treewide: Use fallthrough pseudo-keyword
[linux-block.git] / include / linux / skbuff.h
CommitLineData
2874c5fd 1/* SPDX-License-Identifier: GPL-2.0-or-later */
1da177e4
LT
2/*
3 * Definitions for the 'struct sk_buff' memory handlers.
4 *
5 * Authors:
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
1da177e4
LT
8 */
9
10#ifndef _LINUX_SKBUFF_H
11#define _LINUX_SKBUFF_H
12
1da177e4
LT
13#include <linux/kernel.h>
14#include <linux/compiler.h>
15#include <linux/time.h>
187f1882 16#include <linux/bug.h>
8842d285 17#include <linux/bvec.h>
1da177e4 18#include <linux/cache.h>
56b17425 19#include <linux/rbtree.h>
51f3d02b 20#include <linux/socket.h>
c1d1b437 21#include <linux/refcount.h>
1da177e4 22
60063497 23#include <linux/atomic.h>
1da177e4
LT
24#include <asm/types.h>
25#include <linux/spinlock.h>
1da177e4 26#include <linux/net.h>
3fc7e8a6 27#include <linux/textsearch.h>
1da177e4 28#include <net/checksum.h>
a80958f4 29#include <linux/rcupdate.h>
b7aa0bf7 30#include <linux/hrtimer.h>
131ea667 31#include <linux/dma-mapping.h>
c8f44aff 32#include <linux/netdev_features.h>
363ec392 33#include <linux/sched.h>
e6017571 34#include <linux/sched/clock.h>
1bd758eb 35#include <net/flow_dissector.h>
a60e3cc7 36#include <linux/splice.h>
72b31f72 37#include <linux/in6.h>
8b10cab6 38#include <linux/if_packet.h>
f70ea018 39#include <net/flow.h>
261db6c2
JS
40#if IS_ENABLED(CONFIG_NF_CONNTRACK)
41#include <linux/netfilter/nf_conntrack_common.h>
42#endif
1da177e4 43
7a6ae71b
TH
44/* The interface for checksum offload between the stack and networking drivers
45 * is as follows...
46 *
47 * A. IP checksum related features
48 *
49 * Drivers advertise checksum offload capabilities in the features of a device.
db1f00fb
DC
50 * From the stack's point of view these are capabilities offered by the driver.
51 * A driver typically only advertises features that it is capable of offloading
7a6ae71b
TH
52 * to its device.
53 *
54 * The checksum related features are:
55 *
56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
57 * IP (one's complement) checksum for any combination
58 * of protocols or protocol layering. The checksum is
59 * computed and set in a packet per the CHECKSUM_PARTIAL
60 * interface (see below).
61 *
62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63 * TCP or UDP packets over IPv4. These are specifically
64 * unencapsulated packets of the form IPv4|TCP or
65 * IPv4|UDP where the Protocol field in the IPv4 header
db1f00fb 66 * is TCP or UDP. The IPv4 header may contain IP options.
7a6ae71b
TH
67 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
70 *
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
74 * IPv4|UDP where the Next Header field in the IPv6
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
db1f00fb 82 * This flag is only used to disable the RX checksum
7a6ae71b
TH
83 * feature for a device. The stack will accept receive
84 * checksum indication in packets received on a device
85 * regardless of whether NETIF_F_RXCSUM is set.
86 *
87 * B. Checksumming of received packets by device. Indication of checksum
db1f00fb 88 * verification is set in skb->ip_summed. Possible values are:
78ea85f1
DB
89 *
90 * CHECKSUM_NONE:
91 *
7a6ae71b 92 * Device did not checksum this packet e.g. due to lack of capabilities.
78ea85f1
DB
93 * The packet contains full (though not verified) checksum in packet but
94 * not in skb->csum. Thus, skb->csum is undefined in this case.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * The hardware you're dealing with doesn't calculate the full checksum
99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101 * if their checksums are okay. skb->csum is still undefined in this case
7a6ae71b
TH
102 * though. A driver or device must never modify the checksum field in the
103 * packet even if checksum is verified.
77cffe23
TH
104 *
105 * CHECKSUM_UNNECESSARY is applicable to following protocols:
106 * TCP: IPv6 and IPv4.
107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108 * zero UDP checksum for either IPv4 or IPv6, the networking stack
109 * may perform further validation in this case.
110 * GRE: only if the checksum is present in the header.
111 * SCTP: indicates the CRC in SCTP header has been validated.
b4759dcd 112 * FCOE: indicates the CRC in FC frame has been validated.
77cffe23
TH
113 *
114 * skb->csum_level indicates the number of consecutive checksums found in
115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
117 * and a device is able to verify the checksums for UDP (possibly zero),
db1f00fb 118 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to
77cffe23 119 * two. If the device were only able to verify the UDP checksum and not
db1f00fb 120 * GRE, either because it doesn't support GRE checksum or because GRE
77cffe23
TH
121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
122 * not considered in this case).
78ea85f1
DB
123 *
124 * CHECKSUM_COMPLETE:
125 *
126 * This is the most generic way. The device supplied checksum of the _whole_
db1f00fb 127 * packet as seen by netif_rx() and fills in skb->csum. This means the
78ea85f1
DB
128 * hardware doesn't need to parse L3/L4 headers to implement this.
129 *
b4759dcd
DC
130 * Notes:
131 * - Even if device supports only some protocols, but is able to produce
132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
78ea85f1
DB
134 *
135 * CHECKSUM_PARTIAL:
136 *
6edec0e6
TH
137 * A checksum is set up to be offloaded to a device as described in the
138 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 139 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
140 * on the same host, or it may be set in the input path in GRO or remote
141 * checksum offload. For the purposes of checksum verification, the checksum
142 * referred to by skb->csum_start + skb->csum_offset and any preceding
143 * checksums in the packet are considered verified. Any checksums in the
144 * packet that are after the checksum being offloaded are not considered to
145 * be verified.
