]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/skbuff.h
netfilter: bridge: detect NAT66 correctly and change MAC address
[mirror_ubuntu-bionic-kernel.git] / include / linux / skbuff.h
CommitLineData
1da177e4
LT
1/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
1da177e4 17#include <linux/kernel.h>
fe55f6d5 18#include <linux/kmemcheck.h>
1da177e4
LT
19#include <linux/compiler.h>
20#include <linux/time.h>
187f1882 21#include <linux/bug.h>
1da177e4 22#include <linux/cache.h>
56b17425 23#include <linux/rbtree.h>
51f3d02b 24#include <linux/socket.h>
1da177e4 25
60063497 26#include <linux/atomic.h>
1da177e4
LT
27#include <asm/types.h>
28#include <linux/spinlock.h>
1da177e4 29#include <linux/net.h>
3fc7e8a6 30#include <linux/textsearch.h>
1da177e4 31#include <net/checksum.h>
a80958f4 32#include <linux/rcupdate.h>
b7aa0bf7 33#include <linux/hrtimer.h>
131ea667 34#include <linux/dma-mapping.h>
c8f44aff 35#include <linux/netdev_features.h>
363ec392 36#include <linux/sched.h>
1bd758eb 37#include <net/flow_dissector.h>
a60e3cc7 38#include <linux/splice.h>
72b31f72 39#include <linux/in6.h>
1da177e4 40
78ea85f1
DB
41/* A. Checksumming of received packets by device.
42 *
43 * CHECKSUM_NONE:
44 *
45 * Device failed to checksum this packet e.g. due to lack of capabilities.
46 * The packet contains full (though not verified) checksum in packet but
47 * not in skb->csum. Thus, skb->csum is undefined in this case.
48 *
49 * CHECKSUM_UNNECESSARY:
50 *
51 * The hardware you're dealing with doesn't calculate the full checksum
52 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
53 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
54 * if their checksums are okay. skb->csum is still undefined in this case
55 * though. It is a bad option, but, unfortunately, nowadays most vendors do
56 * this. Apparently with the secret goal to sell you new devices, when you
57 * will add new protocol to your host, f.e. IPv6 8)
58 *
59 * CHECKSUM_UNNECESSARY is applicable to following protocols:
60 * TCP: IPv6 and IPv4.
61 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
62 * zero UDP checksum for either IPv4 or IPv6, the networking stack
63 * may perform further validation in this case.
64 * GRE: only if the checksum is present in the header.
65 * SCTP: indicates the CRC in SCTP header has been validated.
66 *
67 * skb->csum_level indicates the number of consecutive checksums found in
68 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
69 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
70 * and a device is able to verify the checksums for UDP (possibly zero),
71 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
72 * two. If the device were only able to verify the UDP checksum and not
73 * GRE, either because it doesn't support GRE checksum of because GRE
74 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
75 * not considered in this case).
78ea85f1
DB
76 *
77 * CHECKSUM_COMPLETE:
78 *
79 * This is the most generic way. The device supplied checksum of the _whole_
80 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
81 * hardware doesn't need to parse L3/L4 headers to implement this.
82 *
83 * Note: Even if device supports only some protocols, but is able to produce
84 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
85 *
86 * CHECKSUM_PARTIAL:
87 *
6edec0e6
TH
88 * A checksum is set up to be offloaded to a device as described in the
89 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 90 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
91 * on the same host, or it may be set in the input path in GRO or remote
92 * checksum offload. For the purposes of checksum verification, the checksum
93 * referred to by skb->csum_start + skb->csum_offset and any preceding
94 * checksums in the packet are considered verified. Any checksums in the
95 * packet that are after the checksum being offloaded are not considered to
96 * be verified.
78ea85f1
DB
97 *
98 * B. Checksumming on output.
99 *
100 * CHECKSUM_NONE:
101 *
102 * The skb was already checksummed by the protocol, or a checksum is not
103 * required.
104 *
105 * CHECKSUM_PARTIAL:
106 *
107 * The device is required to checksum the packet as seen by hard_start_xmit()
108 * from skb->csum_start up to the end, and to record/write the checksum at
109 * offset skb->csum_start + skb->csum_offset.
110 *
111 * The device must show its capabilities in dev->features, set up at device
112 * setup time, e.g. netdev_features.h:
113 *
114 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
115 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
116 * IPv4. Sigh. Vendors like this way for an unknown reason.
117 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
118 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
119 * NETIF_F_... - Well, you get the picture.
120 *
121 * CHECKSUM_UNNECESSARY:
122 *
123 * Normally, the device will do per protocol specific checksumming. Protocol
124 * implementations that do not want the NIC to perform the checksum
125 * calculation should use this flag in their outgoing skbs.
126 *
127 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
128 * offload. Correspondingly, the FCoE protocol driver
129 * stack should use CHECKSUM_UNNECESSARY.
130 *
131 * Any questions? No questions, good. --ANK
132 */
133
60476372 134/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
135#define CHECKSUM_NONE 0
136#define CHECKSUM_UNNECESSARY 1
137#define CHECKSUM_COMPLETE 2
138#define CHECKSUM_PARTIAL 3
1da177e4 139
77cffe23
TH
140/* Maximum value in skb->csum_level */
141#define SKB_MAX_CSUM_LEVEL 3
142
0bec8c88 143#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 144#define SKB_WITH_OVERHEAD(X) \
deea84b0 145 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
146#define SKB_MAX_ORDER(X, ORDER) \
147 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
148#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
149#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
150
87fb4b7b
ED
151/* return minimum truesize of one skb containing X bytes of data */
152#define SKB_TRUESIZE(X) ((X) + \
153 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
154 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
155
1da177e4 156struct net_device;
716ea3a7 157struct scatterlist;
9c55e01c 158struct pipe_inode_info;
a8f820aa 159struct iov_iter;
fd11a83d 160struct napi_struct;
1da177e4 161
5f79e0f9 162#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
163struct nf_conntrack {
164 atomic_t use;
1da177e4 165};
5f79e0f9 166#endif
1da177e4 167
34666d46 168#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 169struct nf_bridge_info {
bf1ac5ca 170 atomic_t use;
3eaf4025
FW
171 enum {
172 BRNF_PROTO_UNCHANGED,
173 BRNF_PROTO_8021Q,
174 BRNF_PROTO_PPPOE
7fb48c5b 175 } orig_proto:8;
a1e67951 176 bool pkt_otherhost;
bf1ac5ca
ED
177 unsigned int mask;
178 struct net_device *physindev;
7fb48c5b
FW
179 union {
180 struct net_device *physoutdev;
181 char neigh_header[8];
182 };
72b31f72
BT
183 union {
184 __be32 ipv4_daddr;
185 struct in6_addr ipv6_daddr;
186 };
1da177e4
LT
187};
188#endif
189
1da177e4
LT
190struct sk_buff_head {
191 /* These two members must be first. */
192 struct sk_buff *next;
193 struct sk_buff *prev;
194
195 __u32 qlen;
196 spinlock_t lock;
197};
198
199struct sk_buff;
200
9d4dde52
IC
201/* To allow 64K frame to be packed as single skb without frag_list we
202 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
203 * buffers which do not start on a page boundary.
204 *
205 * Since GRO uses frags we allocate at least 16 regardless of page
206 * size.
a715dea3 207 */
9d4dde52 208#if (65536/PAGE_SIZE + 1) < 16
eec00954 209#define MAX_SKB_FRAGS 16UL
a715dea3 210#else
9d4dde52 211#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 212#endif
1da177e4
LT
213
214typedef struct skb_frag_struct skb_frag_t;
215
216struct skb_frag_struct {
a8605c60
IC
217 struct {
218 struct page *p;
219 } page;
cb4dfe56 220#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
221 __u32 page_offset;
222 __u32 size;
cb4dfe56
ED
223#else
224 __u16 page_offset;
225 __u16 size;
226#endif
1da177e4
LT
227};
228
9e903e08
ED
229static inline unsigned int skb_frag_size(const skb_frag_t *frag)
230{
231 return frag->size;
232}
233
234static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
235{
236 frag->size = size;
237}
238
239static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
240{
241 frag->size += delta;
242}
243
244static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
245{
246 frag->size -= delta;
247}
248
ac45f602
PO
249#define HAVE_HW_TIME_STAMP
250
251/**
d3a21be8 252 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
253 * @hwtstamp: hardware time stamp transformed into duration
254 * since arbitrary point in time
ac45f602
PO
255 *
256 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 257 * skb->tstamp.
ac45f602
PO
258 *
259 * hwtstamps can only be compared against other hwtstamps from
260 * the same device.
261 *
262 * This structure is attached to packets as part of the
263 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
264 */
265struct skb_shared_hwtstamps {
266 ktime_t hwtstamp;
ac45f602
PO
267};
268
2244d07b
OH
269/* Definitions for tx_flags in struct skb_shared_info */
270enum {
271 /* generate hardware time stamp */
272 SKBTX_HW_TSTAMP = 1 << 0,
273
e7fd2885 274 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
275 SKBTX_SW_TSTAMP = 1 << 1,
276
277 /* device driver is going to provide hardware time stamp */
278 SKBTX_IN_PROGRESS = 1 << 2,
279
a6686f2f 280 /* device driver supports TX zero-copy buffers */
62b1a8ab 281 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
282
283 /* generate wifi status information (where possible) */
62b1a8ab 284 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
285
286 /* This indicates at least one fragment might be overwritten
287 * (as in vmsplice(), sendfile() ...)
288 * If we need to compute a TX checksum, we'll need to copy
289 * all frags to avoid possible bad checksum
290 */
291 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
292
293 /* generate software time stamp when entering packet scheduling */
294 SKBTX_SCHED_TSTAMP = 1 << 6,
e1c8a607
WB
295
296 /* generate software timestamp on peer data acknowledgment */
297 SKBTX_ACK_TSTAMP = 1 << 7,
a6686f2f
SM
298};
299
e1c8a607
WB
300#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
301 SKBTX_SCHED_TSTAMP | \
302 SKBTX_ACK_TSTAMP)
f24b9be5
WB
303#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
304
a6686f2f
SM
305/*
306 * The callback notifies userspace to release buffers when skb DMA is done in
307 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
308 * The zerocopy_success argument is true if zero copy transmit occurred,
309 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
310 * The ctx field is used to track device context.
311 * The desc field is used to track userspace buffer index.
a6686f2f
SM
312 */
313struct ubuf_info {
e19d6763 314 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 315 void *ctx;
a6686f2f 316 unsigned long desc;
ac45f602
PO
317};
318
1da177e4
LT
319/* This data is invariant across clones and lives at
320 * the end of the header data, ie. at skb->end.
321 */
322struct skb_shared_info {
9f42f126
IC
323 unsigned char nr_frags;
324 __u8 tx_flags;
7967168c
HX
325 unsigned short gso_size;
326 /* Warning: this field is not always filled in (UFO)! */
327 unsigned short gso_segs;
328 unsigned short gso_type;
1da177e4 329 struct sk_buff *frag_list;
ac45f602 330 struct skb_shared_hwtstamps hwtstamps;
09c2d251 331 u32 tskey;
9f42f126 332 __be32 ip6_frag_id;
ec7d2f2c
ED
333
334 /*
335 * Warning : all fields before dataref are cleared in __alloc_skb()
336 */
337 atomic_t dataref;
338
69e3c75f
JB
339 /* Intermediate layers must ensure that destructor_arg
340 * remains valid until skb destructor */
341 void * destructor_arg;
a6686f2f 342
fed66381
ED
343 /* must be last field, see pskb_expand_head() */
344 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
345};
346
347/* We divide dataref into two halves. The higher 16 bits hold references
348 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
349 * the entire skb->data. A clone of a headerless skb holds the length of
350 * the header in skb->hdr_len.
1da177e4
LT
351 *
352 * All users must obey the rule that the skb->data reference count must be
353 * greater than or equal to the payload reference count.
354 *
355 * Holding a reference to the payload part means that the user does not
356 * care about modifications to the header part of skb->data.
