
830struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
831 unsigned int headroom);
832struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
833 int newtailroom, gfp_t priority);
25a91d8d
FD
834int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
835 int offset, int len);
7965bd4d
JP
836int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
837 int len);
838int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
839int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 840#define dev_kfree_skb(a) consume_skb(a)
1da177e4 841
7965bd4d
JP
842int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
843 int getfrag(void *from, char *to, int offset,
844 int len, int odd, struct sk_buff *skb),
845 void *from, int length);
e89e9cf5 846
d94d9fee 847struct skb_seq_state {
677e90ed
TG
848 __u32 lower_offset;
849 __u32 upper_offset;
850 __u32 frag_idx;
851 __u32 stepped_offset;
852 struct sk_buff *root_skb;
853 struct sk_buff *cur_skb;
854 __u8 *frag_data;
855};
856
7965bd4d
JP
857void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
858 unsigned int to, struct skb_seq_state *st);
859unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
860 struct skb_seq_state *st);
861void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 862
7965bd4d
JP
863unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
864 unsigned int to, struct ts_config *config,
865 struct ts_state *state);
3fc7e8a6 866
09323cc4
TH
867/*
868 * Packet hash types specify the type of hash in skb_set_hash.
869 *
870 * Hash types refer to the protocol layer addresses which are used to
871 * construct a packet's hash. The hashes are used to differentiate or identify
872 * flows of the protocol layer for the hash type. Hash types are either
873 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
874 *
875 * Properties of hashes:
876 *
877 * 1) Two packets in different flows have different hash values
878 * 2) Two packets in the same flow should have the same hash value
879 *
880 * A hash at a higher layer is considered to be more specific. A driver should
881 * set the most specific hash possible.
882 *
883 * A driver cannot indicate a more specific hash than the layer at which a hash
884 * was computed. For instance an L3 hash cannot be set as an L4 hash.
885 *
886 * A driver may indicate a hash level which is less specific than the
887 * actual layer the hash was computed on. For instance, a hash computed
888 * at L4 may be considered an L3 hash. This should only be done if the
889 * driver can't unambiguously determine that the HW computed the hash at
890 * the higher layer. Note that the "should" in the second property above
891 * permits this.
892 */
893enum pkt_hash_types {
894 PKT_HASH_TYPE_NONE, /* Undefined type */
895 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
896 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
897 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
898};
899
900static inline void
901skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
902{
61b905da 903 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 904 skb->sw_hash = 0;
61b905da 905 skb->hash = hash;
09323cc4
TH
906}
907
3958afa1
TH
908void __skb_get_hash(struct sk_buff *skb);
909static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 910{
a3b18ddb 911 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 912 __skb_get_hash(skb);
bfb564e7 913
61b905da 914 return skb->hash;
bfb564e7
KK
915}
916
57bdf7f4
TH
917static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
918{
61b905da 919 return skb->hash;
57bdf7f4
TH
920}
921
7539fadc
TH
922static inline void skb_clear_hash(struct sk_buff *skb)
923{
61b905da 924 skb->hash = 0;
a3b18ddb 925 skb->sw_hash = 0;
61b905da 926 skb->l4_hash = 0;
7539fadc
TH
927}
928
929static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
930{
61b905da 931 if (!skb->l4_hash)
7539fadc
TH
932 skb_clear_hash(skb);
933}
934
3df7a74e
TH
935static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
936{
61b905da 937 to->hash = from->hash;
a3b18ddb 938 to->sw_hash = from->sw_hash;
61b905da 939 to->l4_hash = from->l4_hash;
3df7a74e
TH
940};
941
4305b541
ACM
942#ifdef NET_SKBUFF_DATA_USES_OFFSET
943static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
944{
945 return skb->head + skb->end;
946}
ec47ea82
AD
947
948static inline unsigned int skb_end_offset(const struct sk_buff *skb)
949{
950 return skb->end;
951}
4305b541
ACM
952#else
953static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
954{
955 return skb->end;
956}
ec47ea82
AD
957
958static inline unsigned int skb_end_offset(const struct sk_buff *skb)
959{
960 return skb->end - skb->head;
961}
4305b541
ACM
962#endif
963
1da177e4 964/* Internal */
4305b541 965#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 966
ac45f602
PO
967static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
968{
969 return &skb_shinfo(skb)->hwtstamps;
970}
971
1da177e4
LT
972/**
973 * skb_queue_empty - check if a queue is empty
974 * @list: queue head
975 *
976 * Returns true if the queue is empty, false otherwise.
977 */
978static inline int skb_queue_empty(const struct sk_buff_head *list)
979{
fd44b93c 980 return list->next == (const struct sk_buff *) list;
1da177e4
LT
981}
982
fc7ebb21
DM
983/**
984 * skb_queue_is_last - check if skb is the last entry in the queue
985 * @list: queue head
986 * @skb: buffer
987 *
988 * Returns true if @skb is the last buffer on the list.
989 */
990static inline bool skb_queue_is_last(const struct sk_buff_head *list,
991 const struct sk_buff *skb)
992{
fd44b93c 993 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
994}
995
832d11c5
IJ
996/**
997 * skb_queue_is_first - check if skb is the first entry in the queue
998 * @list: queue head
999 * @skb: buffer
1000 *
1001 * Returns true if @skb is the first buffer on the list.
1002 */
1003static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1004 const struct sk_buff *skb)
1005{
fd44b93c 1006 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1007}
1008
249c8b42
DM
1009/**
1010 * skb_queue_next - return the next packet in the queue
1011 * @list: queue head
1012 * @skb: current buffer
1013 *
1014 * Return the next packet in @list after @skb. It is only valid to
1015 * call this if skb_queue_is_last() evaluates to false.
1016 */
1017static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1018 const struct sk_buff *skb)
1019{
1020 /* This BUG_ON may seem severe, but if we just return then we
1021 * are going to dereference garbage.
1022 */
1023 BUG_ON(skb_queue_is_last(list, skb));
1024 return skb->next;
1025}
1026
832d11c5
IJ
1027/**
1028 * skb_queue_prev - return the prev packet in the queue
1029 * @list: queue head
1030 * @skb: current buffer
1031 *
1032 * Return the prev packet in @list before @skb. It is only valid to
1033 * call this if skb_queue_is_first() evaluates to false.
1034 */
1035static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1036 const struct sk_buff *skb)
1037{
1038 /* This BUG_ON may seem severe, but if we just return then we
1039 * are going to dereference garbage.
1040 */
1041 BUG_ON(skb_queue_is_first(list, skb));
1042 return skb->prev;
1043}
1044
1da177e4
LT
1045/**
1046 * skb_get - reference buffer
1047 * @skb: buffer to reference
1048 *
1049 * Makes another reference to a socket buffer and returns a pointer
1050 * to the buffer.
1051 */
1052static inline struct sk_buff *skb_get(struct sk_buff *skb)
1053{
1054 atomic_inc(&skb->users);
1055 return skb;
1056}
1057
1058/*
1059 * If users == 1, we are the only owner and are can avoid redundant
1060 * atomic change.
1061 */
1062
1da177e4
LT
1063/**
1064 * skb_cloned - is the buffer a clone
1065 * @skb: buffer to check
1066 *
1067 * Returns true if the buffer was generated with skb_clone() and is
1068 * one of multiple shared copies of the buffer. Cloned buffers are
1069 * shared data so must not be written to under normal circumstances.
1070 */
1071static inline int skb_cloned(const struct sk_buff *skb)
1072{
1073 return skb->cloned &&
1074 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1075}
1076
14bbd6a5
PS
1077static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1078{
1079 might_sleep_if(pri & __GFP_WAIT);
1080
1081 if (skb_cloned(skb))
1082 return pskb_expand_head(skb, 0, 0, pri);
1083
1084 return 0;
1085}
1086
1da177e4
LT
1087/**
1088 * skb_header_cloned - is the header a clone
1089 * @skb: buffer to check
1090 *
1091 * Returns true if modifying the header part of the buffer requires
1092 * the data to be copied.
1093 */
1094static inline int skb_header_cloned(const struct sk_buff *skb)
1095{
1096 int dataref;
1097
1098 if (!skb->cloned)
1099 return 0;
1100
1101 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1102 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1103 return dataref != 1;
1104}
1105
1106/**
1107 * skb_header_release - release reference to header
1108 * @skb: buffer to operate on
1109 *
1110 * Drop a reference to the header part of the buffer. This is done
1111 * by acquiring a payload reference. You must not read from the header
1112 * part of skb->data after this.
f4a775d1 1113 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1114 */
1115static inline void skb_header_release(struct sk_buff *skb)
1116{
1117 BUG_ON(skb->nohdr);
1118 skb->nohdr = 1;
1119 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1120}
1121
f4a775d1
ED
1122/**
1123 * __skb_header_release - release reference to header
1124 * @skb: buffer to operate on
1125 *
1126 * Variant of skb_header_release() assuming skb is private to caller.
1127 * We can avoid one atomic operation.
1128 */
1129static inline void __skb_header_release(struct sk_buff *skb)
1130{
1131 skb->nohdr = 1;
1132 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1133}
1134
1135
1da177e4
LT
1136/**
1137 * skb_shared - is the buffer shared
1138 * @skb: buffer to check
1139 *
1140 * Returns true if more than one person has a reference to this
1141 * buffer.