78ea85f1 146 *
7a6ae71b
TH
147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
148 * in the skb->ip_summed for a packet. Values are:
78ea85f1
DB
149 *
150 * CHECKSUM_PARTIAL:
151 *
7a6ae71b 152 * The driver is required to checksum the packet as seen by hard_start_xmit()
78ea85f1 153 * from skb->csum_start up to the end, and to record/write the checksum at
7a6ae71b
TH
154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
155 * csum_start and csum_offset values are valid values given the length and
db1f00fb
DC
156 * offset of the packet, but it should not attempt to validate that the
157 * checksum refers to a legitimate transport layer checksum -- it is the
7a6ae71b
TH
158 * purview of the stack to validate that csum_start and csum_offset are set
159 * correctly.
160 *
161 * When the stack requests checksum offload for a packet, the driver MUST
162 * ensure that the checksum is set correctly. A driver can either offload the
163 * checksum calculation to the device, or call skb_checksum_help (in the case
164 * that the device does not support offload for a particular checksum).
165 *
166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
43c26a1a
DC
168 * checksum offload capability.
169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
170 * on network device checksumming capabilities: if a packet does not match
171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
172 * csum_not_inet, see item D.) is called to resolve the checksum.
78ea85f1 173 *
7a6ae71b 174 * CHECKSUM_NONE:
78ea85f1 175 *
7a6ae71b
TH
176 * The skb was already checksummed by the protocol, or a checksum is not
177 * required.
78ea85f1
DB
178 *
179 * CHECKSUM_UNNECESSARY:
180 *
db1f00fb 181 * This has the same meaning as CHECKSUM_NONE for checksum offload on
7a6ae71b 182 * output.
78ea85f1 183 *
7a6ae71b
TH
184 * CHECKSUM_COMPLETE:
185 * Not used in checksum output. If a driver observes a packet with this value
db1f00fb 186 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set.
7a6ae71b
TH
187 *
188 * D. Non-IP checksum (CRC) offloads
189 *
190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191 * offloading the SCTP CRC in a packet. To perform this offload the stack
db1f00fb 192 * will set csum_start and csum_offset accordingly, set ip_summed to
dba00306
DC
193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195 * A driver that supports both IP checksum offload and SCTP CRC32c offload
196 * must verify which offload is configured for a packet by testing the
197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
7a6ae71b
TH
199 *
200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201 * offloading the FCOE CRC in a packet. To perform this offload the stack
202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
db1f00fb
DC
203 * accordingly. Note that there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
7a6ae71b 205 * both IP checksum offload and FCOE CRC offload must verify which offload
db1f00fb 206 * is configured for a packet, presumably by inspecting packet headers.
7a6ae71b
TH
207 *
208 * E. Checksumming on output with GSO.
209 *
210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213 * part of the GSO operation is implied. If a checksum is being offloaded
db1f00fb
DC
214 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and
215 * csum_offset are set to refer to the outermost checksum being offloaded
216 * (two offloaded checksums are possible with UDP encapsulation).
78ea85f1
DB
217 */
218
60476372 219/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
220#define CHECKSUM_NONE 0
221#define CHECKSUM_UNNECESSARY 1
222#define CHECKSUM_COMPLETE 2
223#define CHECKSUM_PARTIAL 3
1da177e4 224
77cffe23
TH
225/* Maximum value in skb->csum_level */
226#define SKB_MAX_CSUM_LEVEL 3
227
0bec8c88 228#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 229#define SKB_WITH_OVERHEAD(X) \
deea84b0 230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
231#define SKB_MAX_ORDER(X, ORDER) \
232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
233#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
234#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
235
87fb4b7b
ED
236/* return minimum truesize of one skb containing X bytes of data */
237#define SKB_TRUESIZE(X) ((X) + \
238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240
7999096f 241struct ahash_request;
1da177e4 242struct net_device;
716ea3a7 243struct scatterlist;
9c55e01c 244struct pipe_inode_info;
a8f820aa 245struct iov_iter;
fd11a83d 246struct napi_struct;
d58e468b
PP
247struct bpf_prog;
248union bpf_attr;
df5042f4 249struct skb_ext;
1da177e4 250
34666d46 251#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 252struct nf_bridge_info {
3eaf4025
FW
253 enum {
254 BRNF_PROTO_UNCHANGED,
255 BRNF_PROTO_8021Q,
256 BRNF_PROTO_PPPOE
7fb48c5b 257 } orig_proto:8;
72b1e5e4
FW
258 u8 pkt_otherhost:1;
259 u8 in_prerouting:1;
260 u8 bridged_dnat:1;
411ffb4f 261 __u16 frag_max_size;
bf1ac5ca 262 struct net_device *physindev;
63cdbc06
FW
263
264 /* always valid & non-NULL from FORWARD on, for physdev match */
265 struct net_device *physoutdev;
7fb48c5b 266 union {
72b1e5e4 267 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
268 __be32 ipv4_daddr;
269 struct in6_addr ipv6_daddr;
72b1e5e4
FW
270
271 /* after prerouting + nat detected: store original source
272 * mac since neigh resolution overwrites it, only used while
273 * skb is out in neigh layer.
274 */
275 char neigh_header[8];
72b31f72 276 };
1da177e4
LT
277};
278#endif
279
95a7233c
PB
280#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
281/* Chain in tc_skb_ext will be used to share the tc chain with
282 * ovs recirc_id. It will be set to the current chain by tc
283 * and read by ovs to recirc_id.