357 */
358#define SKB_DATAREF_SHIFT 16
359#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
360
d179cd12
DM
361
362enum {
c8753d55
VS
363 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
364 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
365 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
366};
367
7967168c
HX
368enum {
369 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 370 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
371
372 /* This indicates the skb is from an untrusted source. */
373 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
374
375 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
376 SKB_GSO_TCP_ECN = 1 << 3,
377
378 SKB_GSO_TCPV6 = 1 << 4,
01d5b2fc
CL
379
380 SKB_GSO_FCOE = 1 << 5,
68c33163
PS
381
382 SKB_GSO_GRE = 1 << 6,
73136267 383
4b28252c 384 SKB_GSO_GRE_CSUM = 1 << 7,
0d89d203 385
4b28252c 386 SKB_GSO_IPIP = 1 << 8,
cb32f511 387
4b28252c 388 SKB_GSO_SIT = 1 << 9,
61c1db7f 389
4b28252c 390 SKB_GSO_UDP_TUNNEL = 1 << 10,
0f4f4ffa
TH
391
392 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
4749c09c 393
59b93b41 394 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
7967168c
HX
395};
396
2e07fa9c
ACM
397#if BITS_PER_LONG > 32
398#define NET_SKBUFF_DATA_USES_OFFSET 1
399#endif
400
401#ifdef NET_SKBUFF_DATA_USES_OFFSET
402typedef unsigned int sk_buff_data_t;
403#else
404typedef unsigned char *sk_buff_data_t;
405#endif
406
363ec392
ED
407/**
408 * struct skb_mstamp - multi resolution time stamps
409 * @stamp_us: timestamp in us resolution
410 * @stamp_jiffies: timestamp in jiffies
411 */
412struct skb_mstamp {
413 union {
414 u64 v64;
415 struct {
416 u32 stamp_us;
417 u32 stamp_jiffies;
418 };
419 };
420};
421
422/**
423 * skb_mstamp_get - get current timestamp
424 * @cl: place to store timestamps
425 */
426static inline void skb_mstamp_get(struct skb_mstamp *cl)
427{
428 u64 val = local_clock();
429
430 do_div(val, NSEC_PER_USEC);
431 cl->stamp_us = (u32)val;
432 cl->stamp_jiffies = (u32)jiffies;
433}
434
435/**
436 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
437 * @t1: pointer to newest sample
438 * @t0: pointer to oldest sample
439 */
440static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
441 const struct skb_mstamp *t0)
442{
443 s32 delta_us = t1->stamp_us - t0->stamp_us;
444 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
445
446 /* If delta_us is negative, this might be because interval is too big,
447 * or local_clock() drift is too big : fallback using jiffies.
448 */
449 if (delta_us <= 0 ||
450 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
451
452 delta_us = jiffies_to_usecs(delta_jiffies);
453
454 return delta_us;
455}
456
457
1da177e4
LT
458/**
459 * struct sk_buff - socket buffer
460 * @next: Next buffer in list
461 * @prev: Previous buffer in list
363ec392 462 * @tstamp: Time we arrived/left
56b17425 463 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 464 * @sk: Socket we are owned by
1da177e4 465 * @dev: Device we arrived on/are leaving by
d84e0bd7 466 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 467 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 468 * @sp: the security path, used for xfrm
1da177e4
LT
469 * @len: Length of actual data
470 * @data_len: Data length
471 * @mac_len: Length of link layer header
334a8132 472 * @hdr_len: writable header length of cloned skb
663ead3b
HX
473 * @csum: Checksum (must include start/offset pair)
474 * @csum_start: Offset from skb->head where checksumming should start
475 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 476 * @priority: Packet queueing priority
60ff7467 477 * @ignore_df: allow local fragmentation
1da177e4 478 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 479 * @ip_summed: Driver fed us an IP checksum
1da177e4 480 * @nohdr: Payload reference only, must not modify header
d84e0bd7 481 * @nfctinfo: Relationship of this skb to the connection
1da177e4 482 * @pkt_type: Packet class
c83c2486 483 * @fclone: skbuff clone status
c83c2486 484 * @ipvs_property: skbuff is owned by ipvs
31729363
RD
485 * @peeked: this packet has been seen already, so stats have been
486 * done for it, don't do them again
ba9dda3a 487 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
488 * @protocol: Packet protocol from driver
489 * @destructor: Destruct function
490 * @nfct: Associated connection, if any
1da177e4 491 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 492 * @skb_iif: ifindex of device we arrived on
1da177e4
LT
493 * @tc_index: Traffic control index
494 * @tc_verd: traffic control verdict
61b905da 495 * @hash: the packet hash
d84e0bd7 496 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 497 * @xmit_more: More SKBs are pending for this queue
553a5672 498 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 499 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 500 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 501 * ports.
a3b18ddb 502 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
503 * @wifi_acked_valid: wifi_acked was set
504 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 505 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
06021292 506 * @napi_id: id of the NAPI struct this skb came from
984bc16c 507 * @secmark: security marking
d84e0bd7 508 * @mark: Generic packet mark
86a9bad3 509 * @vlan_proto: vlan encapsulation protocol
6aa895b0 510 * @vlan_tci: vlan tag control information
0d89d203 511 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
512 * @inner_transport_header: Inner transport layer header (encapsulation)
513 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 514 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
515 * @transport_header: Transport layer header
516 * @network_header: Network layer header
517 * @mac_header: Link layer header
518 * @tail: Tail pointer
519 * @end: End pointer
520 * @head: Head of buffer
521 * @data: Data head pointer
522 * @truesize: Buffer size
523 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
524 */
525
526struct sk_buff {
363ec392 527 union {
56b17425
ED
528 struct {
529 /* These two members must be first. */
530 struct sk_buff *next;
531 struct sk_buff *prev;
532
533 union {
534 ktime_t tstamp;
535 struct skb_mstamp skb_mstamp;
536 };
537 };
538 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 539 };
da3f5cf1 540 struct sock *sk;
1da177e4 541 struct net_device *dev;
1da177e4 542
1da177e4
LT
543 /*
544 * This is the control buffer. It is free to use for every
545 * layer. Please put your private variables there. If you
546 * want to keep them across layers you have to do a skb_clone()
547 * first. This is owned by whoever has the skb queued ATM.
548 */
da3f5cf1 549 char cb[48] __aligned(8);
1da177e4 550
7fee226a 551 unsigned long _skb_refdst;
b1937227 552 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
553#ifdef CONFIG_XFRM
554 struct sec_path *sp;
b1937227
ED
555#endif
556#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
557 struct nf_conntrack *nfct;
558#endif
85224844 559#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 560 struct nf_bridge_info *nf_bridge;
da3f5cf1 561#endif
1da177e4 562 unsigned int len,
334a8132
PM
563 data_len;
564 __u16 mac_len,
565 hdr_len;
b1937227
ED
566
567 /* Following fields are _not_ copied in __copy_skb_header()
568 * Note that queue_mapping is here mostly to fill a hole.
569 */
fe55f6d5 570 kmemcheck_bitfield_begin(flags1);
b1937227
ED
571 __u16 queue_mapping;
572 __u8 cloned:1,
6869c4d8 573 nohdr:1,
b84f4cc9 574 fclone:2,
a59322be 575 peeked:1,
b1937227
ED
576 head_frag:1,
577 xmit_more:1;
578 /* one bit hole */
fe55f6d5 579 kmemcheck_bitfield_end(flags1);
4031ae6e 580
b1937227
ED
581 /* fields enclosed in headers_start/headers_end are copied
582 * using a single memcpy() in __copy_skb_header()
583 */
ebcf34f3 584 /* private: */
b1937227 585 __u32 headers_start[0];
ebcf34f3 586 /* public: */
4031ae6e 587
233577a2
HFS
588/* if you move pkt_type around you also must adapt those constants */
589#ifdef __BIG_ENDIAN_BITFIELD
590#define PKT_TYPE_MAX (7 << 5)
591#else
592#define PKT_TYPE_MAX 7
1da177e4 593#endif
233577a2 594#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 595
233577a2 596 __u8 __pkt_type_offset[0];
b1937227 597 __u8 pkt_type:3;
c93bdd0e 598 __u8 pfmemalloc:1;
b1937227
ED
599 __u8 ignore_df:1;
600 __u8 nfctinfo:3;
601
602 __u8 nf_trace:1;
603 __u8 ip_summed:2;
3853b584 604 __u8 ooo_okay:1;
61b905da 605 __u8 l4_hash:1;
a3b18ddb 606 __u8 sw_hash:1;
6e3e939f
JB
607 __u8 wifi_acked_valid:1;
608 __u8 wifi_acked:1;
b1937227 609
3bdc0eba 610 __u8 no_fcs:1;
77cffe23 611 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 612 __u8 encapsulation:1;
7e2b10c1 613 __u8 encap_hdr_csum:1;
5d0c2b95 614 __u8 csum_valid:1;
7e3cead5 615 __u8 csum_complete_sw:1;
b1937227
ED
616 __u8 csum_level:2;
617 __u8 csum_bad:1;
fe55f6d5 618
b1937227
ED
619#ifdef CONFIG_IPV6_NDISC_NODETYPE
620 __u8 ndisc_nodetype:2;
621#endif
622 __u8 ipvs_property:1;
8bce6d7d 623 __u8 inner_protocol_type:1;
e585f236
TH
624 __u8 remcsum_offload:1;
625 /* 3 or 5 bit hole */
b1937227
ED
626
627#ifdef CONFIG_NET_SCHED
628 __u16 tc_index; /* traffic control index */
629#ifdef CONFIG_NET_CLS_ACT
630 __u16 tc_verd; /* traffic control verdict */
631#endif
632#endif
fe55f6d5 633
b1937227
ED
634 union {
635 __wsum csum;
636 struct {
637 __u16 csum_start;
638 __u16 csum_offset;
639 };
640 };
641 __u32 priority;
642 int skb_iif;
643 __u32 hash;
644 __be16 vlan_proto;
645 __u16 vlan_tci;
2bd82484
ED
646#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
647 union {
648 unsigned int napi_id;
649 unsigned int sender_cpu;
650 };
97fc2f08 651#endif
984bc16c
JM
652#ifdef CONFIG_NETWORK_SECMARK
653 __u32 secmark;
654#endif
3b885787
NH
655 union {
656 __u32 mark;
16fad69c 657 __u32 reserved_tailroom;
3b885787 658 };
1da177e4 659
8bce6d7d
TH
660 union {
661 __be16 inner_protocol;
662 __u8 inner_ipproto;
663 };
664
1a37e412
SH
665 __u16 inner_transport_header;
666 __u16 inner_network_header;
667 __u16 inner_mac_header;
b1937227
ED
668
669 __be16 protocol;
1a37e412
SH
670 __u16 transport_header;
671 __u16 network_header;
672 __u16 mac_header;
b1937227 673
ebcf34f3 674 /* private: */
b1937227 675 __u32 headers_end[0];
ebcf34f3 676 /* public: */
b1937227 677
1da177e4 678 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 679 sk_buff_data_t tail;
4305b541 680 sk_buff_data_t end;
1da177e4 681 unsigned char *head,
4305b541 682 *data;
27a884dc
ACM
683 unsigned int truesize;
684 atomic_t users;
1da177e4
LT
685};
686
687#ifdef __KERNEL__
688/*
689 * Handling routines are only of interest to the kernel
690 */
691#include <linux/slab.h>
692
1da177e4 693
c93bdd0e
MG
694#define SKB_ALLOC_FCLONE 0x01
695#define SKB_ALLOC_RX 0x02
fd11a83d 696#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
697
698/* Returns true if the skb was allocated from PFMEMALLOC reserves */
699static inline bool skb_pfmemalloc(const struct sk_buff *skb)
700{
701 return unlikely(skb->pfmemalloc);
702}
703
7fee226a
ED
704/*
705 * skb might have a dst pointer attached, refcounted or not.
706 * _skb_refdst low order bit is set if refcount was _not_ taken
707 */
708#define SKB_DST_NOREF 1UL
709#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
710
711/**
712 * skb_dst - returns skb dst_entry
713 * @skb: buffer
714 *
715 * Returns skb dst_entry, regardless of reference taken or not.
716 */
adf30907
ED
717static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
718{
7fee226a
ED
719 /* If refdst was not refcounted, check we still are in a
720 * rcu_read_lock section
721 */
722 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
723 !rcu_read_lock_held() &&
724 !rcu_read_lock_bh_held());
725 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
726}
727
7fee226a
ED
728/**
729 * skb_dst_set - sets skb dst
730 * @skb: buffer
731 * @dst: dst entry
732 *
733 * Sets skb dst, assuming a reference was taken on dst and should
734 * be released by skb_dst_drop()
735 */
adf30907
ED
736static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
737{
7fee226a
ED
738 skb->_skb_refdst = (unsigned long)dst;
739}
740
932bc4d7
JA
741/**
742 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
743 * @skb: buffer
744 * @dst: dst entry
745 *
746 * Sets skb dst, assuming a reference was not taken on dst.
747 * If dst entry is cached, we do not take reference and dst_release
748 * will be avoided by refdst_drop. If dst entry is not cached, we take
749 * reference, so that last dst_release can destroy the dst immediately.
750 */
751static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
752{
dbfc4fb7
HFS
753 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
754 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 755}
7fee226a
ED
756
757/**
25985edc 758 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
759 * @skb: buffer
760 */
761static inline bool skb_dst_is_noref(const struct sk_buff *skb)
762{
763 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
764}
765
511c3f92
ED
766static inline struct rtable *skb_rtable(const struct sk_buff *skb)
767{
adf30907 768 return (struct rtable *)skb_dst(skb);
511c3f92
ED
769}
770
7965bd4d
JP
771void kfree_skb(struct sk_buff *skb);
772void kfree_skb_list(struct sk_buff *segs);
773void skb_tx_error(struct sk_buff *skb);
774void consume_skb(struct sk_buff *skb);
775void __kfree_skb(struct sk_buff *skb);
d7e8883c 776extern struct kmem_cache *skbuff_head_cache;
bad43ca8 777
7965bd4d
JP
778void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
779bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
780 bool *fragstolen, int *delta_truesize);
bad43ca8 781
7965bd4d
JP
782struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
783 int node);
2ea2f62c 784struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 785struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 786static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 787 gfp_t priority)
d179cd12 788{
564824b0 789 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
790}
791
2e4e4410
ED
792struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
793 unsigned long data_len,
794 int max_page_order,
795 int *errcode,
796 gfp_t gfp_mask);
797
d0bf4a9e
ED
798/* Layout of fast clones : [skb1][skb2][fclone_ref] */
799struct sk_buff_fclones {
800 struct sk_buff skb1;
801
802 struct sk_buff skb2;
803
804 atomic_t fclone_ref;
805};
806
807/**
808 * skb_fclone_busy - check if fclone is busy
809 * @skb: buffer
810 *
811 * Returns true is skb is a fast clone, and its clone is not freed.