1142 */
1143static inline int skb_shared(const struct sk_buff *skb)
1144{
1145 return atomic_read(&skb->users) != 1;
1146}
1147
1148/**
1149 * skb_share_check - check if buffer is shared and if so clone it
1150 * @skb: buffer to check
1151 * @pri: priority for memory allocation
1152 *
1153 * If the buffer is shared the buffer is cloned and the old copy
1154 * drops a reference. A new clone with a single reference is returned.
1155 * If the buffer is not shared the original buffer is returned. When
1156 * being called from interrupt status or with spinlocks held pri must
1157 * be GFP_ATOMIC.
1158 *
1159 * NULL is returned on a memory allocation failure.
1160 */
47061bc4 1161static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1162{
1163 might_sleep_if(pri & __GFP_WAIT);
1164 if (skb_shared(skb)) {
1165 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1166
1167 if (likely(nskb))
1168 consume_skb(skb);
1169 else
1170 kfree_skb(skb);
1da177e4
LT
1171 skb = nskb;
1172 }
1173 return skb;
1174}
1175
1176/*
1177 * Copy shared buffers into a new sk_buff. We effectively do COW on
1178 * packets to handle cases where we have a local reader and forward
1179 * and a couple of other messy ones. The normal one is tcpdumping
1180 * a packet thats being forwarded.
1181 */
1182
1183/**
1184 * skb_unshare - make a copy of a shared buffer
1185 * @skb: buffer to check
1186 * @pri: priority for memory allocation
1187 *
1188 * If the socket buffer is a clone then this function creates a new
1189 * copy of the data, drops a reference count on the old copy and returns
1190 * the new copy with the reference count at 1. If the buffer is not a clone
1191 * the original buffer is returned. When called with a spinlock held or
1192 * from interrupt state @pri must be %GFP_ATOMIC
1193 *
1194 * %NULL is returned on a memory allocation failure.
1195 */
e2bf521d 1196static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1197 gfp_t pri)
1da177e4
LT
1198{
1199 might_sleep_if(pri & __GFP_WAIT);
1200 if (skb_cloned(skb)) {
1201 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1202
1203 /* Free our shared copy */
1204 if (likely(nskb))
1205 consume_skb(skb);
1206 else
1207 kfree_skb(skb);
1da177e4
LT
1208 skb = nskb;
1209 }
1210 return skb;
1211}
1212
1213/**
1a5778aa 1214 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1215 * @list_: list to peek at
1216 *
1217 * Peek an &sk_buff. Unlike most other operations you _MUST_
1218 * be careful with this one. A peek leaves the buffer on the
1219 * list and someone else may run off with it. You must hold
1220 * the appropriate locks or have a private queue to do this.
1221 *
1222 * Returns %NULL for an empty list or a pointer to the head element.
1223 * The reference count is not incremented and the reference is therefore
1224 * volatile. Use with caution.
1225 */
05bdd2f1 1226static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1227{
18d07000
ED
1228 struct sk_buff *skb = list_->next;
1229
1230 if (skb == (struct sk_buff *)list_)
1231 skb = NULL;
1232 return skb;
1da177e4
LT
1233}
1234
da5ef6e5
PE
1235/**
1236 * skb_peek_next - peek skb following the given one from a queue
1237 * @skb: skb to start from
1238 * @list_: list to peek at
1239 *
1240 * Returns %NULL when the end of the list is met or a pointer to the
1241 * next element. The reference count is not incremented and the
1242 * reference is therefore volatile. Use with caution.
1243 */
1244static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1245 const struct sk_buff_head *list_)
1246{
1247 struct sk_buff *next = skb->next;
18d07000 1248
da5ef6e5
PE
1249 if (next == (struct sk_buff *)list_)
1250 next = NULL;
1251 return next;
1252}
1253
1da177e4 1254/**
1a5778aa 1255 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1256 * @list_: list to peek at
1257 *
1258 * Peek an &sk_buff. Unlike most other operations you _MUST_
1259 * be careful with this one. A peek leaves the buffer on the
1260 * list and someone else may run off with it. You must hold
1261 * the appropriate locks or have a private queue to do this.
1262 *
1263 * Returns %NULL for an empty list or a pointer to the tail element.
1264 * The reference count is not incremented and the reference is therefore
1265 * volatile. Use with caution.
1266 */
05bdd2f1 1267static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1268{
18d07000
ED
1269 struct sk_buff *skb = list_->prev;
1270
1271 if (skb == (struct sk_buff *)list_)
1272 skb = NULL;
1273 return skb;
1274
1da177e4
LT
1275}
1276
1277/**
1278 * skb_queue_len - get queue length
1279 * @list_: list to measure
1280 *
1281 * Return the length of an &sk_buff queue.
1282 */
1283static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1284{
1285 return list_->qlen;
1286}
1287
67fed459
DM
1288/**
1289 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1290 * @list: queue to initialize
1291 *
1292 * This initializes only the list and queue length aspects of
1293 * an sk_buff_head object. This allows to initialize the list
1294 * aspects of an sk_buff_head without reinitializing things like
1295 * the spinlock. It can also be used for on-stack sk_buff_head
1296 * objects where the spinlock is known to not be used.
1297 */
1298static inline void __skb_queue_head_init(struct sk_buff_head *list)
1299{
1300 list->prev = list->next = (struct sk_buff *)list;
1301 list->qlen = 0;
1302}
1303
76f10ad0
AV
1304/*
1305 * This function creates a split out lock class for each invocation;
1306 * this is needed for now since a whole lot of users of the skb-queue
1307 * infrastructure in drivers have different locking usage (in hardirq)
1308 * than the networking core (in softirq only). In the long run either the
1309 * network layer or drivers should need annotation to consolidate the
1310 * main types of usage into 3 classes.
1311 */
1da177e4
LT
1312static inline void skb_queue_head_init(struct sk_buff_head *list)
1313{
1314 spin_lock_init(&list->lock);
67fed459 1315 __skb_queue_head_init(list);
1da177e4
LT
1316}
1317
c2ecba71
PE
1318static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1319 struct lock_class_key *class)
1320{
1321 skb_queue_head_init(list);
1322 lockdep_set_class(&list->lock, class);
1323}
1324
1da177e4 1325/*
bf299275 1326 * Insert an sk_buff on a list.
1da177e4
LT
1327 *
1328 * The "__skb_xxxx()" functions are the non-atomic ones that
1329 * can only be called with interrupts disabled.
1330 */
7965bd4d
JP
1331void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1332 struct sk_buff_head *list);
bf299275
GR
1333static inline void __skb_insert(struct sk_buff *newsk,
1334 struct sk_buff *prev, struct sk_buff *next,
1335 struct sk_buff_head *list)
1336{
1337 newsk->next = next;
1338 newsk->prev = prev;
1339 next->prev = prev->next = newsk;
1340 list->qlen++;
1341}
1da177e4 1342
67fed459
DM
1343static inline void __skb_queue_splice(const struct sk_buff_head *list,
1344 struct sk_buff *prev,
1345 struct sk_buff *next)
1346{
1347 struct sk_buff *first = list->next;
1348 struct sk_buff *last = list->prev;
1349
1350 first->prev = prev;
1351 prev->next = first;
1352
1353 last->next = next;
1354 next->prev = last;
1355}
1356
1357/**
1358 * skb_queue_splice - join two skb lists, this is designed for stacks
1359 * @list: the new list to add
1360 * @head: the place to add it in the first list
1361 */
1362static inline void skb_queue_splice(const struct sk_buff_head *list,
1363 struct sk_buff_head *head)
1364{
1365 if (!skb_queue_empty(list)) {
1366 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1367 head->qlen += list->qlen;
67fed459
DM
1368 }
1369}
1370
1371/**
d9619496 1372 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1373 * @list: the new list to add
1374 * @head: the place to add it in the first list
1375 *
1376 * The list at @list is reinitialised
1377 */
1378static inline void skb_queue_splice_init(struct sk_buff_head *list,
1379 struct sk_buff_head *head)
1380{
1381 if (!skb_queue_empty(list)) {
1382 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1383 head->qlen += list->qlen;
67fed459
DM
1384 __skb_queue_head_init(list);
1385 }
1386}
1387
1388/**
1389 * skb_queue_splice_tail - join two skb lists, each list being a queue
1390 * @list: the new list to add
1391 * @head: the place to add it in the first list
1392 */
1393static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1394 struct sk_buff_head *head)
1395{
1396 if (!skb_queue_empty(list)) {
1397 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1398 head->qlen += list->qlen;
67fed459
DM
1399 }
1400}
1401
1402/**
d9619496 1403 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1404 * @list: the new list to add
1405 * @head: the place to add it in the first list
1406 *
1407 * Each of the lists is a queue.
1408 * The list at @list is reinitialised
1409 */
1410static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1411 struct sk_buff_head *head)
1412{
1413 if (!skb_queue_empty(list)) {
1414 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1415 head->qlen += list->qlen;
67fed459
DM
1416 __skb_queue_head_init(list);
1417 }
1418}
1419
1da177e4 1420/**
300ce174 1421 * __skb_queue_after - queue a buffer at the list head
1da177e4 1422 * @list: list to use
300ce174 1423 * @prev: place after this buffer
1da177e4
LT
1424 * @newsk: buffer to queue
1425 *
300ce174 1426 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1427 * and you must therefore hold required locks before calling it.
1428 *
1429 * A buffer cannot be placed on two lists at the same time.