284 */
285struct tc_skb_ext {
286 __u32 chain;
038ebb1a 287 __u16 mru;
95a7233c
PB
288};
289#endif
290
1da177e4
LT
291struct sk_buff_head {
292 /* These two members must be first. */
293 struct sk_buff *next;
294 struct sk_buff *prev;
295
296 __u32 qlen;
297 spinlock_t lock;
298};
299
300struct sk_buff;
301
9d4dde52
IC
302/* To allow 64K frame to be packed as single skb without frag_list we
303 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
304 * buffers which do not start on a page boundary.
305 *
306 * Since GRO uses frags we allocate at least 16 regardless of page
307 * size.
a715dea3 308 */
9d4dde52 309#if (65536/PAGE_SIZE + 1) < 16
eec00954 310#define MAX_SKB_FRAGS 16UL
a715dea3 311#else
9d4dde52 312#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 313#endif
5f74f82e 314extern int sysctl_max_skb_frags;
1da177e4 315
3953c46c
MRL
316/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
317 * segment using its current segmentation instead.
318 */
319#define GSO_BY_FRAGS 0xFFFF
320
8842d285 321typedef struct bio_vec skb_frag_t;
1da177e4 322
161e6137 323/**
7240b60c 324 * skb_frag_size() - Returns the size of a skb fragment
161e6137
PT
325 * @frag: skb fragment
326 */
9e903e08
ED
327static inline unsigned int skb_frag_size(const skb_frag_t *frag)
328{
b8b576a1 329 return frag->bv_len;
9e903e08
ED
330}
331
161e6137 332/**
7240b60c 333 * skb_frag_size_set() - Sets the size of a skb fragment
161e6137
PT
334 * @frag: skb fragment
335 * @size: size of fragment
336 */
9e903e08
ED
337static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
338{
b8b576a1 339 frag->bv_len = size;
9e903e08
ED
340}
341
161e6137 342/**
7240b60c 343 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
161e6137
PT
344 * @frag: skb fragment
345 * @delta: value to add
346 */
9e903e08
ED
347static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
348{
b8b576a1 349 frag->bv_len += delta;
9e903e08
ED
350}
351
161e6137 352/**
7240b60c 353 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
161e6137
PT
354 * @frag: skb fragment
355 * @delta: value to subtract
356 */
9e903e08
ED
357static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
358{
b8b576a1 359 frag->bv_len -= delta;
9e903e08
ED
360}
361
161e6137
PT
362/**
363 * skb_frag_must_loop - Test if %p is a high memory page
364 * @p: fragment's page
365 */
c613c209
WB
366static inline bool skb_frag_must_loop(struct page *p)
367{
368#if defined(CONFIG_HIGHMEM)
369 if (PageHighMem(p))
370 return true;
371#endif
372 return false;
373}
374
375/**
376 * skb_frag_foreach_page - loop over pages in a fragment
377 *
378 * @f: skb frag to operate on
1dfa5bd3 379 * @f_off: offset from start of f->bv_page
c613c209
WB
380 * @f_len: length from f_off to loop over
381 * @p: (temp var) current page
382 * @p_off: (temp var) offset from start of current page,
383 * non-zero only on first page.
384 * @p_len: (temp var) length in current page,
385 * < PAGE_SIZE only on first and last page.
386 * @copied: (temp var) length so far, excluding current p_len.
387 *
388 * A fragment can hold a compound page, in which case per-page
389 * operations, notably kmap_atomic, must be called for each
390 * regular page.
391 */
392#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
393 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
394 p_off = (f_off) & (PAGE_SIZE - 1), \
395 p_len = skb_frag_must_loop(p) ? \
396 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
397 copied = 0; \
398 copied < f_len; \
399 copied += p_len, p++, p_off = 0, \
400 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
401
ac45f602
PO
402#define HAVE_HW_TIME_STAMP
403
404/**
d3a21be8 405 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
406 * @hwtstamp: hardware time stamp transformed into duration
407 * since arbitrary point in time
ac45f602
PO
408 *
409 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 410 * skb->tstamp.
ac45f602
PO
411 *
412 * hwtstamps can only be compared against other hwtstamps from
413 * the same device.
414 *
415 * This structure is attached to packets as part of the
416 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
417 */
418struct skb_shared_hwtstamps {
419 ktime_t hwtstamp;
ac45f602
PO
420};
421
2244d07b
OH
422/* Definitions for tx_flags in struct skb_shared_info */
423enum {
424 /* generate hardware time stamp */
425 SKBTX_HW_TSTAMP = 1 << 0,
426
e7fd2885 427 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
428 SKBTX_SW_TSTAMP = 1 << 1,
429
430 /* device driver is going to provide hardware time stamp */
431 SKBTX_IN_PROGRESS = 1 << 2,
432
a6686f2f 433 /* device driver supports TX zero-copy buffers */
62b1a8ab 434 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
435
436 /* generate wifi status information (where possible) */
62b1a8ab 437 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
438
439 /* This indicates at least one fragment might be overwritten
440 * (as in vmsplice(), sendfile() ...)
441 * If we need to compute a TX checksum, we'll need to copy
442 * all frags to avoid possible bad checksum
443 */
444 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
445
446 /* generate software time stamp when entering packet scheduling */
447 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
448};
449
52267790 450#define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
e1c8a607 451#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 452 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
453#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
454
a6686f2f
SM
455/*
456 * The callback notifies userspace to release buffers when skb DMA is done in
457 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
458 * The zerocopy_success argument is true if zero copy transmit occurred,
459 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
460 * The ctx field is used to track device context.