39bb5e62
ED
812 * Some drivers call skb_orphan() in their ndo_start_xmit(),
813 * so we also check that this didnt happen.
d0bf4a9e 814 */
39bb5e62
ED
815static inline bool skb_fclone_busy(const struct sock *sk,
816 const struct sk_buff *skb)
d0bf4a9e
ED
817{
818 const struct sk_buff_fclones *fclones;
819
820 fclones = container_of(skb, struct sk_buff_fclones, skb1);
821
822 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 823 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 824 fclones->skb2.sk == sk;
d0bf4a9e
ED
825}
826
d179cd12 827static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 828 gfp_t priority)
d179cd12 829{
c93bdd0e 830 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
831}
832
7965bd4d 833struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
834static inline struct sk_buff *alloc_skb_head(gfp_t priority)
835{
836 return __alloc_skb_head(priority, -1);
837}
838
7965bd4d
JP
839struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
840int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
841struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
842struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
843struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
844 gfp_t gfp_mask, bool fclone);
845static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
846 gfp_t gfp_mask)
847{
848 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
849}
7965bd4d
JP
850
851int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
852struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
853 unsigned int headroom);
854struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
855 int newtailroom, gfp_t priority);
25a91d8d
FD
856int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
857 int offset, int len);
7965bd4d
JP
858int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
859 int len);
860int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
861int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 862#define dev_kfree_skb(a) consume_skb(a)
1da177e4 863
7965bd4d
JP
864int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
865 int getfrag(void *from, char *to, int offset,
866 int len, int odd, struct sk_buff *skb),
867 void *from, int length);
e89e9cf5 868
be12a1fe
HFS
869int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
870 int offset, size_t size);
871
d94d9fee 872struct skb_seq_state {
677e90ed
TG
873 __u32 lower_offset;
874 __u32 upper_offset;
875 __u32 frag_idx;
876 __u32 stepped_offset;
877 struct sk_buff *root_skb;
878 struct sk_buff *cur_skb;
879 __u8 *frag_data;
880};
881
7965bd4d
JP
882void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
883 unsigned int to, struct skb_seq_state *st);
884unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
885 struct skb_seq_state *st);
886void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 887
7965bd4d 888unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 889 unsigned int to, struct ts_config *config);
3fc7e8a6 890
09323cc4
TH
891/*
892 * Packet hash types specify the type of hash in skb_set_hash.
893 *
894 * Hash types refer to the protocol layer addresses which are used to
895 * construct a packet's hash. The hashes are used to differentiate or identify
896 * flows of the protocol layer for the hash type. Hash types are either
897 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
898 *
899 * Properties of hashes:
900 *
901 * 1) Two packets in different flows have different hash values
902 * 2) Two packets in the same flow should have the same hash value
903 *
904 * A hash at a higher layer is considered to be more specific. A driver should
905 * set the most specific hash possible.
906 *
907 * A driver cannot indicate a more specific hash than the layer at which a hash
908 * was computed. For instance an L3 hash cannot be set as an L4 hash.
909 *
910 * A driver may indicate a hash level which is less specific than the
911 * actual layer the hash was computed on. For instance, a hash computed
912 * at L4 may be considered an L3 hash. This should only be done if the
913 * driver can't unambiguously determine that the HW computed the hash at
914 * the higher layer. Note that the "should" in the second property above
915 * permits this.
916 */
917enum pkt_hash_types {
918 PKT_HASH_TYPE_NONE, /* Undefined type */
919 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
920 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
921 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
922};
923
924static inline void
925skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
926{
61b905da 927 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
a3b18ddb 928 skb->sw_hash = 0;
61b905da 929 skb->hash = hash;
09323cc4
TH
930}
931
3958afa1 932static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 933{
a3b18ddb 934 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 935 __skb_get_hash(skb);
bfb564e7 936
61b905da 937 return skb->hash;
bfb564e7
KK
938}
939
50fb7992
TH
940__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
941
57bdf7f4
TH
942static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
943{
61b905da 944 return skb->hash;
57bdf7f4
TH
945}
946
7539fadc
TH
947static inline void skb_clear_hash(struct sk_buff *skb)
948{
61b905da 949 skb->hash = 0;
a3b18ddb 950 skb->sw_hash = 0;
61b905da 951 skb->l4_hash = 0;
7539fadc
TH
952}
953
954static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
955{
61b905da 956 if (!skb->l4_hash)
7539fadc
TH
957 skb_clear_hash(skb);
958}
959
3df7a74e
TH
960static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
961{
61b905da 962 to->hash = from->hash;
a3b18ddb 963 to->sw_hash = from->sw_hash;
61b905da 964 to->l4_hash = from->l4_hash;
3df7a74e
TH
965};
966
c29390c6
ED
967static inline void skb_sender_cpu_clear(struct sk_buff *skb)
968{
969#ifdef CONFIG_XPS
970 skb->sender_cpu = 0;
971#endif
972}
973
4305b541
ACM
974#ifdef NET_SKBUFF_DATA_USES_OFFSET
975static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
976{
977 return skb->head + skb->end;
978}
ec47ea82
AD
979
980static inline unsigned int skb_end_offset(const struct sk_buff *skb)
981{
982 return skb->end;
983}
4305b541
ACM
984#else
985static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
986{
987 return skb->end;
988}
ec47ea82
AD
989
990static inline unsigned int skb_end_offset(const struct sk_buff *skb)
991{
992 return skb->end - skb->head;
993}
4305b541
ACM
994#endif
995
1da177e4 996/* Internal */
4305b541 997#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 998
ac45f602
PO
999static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1000{
1001 return &skb_shinfo(skb)->hwtstamps;
1002}
1003
1da177e4
LT
1004/**
1005 * skb_queue_empty - check if a queue is empty
1006 * @list: queue head
1007 *
1008 * Returns true if the queue is empty, false otherwise.
1009 */
1010static inline int skb_queue_empty(const struct sk_buff_head *list)
1011{
fd44b93c 1012 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1013}
1014
fc7ebb21
DM
1015/**
1016 * skb_queue_is_last - check if skb is the last entry in the queue
1017 * @list: queue head
1018 * @skb: buffer
1019 *
1020 * Returns true if @skb is the last buffer on the list.
1021 */
1022static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1023 const struct sk_buff *skb)
1024{
fd44b93c 1025 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1026}
1027
832d11c5
IJ
1028/**
1029 * skb_queue_is_first - check if skb is the first entry in the queue
1030 * @list: queue head
1031 * @skb: buffer
1032 *
1033 * Returns true if @skb is the first buffer on the list.
1034 */
1035static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1036 const struct sk_buff *skb)
1037{
fd44b93c 1038 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1039}
1040
249c8b42
DM
1041/**
1042 * skb_queue_next - return the next packet in the queue
1043 * @list: queue head
1044 * @skb: current buffer
1045 *
1046 * Return the next packet in @list after @skb. It is only valid to
1047 * call this if skb_queue_is_last() evaluates to false.
1048 */
1049static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1050 const struct sk_buff *skb)
1051{
1052 /* This BUG_ON may seem severe, but if we just return then we
1053 * are going to dereference garbage.
1054 */
1055 BUG_ON(skb_queue_is_last(list, skb));
1056 return skb->next;
1057}
1058
832d11c5
IJ
1059/**
1060 * skb_queue_prev - return the prev packet in the queue
1061 * @list: queue head
1062 * @skb: current buffer
1063 *
1064 * Return the prev packet in @list before @skb. It is only valid to
1065 * call this if skb_queue_is_first() evaluates to false.
1066 */
1067static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1068 const struct sk_buff *skb)
1069{
1070 /* This BUG_ON may seem severe, but if we just return then we
1071 * are going to dereference garbage.
1072 */
1073 BUG_ON(skb_queue_is_first(list, skb));
1074 return skb->prev;
1075}
1076
1da177e4
LT
1077/**
1078 * skb_get - reference buffer
1079 * @skb: buffer to reference
1080 *
1081 * Makes another reference to a socket buffer and returns a pointer
1082 * to the buffer.
1083 */
1084static inline struct sk_buff *skb_get(struct sk_buff *skb)
1085{
1086 atomic_inc(&skb->users);
1087 return skb;
1088}
1089
1090/*
1091 * If users == 1, we are the only owner and are can avoid redundant
1092 * atomic change.
1093 */
1094
1da177e4
LT
1095/**
1096 * skb_cloned - is the buffer a clone
1097 * @skb: buffer to check
1098 *
1099 * Returns true if the buffer was generated with skb_clone() and is
1100 * one of multiple shared copies of the buffer. Cloned buffers are
1101 * shared data so must not be written to under normal circumstances.
1102 */
1103static inline int skb_cloned(const struct sk_buff *skb)
1104{
1105 return skb->cloned &&
1106 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1107}
1108
14bbd6a5
PS
1109static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1110{
1111 might_sleep_if(pri & __GFP_WAIT);
1112
1113 if (skb_cloned(skb))
1114 return pskb_expand_head(skb, 0, 0, pri);
1115
1116 return 0;
1117}
1118
1da177e4
LT
1119/**
1120 * skb_header_cloned - is the header a clone
1121 * @skb: buffer to check
1122 *
1123 * Returns true if modifying the header part of the buffer requires
1124 * the data to be copied.
1125 */
1126static inline int skb_header_cloned(const struct sk_buff *skb)
1127{
1128 int dataref;
1129
1130 if (!skb->cloned)
1131 return 0;
1132
1133 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1134 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1135 return dataref != 1;
1136}
1137
1138/**
1139 * skb_header_release - release reference to header
1140 * @skb: buffer to operate on
1141 *
1142 * Drop a reference to the header part of the buffer. This is done
1143 * by acquiring a payload reference. You must not read from the header
1144 * part of skb->data after this.
f4a775d1 1145 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1146 */
1147static inline void skb_header_release(struct sk_buff *skb)
1148{
1149 BUG_ON(skb->nohdr);
1150 skb->nohdr = 1;
1151 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1152}
1153
f4a775d1
ED
1154/**
1155 * __skb_header_release - release reference to header
1156 * @skb: buffer to operate on
1157 *
1158 * Variant of skb_header_release() assuming skb is private to caller.
1159 * We can avoid one atomic operation.
1160 */
1161static inline void __skb_header_release(struct sk_buff *skb)
1162{
1163 skb->nohdr = 1;
1164 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1165}
1166
1167
1da177e4
LT
1168/**
1169 * skb_shared - is the buffer shared
1170 * @skb: buffer to check
1171 *
1172 * Returns true if more than one person has a reference to this
1173 * buffer.
1174 */
1175static inline int skb_shared(const struct sk_buff *skb)
1176{
1177 return atomic_read(&skb->users) != 1;
1178}
1179
1180/**
1181 * skb_share_check - check if buffer is shared and if so clone it
1182 * @skb: buffer to check
1183 * @pri: priority for memory allocation
1184 *
1185 * If the buffer is shared the buffer is cloned and the old copy
1186 * drops a reference. A new clone with a single reference is returned.
1187 * If the buffer is not shared the original buffer is returned. When
1188 * being called from interrupt status or with spinlocks held pri must
1189 * be GFP_ATOMIC.
1190 *
1191 * NULL is returned on a memory allocation failure.
1192 */
47061bc4 1193static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4
LT
1194{
1195 might_sleep_if(pri & __GFP_WAIT);
1196 if (skb_shared(skb)) {
1197 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1198
1199 if (likely(nskb))
1200 consume_skb(skb);
1201 else
1202 kfree_skb(skb);
1da177e4
LT
1203 skb = nskb;
1204 }
1205 return skb;
1206}
1207
1208/*
1209 * Copy shared buffers into a new sk_buff. We effectively do COW on
1210 * packets to handle cases where we have a local reader and forward
1211 * and a couple of other messy ones. The normal one is tcpdumping
1212 * a packet thats being forwarded.
1213 */
1214
1215/**
1216 * skb_unshare - make a copy of a shared buffer
1217 * @skb: buffer to check
1218 * @pri: priority for memory allocation
1219 *
1220 * If the socket buffer is a clone then this function creates a new
1221 * copy of the data, drops a reference count on the old copy and returns
1222 * the new copy with the reference count at 1. If the buffer is not a clone
1223 * the original buffer is returned. When called with a spinlock held or
1224 * from interrupt state @pri must be %GFP_ATOMIC
1225 *
1226 * %NULL is returned on a memory allocation failure.