1430 */
300ce174
SH
1431static inline void __skb_queue_after(struct sk_buff_head *list,
1432 struct sk_buff *prev,
1433 struct sk_buff *newsk)
1da177e4 1434{
bf299275 1435 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1436}
1437
7965bd4d
JP
1438void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1439 struct sk_buff_head *list);
7de6c033 1440
f5572855
GR
1441static inline void __skb_queue_before(struct sk_buff_head *list,
1442 struct sk_buff *next,
1443 struct sk_buff *newsk)
1444{
1445 __skb_insert(newsk, next->prev, next, list);
1446}
1447
300ce174
SH
1448/**
1449 * __skb_queue_head - queue a buffer at the list head
1450 * @list: list to use
1451 * @newsk: buffer to queue
1452 *
1453 * Queue a buffer at the start of a list. This function takes no locks
1454 * and you must therefore hold required locks before calling it.
1455 *
1456 * A buffer cannot be placed on two lists at the same time.
1457 */
7965bd4d 1458void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1459static inline void __skb_queue_head(struct sk_buff_head *list,
1460 struct sk_buff *newsk)
1461{
1462 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1463}
1464
1da177e4
LT
1465/**
1466 * __skb_queue_tail - queue a buffer at the list tail
1467 * @list: list to use
1468 * @newsk: buffer to queue
1469 *
1470 * Queue a buffer at the end of a list. This function takes no locks
1471 * and you must therefore hold required locks before calling it.
1472 *
1473 * A buffer cannot be placed on two lists at the same time.
1474 */
7965bd4d 1475void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1476static inline void __skb_queue_tail(struct sk_buff_head *list,
1477 struct sk_buff *newsk)
1478{
f5572855 1479 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1480}
1481
1da177e4
LT
1482/*
1483 * remove sk_buff from list. _Must_ be called atomically, and with
1484 * the list known..
1485 */
7965bd4d 1486void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1487static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1488{
1489 struct sk_buff *next, *prev;
1490
1491 list->qlen--;
1492 next = skb->next;
1493 prev = skb->prev;
1494 skb->next = skb->prev = NULL;
1da177e4
LT
1495 next->prev = prev;
1496 prev->next = next;
1497}
1498
f525c06d
GR
1499/**
1500 * __skb_dequeue - remove from the head of the queue
1501 * @list: list to dequeue from
1502 *
1503 * Remove the head of the list. This function does not take any locks
1504 * so must be used with appropriate locks held only. The head item is
1505 * returned or %NULL if the list is empty.
1506 */
7965bd4d 1507struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1508static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1509{
1510 struct sk_buff *skb = skb_peek(list);
1511 if (skb)
1512 __skb_unlink(skb, list);
1513 return skb;
1514}
1da177e4
LT
1515
1516/**
1517 * __skb_dequeue_tail - remove from the tail of the queue
1518 * @list: list to dequeue from
1519 *
1520 * Remove the tail of the list. This function does not take any locks
1521 * so must be used with appropriate locks held only. The tail item is
1522 * returned or %NULL if the list is empty.
1523 */
7965bd4d 1524struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1525static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1526{
1527 struct sk_buff *skb = skb_peek_tail(list);
1528 if (skb)
1529 __skb_unlink(skb, list);
1530 return skb;
1531}
1532
1533
bdcc0924 1534static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1535{
1536 return skb->data_len;
1537}
1538
1539static inline unsigned int skb_headlen(const struct sk_buff *skb)
1540{
1541 return skb->len - skb->data_len;
1542}
1543
1544static inline int skb_pagelen(const struct sk_buff *skb)
1545{
1546 int i, len = 0;
1547
1548 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1549 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1550 return len + skb_headlen(skb);
1551}
1552
131ea667
IC
1553/**
1554 * __skb_fill_page_desc - initialise a paged fragment in an skb
1555 * @skb: buffer containing fragment to be initialised
1556 * @i: paged fragment index to initialise
1557 * @page: the page to use for this fragment
1558 * @off: the offset to the data with @page
1559 * @size: the length of the data
1560 *
1561 * Initialises the @i'th fragment of @skb to point to &size bytes at
1562 * offset @off within @page.
1563 *
1564 * Does not take any additional reference on the fragment.
1565 */
1566static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1567 struct page *page, int off, int size)
1da177e4
LT
1568{
1569 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1570
c48a11c7
MG
1571 /*
1572 * Propagate page->pfmemalloc to the skb if we can. The problem is
1573 * that not all callers have unique ownership of the page. If
1574 * pfmemalloc is set, we check the mapping as a mapping implies
1575 * page->index is set (index and pfmemalloc share space).
1576 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1577 * do not lose pfmemalloc information as the pages would not be
1578 * allocated using __GFP_MEMALLOC.
1579 */
a8605c60 1580 frag->page.p = page;
1da177e4 1581 frag->page_offset = off;
9e903e08 1582 skb_frag_size_set(frag, size);
cca7af38
PE
1583
1584 page = compound_head(page);
1585 if (page->pfmemalloc && !page->mapping)
1586 skb->pfmemalloc = true;
131ea667
IC
1587}
1588
1589/**
1590 * skb_fill_page_desc - initialise a paged fragment in an skb
1591 * @skb: buffer containing fragment to be initialised
1592 * @i: paged fragment index to initialise
1593 * @page: the page to use for this fragment
1594 * @off: the offset to the data with @page
1595 * @size: the length of the data
1596 *
1597 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1598 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1599 * addition updates @skb such that @i is the last fragment.
1600 *
1601 * Does not take any additional reference on the fragment.
1602 */
1603static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1604 struct page *page, int off, int size)
1605{
1606 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1607 skb_shinfo(skb)->nr_frags = i + 1;
1608}
1609
7965bd4d
JP
1610void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1611 int size, unsigned int truesize);
654bed16 1612
f8e617e1
JW
1613void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1614 unsigned int truesize);
1615
1da177e4 1616#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1617#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1618#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1619
27a884dc
ACM
1620#ifdef NET_SKBUFF_DATA_USES_OFFSET
1621static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1622{
1623 return skb->head + skb->tail;
1624}
1625
1626static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1627{
1628 skb->tail = skb->data - skb->head;
1629}
1630
1631static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1632{
1633 skb_reset_tail_pointer(skb);
1634 skb->tail += offset;
1635}
7cc46190 1636
27a884dc
ACM
1637#else /* NET_SKBUFF_DATA_USES_OFFSET */
1638static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1639{
1640 return skb->tail;
1641}
1642
1643static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1644{
1645 skb->tail = skb->data;
1646}
1647
1648static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1649{
1650 skb->tail = skb->data + offset;
1651}
4305b541 1652
27a884dc
ACM
1653#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1654
1da177e4
LT
1655/*
1656 * Add data to an sk_buff
1657 */
0c7ddf36 1658unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1659unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1660static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1661{
27a884dc 1662 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1663 SKB_LINEAR_ASSERT(skb);
1664 skb->tail += len;
1665 skb->len += len;
1666 return tmp;
1667}
1668
7965bd4d 1669unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1670static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1671{
1672 skb->data -= len;
1673 skb->len += len;
1674 return skb->data;
1675}
1676
7965bd4d 1677unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1678static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1679{
1680 skb->len -= len;
1681 BUG_ON(skb->len < skb->data_len);
1682 return skb->data += len;
1683}
1684
47d29646
DM
1685static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1686{
1687 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1688}
1689
7965bd4d 1690unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1691
1692static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1693{
1694 if (len > skb_headlen(skb) &&
987c402a 1695 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1696 return NULL;
1697 skb->len -= len;
1698 return skb->data += len;
1699}
1700
1701static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1702{
1703 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1704}
1705
1706static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1707{
1708 if (likely(len <= skb_headlen(skb)))
1709 return 1;
1710 if (unlikely(len > skb->len))
1711 return 0;
987c402a 1712 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1713}
1714
1715/**
1716 * skb_headroom - bytes at buffer head
1717 * @skb: buffer to check
1718 *
1719 * Return the number of bytes of free space at the head of an &sk_buff.
1720 */
c2636b4d 1721static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1722{
1723 return skb->data - skb->head;
1724}
1725
1726/**
1727 * skb_tailroom - bytes at buffer end
1728 * @skb: buffer to check
1729 *
1730 * Return the number of bytes of free space at the tail of an sk_buff
1731 */
1732static inline int skb_tailroom(const struct sk_buff *skb)
1733{
4305b541 1734 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1735}
1736
a21d4572
ED
1737/**
1738 * skb_availroom - bytes at buffer end
1739 * @skb: buffer to check
1740 *
1741 * Return the number of bytes of free space at the tail of an sk_buff
1742 * allocated by sk_stream_alloc()
1743 */
1744static inline int skb_availroom(const struct sk_buff *skb)
1745{
16fad69c
ED
1746 if (skb_is_nonlinear(skb))
1747 return 0;
1748
1749 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1750}
1751
1da177e4
LT
1752/**
1753 * skb_reserve - adjust headroom
1754 * @skb: buffer to alter
1755 * @len: bytes to move
1756 *
1757 * Increase the headroom of an empty &sk_buff by reducing the tail
1758 * room. This is only allowed for an empty buffer.