461 * The desc field is used to track userspace buffer index.
a6686f2f
SM
462 */
463struct ubuf_info {
e19d6763 464 void (*callback)(struct ubuf_info *, bool zerocopy_success);
4ab6c99d
WB
465 union {
466 struct {
467 unsigned long desc;
468 void *ctx;
469 };
470 struct {
471 u32 id;
472 u16 len;
473 u16 zerocopy:1;
474 u32 bytelen;
475 };
476 };
c1d1b437 477 refcount_t refcnt;
a91dbff5
WB
478
479 struct mmpin {
480 struct user_struct *user;
481 unsigned int num_pg;
482 } mmp;
ac45f602
PO
483};
484
52267790
WB
485#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
486
6f89dbce
SV
487int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
488void mm_unaccount_pinned_pages(struct mmpin *mmp);
489
52267790 490struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
4ab6c99d
WB
491struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
492 struct ubuf_info *uarg);
52267790
WB
493
494static inline void sock_zerocopy_get(struct ubuf_info *uarg)
495{
c1d1b437 496 refcount_inc(&uarg->refcnt);
52267790
WB
497}
498
499void sock_zerocopy_put(struct ubuf_info *uarg);
52900d22 500void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
52267790
WB
501
502void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
503
b5947e5d 504int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
52267790
WB
505int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
506 struct msghdr *msg, int len,
507 struct ubuf_info *uarg);
508
1da177e4
LT
509/* This data is invariant across clones and lives at
510 * the end of the header data, ie. at skb->end.
511 */
512struct skb_shared_info {
de8f3a83
DB
513 __u8 __unused;
514 __u8 meta_len;
515 __u8 nr_frags;
9f42f126 516 __u8 tx_flags;
7967168c
HX
517 unsigned short gso_size;
518 /* Warning: this field is not always filled in (UFO)! */
519 unsigned short gso_segs;
1da177e4 520 struct sk_buff *frag_list;
ac45f602 521 struct skb_shared_hwtstamps hwtstamps;
7f564528 522 unsigned int gso_type;
09c2d251 523 u32 tskey;
ec7d2f2c
ED
524
525 /*
526 * Warning : all fields before dataref are cleared in __alloc_skb()
527 */
528 atomic_t dataref;
529
69e3c75f
JB
530 /* Intermediate layers must ensure that destructor_arg
531 * remains valid until skb destructor */
532 void * destructor_arg;
a6686f2f 533
fed66381
ED
534 /* must be last field, see pskb_expand_head() */
535 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
536};
537
538/* We divide dataref into two halves. The higher 16 bits hold references
539 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
540 * the entire skb->data. A clone of a headerless skb holds the length of
541 * the header in skb->hdr_len.
1da177e4
LT
542 *
543 * All users must obey the rule that the skb->data reference count must be
544 * greater than or equal to the payload reference count.
545 *
546 * Holding a reference to the payload part means that the user does not
547 * care about modifications to the header part of skb->data.
548 */
549#define SKB_DATAREF_SHIFT 16
550#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
551
d179cd12
DM
552
553enum {
c8753d55
VS
554 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
555 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
556 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
557};
558
7967168c
HX
559enum {
560 SKB_GSO_TCPV4 = 1 << 0,
576a30eb
HX
561
562 /* This indicates the skb is from an untrusted source. */
d9d30adf 563 SKB_GSO_DODGY = 1 << 1,
b0da8537
MC
564
565 /* This indicates the tcp segment has CWR set. */
d9d30adf 566 SKB_GSO_TCP_ECN = 1 << 2,
f83ef8c0 567
d9d30adf 568 SKB_GSO_TCP_FIXEDID = 1 << 3,
01d5b2fc 569
d9d30adf 570 SKB_GSO_TCPV6 = 1 << 4,
68c33163 571
d9d30adf 572 SKB_GSO_FCOE = 1 << 5,
73136267 573
d9d30adf 574 SKB_GSO_GRE = 1 << 6,
0d89d203 575
d9d30adf 576 SKB_GSO_GRE_CSUM = 1 << 7,
cb32f511 577
d9d30adf 578 SKB_GSO_IPXIP4 = 1 << 8,
61c1db7f 579
d9d30adf 580 SKB_GSO_IPXIP6 = 1 << 9,
0f4f4ffa 581
d9d30adf 582 SKB_GSO_UDP_TUNNEL = 1 << 10,
4749c09c 583
d9d30adf 584 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
cbc53e08 585
d9d30adf 586 SKB_GSO_PARTIAL = 1 << 12,
802ab55a 587
d9d30adf 588 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
90017acc 589
d9d30adf 590 SKB_GSO_SCTP = 1 << 14,
c7ef8f0c 591
d9d30adf 592 SKB_GSO_ESP = 1 << 15,
0c19f846
WB
593
594 SKB_GSO_UDP = 1 << 16,
ee80d1eb
WB
595
596 SKB_GSO_UDP_L4 = 1 << 17,
3b335832
SK
597
598 SKB_GSO_FRAGLIST = 1 << 18,
7967168c
HX
599};
600
2e07fa9c
ACM
601#if BITS_PER_LONG > 32
602#define NET_SKBUFF_DATA_USES_OFFSET 1
603#endif
604
605#ifdef NET_SKBUFF_DATA_USES_OFFSET
606typedef unsigned int sk_buff_data_t;
607#else
608typedef unsigned char *sk_buff_data_t;
609#endif
610
161e6137 611/**
1da177e4
LT
612 * struct sk_buff - socket buffer
613 * @next: Next buffer in list
614 * @prev: Previous buffer in list
363ec392 615 * @tstamp: Time we arrived/left
d2f273f0
RD
616 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
617 * for retransmit timer
56b17425 618 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d2f273f0 619 * @list: queue head
d84e0bd7 620 * @sk: Socket we are owned by
d2f273f0
RD
621 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
622 * fragmentation management
1da177e4 623 * @dev: Device we arrived on/are leaving by
d2f273f0 624 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
d84e0bd7 625 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 626 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 627 * @sp: the security path, used for xfrm
1da177e4
LT
628 * @len: Length of actual data
629 * @data_len: Data length
630 * @mac_len: Length of link layer header
334a8132 631 * @hdr_len: writable header length of cloned skb
663ead3b
HX
632 * @csum: Checksum (must include start/offset pair)
633 * @csum_start: Offset from skb->head where checksumming should start
634 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 635 * @priority: Packet queueing priority