1227 */
e2bf521d 1228static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1229 gfp_t pri)
1da177e4
LT
1230{
1231 might_sleep_if(pri & __GFP_WAIT);
1232 if (skb_cloned(skb)) {
1233 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1234
1235 /* Free our shared copy */
1236 if (likely(nskb))
1237 consume_skb(skb);
1238 else
1239 kfree_skb(skb);
1da177e4
LT
1240 skb = nskb;
1241 }
1242 return skb;
1243}
1244
1245/**
1a5778aa 1246 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1247 * @list_: list to peek at
1248 *
1249 * Peek an &sk_buff. Unlike most other operations you _MUST_
1250 * be careful with this one. A peek leaves the buffer on the
1251 * list and someone else may run off with it. You must hold
1252 * the appropriate locks or have a private queue to do this.
1253 *
1254 * Returns %NULL for an empty list or a pointer to the head element.
1255 * The reference count is not incremented and the reference is therefore
1256 * volatile. Use with caution.
1257 */
05bdd2f1 1258static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1259{
18d07000
ED
1260 struct sk_buff *skb = list_->next;
1261
1262 if (skb == (struct sk_buff *)list_)
1263 skb = NULL;
1264 return skb;
1da177e4
LT
1265}
1266
da5ef6e5
PE
1267/**
1268 * skb_peek_next - peek skb following the given one from a queue
1269 * @skb: skb to start from
1270 * @list_: list to peek at
1271 *
1272 * Returns %NULL when the end of the list is met or a pointer to the
1273 * next element. The reference count is not incremented and the
1274 * reference is therefore volatile. Use with caution.
1275 */
1276static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1277 const struct sk_buff_head *list_)
1278{
1279 struct sk_buff *next = skb->next;
18d07000 1280
da5ef6e5
PE
1281 if (next == (struct sk_buff *)list_)
1282 next = NULL;
1283 return next;
1284}
1285
1da177e4 1286/**
1a5778aa 1287 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1288 * @list_: list to peek at
1289 *
1290 * Peek an &sk_buff. Unlike most other operations you _MUST_
1291 * be careful with this one. A peek leaves the buffer on the
1292 * list and someone else may run off with it. You must hold
1293 * the appropriate locks or have a private queue to do this.
1294 *
1295 * Returns %NULL for an empty list or a pointer to the tail element.
1296 * The reference count is not incremented and the reference is therefore
1297 * volatile. Use with caution.
1298 */
05bdd2f1 1299static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1300{
18d07000
ED
1301 struct sk_buff *skb = list_->prev;
1302
1303 if (skb == (struct sk_buff *)list_)
1304 skb = NULL;
1305 return skb;
1306
1da177e4
LT
1307}
1308
1309/**
1310 * skb_queue_len - get queue length
1311 * @list_: list to measure
1312 *
1313 * Return the length of an &sk_buff queue.
1314 */
1315static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1316{
1317 return list_->qlen;
1318}
1319
67fed459
DM
1320/**
1321 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1322 * @list: queue to initialize
1323 *
1324 * This initializes only the list and queue length aspects of
1325 * an sk_buff_head object. This allows to initialize the list
1326 * aspects of an sk_buff_head without reinitializing things like
1327 * the spinlock. It can also be used for on-stack sk_buff_head
1328 * objects where the spinlock is known to not be used.
1329 */
1330static inline void __skb_queue_head_init(struct sk_buff_head *list)
1331{
1332 list->prev = list->next = (struct sk_buff *)list;
1333 list->qlen = 0;
1334}
1335
76f10ad0
AV
1336/*
1337 * This function creates a split out lock class for each invocation;
1338 * this is needed for now since a whole lot of users of the skb-queue
1339 * infrastructure in drivers have different locking usage (in hardirq)
1340 * than the networking core (in softirq only). In the long run either the
1341 * network layer or drivers should need annotation to consolidate the
1342 * main types of usage into 3 classes.
1343 */
1da177e4
LT
1344static inline void skb_queue_head_init(struct sk_buff_head *list)
1345{
1346 spin_lock_init(&list->lock);
67fed459 1347 __skb_queue_head_init(list);
1da177e4
LT
1348}
1349
c2ecba71
PE
1350static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1351 struct lock_class_key *class)
1352{
1353 skb_queue_head_init(list);
1354 lockdep_set_class(&list->lock, class);
1355}
1356
1da177e4 1357/*
bf299275 1358 * Insert an sk_buff on a list.
1da177e4
LT
1359 *
1360 * The "__skb_xxxx()" functions are the non-atomic ones that
1361 * can only be called with interrupts disabled.
1362 */
7965bd4d
JP
1363void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1364 struct sk_buff_head *list);
bf299275
GR
1365static inline void __skb_insert(struct sk_buff *newsk,
1366 struct sk_buff *prev, struct sk_buff *next,
1367 struct sk_buff_head *list)
1368{
1369 newsk->next = next;
1370 newsk->prev = prev;
1371 next->prev = prev->next = newsk;
1372 list->qlen++;
1373}
1da177e4 1374
67fed459
DM
1375static inline void __skb_queue_splice(const struct sk_buff_head *list,
1376 struct sk_buff *prev,
1377 struct sk_buff *next)
1378{
1379 struct sk_buff *first = list->next;
1380 struct sk_buff *last = list->prev;
1381
1382 first->prev = prev;
1383 prev->next = first;
1384
1385 last->next = next;
1386 next->prev = last;
1387}
1388
1389/**
1390 * skb_queue_splice - join two skb lists, this is designed for stacks
1391 * @list: the new list to add
1392 * @head: the place to add it in the first list
1393 */
1394static inline void skb_queue_splice(const struct sk_buff_head *list,
1395 struct sk_buff_head *head)
1396{
1397 if (!skb_queue_empty(list)) {
1398 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1399 head->qlen += list->qlen;
67fed459
DM
1400 }
1401}
1402
1403/**
d9619496 1404 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1405 * @list: the new list to add
1406 * @head: the place to add it in the first list
1407 *
1408 * The list at @list is reinitialised
1409 */
1410static inline void skb_queue_splice_init(struct sk_buff_head *list,
1411 struct sk_buff_head *head)
1412{
1413 if (!skb_queue_empty(list)) {
1414 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1415 head->qlen += list->qlen;
67fed459
DM
1416 __skb_queue_head_init(list);
1417 }
1418}
1419
1420/**
1421 * skb_queue_splice_tail - join two skb lists, each list being a queue
1422 * @list: the new list to add
1423 * @head: the place to add it in the first list
1424 */
1425static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1426 struct sk_buff_head *head)
1427{
1428 if (!skb_queue_empty(list)) {
1429 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1430 head->qlen += list->qlen;
67fed459
DM
1431 }
1432}
1433
1434/**
d9619496 1435 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1436 * @list: the new list to add
1437 * @head: the place to add it in the first list
1438 *
1439 * Each of the lists is a queue.
1440 * The list at @list is reinitialised
1441 */
1442static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1443 struct sk_buff_head *head)
1444{
1445 if (!skb_queue_empty(list)) {
1446 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1447 head->qlen += list->qlen;
67fed459
DM
1448 __skb_queue_head_init(list);
1449 }
1450}
1451
1da177e4 1452/**
300ce174 1453 * __skb_queue_after - queue a buffer at the list head
1da177e4 1454 * @list: list to use
300ce174 1455 * @prev: place after this buffer
1da177e4
LT
1456 * @newsk: buffer to queue
1457 *
300ce174 1458 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1459 * and you must therefore hold required locks before calling it.
1460 *
1461 * A buffer cannot be placed on two lists at the same time.
1462 */
300ce174
SH
1463static inline void __skb_queue_after(struct sk_buff_head *list,
1464 struct sk_buff *prev,
1465 struct sk_buff *newsk)
1da177e4 1466{
bf299275 1467 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1468}
1469
7965bd4d
JP
1470void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1471 struct sk_buff_head *list);
7de6c033 1472
f5572855
GR
1473static inline void __skb_queue_before(struct sk_buff_head *list,
1474 struct sk_buff *next,
1475 struct sk_buff *newsk)
1476{
1477 __skb_insert(newsk, next->prev, next, list);
1478}
1479
300ce174
SH
1480/**
1481 * __skb_queue_head - queue a buffer at the list head
1482 * @list: list to use
1483 * @newsk: buffer to queue
1484 *
1485 * Queue a buffer at the start of a list. This function takes no locks
1486 * and you must therefore hold required locks before calling it.
1487 *
1488 * A buffer cannot be placed on two lists at the same time.
1489 */
7965bd4d 1490void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1491static inline void __skb_queue_head(struct sk_buff_head *list,
1492 struct sk_buff *newsk)
1493{
1494 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1495}
1496
1da177e4
LT
1497/**
1498 * __skb_queue_tail - queue a buffer at the list tail
1499 * @list: list to use
1500 * @newsk: buffer to queue
1501 *
1502 * Queue a buffer at the end of a list. This function takes no locks
1503 * and you must therefore hold required locks before calling it.
1504 *
1505 * A buffer cannot be placed on two lists at the same time.
1506 */
7965bd4d 1507void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1508static inline void __skb_queue_tail(struct sk_buff_head *list,
1509 struct sk_buff *newsk)
1510{
f5572855 1511 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1512}
1513
1da177e4
LT
1514/*
1515 * remove sk_buff from list. _Must_ be called atomically, and with
1516 * the list known..
1517 */
7965bd4d 1518void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1519static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1520{
1521 struct sk_buff *next, *prev;
1522
1523 list->qlen--;
1524 next = skb->next;
1525 prev = skb->prev;
1526 skb->next = skb->prev = NULL;
1da177e4
LT
1527 next->prev = prev;
1528 prev->next = next;
1529}
1530
f525c06d
GR
1531/**
1532 * __skb_dequeue - remove from the head of the queue
1533 * @list: list to dequeue from
1534 *
1535 * Remove the head of the list. This function does not take any locks
1536 * so must be used with appropriate locks held only. The head item is
1537 * returned or %NULL if the list is empty.
1538 */
7965bd4d 1539struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1540static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1541{
1542 struct sk_buff *skb = skb_peek(list);
1543 if (skb)
1544 __skb_unlink(skb, list);
1545 return skb;
1546}
1da177e4
LT
1547
1548/**
1549 * __skb_dequeue_tail - remove from the tail of the queue
1550 * @list: list to dequeue from
1551 *
1552 * Remove the tail of the list. This function does not take any locks
1553 * so must be used with appropriate locks held only. The tail item is
1554 * returned or %NULL if the list is empty.
1555 */
7965bd4d 1556struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1557static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1558{
1559 struct sk_buff *skb = skb_peek_tail(list);
1560 if (skb)
1561 __skb_unlink(skb, list);
1562 return skb;
1563}
1564
1565
bdcc0924 1566static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1567{
1568 return skb->data_len;
1569}
1570
1571static inline unsigned int skb_headlen(const struct sk_buff *skb)
1572{
1573 return skb->len - skb->data_len;
1574}
1575
1576static inline int skb_pagelen(const struct sk_buff *skb)
1577{
1578 int i, len = 0;
1579
1580 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
9e903e08 1581 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1582 return len + skb_headlen(skb);
1583}
1584
131ea667
IC
1585/**
1586 * __skb_fill_page_desc - initialise a paged fragment in an skb
1587 * @skb: buffer containing fragment to be initialised
1588 * @i: paged fragment index to initialise
1589 * @page: the page to use for this fragment
1590 * @off: the offset to the data with @page
1591 * @size: the length of the data
1592 *
1593 * Initialises the @i'th fragment of @skb to point to &size bytes at
1594 * offset @off within @page.
1595 *
1596 * Does not take any additional reference on the fragment.
1597 */
1598static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1599 struct page *page, int off, int size)
1da177e4
LT
1600{
1601 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1602
c48a11c7
MG
1603 /*
1604 * Propagate page->pfmemalloc to the skb if we can. The problem is
1605 * that not all callers have unique ownership of the page. If
1606 * pfmemalloc is set, we check the mapping as a mapping implies
1607 * page->index is set (index and pfmemalloc share space).
1608 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1609 * do not lose pfmemalloc information as the pages would not be
1610 * allocated using __GFP_MEMALLOC.
1611 */
a8605c60 1612 frag->page.p = page;
1da177e4 1613 frag->page_offset = off;
9e903e08 1614 skb_frag_size_set(frag, size);
cca7af38
PE
1615
1616 page = compound_head(page);
1617 if (page->pfmemalloc && !page->mapping)
1618 skb->pfmemalloc = true;
131ea667
IC
1619}
1620
1621/**
1622 * skb_fill_page_desc - initialise a paged fragment in an skb
1623 * @skb: buffer containing fragment to be initialised
1624 * @i: paged fragment index to initialise
1625 * @page: the page to use for this fragment
1626 * @off: the offset to the data with @page
1627 * @size: the length of the data
1628 *
1629 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1630 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1631 * addition updates @skb such that @i is the last fragment.
1632 *
1633 * Does not take any additional reference on the fragment.