1759 */
8243126c 1760static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1761{
1762 skb->data += len;
1763 skb->tail += len;
1764}
1765
8bce6d7d
TH
1766#define ENCAP_TYPE_ETHER 0
1767#define ENCAP_TYPE_IPPROTO 1
1768
1769static inline void skb_set_inner_protocol(struct sk_buff *skb,
1770 __be16 protocol)
1771{
1772 skb->inner_protocol = protocol;
1773 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1774}
1775
1776static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1777 __u8 ipproto)
1778{
1779 skb->inner_ipproto = ipproto;
1780 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1781}
1782
6a674e9c
JG
1783static inline void skb_reset_inner_headers(struct sk_buff *skb)
1784{
aefbd2b3 1785 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1786 skb->inner_network_header = skb->network_header;
1787 skb->inner_transport_header = skb->transport_header;
1788}
1789
0b5c9db1
JP
1790static inline void skb_reset_mac_len(struct sk_buff *skb)
1791{
1792 skb->mac_len = skb->network_header - skb->mac_header;
1793}
1794
6a674e9c
JG
1795static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1796 *skb)
1797{
1798 return skb->head + skb->inner_transport_header;
1799}
1800
1801static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1802{
1803 skb->inner_transport_header = skb->data - skb->head;
1804}
1805
1806static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1807 const int offset)
1808{
1809 skb_reset_inner_transport_header(skb);
1810 skb->inner_transport_header += offset;
1811}
1812
1813static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1814{
1815 return skb->head + skb->inner_network_header;
1816}
1817
1818static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1819{
1820 skb->inner_network_header = skb->data - skb->head;
1821}
1822
1823static inline void skb_set_inner_network_header(struct sk_buff *skb,
1824 const int offset)
1825{
1826 skb_reset_inner_network_header(skb);
1827 skb->inner_network_header += offset;
1828}
1829
aefbd2b3
PS
1830static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1831{
1832 return skb->head + skb->inner_mac_header;
1833}
1834
1835static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1836{
1837 skb->inner_mac_header = skb->data - skb->head;
1838}
1839
1840static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1841 const int offset)
1842{
1843 skb_reset_inner_mac_header(skb);
1844 skb->inner_mac_header += offset;
1845}
fda55eca
ED
1846static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1847{
35d04610 1848 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1849}
1850
9c70220b
ACM
1851static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1852{
2e07fa9c 1853 return skb->head + skb->transport_header;
9c70220b
ACM
1854}
1855
badff6d0
ACM
1856static inline void skb_reset_transport_header(struct sk_buff *skb)
1857{
2e07fa9c 1858 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1859}
1860
967b05f6
ACM
1861static inline void skb_set_transport_header(struct sk_buff *skb,
1862 const int offset)
1863{
2e07fa9c
ACM
1864 skb_reset_transport_header(skb);
1865 skb->transport_header += offset;
ea2ae17d
ACM
1866}
1867
d56f90a7
ACM
1868static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1869{
2e07fa9c 1870 return skb->head + skb->network_header;
d56f90a7
ACM
1871}
1872
c1d2bbe1
ACM
1873static inline void skb_reset_network_header(struct sk_buff *skb)
1874{
2e07fa9c 1875 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1876}
1877
c14d2450
ACM
1878static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1879{
2e07fa9c
ACM
1880 skb_reset_network_header(skb);
1881 skb->network_header += offset;
c14d2450
ACM
1882}
1883
2e07fa9c 1884static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1885{
2e07fa9c 1886 return skb->head + skb->mac_header;
bbe735e4
ACM
1887}
1888
2e07fa9c 1889static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1890{
35d04610 1891 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1892}
1893
1894static inline void skb_reset_mac_header(struct sk_buff *skb)
1895{
1896 skb->mac_header = skb->data - skb->head;
1897}
1898
1899static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1900{
1901 skb_reset_mac_header(skb);
1902 skb->mac_header += offset;
1903}
1904
0e3da5bb
TT
1905static inline void skb_pop_mac_header(struct sk_buff *skb)
1906{
1907 skb->mac_header = skb->network_header;
1908}
1909
fbbdb8f0
YX
1910static inline void skb_probe_transport_header(struct sk_buff *skb,
1911 const int offset_hint)
1912{
1913 struct flow_keys keys;
1914
1915 if (skb_transport_header_was_set(skb))
1916 return;
1917 else if (skb_flow_dissect(skb, &keys))
1918 skb_set_transport_header(skb, keys.thoff);
1919 else
1920 skb_set_transport_header(skb, offset_hint);
1921}
1922
03606895
ED
1923static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1924{
1925 if (skb_mac_header_was_set(skb)) {
1926 const unsigned char *old_mac = skb_mac_header(skb);
1927
1928 skb_set_mac_header(skb, -skb->mac_len);
1929 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1930 }
1931}
1932
04fb451e
MM
1933static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1934{
1935 return skb->csum_start - skb_headroom(skb);
1936}
1937
2e07fa9c
ACM
1938static inline int skb_transport_offset(const struct sk_buff *skb)
1939{
1940 return skb_transport_header(skb) - skb->data;
1941}
1942
1943static inline u32 skb_network_header_len(const struct sk_buff *skb)
1944{
1945 return skb->transport_header - skb->network_header;
1946}
1947
6a674e9c
JG
1948static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1949{
1950 return skb->inner_transport_header - skb->inner_network_header;
1951}
1952
2e07fa9c
ACM
1953static inline int skb_network_offset(const struct sk_buff *skb)
1954{
1955 return skb_network_header(skb) - skb->data;
1956}
48d49d0c 1957
6a674e9c
JG
1958static inline int skb_inner_network_offset(const struct sk_buff *skb)
1959{
1960 return skb_inner_network_header(skb) - skb->data;
1961}
1962
f9599ce1
CG
1963static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1964{
1965 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1966}
1967
1da177e4
LT
1968/*
1969 * CPUs often take a performance hit when accessing unaligned memory
1970 * locations. The actual performance hit varies, it can be small if the
1971 * hardware handles it or large if we have to take an exception and fix it
1972 * in software.
1973 *
1974 * Since an ethernet header is 14 bytes network drivers often end up with
1975 * the IP header at an unaligned offset. The IP header can be aligned by
1976 * shifting the start of the packet by 2 bytes. Drivers should do this
1977 * with:
1978 *
8660c124 1979 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
1980 *
1981 * The downside to this alignment of the IP header is that the DMA is now
1982 * unaligned. On some architectures the cost of an unaligned DMA is high
1983 * and this cost outweighs the gains made by aligning the IP header.
8660c124 1984 *
1da177e4
LT
1985 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1986 * to be overridden.
1987 */
1988#ifndef NET_IP_ALIGN
1989#define NET_IP_ALIGN 2
1990#endif
1991
025be81e
AB
1992/*
1993 * The networking layer reserves some headroom in skb data (via
1994 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1995 * the header has to grow. In the default case, if the header has to grow
d6301d3d 1996 * 32 bytes or less we avoid the reallocation.
025be81e
AB
1997 *
1998 * Unfortunately this headroom changes the DMA alignment of the resulting
1999 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2000 * on some architectures. An architecture can override this value,
2001 * perhaps setting it to a cacheline in size (since that will maintain
2002 * cacheline alignment of the DMA). It must be a power of 2.
2003 *
d6301d3d 2004 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2005 * headroom, you should not reduce this.
5933dd2f
ED
2006 *
2007 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2008 * to reduce average number of cache lines per packet.
2009 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2010 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2011 */
2012#ifndef NET_SKB_PAD
5933dd2f 2013#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2014#endif
2015
7965bd4d 2016int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2017
2018static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2019{
c4264f27 2020 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2021 WARN_ON(1);
2022 return;
2023 }
27a884dc
ACM
2024 skb->len = len;
2025 skb_set_tail_pointer(skb, len);
1da177e4
LT
2026}
2027
7965bd4d 2028void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2029
2030static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2031{
3cc0e873
HX
2032 if (skb->data_len)
2033 return ___pskb_trim(skb, len);
2034 __skb_trim(skb, len);
2035 return 0;
1da177e4
LT
2036}
2037
2038static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2039{
2040 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2041}
2042
e9fa4f7b
HX
2043/**
2044 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2045 * @skb: buffer to alter
2046 * @len: new length
2047 *
2048 * This is identical to pskb_trim except that the caller knows that
2049 * the skb is not cloned so we should never get an error due to out-
2050 * of-memory.
2051 */
2052static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2053{
2054 int err = pskb_trim(skb, len);
2055 BUG_ON(err);
2056}
2057
1da177e4
LT
2058/**
2059 * skb_orphan - orphan a buffer
2060 * @skb: buffer to orphan
2061 *
2062 * If a buffer currently has an owner then we call the owner's
2063 * destructor function and make the @skb unowned. The buffer continues
2064 * to exist but is no longer charged to its former owner.
2065 */
2066static inline void skb_orphan(struct sk_buff *skb)
2067{
c34a7612 2068 if (skb->destructor) {
1da177e4 2069 skb->destructor(skb);
c34a7612
ED
2070 skb->destructor = NULL;
2071 skb->sk = NULL;
376c7311
ED
2072 } else {
2073 BUG_ON(skb->sk);
c34a7612 2074 }
1da177e4
LT
2075}
2076
a353e0ce
MT
2077/**
2078 * skb_orphan_frags - orphan the frags contained in a buffer
2079 * @skb: buffer to orphan frags from
2080 * @gfp_mask: allocation mask for replacement pages
2081 *
2082 * For each frag in the SKB which needs a destructor (i.e. has an
2083 * owner) create a copy of that frag and release the original
2084 * page by calling the destructor.