60ff7467 636 * @ignore_df: allow local fragmentation
1da177e4 637 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 638 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
639 * @nohdr: Payload reference only, must not modify header
640 * @pkt_type: Packet class
c83c2486 641 * @fclone: skbuff clone status
c83c2486 642 * @ipvs_property: skbuff is owned by ipvs
d2f273f0
RD
643 * @inner_protocol_type: whether the inner protocol is
644 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
645 * @remcsum_offload: remote checksum offload is enabled
875e8939
IS
646 * @offload_fwd_mark: Packet was L2-forwarded in hardware
647 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
e7246e12 648 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 649 * @tc_at_ingress: used within tc_classify to distinguish in/egress
2c64605b
PNA
650 * @redirected: packet was redirected by packet classifier
651 * @from_ingress: packet was redirected from the ingress path
31729363
RD
652 * @peeked: this packet has been seen already, so stats have been
653 * done for it, don't do them again
ba9dda3a 654 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
655 * @protocol: Packet protocol from driver
656 * @destructor: Destruct function
e2080072 657 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
a9e419dc 658 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 659 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 660 * @skb_iif: ifindex of device we arrived on
1da177e4 661 * @tc_index: Traffic control index
61b905da 662 * @hash: the packet hash
d84e0bd7 663 * @queue_mapping: Queue mapping for multiqueue devices
d2f273f0
RD
664 * @head_frag: skb was allocated from page fragments,
665 * not allocated by kmalloc() or vmalloc().
8b700862 666 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
df5042f4 667 * @active_extensions: active extensions (skb_ext_id types)
553a5672 668 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 669 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 670 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 671 * ports.
a3b18ddb 672 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
673 * @wifi_acked_valid: wifi_acked was set
674 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 675 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
d2f273f0
RD
676 * @encapsulation: indicates the inner headers in the skbuff are valid
677 * @encap_hdr_csum: software checksum is needed
678 * @csum_valid: checksum is already valid
dba00306 679 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
d2f273f0
RD
680 * @csum_complete_sw: checksum was completed by software
681 * @csum_level: indicates the number of consecutive checksums found in
682 * the packet minus one that have been verified as
683 * CHECKSUM_UNNECESSARY (max 3)
4ff06203 684 * @dst_pending_confirm: need to confirm neighbour
a48d189e 685 * @decrypted: Decrypted SKB
161e6137 686 * @napi_id: id of the NAPI struct this skb came from
d2f273f0 687 * @sender_cpu: (aka @napi_id) source CPU in XPS
984bc16c 688 * @secmark: security marking
d84e0bd7 689 * @mark: Generic packet mark
d2f273f0
RD
690 * @reserved_tailroom: (aka @mark) number of bytes of free space available
691 * at the tail of an sk_buff
692 * @vlan_present: VLAN tag is present
86a9bad3 693 * @vlan_proto: vlan encapsulation protocol
6aa895b0 694 * @vlan_tci: vlan tag control information
0d89d203 695 * @inner_protocol: Protocol (encapsulation)
d2f273f0
RD
696 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
697 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
6a674e9c
JG
698 * @inner_transport_header: Inner transport layer header (encapsulation)
699 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 700 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
701 * @transport_header: Transport layer header
702 * @network_header: Network layer header
703 * @mac_header: Link layer header
704 * @tail: Tail pointer
705 * @end: End pointer
706 * @head: Head of buffer
707 * @data: Data head pointer
708 * @truesize: Buffer size
709 * @users: User count - see {datagram,tcp}.c
df5042f4 710 * @extensions: allocated extensions, valid if active_extensions is nonzero
1da177e4
LT
711 */
712
713struct sk_buff {
363ec392 714 union {
56b17425
ED
715 struct {
716 /* These two members must be first. */
717 struct sk_buff *next;
718 struct sk_buff *prev;
719
720 union {
bffa72cf
ED
721 struct net_device *dev;
722 /* Some protocols might use this space to store information,
723 * while device pointer would be NULL.
724 * UDP receive path is one user.
725 */
726 unsigned long dev_scratch;
56b17425
ED
727 };
728 };
fa0f5273 729 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
d4546c25 730 struct list_head list;
363ec392 731 };
fa0f5273
PO
732
733 union {
734 struct sock *sk;
735 int ip_defrag_offset;
736 };
1da177e4 737
c84d9490 738 union {
bffa72cf 739 ktime_t tstamp;
d3edd06e 740 u64 skb_mstamp_ns; /* earliest departure time */
c84d9490 741 };
1da177e4
LT
742 /*
743 * This is the control buffer. It is free to use for every
744 * layer. Please put your private variables there. If you
745 * want to keep them across layers you have to do a skb_clone()
746 * first. This is owned by whoever has the skb queued ATM.
747 */
da3f5cf1 748 char cb[48] __aligned(8);
1da177e4 749
e2080072
ED
750 union {
751 struct {
752 unsigned long _skb_refdst;
753 void (*destructor)(struct sk_buff *skb);
754 };
755 struct list_head tcp_tsorted_anchor;
756 };
757
b1937227 758#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 759 unsigned long _nfct;
da3f5cf1 760#endif
1da177e4 761 unsigned int len,
334a8132
PM
762 data_len;
763 __u16 mac_len,
764 hdr_len;
b1937227
ED
765
766 /* Following fields are _not_ copied in __copy_skb_header()
767 * Note that queue_mapping is here mostly to fill a hole.