1634 */
1635static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1636 struct page *page, int off, int size)
1637{
1638 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1639 skb_shinfo(skb)->nr_frags = i + 1;
1640}
1641
7965bd4d
JP
1642void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1643 int size, unsigned int truesize);
654bed16 1644
f8e617e1
JW
1645void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1646 unsigned int truesize);
1647
1da177e4 1648#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1649#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1650#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1651
27a884dc
ACM
1652#ifdef NET_SKBUFF_DATA_USES_OFFSET
1653static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1654{
1655 return skb->head + skb->tail;
1656}
1657
1658static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1659{
1660 skb->tail = skb->data - skb->head;
1661}
1662
1663static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1664{
1665 skb_reset_tail_pointer(skb);
1666 skb->tail += offset;
1667}
7cc46190 1668
27a884dc
ACM
1669#else /* NET_SKBUFF_DATA_USES_OFFSET */
1670static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1671{
1672 return skb->tail;
1673}
1674
1675static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1676{
1677 skb->tail = skb->data;
1678}
1679
1680static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1681{
1682 skb->tail = skb->data + offset;
1683}
4305b541 1684
27a884dc
ACM
1685#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1686
1da177e4
LT
1687/*
1688 * Add data to an sk_buff
1689 */
0c7ddf36 1690unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1691unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1692static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1693{
27a884dc 1694 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1695 SKB_LINEAR_ASSERT(skb);
1696 skb->tail += len;
1697 skb->len += len;
1698 return tmp;
1699}
1700
7965bd4d 1701unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1702static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1703{
1704 skb->data -= len;
1705 skb->len += len;
1706 return skb->data;
1707}
1708
7965bd4d 1709unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1710static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1711{
1712 skb->len -= len;
1713 BUG_ON(skb->len < skb->data_len);
1714 return skb->data += len;
1715}
1716
47d29646
DM
1717static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1718{
1719 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1720}
1721
7965bd4d 1722unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1723
1724static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1725{
1726 if (len > skb_headlen(skb) &&
987c402a 1727 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1728 return NULL;
1729 skb->len -= len;
1730 return skb->data += len;
1731}
1732
1733static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1734{
1735 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1736}
1737
1738static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1739{
1740 if (likely(len <= skb_headlen(skb)))
1741 return 1;
1742 if (unlikely(len > skb->len))
1743 return 0;
987c402a 1744 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1745}
1746
1747/**
1748 * skb_headroom - bytes at buffer head
1749 * @skb: buffer to check
1750 *
1751 * Return the number of bytes of free space at the head of an &sk_buff.
1752 */
c2636b4d 1753static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1754{
1755 return skb->data - skb->head;
1756}
1757
1758/**
1759 * skb_tailroom - bytes at buffer end
1760 * @skb: buffer to check
1761 *
1762 * Return the number of bytes of free space at the tail of an sk_buff
1763 */
1764static inline int skb_tailroom(const struct sk_buff *skb)
1765{
4305b541 1766 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1767}
1768
a21d4572
ED
1769/**
1770 * skb_availroom - bytes at buffer end
1771 * @skb: buffer to check
1772 *
1773 * Return the number of bytes of free space at the tail of an sk_buff
1774 * allocated by sk_stream_alloc()
1775 */
1776static inline int skb_availroom(const struct sk_buff *skb)
1777{
16fad69c
ED
1778 if (skb_is_nonlinear(skb))
1779 return 0;
1780
1781 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1782}
1783
1da177e4
LT
1784/**
1785 * skb_reserve - adjust headroom
1786 * @skb: buffer to alter
1787 * @len: bytes to move
1788 *
1789 * Increase the headroom of an empty &sk_buff by reducing the tail
1790 * room. This is only allowed for an empty buffer.
1791 */
8243126c 1792static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1793{
1794 skb->data += len;
1795 skb->tail += len;
1796}
1797
8bce6d7d
TH
1798#define ENCAP_TYPE_ETHER 0
1799#define ENCAP_TYPE_IPPROTO 1
1800
1801static inline void skb_set_inner_protocol(struct sk_buff *skb,
1802 __be16 protocol)
1803{
1804 skb->inner_protocol = protocol;
1805 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1806}
1807
1808static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1809 __u8 ipproto)
1810{
1811 skb->inner_ipproto = ipproto;
1812 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1813}
1814
6a674e9c
JG
1815static inline void skb_reset_inner_headers(struct sk_buff *skb)
1816{
aefbd2b3 1817 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
1818 skb->inner_network_header = skb->network_header;
1819 skb->inner_transport_header = skb->transport_header;
1820}
1821
0b5c9db1
JP
1822static inline void skb_reset_mac_len(struct sk_buff *skb)
1823{
1824 skb->mac_len = skb->network_header - skb->mac_header;
1825}
1826
6a674e9c
JG
1827static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1828 *skb)
1829{
1830 return skb->head + skb->inner_transport_header;
1831}
1832
1833static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1834{
1835 skb->inner_transport_header = skb->data - skb->head;
1836}
1837
1838static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1839 const int offset)
1840{
1841 skb_reset_inner_transport_header(skb);
1842 skb->inner_transport_header += offset;
1843}
1844
1845static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1846{
1847 return skb->head + skb->inner_network_header;
1848}
1849
1850static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1851{
1852 skb->inner_network_header = skb->data - skb->head;
1853}
1854
1855static inline void skb_set_inner_network_header(struct sk_buff *skb,
1856 const int offset)
1857{
1858 skb_reset_inner_network_header(skb);
1859 skb->inner_network_header += offset;
1860}
1861
aefbd2b3
PS
1862static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1863{
1864 return skb->head + skb->inner_mac_header;
1865}
1866
1867static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1868{
1869 skb->inner_mac_header = skb->data - skb->head;
1870}
1871
1872static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1873 const int offset)
1874{
1875 skb_reset_inner_mac_header(skb);
1876 skb->inner_mac_header += offset;
1877}
fda55eca
ED
1878static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1879{
35d04610 1880 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
1881}
1882
9c70220b
ACM
1883static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1884{
2e07fa9c 1885 return skb->head + skb->transport_header;
9c70220b
ACM
1886}
1887
badff6d0
ACM
1888static inline void skb_reset_transport_header(struct sk_buff *skb)
1889{
2e07fa9c 1890 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
1891}
1892
967b05f6
ACM
1893static inline void skb_set_transport_header(struct sk_buff *skb,
1894 const int offset)
1895{
2e07fa9c
ACM
1896 skb_reset_transport_header(skb);
1897 skb->transport_header += offset;
ea2ae17d
ACM
1898}
1899
d56f90a7
ACM
1900static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1901{
2e07fa9c 1902 return skb->head + skb->network_header;
d56f90a7
ACM
1903}
1904
c1d2bbe1
ACM
1905static inline void skb_reset_network_header(struct sk_buff *skb)
1906{
2e07fa9c 1907 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
1908}
1909
c14d2450
ACM
1910static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1911{
2e07fa9c
ACM
1912 skb_reset_network_header(skb);
1913 skb->network_header += offset;
c14d2450
ACM
1914}
1915
2e07fa9c 1916static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 1917{
2e07fa9c 1918 return skb->head + skb->mac_header;
bbe735e4
ACM
1919}
1920
2e07fa9c 1921static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 1922{
35d04610 1923 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
1924}
1925
1926static inline void skb_reset_mac_header(struct sk_buff *skb)
1927{
1928 skb->mac_header = skb->data - skb->head;
1929}
1930
1931static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1932{
1933 skb_reset_mac_header(skb);
1934 skb->mac_header += offset;
1935}
1936
0e3da5bb
TT
1937static inline void skb_pop_mac_header(struct sk_buff *skb)
1938{
1939 skb->mac_header = skb->network_header;
1940}
1941
fbbdb8f0
YX
1942static inline void skb_probe_transport_header(struct sk_buff *skb,
1943 const int offset_hint)
1944{
1945 struct flow_keys keys;
1946
1947 if (skb_transport_header_was_set(skb))
1948 return;
06635a35 1949 else if (skb_flow_dissect_flow_keys(skb, &keys))
42aecaa9 1950 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
1951 else
1952 skb_set_transport_header(skb, offset_hint);
1953}
1954
03606895
ED
1955static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1956{
1957 if (skb_mac_header_was_set(skb)) {
1958 const unsigned char *old_mac = skb_mac_header(skb);
1959
1960 skb_set_mac_header(skb, -skb->mac_len);
1961 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1962 }
1963}
1964
04fb451e
MM
1965static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1966{
1967 return skb->csum_start - skb_headroom(skb);
1968}
1969
2e07fa9c
ACM
1970static inline int skb_transport_offset(const struct sk_buff *skb)
1971{
1972 return skb_transport_header(skb) - skb->data;
1973}
1974
1975static inline u32 skb_network_header_len(const struct sk_buff *skb)
1976{
1977 return skb->transport_header - skb->network_header;
1978}
1979
6a674e9c
JG
1980static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1981{
1982 return skb->inner_transport_header - skb->inner_network_header;
1983}
1984
2e07fa9c
ACM
1985static inline int skb_network_offset(const struct sk_buff *skb)
1986{
1987 return skb_network_header(skb) - skb->data;
1988}
48d49d0c 1989
6a674e9c
JG
1990static inline int skb_inner_network_offset(const struct sk_buff *skb)
1991{
1992 return skb_inner_network_header(skb) - skb->data;
1993}
1994
f9599ce1
CG
1995static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1996{
1997 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1998}
1999
1da177e4
LT
2000/*
2001 * CPUs often take a performance hit when accessing unaligned memory
2002 * locations. The actual performance hit varies, it can be small if the
2003 * hardware handles it or large if we have to take an exception and fix it
2004 * in software.
2005 *
2006 * Since an ethernet header is 14 bytes network drivers often end up with
2007 * the IP header at an unaligned offset. The IP header can be aligned by
2008 * shifting the start of the packet by 2 bytes. Drivers should do this
2009 * with:
2010 *
8660c124 2011 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2012 *
2013 * The downside to this alignment of the IP header is that the DMA is now
2014 * unaligned. On some architectures the cost of an unaligned DMA is high
2015 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2016 *
1da177e4
LT
2017 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2018 * to be overridden.
2019 */
2020#ifndef NET_IP_ALIGN
2021#define NET_IP_ALIGN 2
2022#endif
2023
025be81e
AB
2024/*
2025 * The networking layer reserves some headroom in skb data (via
2026 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2027 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2028 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2029 *
2030 * Unfortunately this headroom changes the DMA alignment of the resulting
2031 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2032 * on some architectures. An architecture can override this value,
2033 * perhaps setting it to a cacheline in size (since that will maintain
2034 * cacheline alignment of the DMA). It must be a power of 2.
2035 *
d6301d3d 2036 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2037 * headroom, you should not reduce this.
5933dd2f
ED
2038 *
2039 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2040 * to reduce average number of cache lines per packet.
2041 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2042 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2043 */
2044#ifndef NET_SKB_PAD
5933dd2f 2045#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2046#endif
2047
7965bd4d 2048int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2049
2050static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2051{
c4264f27 2052 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2053 WARN_ON(1);
2054 return;
2055 }
27a884dc
ACM
2056 skb->len = len;
2057 skb_set_tail_pointer(skb, len);
1da177e4
LT
2058}
2059
7965bd4d 2060void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2061
2062static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2063{
3cc0e873
HX
2064 if (skb->data_len)
2065 return ___pskb_trim(skb, len);
2066 __skb_trim(skb, len);
2067 return 0;
1da177e4
LT
2068}
2069
2070static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2071{
2072 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2073}
2074
e9fa4f7b
HX
2075/**
2076 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2077 * @skb: buffer to alter
2078 * @len: new length
2079 *
2080 * This is identical to pskb_trim except that the caller knows that
2081 * the skb is not cloned so we should never get an error due to out-
2082 * of-memory.
2083 */
2084static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2085{
2086 int err = pskb_trim(skb, len);
2087 BUG_ON(err);
2088}
2089
1da177e4
LT
2090/**
2091 * skb_orphan - orphan a buffer
2092 * @skb: buffer to orphan
2093 *
2094 * If a buffer currently has an owner then we call the owner's
2095 * destructor function and make the @skb unowned. The buffer continues
2096 * to exist but is no longer charged to its former owner.
2097 */
2098static inline void skb_orphan(struct sk_buff *skb)
2099{
c34a7612 2100 if (skb->destructor) {
1da177e4 2101 skb->destructor(skb);
c34a7612
ED
2102 skb->destructor = NULL;
2103 skb->sk = NULL;
376c7311
ED
2104 } else {
2105 BUG_ON(skb->sk);
c34a7612 2106 }
1da177e4
LT
2107}
2108
a353e0ce
MT
2109/**
2110 * skb_orphan_frags - orphan the frags contained in a buffer
2111 * @skb: buffer to orphan frags from
2112 * @gfp_mask: allocation mask for replacement pages
2113 *
2114 * For each frag in the SKB which needs a destructor (i.e. has an
2115 * owner) create a copy of that frag and release the original
2116 * page by calling the destructor.
2117 */
2118static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2119{
2120 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2121 return 0;
2122 return skb_copy_ubufs(skb, gfp_mask);
2123}
2124
1da177e4
LT
2125/**
2126 * __skb_queue_purge - empty a list
2127 * @list: list to empty
2128 *
2129 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2130 * the list and one reference dropped. This function does not take the
2131 * list lock and the caller must hold the relevant locks to use it.