2085 */
2086static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2087{
2088 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2089 return 0;
2090 return skb_copy_ubufs(skb, gfp_mask);
2091}
2092
1da177e4
LT
2093/**
2094 * __skb_queue_purge - empty a list
2095 * @list: list to empty
2096 *
2097 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2098 * the list and one reference dropped. This function does not take the
2099 * list lock and the caller must hold the relevant locks to use it.
2100 */
7965bd4d 2101void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2102static inline void __skb_queue_purge(struct sk_buff_head *list)
2103{
2104 struct sk_buff *skb;
2105 while ((skb = __skb_dequeue(list)) != NULL)
2106 kfree_skb(skb);
2107}
2108
e5e67305
AD
2109#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2110#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2111#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2112
7965bd4d 2113void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2114
7965bd4d
JP
2115struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2116 gfp_t gfp_mask);
8af27456
CH
2117
2118/**
2119 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2120 * @dev: network device to receive on
2121 * @length: length to allocate
2122 *
2123 * Allocate a new &sk_buff and assign it a usage count of one. The
2124 * buffer has unspecified headroom built in. Users should allocate
2125 * the headroom they think they need without accounting for the
2126 * built in space. The built in space is used for optimisations.
2127 *
2128 * %NULL is returned if there is no free memory. Although this function
2129 * allocates memory it can be called from an interrupt.
2130 */
2131static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2132 unsigned int length)
8af27456
CH
2133{
2134 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2135}
2136
6f532612
ED
2137/* legacy helper around __netdev_alloc_skb() */
2138static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2139 gfp_t gfp_mask)
2140{
2141 return __netdev_alloc_skb(NULL, length, gfp_mask);
2142}
2143
2144/* legacy helper around netdev_alloc_skb() */
2145static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2146{
2147 return netdev_alloc_skb(NULL, length);
2148}
2149
2150
4915a0de
ED
2151static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2152 unsigned int length, gfp_t gfp)
61321bbd 2153{
4915a0de 2154 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2155
2156 if (NET_IP_ALIGN && skb)
2157 skb_reserve(skb, NET_IP_ALIGN);
2158 return skb;
2159}
2160
4915a0de
ED
2161static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2162 unsigned int length)
2163{
2164 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2165}
2166
71dfda58
AD
2167/**
2168 * __dev_alloc_pages - allocate page for network Rx
2169 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2170 * @order: size of the allocation
2171 *
2172 * Allocate a new page.
2173 *
2174 * %NULL is returned if there is no free memory.
2175*/
2176static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2177 unsigned int order)
2178{
2179 /* This piece of code contains several assumptions.
2180 * 1. This is for device Rx, therefor a cold page is preferred.
2181 * 2. The expectation is the user wants a compound page.
2182 * 3. If requesting a order 0 page it will not be compound
2183 * due to the check to see if order has a value in prep_new_page
2184 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2185 * code in gfp_to_alloc_flags that should be enforcing this.
2186 */
2187 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2188
2189 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2190}
2191
2192static inline struct page *dev_alloc_pages(unsigned int order)
2193{
2194 return __dev_alloc_pages(GFP_ATOMIC, order);
2195}
2196
2197/**
2198 * __dev_alloc_page - allocate a page for network Rx
2199 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2200 *
2201 * Allocate a new page.
2202 *
2203 * %NULL is returned if there is no free memory.
2204 */
2205static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2206{
2207 return __dev_alloc_pages(gfp_mask, 0);
2208}
2209
2210static inline struct page *dev_alloc_page(void)
2211{
2212 return __dev_alloc_page(GFP_ATOMIC);
2213}
2214
0614002b
MG
2215/**
2216 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2217 * @page: The page that was allocated from skb_alloc_page
2218 * @skb: The skb that may need pfmemalloc set
2219 */
2220static inline void skb_propagate_pfmemalloc(struct page *page,
2221 struct sk_buff *skb)
2222{
2223 if (page && page->pfmemalloc)
2224 skb->pfmemalloc = true;
2225}
2226
131ea667 2227/**
e227867f 2228 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2229 * @frag: the paged fragment
2230 *
2231 * Returns the &struct page associated with @frag.
2232 */
2233static inline struct page *skb_frag_page(const skb_frag_t *frag)
2234{
a8605c60 2235 return frag->page.p;
131ea667
IC
2236}
2237
2238/**
2239 * __skb_frag_ref - take an addition reference on a paged fragment.
2240 * @frag: the paged fragment
2241 *
2242 * Takes an additional reference on the paged fragment @frag.
2243 */
2244static inline void __skb_frag_ref(skb_frag_t *frag)
2245{
2246 get_page(skb_frag_page(frag));
2247}
2248
2249/**
2250 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2251 * @skb: the buffer
2252 * @f: the fragment offset.
2253 *
2254 * Takes an additional reference on the @f'th paged fragment of @skb.
2255 */
2256static inline void skb_frag_ref(struct sk_buff *skb, int f)
2257{
2258 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2259}
2260
2261/**
2262 * __skb_frag_unref - release a reference on a paged fragment.
2263 * @frag: the paged fragment
2264 *
2265 * Releases a reference on the paged fragment @frag.
2266 */
2267static inline void __skb_frag_unref(skb_frag_t *frag)
2268{
2269 put_page(skb_frag_page(frag));
2270}
2271
2272/**
2273 * skb_frag_unref - release a reference on a paged fragment of an skb.
2274 * @skb: the buffer
2275 * @f: the fragment offset
2276 *
2277 * Releases a reference on the @f'th paged fragment of @skb.
2278 */
2279static inline void skb_frag_unref(struct sk_buff *skb, int f)
2280{
2281 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2282}
2283
2284/**
2285 * skb_frag_address - gets the address of the data contained in a paged fragment
2286 * @frag: the paged fragment buffer
2287 *
2288 * Returns the address of the data within @frag. The page must already
2289 * be mapped.
2290 */
2291static inline void *skb_frag_address(const skb_frag_t *frag)
2292{
2293 return page_address(skb_frag_page(frag)) + frag->page_offset;
2294}
2295
2296/**
2297 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2298 * @frag: the paged fragment buffer
2299 *
2300 * Returns the address of the data within @frag. Checks that the page
2301 * is mapped and returns %NULL otherwise.
2302 */
2303static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2304{
2305 void *ptr = page_address(skb_frag_page(frag));
2306 if (unlikely(!ptr))
2307 return NULL;
2308
2309 return ptr + frag->page_offset;
2310}
2311
2312/**
2313 * __skb_frag_set_page - sets the page contained in a paged fragment
2314 * @frag: the paged fragment
2315 * @page: the page to set
2316 *
2317 * Sets the fragment @frag to contain @page.
2318 */
2319static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2320{
a8605c60 2321 frag->page.p = page;
131ea667
IC
2322}
2323
2324/**
2325 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2326 * @skb: the buffer
2327 * @f: the fragment offset
2328 * @page: the page to set
2329 *
2330 * Sets the @f'th fragment of @skb to contain @page.
2331 */
2332static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2333 struct page *page)
2334{
2335 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2336}
2337
400dfd3a
ED
2338bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2339
131ea667
IC
2340/**
2341 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2342 * @dev: the device to map the fragment to
131ea667
IC
2343 * @frag: the paged fragment to map
2344 * @offset: the offset within the fragment (starting at the
2345 * fragment's own offset)
2346 * @size: the number of bytes to map
f83347df 2347 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2348 *
2349 * Maps the page associated with @frag to @device.
2350 */
2351static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2352 const skb_frag_t *frag,
2353 size_t offset, size_t size,
2354 enum dma_data_direction dir)
2355{
2356 return dma_map_page(dev, skb_frag_page(frag),
2357 frag->page_offset + offset, size, dir);
2358}
2359
117632e6
ED
2360static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2361 gfp_t gfp_mask)
2362{
2363 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2364}
2365
bad93e9d
OP
2366
2367static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2368 gfp_t gfp_mask)
2369{
2370 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2371}
2372
2373
334a8132
PM
2374/**
2375 * skb_clone_writable - is the header of a clone writable
2376 * @skb: buffer to check
2377 * @len: length up to which to write
2378 *
2379 * Returns true if modifying the header part of the cloned buffer
2380 * does not requires the data to be copied.
2381 */
05bdd2f1 2382static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2383{
2384 return !skb_header_cloned(skb) &&
2385 skb_headroom(skb) + len <= skb->hdr_len;
2386}
2387
d9cc2048
HX
2388static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2389 int cloned)
2390{
2391 int delta = 0;
2392
d9cc2048
HX
2393 if (headroom > skb_headroom(skb))
2394 delta = headroom - skb_headroom(skb);
2395
2396 if (delta || cloned)
2397 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2398 GFP_ATOMIC);
2399 return 0;
2400}
2401
1da177e4
LT
2402/**
2403 * skb_cow - copy header of skb when it is required
2404 * @skb: buffer to cow
2405 * @headroom: needed headroom
2406 *
2407 * If the skb passed lacks sufficient headroom or its data part
2408 * is shared, data is reallocated. If reallocation fails, an error
2409 * is returned and original skb is not changed.
2410 *
2411 * The result is skb with writable area skb->head...skb->tail
2412 * and at least @headroom of space at head.