768 */
b1937227 769 __u16 queue_mapping;
36bbef52
DB
770
771/* if you move cloned around you also must adapt those constants */
772#ifdef __BIG_ENDIAN_BITFIELD
773#define CLONED_MASK (1 << 7)
774#else
775#define CLONED_MASK 1
776#endif
777#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
778
d2f273f0 779 /* private: */
36bbef52 780 __u8 __cloned_offset[0];
d2f273f0 781 /* public: */
b1937227 782 __u8 cloned:1,
6869c4d8 783 nohdr:1,
b84f4cc9 784 fclone:2,
a59322be 785 peeked:1,
b1937227 786 head_frag:1,
8b700862 787 pfmemalloc:1;
df5042f4
FW
788#ifdef CONFIG_SKB_EXTENSIONS
789 __u8 active_extensions;
790#endif
b1937227
ED
791 /* fields enclosed in headers_start/headers_end are copied
792 * using a single memcpy() in __copy_skb_header()
793 */
ebcf34f3 794 /* private: */
b1937227 795 __u32 headers_start[0];
ebcf34f3 796 /* public: */
4031ae6e 797
233577a2
HFS
798/* if you move pkt_type around you also must adapt those constants */
799#ifdef __BIG_ENDIAN_BITFIELD
800#define PKT_TYPE_MAX (7 << 5)
801#else
802#define PKT_TYPE_MAX 7
1da177e4 803#endif
233577a2 804#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 805
d2f273f0 806 /* private: */
233577a2 807 __u8 __pkt_type_offset[0];
d2f273f0 808 /* public: */
b1937227 809 __u8 pkt_type:3;
b1937227 810 __u8 ignore_df:1;
b1937227
ED
811 __u8 nf_trace:1;
812 __u8 ip_summed:2;
3853b584 813 __u8 ooo_okay:1;
8b700862 814
61b905da 815 __u8 l4_hash:1;
a3b18ddb 816 __u8 sw_hash:1;
6e3e939f
JB
817 __u8 wifi_acked_valid:1;
818 __u8 wifi_acked:1;
3bdc0eba 819 __u8 no_fcs:1;
77cffe23 820 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 821 __u8 encapsulation:1;
7e2b10c1 822 __u8 encap_hdr_csum:1;
5d0c2b95 823 __u8 csum_valid:1;
8b700862 824
0c4b2d37
MM
825#ifdef __BIG_ENDIAN_BITFIELD
826#define PKT_VLAN_PRESENT_BIT 7
827#else
828#define PKT_VLAN_PRESENT_BIT 0
829#endif
830#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
d2f273f0 831 /* private: */
0c4b2d37 832 __u8 __pkt_vlan_present_offset[0];
d2f273f0 833 /* public: */
0c4b2d37 834 __u8 vlan_present:1;
7e3cead5 835 __u8 csum_complete_sw:1;
b1937227 836 __u8 csum_level:2;
dba00306 837 __u8 csum_not_inet:1;
4ff06203 838 __u8 dst_pending_confirm:1;
b1937227
ED
839#ifdef CONFIG_IPV6_NDISC_NODETYPE
840 __u8 ndisc_nodetype:2;
841#endif
8b700862 842
0c4b2d37 843 __u8 ipvs_property:1;
8bce6d7d 844 __u8 inner_protocol_type:1;
e585f236 845 __u8 remcsum_offload:1;
6bc506b4
IS
846#ifdef CONFIG_NET_SWITCHDEV
847 __u8 offload_fwd_mark:1;
875e8939 848 __u8 offload_l3_fwd_mark:1;
6bc506b4 849#endif
e7246e12
WB
850#ifdef CONFIG_NET_CLS_ACT
851 __u8 tc_skip_classify:1;
8dc07fdb 852 __u8 tc_at_ingress:1;
2c64605b
PNA
853#endif
854#ifdef CONFIG_NET_REDIRECT
855 __u8 redirected:1;
856 __u8 from_ingress:1;
e7246e12 857#endif
a48d189e
SB
858#ifdef CONFIG_TLS_DEVICE
859 __u8 decrypted:1;
860#endif
b1937227
ED
861
862#ifdef CONFIG_NET_SCHED
863 __u16 tc_index; /* traffic control index */
b1937227 864#endif
fe55f6d5 865
b1937227
ED
866 union {
867 __wsum csum;
868 struct {
869 __u16 csum_start;
870 __u16 csum_offset;
871 };
872 };
873 __u32 priority;
874 int skb_iif;
875 __u32 hash;
876 __be16 vlan_proto;
877 __u16 vlan_tci;
2bd82484
ED
878#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
879 union {
880 unsigned int napi_id;
881 unsigned int sender_cpu;
882 };
97fc2f08 883#endif
984bc16c 884#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 885 __u32 secmark;
0c4f691f 886#endif
0c4f691f 887
3b885787
NH
888 union {
889 __u32 mark;
16fad69c 890 __u32 reserved_tailroom;
3b885787 891 };
1da177e4 892
8bce6d7d
TH
893 union {
894 __be16 inner_protocol;
895 __u8 inner_ipproto;
896 };
897
1a37e412
SH
898 __u16 inner_transport_header;
899 __u16 inner_network_header;
900 __u16 inner_mac_header;
b1937227
ED
901
902 __be16 protocol;
1a37e412
SH
903 __u16 transport_header;
904 __u16 network_header;
905 __u16 mac_header;
b1937227 906
ebcf34f3 907 /* private: */
b1937227 908 __u32 headers_end[0];
ebcf34f3 909 /* public: */
b1937227 910
1da177e4 911 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 912 sk_buff_data_t tail;
4305b541 913 sk_buff_data_t end;
1da177e4 914 unsigned char *head,
4305b541 915 *data;
27a884dc 916 unsigned int truesize;
63354797 917 refcount_t users;
df5042f4
FW
918
919#ifdef CONFIG_SKB_EXTENSIONS
920 /* only useable after checking ->active_extensions != 0 */
921 struct skb_ext *extensions;
922#endif
1da177e4
LT
923};
924
925#ifdef __KERNEL__
926/*
927 * Handling routines are only of interest to the kernel
928 */
1da177e4 929
c93bdd0e
MG
930#define SKB_ALLOC_FCLONE 0x01
931#define SKB_ALLOC_RX 0x02
fd11a83d 932#define SKB_ALLOC_NAPI 0x04
c93bdd0e 933
161e6137
PT
934/**
935 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
936 * @skb: buffer
937 */
c93bdd0e
MG
938static inline bool skb_pfmemalloc(const struct sk_buff *skb)
939{
940 return unlikely(skb->pfmemalloc);
941}
942
7fee226a
ED
943/*
944 * skb might have a dst pointer attached, refcounted or not.