2132 */
7965bd4d 2133void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2134static inline void __skb_queue_purge(struct sk_buff_head *list)
2135{
2136 struct sk_buff *skb;
2137 while ((skb = __skb_dequeue(list)) != NULL)
2138 kfree_skb(skb);
2139}
2140
7965bd4d 2141void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2142
7965bd4d
JP
2143struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2144 gfp_t gfp_mask);
8af27456
CH
2145
2146/**
2147 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2148 * @dev: network device to receive on
2149 * @length: length to allocate
2150 *
2151 * Allocate a new &sk_buff and assign it a usage count of one. The
2152 * buffer has unspecified headroom built in. Users should allocate
2153 * the headroom they think they need without accounting for the
2154 * built in space. The built in space is used for optimisations.
2155 *
2156 * %NULL is returned if there is no free memory. Although this function
2157 * allocates memory it can be called from an interrupt.
2158 */
2159static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2160 unsigned int length)
8af27456
CH
2161{
2162 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2163}
2164
6f532612
ED
2165/* legacy helper around __netdev_alloc_skb() */
2166static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2167 gfp_t gfp_mask)
2168{
2169 return __netdev_alloc_skb(NULL, length, gfp_mask);
2170}
2171
2172/* legacy helper around netdev_alloc_skb() */
2173static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2174{
2175 return netdev_alloc_skb(NULL, length);
2176}
2177
2178
4915a0de
ED
2179static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2180 unsigned int length, gfp_t gfp)
61321bbd 2181{
4915a0de 2182 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2183
2184 if (NET_IP_ALIGN && skb)
2185 skb_reserve(skb, NET_IP_ALIGN);
2186 return skb;
2187}
2188
4915a0de
ED
2189static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2190 unsigned int length)
2191{
2192 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2193}
2194
181edb2b
AD
2195static inline void skb_free_frag(void *addr)
2196{
2197 __free_page_frag(addr);
2198}
2199
ffde7328 2200void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2201struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2202 unsigned int length, gfp_t gfp_mask);
2203static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2204 unsigned int length)
2205{
2206 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2207}
ffde7328 2208
71dfda58
AD
2209/**
2210 * __dev_alloc_pages - allocate page for network Rx
2211 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2212 * @order: size of the allocation
2213 *
2214 * Allocate a new page.
2215 *
2216 * %NULL is returned if there is no free memory.
2217*/
2218static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2219 unsigned int order)
2220{
2221 /* This piece of code contains several assumptions.
2222 * 1. This is for device Rx, therefor a cold page is preferred.
2223 * 2. The expectation is the user wants a compound page.
2224 * 3. If requesting a order 0 page it will not be compound
2225 * due to the check to see if order has a value in prep_new_page
2226 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2227 * code in gfp_to_alloc_flags that should be enforcing this.
2228 */
2229 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2230
2231 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2232}
2233
2234static inline struct page *dev_alloc_pages(unsigned int order)
2235{
2236 return __dev_alloc_pages(GFP_ATOMIC, order);
2237}
2238
2239/**
2240 * __dev_alloc_page - allocate a page for network Rx
2241 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2242 *
2243 * Allocate a new page.
2244 *
2245 * %NULL is returned if there is no free memory.
2246 */
2247static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2248{
2249 return __dev_alloc_pages(gfp_mask, 0);
2250}
2251
2252static inline struct page *dev_alloc_page(void)
2253{
2254 return __dev_alloc_page(GFP_ATOMIC);
2255}
2256
0614002b
MG
2257/**
2258 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2259 * @page: The page that was allocated from skb_alloc_page
2260 * @skb: The skb that may need pfmemalloc set
2261 */
2262static inline void skb_propagate_pfmemalloc(struct page *page,
2263 struct sk_buff *skb)
2264{
2265 if (page && page->pfmemalloc)
2266 skb->pfmemalloc = true;
2267}
2268
131ea667 2269/**
e227867f 2270 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2271 * @frag: the paged fragment
2272 *
2273 * Returns the &struct page associated with @frag.
2274 */
2275static inline struct page *skb_frag_page(const skb_frag_t *frag)
2276{
a8605c60 2277 return frag->page.p;
131ea667
IC
2278}
2279
2280/**
2281 * __skb_frag_ref - take an addition reference on a paged fragment.
2282 * @frag: the paged fragment
2283 *
2284 * Takes an additional reference on the paged fragment @frag.
2285 */
2286static inline void __skb_frag_ref(skb_frag_t *frag)
2287{
2288 get_page(skb_frag_page(frag));
2289}
2290
2291/**
2292 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2293 * @skb: the buffer
2294 * @f: the fragment offset.
2295 *
2296 * Takes an additional reference on the @f'th paged fragment of @skb.
2297 */
2298static inline void skb_frag_ref(struct sk_buff *skb, int f)
2299{
2300 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2301}
2302
2303/**
2304 * __skb_frag_unref - release a reference on a paged fragment.
2305 * @frag: the paged fragment
2306 *
2307 * Releases a reference on the paged fragment @frag.
2308 */
2309static inline void __skb_frag_unref(skb_frag_t *frag)
2310{
2311 put_page(skb_frag_page(frag));
2312}
2313
2314/**
2315 * skb_frag_unref - release a reference on a paged fragment of an skb.
2316 * @skb: the buffer
2317 * @f: the fragment offset
2318 *
2319 * Releases a reference on the @f'th paged fragment of @skb.
2320 */
2321static inline void skb_frag_unref(struct sk_buff *skb, int f)
2322{
2323 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2324}
2325
2326/**
2327 * skb_frag_address - gets the address of the data contained in a paged fragment
2328 * @frag: the paged fragment buffer
2329 *
2330 * Returns the address of the data within @frag. The page must already
2331 * be mapped.
2332 */
2333static inline void *skb_frag_address(const skb_frag_t *frag)
2334{
2335 return page_address(skb_frag_page(frag)) + frag->page_offset;
2336}
2337
2338/**
2339 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2340 * @frag: the paged fragment buffer
2341 *
2342 * Returns the address of the data within @frag. Checks that the page
2343 * is mapped and returns %NULL otherwise.
2344 */
2345static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2346{
2347 void *ptr = page_address(skb_frag_page(frag));
2348 if (unlikely(!ptr))
2349 return NULL;
2350
2351 return ptr + frag->page_offset;
2352}
2353
2354/**
2355 * __skb_frag_set_page - sets the page contained in a paged fragment
2356 * @frag: the paged fragment
2357 * @page: the page to set
2358 *
2359 * Sets the fragment @frag to contain @page.
2360 */
2361static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2362{
a8605c60 2363 frag->page.p = page;
131ea667
IC
2364}
2365
2366/**
2367 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2368 * @skb: the buffer
2369 * @f: the fragment offset
2370 * @page: the page to set
2371 *
2372 * Sets the @f'th fragment of @skb to contain @page.
2373 */
2374static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2375 struct page *page)
2376{
2377 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2378}
2379
400dfd3a
ED
2380bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2381
131ea667
IC
2382/**
2383 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2384 * @dev: the device to map the fragment to
131ea667
IC
2385 * @frag: the paged fragment to map
2386 * @offset: the offset within the fragment (starting at the
2387 * fragment's own offset)
2388 * @size: the number of bytes to map
f83347df 2389 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2390 *
2391 * Maps the page associated with @frag to @device.
2392 */
2393static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2394 const skb_frag_t *frag,
2395 size_t offset, size_t size,
2396 enum dma_data_direction dir)
2397{
2398 return dma_map_page(dev, skb_frag_page(frag),
2399 frag->page_offset + offset, size, dir);
2400}
2401
117632e6
ED
2402static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2403 gfp_t gfp_mask)
2404{
2405 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2406}
2407
bad93e9d
OP
2408
2409static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2410 gfp_t gfp_mask)
2411{
2412 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2413}
2414
2415
334a8132
PM
2416/**
2417 * skb_clone_writable - is the header of a clone writable
2418 * @skb: buffer to check
2419 * @len: length up to which to write
2420 *
2421 * Returns true if modifying the header part of the cloned buffer
2422 * does not requires the data to be copied.
2423 */
05bdd2f1 2424static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2425{
2426 return !skb_header_cloned(skb) &&
2427 skb_headroom(skb) + len <= skb->hdr_len;
2428}
2429
d9cc2048
HX
2430static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2431 int cloned)
2432{
2433 int delta = 0;
2434
d9cc2048
HX
2435 if (headroom > skb_headroom(skb))
2436 delta = headroom - skb_headroom(skb);
2437
2438 if (delta || cloned)
2439 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2440 GFP_ATOMIC);
2441 return 0;
2442}
2443
1da177e4
LT
2444/**
2445 * skb_cow - copy header of skb when it is required
2446 * @skb: buffer to cow
2447 * @headroom: needed headroom
2448 *
2449 * If the skb passed lacks sufficient headroom or its data part
2450 * is shared, data is reallocated. If reallocation fails, an error
2451 * is returned and original skb is not changed.
2452 *
2453 * The result is skb with writable area skb->head...skb->tail
2454 * and at least @headroom of space at head.
2455 */
2456static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2457{
d9cc2048
HX
2458 return __skb_cow(skb, headroom, skb_cloned(skb));
2459}
1da177e4 2460
d9cc2048
HX
2461/**
2462 * skb_cow_head - skb_cow but only making the head writable
2463 * @skb: buffer to cow
2464 * @headroom: needed headroom
2465 *
2466 * This function is identical to skb_cow except that we replace the
2467 * skb_cloned check by skb_header_cloned. It should be used when
2468 * you only need to push on some header and do not need to modify
2469 * the data.
2470 */
2471static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2472{
2473 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2474}
2475
2476/**
2477 * skb_padto - pad an skbuff up to a minimal size
2478 * @skb: buffer to pad
2479 * @len: minimal length
2480 *
2481 * Pads up a buffer to ensure the trailing bytes exist and are
2482 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2483 * is untouched. Otherwise it is extended. Returns zero on
2484 * success. The skb is freed on error.
1da177e4 2485 */
5b057c6b 2486static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2487{
2488 unsigned int size = skb->len;
2489 if (likely(size >= len))
5b057c6b 2490 return 0;
987c402a 2491 return skb_pad(skb, len - size);
1da177e4
LT
2492}
2493
9c0c1124
AD
2494/**
2495 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2496 * @skb: buffer to pad
2497 * @len: minimal length
2498 *
2499 * Pads up a buffer to ensure the trailing bytes exist and are
2500 * blanked. If the buffer already contains sufficient data it
2501 * is untouched. Otherwise it is extended. Returns zero on
2502 * success. The skb is freed on error.
2503 */
2504static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2505{
2506 unsigned int size = skb->len;
2507
2508 if (unlikely(size < len)) {
2509 len -= size;
2510 if (skb_pad(skb, len))
2511 return -ENOMEM;
2512 __skb_put(skb, len);
2513 }
2514 return 0;
2515}
2516
1da177e4 2517static inline int skb_add_data(struct sk_buff *skb,
af2b040e 2518 struct iov_iter *from, int copy)
1da177e4
LT
2519{
2520 const int off = skb->len;
2521
2522 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e
AV
2523 __wsum csum = 0;
2524 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2525 &csum, from) == copy) {
1da177e4
LT
2526 skb->csum = csum_block_add(skb->csum, csum, off);
2527 return 0;
2528 }
af2b040e 2529 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
1da177e4
LT
2530 return 0;
2531
2532 __skb_trim(skb, off);
2533 return -EFAULT;
2534}
2535
38ba0a65
ED
2536static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2537 const struct page *page, int off)
1da177e4
LT
2538{
2539 if (i) {
9e903e08 2540 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2541
ea2ab693 2542 return page == skb_frag_page(frag) &&
9e903e08 2543 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2544 }
38ba0a65 2545 return false;
1da177e4
LT
2546}
2547
364c6bad
HX
2548static inline int __skb_linearize(struct sk_buff *skb)
2549{
2550 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2551}
2552
1da177e4
LT
2553/**
2554 * skb_linearize - convert paged skb to linear one
2555 * @skb: buffer to linarize
1da177e4
LT
2556 *
2557 * If there is no free memory -ENOMEM is returned, otherwise zero
2558 * is returned and the old skb data released.
2559 */
364c6bad
HX
2560static inline int skb_linearize(struct sk_buff *skb)
2561{
2562 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2563}
2564
cef401de
ED
2565/**
2566 * skb_has_shared_frag - can any frag be overwritten
2567 * @skb: buffer to test
2568 *
2569 * Return true if the skb has at least one frag that might be modified
2570 * by an external entity (as in vmsplice()/sendfile())
2571 */
2572static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2573{
c9af6db4
PS
2574 return skb_is_nonlinear(skb) &&
2575 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2576}
2577
364c6bad
HX
2578/**
2579 * skb_linearize_cow - make sure skb is linear and writable
2580 * @skb: buffer to process
2581 *
2582 * If there is no free memory -ENOMEM is returned, otherwise zero
2583 * is returned and the old skb data released.