2413 */
2414static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2415{
d9cc2048
HX
2416 return __skb_cow(skb, headroom, skb_cloned(skb));
2417}
1da177e4 2418
d9cc2048
HX
2419/**
2420 * skb_cow_head - skb_cow but only making the head writable
2421 * @skb: buffer to cow
2422 * @headroom: needed headroom
2423 *
2424 * This function is identical to skb_cow except that we replace the
2425 * skb_cloned check by skb_header_cloned. It should be used when
2426 * you only need to push on some header and do not need to modify
2427 * the data.
2428 */
2429static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2430{
2431 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2432}
2433
2434/**
2435 * skb_padto - pad an skbuff up to a minimal size
2436 * @skb: buffer to pad
2437 * @len: minimal length
2438 *
2439 * Pads up a buffer to ensure the trailing bytes exist and are
2440 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2441 * is untouched. Otherwise it is extended. Returns zero on
2442 * success. The skb is freed on error.
1da177e4 2443 */
5b057c6b 2444static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2445{
2446 unsigned int size = skb->len;
2447 if (likely(size >= len))
5b057c6b 2448 return 0;
987c402a 2449 return skb_pad(skb, len - size);
1da177e4
LT
2450}
2451
9c0c1124
AD
2452/**
2453 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2454 * @skb: buffer to pad
2455 * @len: minimal length
2456 *
2457 * Pads up a buffer to ensure the trailing bytes exist and are
2458 * blanked. If the buffer already contains sufficient data it
2459 * is untouched. Otherwise it is extended. Returns zero on
2460 * success. The skb is freed on error.
2461 */
2462static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2463{
2464 unsigned int size = skb->len;
2465
2466 if (unlikely(size < len)) {
2467 len -= size;
2468 if (skb_pad(skb, len))
2469 return -ENOMEM;
2470 __skb_put(skb, len);
2471 }
2472 return 0;
2473}
2474
1da177e4
LT
2475static inline int skb_add_data(struct sk_buff *skb,
2476 char __user *from, int copy)
2477{
2478 const int off = skb->len;
2479
2480 if (skb->ip_summed == CHECKSUM_NONE) {
2481 int err = 0;
5084205f 2482 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1da177e4
LT
2483 copy, 0, &err);
2484 if (!err) {
2485 skb->csum = csum_block_add(skb->csum, csum, off);
2486 return 0;
2487 }
2488 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2489 return 0;
2490
2491 __skb_trim(skb, off);
2492 return -EFAULT;
2493}
2494
38ba0a65
ED
2495static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2496 const struct page *page, int off)
1da177e4
LT
2497{
2498 if (i) {
9e903e08 2499 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2500
ea2ab693 2501 return page == skb_frag_page(frag) &&
9e903e08 2502 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2503 }
38ba0a65 2504 return false;
1da177e4
LT
2505}
2506
364c6bad
HX
2507static inline int __skb_linearize(struct sk_buff *skb)
2508{
2509 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2510}
2511
1da177e4
LT
2512/**
2513 * skb_linearize - convert paged skb to linear one
2514 * @skb: buffer to linarize
1da177e4
LT
2515 *
2516 * If there is no free memory -ENOMEM is returned, otherwise zero
2517 * is returned and the old skb data released.
2518 */
364c6bad
HX
2519static inline int skb_linearize(struct sk_buff *skb)
2520{
2521 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2522}
2523
cef401de
ED
2524/**
2525 * skb_has_shared_frag - can any frag be overwritten
2526 * @skb: buffer to test
2527 *
2528 * Return true if the skb has at least one frag that might be modified
2529 * by an external entity (as in vmsplice()/sendfile())
2530 */
2531static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2532{
c9af6db4
PS
2533 return skb_is_nonlinear(skb) &&
2534 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2535}
2536
364c6bad
HX
2537/**
2538 * skb_linearize_cow - make sure skb is linear and writable
2539 * @skb: buffer to process
2540 *
2541 * If there is no free memory -ENOMEM is returned, otherwise zero
2542 * is returned and the old skb data released.
2543 */
2544static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2545{
364c6bad
HX
2546 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2547 __skb_linearize(skb) : 0;
1da177e4
LT
2548}
2549
2550/**
2551 * skb_postpull_rcsum - update checksum for received skb after pull
2552 * @skb: buffer to update
2553 * @start: start of data before pull
2554 * @len: length of data pulled
2555 *
2556 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2557 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2558 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2559 */
2560
2561static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2562 const void *start, unsigned int len)
1da177e4 2563{
84fa7933 2564 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2565 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2566}
2567
cbb042f9
HX
2568unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2569
7ce5a27f
DM
2570/**
2571 * pskb_trim_rcsum - trim received skb and update checksum
2572 * @skb: buffer to trim
2573 * @len: new length
2574 *
2575 * This is exactly the same as pskb_trim except that it ensures the
2576 * checksum of received packets are still valid after the operation.
2577 */
2578
2579static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2580{
2581 if (likely(len >= skb->len))
2582 return 0;
2583 if (skb->ip_summed == CHECKSUM_COMPLETE)
2584 skb->ip_summed = CHECKSUM_NONE;
2585 return __pskb_trim(skb, len);
2586}
2587
1da177e4
LT
2588#define skb_queue_walk(queue, skb) \
2589 for (skb = (queue)->next; \
a1e4891f 2590 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2591 skb = skb->next)
2592
46f8914e
JC
2593#define skb_queue_walk_safe(queue, skb, tmp) \
2594 for (skb = (queue)->next, tmp = skb->next; \
2595 skb != (struct sk_buff *)(queue); \
2596 skb = tmp, tmp = skb->next)
2597
1164f52a 2598#define skb_queue_walk_from(queue, skb) \
a1e4891f 2599 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2600 skb = skb->next)
2601
2602#define skb_queue_walk_from_safe(queue, skb, tmp) \
2603 for (tmp = skb->next; \
2604 skb != (struct sk_buff *)(queue); \
2605 skb = tmp, tmp = skb->next)
2606
300ce174
SH
2607#define skb_queue_reverse_walk(queue, skb) \
2608 for (skb = (queue)->prev; \
a1e4891f 2609 skb != (struct sk_buff *)(queue); \
300ce174
SH
2610 skb = skb->prev)
2611
686a2955
DM
2612#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2613 for (skb = (queue)->prev, tmp = skb->prev; \
2614 skb != (struct sk_buff *)(queue); \
2615 skb = tmp, tmp = skb->prev)
2616
2617#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2618 for (tmp = skb->prev; \
2619 skb != (struct sk_buff *)(queue); \
2620 skb = tmp, tmp = skb->prev)
1da177e4 2621
21dc3301 2622static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2623{
2624 return skb_shinfo(skb)->frag_list != NULL;
2625}
2626
2627static inline void skb_frag_list_init(struct sk_buff *skb)
2628{
2629 skb_shinfo(skb)->frag_list = NULL;
2630}
2631
2632static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2633{
2634 frag->next = skb_shinfo(skb)->frag_list;
2635 skb_shinfo(skb)->frag_list = frag;
2636}
2637
2638#define skb_walk_frags(skb, iter) \
2639 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2640
7965bd4d
JP
2641struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2642 int *peeked, int *off, int *err);
2643struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2644 int *err);
2645unsigned int datagram_poll(struct file *file, struct socket *sock,
2646 struct poll_table_struct *wait);
c0371da6
AV
2647int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2648 struct iov_iter *to, int size);
51f3d02b
DM
2649static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2650 struct msghdr *msg, int size)
2651{
e5a4b0bb 2652 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 2653}
e5a4b0bb
AV
2654int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2655 struct msghdr *msg);
3a654f97
AV
2656int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2657 struct iov_iter *from, int len);
3a654f97 2658int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d
JP
2659void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2660void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2661int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2662int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2663int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2664__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2665 int len, __wsum csum);
2666int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2667 struct pipe_inode_info *pipe, unsigned int len,
2668 unsigned int flags);
2669void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2670unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2671int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2672 int len, int hlen);
7965bd4d
JP
2673void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2674int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2675void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2676unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2677struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2678struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 2679int skb_ensure_writable(struct sk_buff *skb, int write_len);
93515d53
JP
2680int skb_vlan_pop(struct sk_buff *skb);
2681int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
20380731 2682
6ce8e9ce
AV
2683static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2684{
c0371da6
AV
2685 /* XXX: stripping const */
2686 return memcpy_fromiovec(data, (struct iovec *)msg->msg_iter.iov, len);
6ce8e9ce
AV
2687}
2688
7eab8d9e
AV
2689static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2690{
e5a4b0bb 2691 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
2692}
2693
2817a336
DB
2694struct skb_checksum_ops {
2695 __wsum (*update)(const void *mem, int len, __wsum wsum);
2696 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2697};
2698
2699__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2700 __wsum csum, const struct skb_checksum_ops *ops);
2701__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2702 __wsum csum);
2703
690e36e7
DM
2704static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2705 int len, void *data, int hlen, void *buffer)
1da177e4 2706{
55820ee2 2707 if (hlen - offset >= len)
690e36e7 2708 return data + offset;
1da177e4 2709
690e36e7
DM
2710 if (!skb ||
2711 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2712 return NULL;
2713
2714 return buffer;
2715}
2716
690e36e7
DM
2717static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2718 int len, void *buffer)
2719{
2720 return __skb_header_pointer(skb, offset, len, skb->data,
2721 skb_headlen(skb), buffer);
2722}
2723
4262e5cc
DB
2724/**
2725 * skb_needs_linearize - check if we need to linearize a given skb
2726 * depending on the given device features.