945 * _skb_refdst low order bit is set if refcount was _not_ taken
946 */
947#define SKB_DST_NOREF 1UL
948#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
949
950/**
951 * skb_dst - returns skb dst_entry
952 * @skb: buffer
953 *
954 * Returns skb dst_entry, regardless of reference taken or not.
955 */
adf30907
ED
956static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
957{
161e6137 958 /* If refdst was not refcounted, check we still are in a
7fee226a
ED
959 * rcu_read_lock section
960 */
961 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
962 !rcu_read_lock_held() &&
963 !rcu_read_lock_bh_held());
964 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
965}
966
7fee226a
ED
967/**
968 * skb_dst_set - sets skb dst
969 * @skb: buffer
970 * @dst: dst entry
971 *
972 * Sets skb dst, assuming a reference was taken on dst and should
973 * be released by skb_dst_drop()
974 */
adf30907
ED
975static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
976{
7fee226a
ED
977 skb->_skb_refdst = (unsigned long)dst;
978}
979
932bc4d7
JA
980/**
981 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
982 * @skb: buffer
983 * @dst: dst entry
984 *
985 * Sets skb dst, assuming a reference was not taken on dst.
986 * If dst entry is cached, we do not take reference and dst_release
987 * will be avoided by refdst_drop. If dst entry is not cached, we take
988 * reference, so that last dst_release can destroy the dst immediately.
989 */
990static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
991{
dbfc4fb7
HFS
992 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
993 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 994}
7fee226a
ED
995
996/**
25985edc 997 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
998 * @skb: buffer
999 */
1000static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1001{
1002 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
1003}
1004
161e6137
PT
1005/**
1006 * skb_rtable - Returns the skb &rtable
1007 * @skb: buffer
1008 */
511c3f92
ED
1009static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1010{
adf30907 1011 return (struct rtable *)skb_dst(skb);
511c3f92
ED
1012}
1013
8b10cab6
JHS
1014/* For mangling skb->pkt_type from user space side from applications
1015 * such as nft, tc, etc, we only allow a conservative subset of
1016 * possible pkt_types to be set.
1017*/
1018static inline bool skb_pkt_type_ok(u32 ptype)
1019{
1020 return ptype <= PACKET_OTHERHOST;
1021}
1022
161e6137
PT
1023/**
1024 * skb_napi_id - Returns the skb's NAPI id
1025 * @skb: buffer
1026 */
90b602f8
ML
1027static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1028{
1029#ifdef CONFIG_NET_RX_BUSY_POLL
1030 return skb->napi_id;
1031#else
1032 return 0;
1033#endif
1034}
1035
161e6137
PT
1036/**
1037 * skb_unref - decrement the skb's reference count
1038 * @skb: buffer
1039 *
1040 * Returns true if we can free the skb.
1041 */
3889a803
PA
1042static inline bool skb_unref(struct sk_buff *skb)
1043{
1044 if (unlikely(!skb))
1045 return false;
63354797 1046 if (likely(refcount_read(&skb->users) == 1))
3889a803 1047 smp_rmb();
63354797 1048 else if (likely(!refcount_dec_and_test(&skb->users)))
3889a803
PA
1049 return false;
1050
1051 return true;
1052}
1053
0a463c78 1054void skb_release_head_state(struct sk_buff *skb);
7965bd4d
JP
1055void kfree_skb(struct sk_buff *skb);
1056void kfree_skb_list(struct sk_buff *segs);
6413139d 1057void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
7965bd4d
JP
1058void skb_tx_error(struct sk_buff *skb);
1059void consume_skb(struct sk_buff *skb);
ca2c1418 1060void __consume_stateless_skb(struct sk_buff *skb);
7965bd4d 1061void __kfree_skb(struct sk_buff *skb);
d7e8883c 1062extern struct kmem_cache *skbuff_head_cache;
bad43ca8 1063
7965bd4d
JP
1064void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1065bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1066 bool *fragstolen, int *delta_truesize);
bad43ca8 1067
7965bd4d
JP
1068struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1069 int node);
2ea2f62c 1070struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 1071struct sk_buff *build_skb(void *data, unsigned int frag_size);
ba0509b6
JDB
1072struct sk_buff *build_skb_around(struct sk_buff *skb,
1073 void *data, unsigned int frag_size);
161e6137
PT
1074
1075/**
1076 * alloc_skb - allocate a network buffer
1077 * @size: size to allocate
1078 * @priority: allocation mask
1079 *
1080 * This function is a convenient wrapper around __alloc_skb().