2584 */
2585static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2586{
364c6bad
HX
2587 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2588 __skb_linearize(skb) : 0;
1da177e4
LT
2589}
2590
2591/**
2592 * skb_postpull_rcsum - update checksum for received skb after pull
2593 * @skb: buffer to update
2594 * @start: start of data before pull
2595 * @len: length of data pulled
2596 *
2597 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2598 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2599 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4
LT
2600 */
2601
2602static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2603 const void *start, unsigned int len)
1da177e4 2604{
84fa7933 2605 if (skb->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
2606 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2607}
2608
cbb042f9
HX
2609unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2610
7ce5a27f
DM
2611/**
2612 * pskb_trim_rcsum - trim received skb and update checksum
2613 * @skb: buffer to trim
2614 * @len: new length
2615 *
2616 * This is exactly the same as pskb_trim except that it ensures the
2617 * checksum of received packets are still valid after the operation.
2618 */
2619
2620static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2621{
2622 if (likely(len >= skb->len))
2623 return 0;
2624 if (skb->ip_summed == CHECKSUM_COMPLETE)
2625 skb->ip_summed = CHECKSUM_NONE;
2626 return __pskb_trim(skb, len);
2627}
2628
1da177e4
LT
2629#define skb_queue_walk(queue, skb) \
2630 for (skb = (queue)->next; \
a1e4891f 2631 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2632 skb = skb->next)
2633
46f8914e
JC
2634#define skb_queue_walk_safe(queue, skb, tmp) \
2635 for (skb = (queue)->next, tmp = skb->next; \
2636 skb != (struct sk_buff *)(queue); \
2637 skb = tmp, tmp = skb->next)
2638
1164f52a 2639#define skb_queue_walk_from(queue, skb) \
a1e4891f 2640 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2641 skb = skb->next)
2642
2643#define skb_queue_walk_from_safe(queue, skb, tmp) \
2644 for (tmp = skb->next; \
2645 skb != (struct sk_buff *)(queue); \
2646 skb = tmp, tmp = skb->next)
2647
300ce174
SH
2648#define skb_queue_reverse_walk(queue, skb) \
2649 for (skb = (queue)->prev; \
a1e4891f 2650 skb != (struct sk_buff *)(queue); \
300ce174
SH
2651 skb = skb->prev)
2652
686a2955
DM
2653#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2654 for (skb = (queue)->prev, tmp = skb->prev; \
2655 skb != (struct sk_buff *)(queue); \
2656 skb = tmp, tmp = skb->prev)
2657
2658#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2659 for (tmp = skb->prev; \
2660 skb != (struct sk_buff *)(queue); \
2661 skb = tmp, tmp = skb->prev)
1da177e4 2662
21dc3301 2663static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2664{
2665 return skb_shinfo(skb)->frag_list != NULL;
2666}
2667
2668static inline void skb_frag_list_init(struct sk_buff *skb)
2669{
2670 skb_shinfo(skb)->frag_list = NULL;
2671}
2672
2673static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2674{
2675 frag->next = skb_shinfo(skb)->frag_list;
2676 skb_shinfo(skb)->frag_list = frag;
2677}
2678
2679#define skb_walk_frags(skb, iter) \
2680 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2681
7965bd4d
JP
2682struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2683 int *peeked, int *off, int *err);
2684struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2685 int *err);
2686unsigned int datagram_poll(struct file *file, struct socket *sock,
2687 struct poll_table_struct *wait);
c0371da6
AV
2688int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2689 struct iov_iter *to, int size);
51f3d02b
DM
2690static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2691 struct msghdr *msg, int size)
2692{
e5a4b0bb 2693 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 2694}
e5a4b0bb
AV
2695int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2696 struct msghdr *msg);
3a654f97
AV
2697int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2698 struct iov_iter *from, int len);
3a654f97 2699int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d
JP
2700void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2701void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2702int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
2703int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2704int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2705__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2706 int len, __wsum csum);
a60e3cc7
HFS
2707ssize_t skb_socket_splice(struct sock *sk,
2708 struct pipe_inode_info *pipe,
2709 struct splice_pipe_desc *spd);
2710int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 2711 struct pipe_inode_info *pipe, unsigned int len,
a60e3cc7
HFS
2712 unsigned int flags,
2713 ssize_t (*splice_cb)(struct sock *,
2714 struct pipe_inode_info *,
2715 struct splice_pipe_desc *));
7965bd4d 2716void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 2717unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
2718int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2719 int len, int hlen);
7965bd4d
JP
2720void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2721int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2722void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 2723unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
7965bd4d 2724struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 2725struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 2726int skb_ensure_writable(struct sk_buff *skb, int write_len);
93515d53
JP
2727int skb_vlan_pop(struct sk_buff *skb);
2728int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
20380731 2729
6ce8e9ce
AV
2730static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2731{
21226abb 2732 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
6ce8e9ce
AV
2733}
2734
7eab8d9e
AV
2735static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2736{
e5a4b0bb 2737 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
2738}
2739
2817a336
DB
2740struct skb_checksum_ops {
2741 __wsum (*update)(const void *mem, int len, __wsum wsum);
2742 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2743};
2744
2745__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2746 __wsum csum, const struct skb_checksum_ops *ops);
2747__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2748 __wsum csum);
2749
690e36e7
DM
2750static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2751 int len, void *data, int hlen, void *buffer)
1da177e4 2752{
55820ee2 2753 if (hlen - offset >= len)
690e36e7 2754 return data + offset;
1da177e4 2755
690e36e7
DM
2756 if (!skb ||
2757 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
2758 return NULL;
2759
2760 return buffer;
2761}
2762
690e36e7
DM
2763static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2764 int len, void *buffer)
2765{
2766 return __skb_header_pointer(skb, offset, len, skb->data,
2767 skb_headlen(skb), buffer);
2768}
2769
4262e5cc
DB
2770/**
2771 * skb_needs_linearize - check if we need to linearize a given skb
2772 * depending on the given device features.
2773 * @skb: socket buffer to check
2774 * @features: net device features
2775 *
2776 * Returns true if either:
2777 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2778 * 2. skb is fragmented and the device does not support SG.
2779 */
2780static inline bool skb_needs_linearize(struct sk_buff *skb,
2781 netdev_features_t features)
2782{
2783 return skb_is_nonlinear(skb) &&
2784 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2785 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2786}
2787
d626f62b
ACM
2788static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2789 void *to,
2790 const unsigned int len)
2791{
2792 memcpy(to, skb->data, len);
2793}
2794
2795static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2796 const int offset, void *to,
2797 const unsigned int len)
2798{
2799 memcpy(to, skb->data + offset, len);
2800}
2801
27d7ff46
ACM
2802static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2803 const void *from,
2804 const unsigned int len)
2805{
2806 memcpy(skb->data, from, len);
2807}
2808
2809static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2810 const int offset,
2811 const void *from,
2812 const unsigned int len)
2813{
2814 memcpy(skb->data + offset, from, len);
2815}
2816
7965bd4d 2817void skb_init(void);
1da177e4 2818
ac45f602
PO
2819static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2820{
2821 return skb->tstamp;
2822}
2823
a61bbcf2
PM
2824/**
2825 * skb_get_timestamp - get timestamp from a skb
2826 * @skb: skb to get stamp from
2827 * @stamp: pointer to struct timeval to store stamp in
2828 *
2829 * Timestamps are stored in the skb as offsets to a base timestamp.
2830 * This function converts the offset back to a struct timeval and stores
2831 * it in stamp.
2832 */
ac45f602
PO
2833static inline void skb_get_timestamp(const struct sk_buff *skb,
2834 struct timeval *stamp)
a61bbcf2 2835{
b7aa0bf7 2836 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
2837}
2838
ac45f602
PO
2839static inline void skb_get_timestampns(const struct sk_buff *skb,
2840 struct timespec *stamp)
2841{
2842 *stamp = ktime_to_timespec(skb->tstamp);
2843}
2844
b7aa0bf7 2845static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 2846{
b7aa0bf7 2847 skb->tstamp = ktime_get_real();
a61bbcf2
PM
2848}
2849
164891aa
SH
2850static inline ktime_t net_timedelta(ktime_t t)
2851{
2852 return ktime_sub(ktime_get_real(), t);
2853}
2854
b9ce204f
IJ
2855static inline ktime_t net_invalid_timestamp(void)
2856{
2857 return ktime_set(0, 0);
2858}
a61bbcf2 2859
62bccb8c
AD
2860struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2861
c1f19b51
RC
2862#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2863
7965bd4d
JP
2864void skb_clone_tx_timestamp(struct sk_buff *skb);
2865bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
2866
2867#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2868
2869static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2870{
2871}
2872
2873static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2874{
2875 return false;
2876}
2877
2878#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2879
2880/**
2881 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2882 *
da92b194
RC
2883 * PHY drivers may accept clones of transmitted packets for
2884 * timestamping via their phy_driver.txtstamp method. These drivers
2885 * must call this function to return the skb back to the stack, with
2886 * or without a timestamp.
2887 *
c1f19b51 2888 * @skb: clone of the the original outgoing packet
da92b194 2889 * @hwtstamps: hardware time stamps, may be NULL if not available
c1f19b51
RC
2890 *
2891 */
2892void skb_complete_tx_timestamp(struct sk_buff *skb,
2893 struct skb_shared_hwtstamps *hwtstamps);
2894
e7fd2885
WB
2895void __skb_tstamp_tx(struct sk_buff *orig_skb,
2896 struct skb_shared_hwtstamps *hwtstamps,
2897 struct sock *sk, int tstype);
2898
ac45f602
PO
2899/**
2900 * skb_tstamp_tx - queue clone of skb with send time stamps
2901 * @orig_skb: the original outgoing packet
2902 * @hwtstamps: hardware time stamps, may be NULL if not available
2903 *
2904 * If the skb has a socket associated, then this function clones the
2905 * skb (thus sharing the actual data and optional structures), stores
2906 * the optional hardware time stamping information (if non NULL) or
2907 * generates a software time stamp (otherwise), then queues the clone
2908 * to the error queue of the socket. Errors are silently ignored.
2909 */
7965bd4d
JP
2910void skb_tstamp_tx(struct sk_buff *orig_skb,
2911 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 2912
4507a715
RC
2913static inline void sw_tx_timestamp(struct sk_buff *skb)
2914{
2244d07b
OH
2915 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2916 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
2917 skb_tstamp_tx(skb, NULL);
2918}
2919
2920/**
2921 * skb_tx_timestamp() - Driver hook for transmit timestamping
2922 *
2923 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 2924 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 2925 *
73409f3b
DM
2926 * Specifically, one should make absolutely sure that this function is
2927 * called before TX completion of this packet can trigger. Otherwise
2928 * the packet could potentially already be freed.
2929 *
4507a715
RC
2930 * @skb: A socket buffer.
2931 */
2932static inline void skb_tx_timestamp(struct sk_buff *skb)
2933{
c1f19b51 2934 skb_clone_tx_timestamp(skb);
4507a715
RC
2935 sw_tx_timestamp(skb);
2936}
2937
6e3e939f
JB
2938/**
2939 * skb_complete_wifi_ack - deliver skb with wifi status
2940 *
2941 * @skb: the original outgoing packet
2942 * @acked: ack status
2943 *
2944 */
2945void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2946
7965bd4d
JP
2947__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2948__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 2949
60476372
HX
2950static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2951{
6edec0e6
TH
2952 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
2953 skb->csum_valid ||
2954 (skb->ip_summed == CHECKSUM_PARTIAL &&
2955 skb_checksum_start_offset(skb) >= 0));
60476372
HX
2956}
2957
fb286bb2
HX
2958/**
2959 * skb_checksum_complete - Calculate checksum of an entire packet
2960 * @skb: packet to process
2961 *
2962 * This function calculates the checksum over the entire packet plus
2963 * the value of skb->csum. The latter can be used to supply the
2964 * checksum of a pseudo header as used by TCP/UDP. It returns the
2965 * checksum.
2966 *
2967 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2968 * this function can be used to verify that checksum on received
2969 * packets. In that case the function should return zero if the
2970 * checksum is correct. In particular, this function will return zero
2971 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2972 * hardware has already verified the correctness of the checksum.
2973 */
4381ca3c 2974static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 2975{
60476372
HX
2976 return skb_csum_unnecessary(skb) ?
2977 0 : __skb_checksum_complete(skb);
fb286bb2
HX
2978}
2979
77cffe23
TH
2980static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2981{
2982 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2983 if (skb->csum_level == 0)
2984 skb->ip_summed = CHECKSUM_NONE;
2985 else
2986 skb->csum_level--;
2987 }
2988}
2989
2990static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2991{
2992 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2993 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2994 skb->csum_level++;
2995 } else if (skb->ip_summed == CHECKSUM_NONE) {
2996 skb->ip_summed = CHECKSUM_UNNECESSARY;
2997 skb->csum_level = 0;
2998 }
2999}
3000
5a212329
TH
3001static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3002{
3003 /* Mark current checksum as bad (typically called from GRO
3004 * path). In the case that ip_summed is CHECKSUM_NONE
3005 * this must be the first checksum encountered in the packet.
3006 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3007 * checksum after the last one validated. For UDP, a zero
3008 * checksum can not be marked as bad.