2727 * @skb: socket buffer to check
2728 * @features: net device features
2729 *
2730 * Returns true if either:
2731 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2732 * 2. skb is fragmented and the device does not support SG.
2733 */
2734static inline bool skb_needs_linearize(struct sk_buff *skb,
2735 netdev_features_t features)
2736{
2737 return skb_is_nonlinear(skb) &&
2738 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2739 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2740}
2741
d626f62b
ACM
2742static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2743 void *to,
2744 const unsigned int len)
2745{
2746 memcpy(to, skb->data, len);
2747}
2748
2749static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2750 const int offset, void *to,
2751 const unsigned int len)
2752{
2753 memcpy(to, skb->data + offset, len);
2754}
2755
27d7ff46
ACM
2756static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2757 const void *from,
2758 const unsigned int len)
2759{
2760 memcpy(skb->data, from, len);
2761}
2762
2763static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2764 const int offset,
2765 const void *from,
2766 const unsigned int len)
2767{
2768 memcpy(skb->data + offset, from, len);
2769}
2770
7965bd4d 2771void skb_init(void);
1da177e4 2772
ac45f602
PO
2773static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2774{
2775 return skb->tstamp;
2776}
2777
a61bbcf2
PM
2778/**
2779 * skb_get_timestamp - get timestamp from a skb
2780 * @skb: skb to get stamp from
2781 * @stamp: pointer to struct timeval to store stamp in
2782 *
2783 * Timestamps are stored in the skb as offsets to a base timestamp.
2784 * This function converts the offset back to a struct timeval and stores
2785 * it in stamp.
2786 */
ac45f602
PO
2787static inline void skb_get_timestamp(const struct sk_buff *skb,
2788 struct timeval *stamp)
a61bbcf2 2789{
b7aa0bf7 2790 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2791}
2792
ac45f602
PO
2793static inline void skb_get_timestampns(const struct sk_buff *skb,
2794 struct timespec *stamp)
2795{
2796 *stamp = ktime_to_timespec(skb->tstamp);
2797}
2798
b7aa0bf7 2799static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2800{
b7aa0bf7 2801 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2802}
2803
164891aa
SH
2804static inline ktime_t net_timedelta(ktime_t t)
2805{
2806 return ktime_sub(ktime_get_real(), t);
2807}
2808
b9ce204f
IJ
2809static inline ktime_t net_invalid_timestamp(void)
2810{
2811 return ktime_set(0, 0);
2812}
a61bbcf2 2813
62bccb8c
AD
2814struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2815
c1f19b51
RC
2816#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2817
7965bd4d
JP
2818void skb_clone_tx_timestamp(struct sk_buff *skb);
2819bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2820
2821#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2822
2823static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2824{
2825}
2826
2827static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2828{
2829 return false;
2830}
2831
2832#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2833
2834/**
2835 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2836 *
da92b194
RC
2837 * PHY drivers may accept clones of transmitted packets for
2838 * timestamping via their phy_driver.txtstamp method. These drivers
2839 * must call this function to return the skb back to the stack, with
2840 * or without a timestamp.
2841 *
c1f19b51 2842 * @skb: clone of the the original outgoing packet
da92b194 2843 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2844 *
2845 */
2846void skb_complete_tx_timestamp(struct sk_buff *skb,
2847 struct skb_shared_hwtstamps *hwtstamps);
2848
e7fd2885
WB
2849void __skb_tstamp_tx(struct sk_buff *orig_skb,
2850 struct skb_shared_hwtstamps *hwtstamps,
2851 struct sock *sk, int tstype);
2852
ac45f602
PO
2853/**
2854 * skb_tstamp_tx - queue clone of skb with send time stamps
2855 * @orig_skb: the original outgoing packet
2856 * @hwtstamps: hardware time stamps, may be NULL if not available
2857 *
2858 * If the skb has a socket associated, then this function clones the
2859 * skb (thus sharing the actual data and optional structures), stores
2860 * the optional hardware time stamping information (if non NULL) or
2861 * generates a software time stamp (otherwise), then queues the clone
2862 * to the error queue of the socket. Errors are silently ignored.
2863 */
7965bd4d
JP
2864void skb_tstamp_tx(struct sk_buff *orig_skb,
2865 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2866
4507a715
RC
2867static inline void sw_tx_timestamp(struct sk_buff *skb)
2868{
2244d07b
OH
2869 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2870 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2871 skb_tstamp_tx(skb, NULL);
2872}
2873
2874/**
2875 * skb_tx_timestamp() - Driver hook for transmit timestamping
2876 *
2877 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2878 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2879 *
73409f3b
DM
2880 * Specifically, one should make absolutely sure that this function is
2881 * called before TX completion of this packet can trigger. Otherwise
2882 * the packet could potentially already be freed.
2883 *
4507a715
RC
2884 * @skb: A socket buffer.
2885 */
2886static inline void skb_tx_timestamp(struct sk_buff *skb)
2887{
c1f19b51 2888 skb_clone_tx_timestamp(skb);
4507a715
RC
2889 sw_tx_timestamp(skb);
2890}
2891
6e3e939f
JB
2892/**
2893 * skb_complete_wifi_ack - deliver skb with wifi status
2894 *
2895 * @skb: the original outgoing packet
2896 * @acked: ack status
2897 *
2898 */
2899void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2900
7965bd4d
JP
2901__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2902__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2903
60476372
HX
2904static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2905{
5d0c2b95 2906 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
60476372
HX
2907}
2908
fb286bb2
HX
2909/**
2910 * skb_checksum_complete - Calculate checksum of an entire packet
2911 * @skb: packet to process
2912 *
2913 * This function calculates the checksum over the entire packet plus
2914 * the value of skb->csum. The latter can be used to supply the
2915 * checksum of a pseudo header as used by TCP/UDP. It returns the
2916 * checksum.
2917 *
2918 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2919 * this function can be used to verify that checksum on received
2920 * packets. In that case the function should return zero if the
2921 * checksum is correct. In particular, this function will return zero
2922 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2923 * hardware has already verified the correctness of the checksum.
2924 */
4381ca3c 2925static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2926{
60476372
HX
2927 return skb_csum_unnecessary(skb) ?
2928 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2929}
2930
77cffe23
TH
2931static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2932{
2933 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2934 if (skb->csum_level == 0)
2935 skb->ip_summed = CHECKSUM_NONE;
2936 else
2937 skb->csum_level--;
2938 }
2939}
2940
2941static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2942{
2943 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2944 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2945 skb->csum_level++;
2946 } else if (skb->ip_summed == CHECKSUM_NONE) {
2947 skb->ip_summed = CHECKSUM_UNNECESSARY;
2948 skb->csum_level = 0;
2949 }
2950}
2951
5a212329
TH
2952static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2953{
2954 /* Mark current checksum as bad (typically called from GRO
2955 * path). In the case that ip_summed is CHECKSUM_NONE
2956 * this must be the first checksum encountered in the packet.
2957 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2958 * checksum after the last one validated. For UDP, a zero
2959 * checksum can not be marked as bad.
2960 */
2961
2962 if (skb->ip_summed == CHECKSUM_NONE ||
2963 skb->ip_summed == CHECKSUM_UNNECESSARY)
2964 skb->csum_bad = 1;
2965}
2966
76ba0aae
TH
2967/* Check if we need to perform checksum complete validation.
2968 *
2969 * Returns true if checksum complete is needed, false otherwise
2970 * (either checksum is unnecessary or zero checksum is allowed).
2971 */
2972static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2973 bool zero_okay,
2974 __sum16 check)
2975{
5d0c2b95
TH
2976 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2977 skb->csum_valid = 1;
77cffe23 2978 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
2979 return false;
2980 }
2981
2982 return true;
2983}
2984
2985/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2986 * in checksum_init.
2987 */
2988#define CHECKSUM_BREAK 76
2989
2990/* Validate (init) checksum based on checksum complete.
2991 *
2992 * Return values:
2993 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2994 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2995 * checksum is stored in skb->csum for use in __skb_checksum_complete
2996 * non-zero: value of invalid checksum
2997 *
2998 */
2999static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3000 bool complete,
3001 __wsum psum)
3002{
3003 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3004 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3005 skb->csum_valid = 1;
76ba0aae
TH
3006 return 0;
3007 }
5a212329
TH
3008 } else if (skb->csum_bad) {
3009 /* ip_summed == CHECKSUM_NONE in this case */
3010 return 1;
76ba0aae
TH
3011 }
3012
3013 skb->csum = psum;
3014
5d0c2b95
TH
3015 if (complete || skb->len <= CHECKSUM_BREAK) {
3016 __sum16 csum;
3017
3018 csum = __skb_checksum_complete(skb);
3019 skb->csum_valid = !csum;
3020 return csum;
3021 }
76ba0aae
TH
3022
3023 return 0;
3024}
3025
3026static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3027{
3028 return 0;
3029}
3030
3031/* Perform checksum validate (init). Note that this is a macro since we only
3032 * want to calculate the pseudo header which is an input function if necessary.
3033 * First we try to validate without any computation (checksum unnecessary) and
3034 * then calculate based on checksum complete calling the function to compute
3035 * pseudo header.
3036 *
3037 * Return values:
3038 * 0: checksum is validated or try to in skb_checksum_complete
3039 * non-zero: value of invalid checksum
3040 */
3041#define __skb_checksum_validate(skb, proto, complete, \
3042 zero_okay, check, compute_pseudo) \
3043({ \
3044 __sum16 __ret = 0; \
5d0c2b95 3045 skb->csum_valid = 0; \
76ba0aae
TH
3046 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3047 __ret = __skb_checksum_validate_complete(skb, \
3048 complete, compute_pseudo(skb, proto)); \
3049 __ret; \
3050})
3051
3052#define skb_checksum_init(skb, proto, compute_pseudo) \
3053 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3054
3055#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3056 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3057
3058#define skb_checksum_validate(skb, proto, compute_pseudo) \
3059 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3060
3061#define skb_checksum_validate_zero_check(skb, proto, check, \
3062 compute_pseudo) \
3063 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
3064
3065#define skb_checksum_simple_validate(skb) \
3066 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3067
d96535a1
TH
3068static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3069{
3070 return (skb->ip_summed == CHECKSUM_NONE &&
3071 skb->csum_valid && !skb->csum_bad);
3072}
3073
3074static inline void __skb_checksum_convert(struct sk_buff *skb,
3075 __sum16 check, __wsum pseudo)
3076{
3077 skb->csum = ~pseudo;
3078 skb->ip_summed = CHECKSUM_COMPLETE;
3079}
3080
3081#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3082do { \
3083 if (__skb_checksum_convert_check(skb)) \
3084 __skb_checksum_convert(skb, check, \
3085 compute_pseudo(skb, proto)); \
3086} while (0)
3087
5f79e0f9 3088#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3089void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3090static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3091{
3092 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3093 nf_conntrack_destroy(nfct);
1da177e4
LT
3094}
3095static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3096{
3097 if (nfct)
3098 atomic_inc(&nfct->use);
3099}
2fc72c7b 3100#endif
34666d46 3101#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3102static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3103{
3104 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3105 kfree(nf_bridge);
3106}
3107static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3108{
3109 if (nf_bridge)
3110 atomic_inc(&nf_bridge->use);
3111}
3112#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3113static inline void nf_reset(struct sk_buff *skb)
3114{
5f79e0f9 3115#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3116 nf_conntrack_put(skb->nfct);
3117 skb->nfct = NULL;
2fc72c7b 3118#endif
34666d46 3119#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3120 nf_bridge_put(skb->nf_bridge);
3121 skb->nf_bridge = NULL;
3122#endif
3123}
3124
124dff01
PM
3125static inline void nf_reset_trace(struct sk_buff *skb)
3126{
478b360a 3127#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3128 skb->nf_trace = 0;
3129#endif
a193a4ab
PM
3130}
3131
edda553c 3132/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3133static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3134 bool copy)
edda553c 3135{
5f79e0f9 3136#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3137 dst->nfct = src->nfct;
3138 nf_conntrack_get(src->nfct);
b1937227
ED
3139 if (copy)
3140 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3141#endif
34666d46 3142#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3143 dst->nf_bridge = src->nf_bridge;
3144 nf_bridge_get(src->nf_bridge);
3145#endif
478b360a 3146#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3147 if (copy)
3148 dst->nf_trace = src->nf_trace;
478b360a 3149#endif
edda553c
YK
3150}
3151
e7ac05f3
YK
3152static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3153{
e7ac05f3 3154#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3155 nf_conntrack_put(dst->nfct);
2fc72c7b 3156#endif
34666d46 3157#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3158 nf_bridge_put(dst->nf_bridge);
3159#endif
b1937227 3160 __nf_copy(dst, src, true);
e7ac05f3
YK
3161}
3162
984bc16c
JM
3163#ifdef CONFIG_NETWORK_SECMARK
3164static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3165{
3166 to->secmark = from->secmark;
3167}
3168
3169static inline void skb_init_secmark(struct sk_buff *skb)
3170{
3171 skb->secmark = 0;
3172}
3173#else
3174static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3175{ }
3176
3177static inline void skb_init_secmark(struct sk_buff *skb)
3178{ }
3179#endif
3180
574f7194
EB
3181static inline bool skb_irq_freeable(const struct sk_buff *skb)
3182{
3183 return !skb->destructor &&
3184#if IS_ENABLED(CONFIG_XFRM)
3185 !skb->sp &&
3186#endif
3187#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3188 !skb->nfct &&
3189#endif
3190 !skb->_skb_refdst &&
3191 !skb_has_frag_list(skb);
3192}
3193
f25f4e44
PWJ
3194static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3195{
f25f4e44 3196 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3197}
3198
9247744e 3199static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3200{
4e3ab47a 3201 return skb->queue_mapping;
4e3ab47a
PE
3202}
3203
f25f4e44
PWJ
3204static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3205{
f25f4e44 3206 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3207}
3208
d5a9e24a
DM
3209static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3210{
3211 skb->queue_mapping = rx_queue + 1;
3212}
3213
9247744e 3214static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3215{
3216 return skb->queue_mapping - 1;
3217}
3218
9247744e 3219static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3220{
a02cec21 3221 return skb->queue_mapping != 0;
d5a9e24a
DM
3222}
3223
0e001614 3224u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
7965bd4d 3225 unsigned int num_tx_queues);
9247744e 3226
def8b4fa
AD
3227static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3228{
0b3d8e08 3229#ifdef CONFIG_XFRM
def8b4fa 3230 return skb->sp;
def8b4fa 3231#else
def8b4fa 3232 return NULL;
def8b4fa 3233#endif
0b3d8e08 3234}
def8b4fa 3235
68c33163
PS
3236/* Keeps track of mac header offset relative to skb->head.
3237 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3238 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3239 * tunnel skb it points to outer mac header.
3240 * Keeps track of level of encapsulation of network headers.
3241 */
68c33163 3242struct skb_gso_cb {
3347c960
ED
3243 int mac_offset;
3244 int encap_level;
7e2b10c1 3245 __u16 csum_start;
68c33163
PS
3246};
3247#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3248
3249static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3250{
3251 return (skb_mac_header(inner_skb) - inner_skb->head) -
3252 SKB_GSO_CB(inner_skb)->mac_offset;
3253}
3254
1e2bd517
PS
3255static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3256{
3257 int new_headroom, headroom;
3258 int ret;
3259
3260 headroom = skb_headroom(skb);
3261 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3262 if (ret)
3263 return ret;
3264
3265 new_headroom = skb_headroom(skb);
3266 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3267 return 0;
3268}
3269
7e2b10c1
TH
3270/* Compute the checksum for a gso segment. First compute the checksum value
3271 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3272 * then add in skb->csum (checksum from csum_start to end of packet).
3273 * skb->csum and csum_start are then updated to reflect the checksum of the
3274 * resultant packet starting from the transport header-- the resultant checksum
3275 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3276 * header.
3277 */
3278static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3279{
3280 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3281 skb_transport_offset(skb);
3282 __u16 csum;
3283
3284 csum = csum_fold(csum_partial(skb_transport_header(skb),
3285 plen, skb->csum));
3286 skb->csum = res;
3287 SKB_GSO_CB(skb)->csum_start -= plen;
3288
3289 return csum;
3290}
3291
bdcc0924 3292static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3293{
3294 return skb_shinfo(skb)->gso_size;
3295}
3296
36a8f39e 3297/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3298static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3299{
3300 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3301}
3302
7965bd4d 3303void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3304
3305static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3306{
3307 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3308 * wanted then gso_type will be set. */
05bdd2f1
ED
3309 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3310
b78462eb
AD
3311 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3312 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3313 __skb_warn_lro_forwarding(skb);
3314 return true;
3315 }
3316 return false;
3317}
3318
35fc92a9
HX
3319static inline void skb_forward_csum(struct sk_buff *skb)
3320{
3321 /* Unfortunately we don't support this one. Any brave souls? */
3322 if (skb->ip_summed == CHECKSUM_COMPLETE)
3323 skb->ip_summed = CHECKSUM_NONE;
3324}
3325
bc8acf2c
ED
3326/**
3327 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3328 * @skb: skb to check
3329 *
3330 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3331 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3332 * use this helper, to document places where we make this assertion.
3333 */
05bdd2f1 3334static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3335{
3336#ifdef DEBUG
3337 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3338#endif
3339}
3340
f35d9d8a 3341bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3342
ed1f50c3
PD
3343int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3344
56193d1b
AD
3345u32 skb_get_poff(const struct sk_buff *skb);
3346u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3347 const struct flow_keys *keys, int hlen);
f77668dc 3348
3a7c1ee4
AD
3349/**
3350 * skb_head_is_locked - Determine if the skb->head is locked down
3351 * @skb: skb to check
3352 *
3353 * The head on skbs build around a head frag can be removed if they are
3354 * not cloned. This function returns true if the skb head is locked down
3355 * due to either being allocated via kmalloc, or by being a clone with
3356 * multiple references to the head.
3357 */
3358static inline bool skb_head_is_locked(const struct sk_buff *skb)
3359{
3360 return !skb->head_frag || skb_cloned(skb);
3361}
fe6cc55f
FW
3362
3363/**
3364 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3365 *
3366 * @skb: GSO skb
3367 *
3368 * skb_gso_network_seglen is used to determine the real size of the
3369 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3370 *
3371 * The MAC/L2 header is not accounted for.
3372 */
3373static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3374{
3375 unsigned int hdr_len = skb_transport_header(skb) -
3376 skb_network_header(skb);
3377 return hdr_len + skb_gso_transport_seglen(skb);
3378}
1da177e4
LT
3379#endif /* __KERNEL__ */
3380#endif /* _LINUX_SKBUFF_H */