3009 */
3010
3011 if (skb->ip_summed == CHECKSUM_NONE ||
3012 skb->ip_summed == CHECKSUM_UNNECESSARY)
3013 skb->csum_bad = 1;
3014}
3015
76ba0aae
TH
3016/* Check if we need to perform checksum complete validation.
3017 *
3018 * Returns true if checksum complete is needed, false otherwise
3019 * (either checksum is unnecessary or zero checksum is allowed).
3020 */
3021static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3022 bool zero_okay,
3023 __sum16 check)
3024{
5d0c2b95
TH
3025 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3026 skb->csum_valid = 1;
77cffe23 3027 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3028 return false;
3029 }
3030
3031 return true;
3032}
3033
3034/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3035 * in checksum_init.
3036 */
3037#define CHECKSUM_BREAK 76
3038
4e18b9ad
TH
3039/* Unset checksum-complete
3040 *
3041 * Unset checksum complete can be done when packet is being modified
3042 * (uncompressed for instance) and checksum-complete value is
3043 * invalidated.
3044 */
3045static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3046{
3047 if (skb->ip_summed == CHECKSUM_COMPLETE)
3048 skb->ip_summed = CHECKSUM_NONE;
3049}
3050
76ba0aae
TH
3051/* Validate (init) checksum based on checksum complete.
3052 *
3053 * Return values:
3054 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3055 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3056 * checksum is stored in skb->csum for use in __skb_checksum_complete
3057 * non-zero: value of invalid checksum
3058 *
3059 */
3060static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3061 bool complete,
3062 __wsum psum)
3063{
3064 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3065 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3066 skb->csum_valid = 1;
76ba0aae
TH
3067 return 0;
3068 }
5a212329
TH
3069 } else if (skb->csum_bad) {
3070 /* ip_summed == CHECKSUM_NONE in this case */
c91d4606 3071 return (__force __sum16)1;
76ba0aae
TH
3072 }
3073
3074 skb->csum = psum;
3075
5d0c2b95
TH
3076 if (complete || skb->len <= CHECKSUM_BREAK) {
3077 __sum16 csum;
3078
3079 csum = __skb_checksum_complete(skb);
3080 skb->csum_valid = !csum;
3081 return csum;
3082 }
76ba0aae
TH
3083
3084 return 0;
3085}
3086
3087static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3088{
3089 return 0;
3090}
3091
3092/* Perform checksum validate (init). Note that this is a macro since we only
3093 * want to calculate the pseudo header which is an input function if necessary.
3094 * First we try to validate without any computation (checksum unnecessary) and
3095 * then calculate based on checksum complete calling the function to compute
3096 * pseudo header.
3097 *
3098 * Return values:
3099 * 0: checksum is validated or try to in skb_checksum_complete
3100 * non-zero: value of invalid checksum
3101 */
3102#define __skb_checksum_validate(skb, proto, complete, \
3103 zero_okay, check, compute_pseudo) \
3104({ \
3105 __sum16 __ret = 0; \
5d0c2b95 3106 skb->csum_valid = 0; \
76ba0aae
TH
3107 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3108 __ret = __skb_checksum_validate_complete(skb, \
3109 complete, compute_pseudo(skb, proto)); \
3110 __ret; \
3111})
3112
3113#define skb_checksum_init(skb, proto, compute_pseudo) \
3114 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3115
3116#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3117 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3118
3119#define skb_checksum_validate(skb, proto, compute_pseudo) \
3120 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3121
3122#define skb_checksum_validate_zero_check(skb, proto, check, \
3123 compute_pseudo) \
096a4cfa 3124 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3125
3126#define skb_checksum_simple_validate(skb) \
3127 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3128
d96535a1
TH
3129static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3130{
3131 return (skb->ip_summed == CHECKSUM_NONE &&
3132 skb->csum_valid && !skb->csum_bad);
3133}
3134
3135static inline void __skb_checksum_convert(struct sk_buff *skb,
3136 __sum16 check, __wsum pseudo)
3137{
3138 skb->csum = ~pseudo;
3139 skb->ip_summed = CHECKSUM_COMPLETE;
3140}
3141
3142#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3143do { \
3144 if (__skb_checksum_convert_check(skb)) \
3145 __skb_checksum_convert(skb, check, \
3146 compute_pseudo(skb, proto)); \
3147} while (0)
3148
15e2396d
TH
3149static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3150 u16 start, u16 offset)
3151{
3152 skb->ip_summed = CHECKSUM_PARTIAL;
3153 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3154 skb->csum_offset = offset - start;
3155}
3156
dcdc8994
TH
3157/* Update skbuf and packet to reflect the remote checksum offload operation.
3158 * When called, ptr indicates the starting point for skb->csum when
3159 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3160 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3161 */
3162static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3163 int start, int offset, bool nopartial)
dcdc8994
TH
3164{
3165 __wsum delta;
3166
15e2396d
TH
3167 if (!nopartial) {
3168 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3169 return;
3170 }
3171
dcdc8994
TH
3172 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3173 __skb_checksum_complete(skb);
3174 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3175 }
3176
3177 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3178
3179 /* Adjust skb->csum since we changed the packet */
3180 skb->csum = csum_add(skb->csum, delta);
3181}
3182
5f79e0f9 3183#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3184void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3185static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3186{
3187 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3188 nf_conntrack_destroy(nfct);
1da177e4
LT
3189}
3190static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3191{
3192 if (nfct)
3193 atomic_inc(&nfct->use);
3194}
2fc72c7b 3195#endif
34666d46 3196#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3197static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3198{
3199 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3200 kfree(nf_bridge);
3201}
3202static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3203{
3204 if (nf_bridge)
3205 atomic_inc(&nf_bridge->use);
3206}
3207#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3208static inline void nf_reset(struct sk_buff *skb)
3209{
5f79e0f9 3210#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a193a4ab
PM
3211 nf_conntrack_put(skb->nfct);
3212 skb->nfct = NULL;
2fc72c7b 3213#endif
34666d46 3214#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3215 nf_bridge_put(skb->nf_bridge);
3216 skb->nf_bridge = NULL;
3217#endif
3218}
3219
124dff01
PM
3220static inline void nf_reset_trace(struct sk_buff *skb)
3221{
478b360a 3222#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3223 skb->nf_trace = 0;
3224#endif
a193a4ab
PM
3225}
3226
edda553c 3227/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3228static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3229 bool copy)
edda553c 3230{
5f79e0f9 3231#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
edda553c
YK
3232 dst->nfct = src->nfct;
3233 nf_conntrack_get(src->nfct);
b1937227
ED
3234 if (copy)
3235 dst->nfctinfo = src->nfctinfo;
2fc72c7b 3236#endif
34666d46 3237#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3238 dst->nf_bridge = src->nf_bridge;
3239 nf_bridge_get(src->nf_bridge);
3240#endif
478b360a 3241#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3242 if (copy)
3243 dst->nf_trace = src->nf_trace;
478b360a 3244#endif
edda553c
YK
3245}
3246
e7ac05f3
YK
3247static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3248{
e7ac05f3 3249#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
5f79e0f9 3250 nf_conntrack_put(dst->nfct);
2fc72c7b 3251#endif
34666d46 3252#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3253 nf_bridge_put(dst->nf_bridge);
3254#endif
b1937227 3255 __nf_copy(dst, src, true);
e7ac05f3
YK
3256}
3257
984bc16c
JM
3258#ifdef CONFIG_NETWORK_SECMARK
3259static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3260{
3261 to->secmark = from->secmark;
3262}
3263
3264static inline void skb_init_secmark(struct sk_buff *skb)
3265{
3266 skb->secmark = 0;
3267}
3268#else
3269static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3270{ }
3271
3272static inline void skb_init_secmark(struct sk_buff *skb)
3273{ }
3274#endif
3275
574f7194
EB
3276static inline bool skb_irq_freeable(const struct sk_buff *skb)
3277{
3278 return !skb->destructor &&
3279#if IS_ENABLED(CONFIG_XFRM)
3280 !skb->sp &&
3281#endif
3282#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3283 !skb->nfct &&
3284#endif
3285 !skb->_skb_refdst &&
3286 !skb_has_frag_list(skb);
3287}
3288
f25f4e44
PWJ
3289static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3290{
f25f4e44 3291 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3292}
3293
9247744e 3294static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3295{
4e3ab47a 3296 return skb->queue_mapping;
4e3ab47a
PE
3297}
3298
f25f4e44
PWJ
3299static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3300{
f25f4e44 3301 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3302}
3303
d5a9e24a
DM
3304static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3305{
3306 skb->queue_mapping = rx_queue + 1;
3307}
3308
9247744e 3309static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3310{
3311 return skb->queue_mapping - 1;
3312}
3313
9247744e 3314static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3315{
a02cec21 3316 return skb->queue_mapping != 0;
d5a9e24a
DM
3317}
3318
def8b4fa
AD
3319static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3320{
0b3d8e08 3321#ifdef CONFIG_XFRM
def8b4fa 3322 return skb->sp;
def8b4fa 3323#else
def8b4fa 3324 return NULL;
def8b4fa 3325#endif
0b3d8e08 3326}
def8b4fa 3327
68c33163
PS
3328/* Keeps track of mac header offset relative to skb->head.
3329 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3330 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3331 * tunnel skb it points to outer mac header.
3332 * Keeps track of level of encapsulation of network headers.
3333 */
68c33163 3334struct skb_gso_cb {
3347c960
ED
3335 int mac_offset;
3336 int encap_level;
7e2b10c1 3337 __u16 csum_start;
68c33163
PS
3338};
3339#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3340
3341static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3342{
3343 return (skb_mac_header(inner_skb) - inner_skb->head) -
3344 SKB_GSO_CB(inner_skb)->mac_offset;
3345}
3346
1e2bd517
PS
3347static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3348{
3349 int new_headroom, headroom;
3350 int ret;
3351
3352 headroom = skb_headroom(skb);
3353 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3354 if (ret)
3355 return ret;
3356
3357 new_headroom = skb_headroom(skb);
3358 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3359 return 0;
3360}
3361
7e2b10c1
TH
3362/* Compute the checksum for a gso segment. First compute the checksum value
3363 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3364 * then add in skb->csum (checksum from csum_start to end of packet).
3365 * skb->csum and csum_start are then updated to reflect the checksum of the
3366 * resultant packet starting from the transport header-- the resultant checksum
3367 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3368 * header.
3369 */
3370static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3371{
3372 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
c91d4606
ED
3373 skb_transport_offset(skb);
3374 __wsum partial;
7e2b10c1 3375
c91d4606 3376 partial = csum_partial(skb_transport_header(skb), plen, skb->csum);
7e2b10c1
TH
3377 skb->csum = res;
3378 SKB_GSO_CB(skb)->csum_start -= plen;
3379
c91d4606 3380 return csum_fold(partial);
7e2b10c1
TH
3381}
3382
bdcc0924 3383static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3384{
3385 return skb_shinfo(skb)->gso_size;
3386}
3387
36a8f39e 3388/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3389static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3390{
3391 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3392}
3393
7965bd4d 3394void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3395
3396static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3397{
3398 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3399 * wanted then gso_type will be set. */
05bdd2f1
ED
3400 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3401
b78462eb
AD
3402 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3403 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3404 __skb_warn_lro_forwarding(skb);
3405 return true;
3406 }
3407 return false;
3408}
3409
35fc92a9
HX
3410static inline void skb_forward_csum(struct sk_buff *skb)
3411{
3412 /* Unfortunately we don't support this one. Any brave souls? */
3413 if (skb->ip_summed == CHECKSUM_COMPLETE)
3414 skb->ip_summed = CHECKSUM_NONE;
3415}
3416
bc8acf2c
ED
3417/**
3418 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3419 * @skb: skb to check
3420 *
3421 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3422 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3423 * use this helper, to document places where we make this assertion.
3424 */
05bdd2f1 3425static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3426{
3427#ifdef DEBUG
3428 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3429#endif
3430}
3431
f35d9d8a 3432bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3433
ed1f50c3 3434int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
3435struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3436 unsigned int transport_len,
3437 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 3438
3a7c1ee4
AD
3439/**
3440 * skb_head_is_locked - Determine if the skb->head is locked down
3441 * @skb: skb to check
3442 *
3443 * The head on skbs build around a head frag can be removed if they are
3444 * not cloned. This function returns true if the skb head is locked down
3445 * due to either being allocated via kmalloc, or by being a clone with
3446 * multiple references to the head.
3447 */
3448static inline bool skb_head_is_locked(const struct sk_buff *skb)
3449{
3450 return !skb->head_frag || skb_cloned(skb);
3451}
fe6cc55f
FW
3452
3453/**
3454 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3455 *
3456 * @skb: GSO skb
3457 *
3458 * skb_gso_network_seglen is used to determine the real size of the
3459 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3460 *
3461 * The MAC/L2 header is not accounted for.
3462 */
3463static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3464{
3465 unsigned int hdr_len = skb_transport_header(skb) -
3466 skb_network_header(skb);
3467 return hdr_len + skb_gso_transport_seglen(skb);
3468}
1da177e4
LT
3469#endif /* __KERNEL__ */
3470#endif /* _LINUX_SKBUFF_H */