1081 */
d179cd12 1082static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 1083 gfp_t priority)
d179cd12 1084{
564824b0 1085 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
1086}
1087
2e4e4410
ED
1088struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1089 unsigned long data_len,
1090 int max_page_order,
1091 int *errcode,
1092 gfp_t gfp_mask);
da29e4b4 1093struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
2e4e4410 1094
d0bf4a9e
ED
1095/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1096struct sk_buff_fclones {
1097 struct sk_buff skb1;
1098
1099 struct sk_buff skb2;
1100
2638595a 1101 refcount_t fclone_ref;
d0bf4a9e
ED
1102};
1103
1104/**
1105 * skb_fclone_busy - check if fclone is busy
293de7de 1106 * @sk: socket
d0bf4a9e
ED
1107 * @skb: buffer
1108 *
bda13fed 1109 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
1110 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1111 * so we also check that this didnt happen.
d0bf4a9e 1112 */
39bb5e62
ED
1113static inline bool skb_fclone_busy(const struct sock *sk,
1114 const struct sk_buff *skb)
d0bf4a9e
ED
1115{
1116 const struct sk_buff_fclones *fclones;
1117
1118 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1119
1120 return skb->fclone == SKB_FCLONE_ORIG &&
2638595a 1121 refcount_read(&fclones->fclone_ref) > 1 &&
39bb5e62 1122 fclones->skb2.sk == sk;
d0bf4a9e
ED
1123}
1124
161e6137
PT
1125/**
1126 * alloc_skb_fclone - allocate a network buffer from fclone cache
1127 * @size: size to allocate
1128 * @priority: allocation mask
1129 *
1130 * This function is a convenient wrapper around __alloc_skb().
1131 */
d179cd12 1132static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 1133 gfp_t priority)
d179cd12 1134{
c93bdd0e 1135 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
1136}
1137
7965bd4d 1138struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
b0768a86 1139void skb_headers_offset_update(struct sk_buff *skb, int off);
7965bd4d
JP
1140int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1141struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
08303c18 1142void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
7965bd4d 1143struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
1144struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1145 gfp_t gfp_mask, bool fclone);
1146static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1147 gfp_t gfp_mask)
1148{
1149 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1150}
7965bd4d
JP
1151
1152int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1153struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1154 unsigned int headroom);
1155struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1156 int newtailroom, gfp_t priority);
48a1df65
JD
1157int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1158 int offset, int len);
1159int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1160 int offset, int len);
7965bd4d 1161int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
cd0a137a
FF
1162int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1163
1164/**
1165 * skb_pad - zero pad the tail of an skb
1166 * @skb: buffer to pad
1167 * @pad: space to pad
1168 *
1169 * Ensure that a buffer is followed by a padding area that is zero
1170 * filled. Used by network drivers which may DMA or transfer data
1171 * beyond the buffer end onto the wire.
1172 *
1173 * May return error in out of memory cases. The skb is freed on error.
1174 */
1175static inline int skb_pad(struct sk_buff *skb, int pad)
1176{
1177 return __skb_pad(skb, pad, true);
1178}
ead2ceb0 1179#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1180
be12a1fe
HFS
1181int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1182 int offset, size_t size);
1183
d94d9fee 1184struct skb_seq_state {
677e90ed
TG
1185 __u32 lower_offset;
1186 __u32 upper_offset;
1187 __u32 frag_idx;
1188 __u32 stepped_offset;
1189 struct sk_buff *root_skb;
1190 struct sk_buff *cur_skb;
1191 __u8 *frag_data;
1192};
1193
7965bd4d
JP
1194void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1195 unsigned int to, struct skb_seq_state *st);
1196unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1197 struct skb_seq_state *st);
1198void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1199
7965bd4d 1200unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1201 unsigned int to, struct ts_config *config);
3fc7e8a6 1202
09323cc4
TH
1203/*
1204 * Packet hash types specify the type of hash in skb_set_hash.
1205 *
1206 * Hash types refer to the protocol layer addresses which are used to
1207 * construct a packet's hash. The hashes are used to differentiate or identify
1208 * flows of the protocol layer for the hash type. Hash types are either
1209 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1210 *
1211 * Properties of hashes:
1212 *
1213 * 1) Two packets in different flows have different hash values
1214 * 2) Two packets in the same flow should have the same hash value
1215 *
1216 * A hash at a higher layer is considered to be more specific. A driver should
1217 * set the most specific hash possible.
1218 *
1219 * A driver cannot indicate a more specific hash than the layer at which a hash
1220 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1221 *
1222 * A driver may indicate a hash level which is less specific than the
1223 * actual layer the hash was computed on. For instance, a hash computed
1224 * at L4 may be considered an L3 hash. This should only be done if the
1225 * driver can't unambiguously determine that the HW computed the hash at
1226 * the higher layer. Note that the "should" in the second property above
1227 * permits this.
1228 */
1229enum pkt_hash_types {
1230 PKT_HASH_TYPE_NONE, /* Undefined type */
1231 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1232 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1233 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1234};
1235
bcc83839 1236static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1237{
bcc83839 1238 skb->hash = 0;
a3b18ddb 1239 skb->sw_hash = 0;
bcc83839
TH
1240 skb->l4_hash = 0;
1241}
1242
1243static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1244{
1245 if (!skb->l4_hash)
1246 skb_clear_hash(skb);
1247}
1248
1249static inline void
1250__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1251{
1252 skb->l4_hash = is_l4;
1253 skb->sw_hash = is_sw;
61b905da 1254 skb->hash = hash;
09323cc4
TH
1255}
1256
bcc83839
TH
1257static inline void
1258skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1259{
1260 /* Used by drivers to set hash from HW */
1261 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1262}
1263
1264static inline void
1265__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1266{
1267 __skb_set_hash(skb, hash, true, is_l4);
1268}
1269
e5276937 1270void __skb_get_hash(struct sk_buff *skb);
b917783c 1271u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937
TH
1272u32 skb_get_poff(const struct sk_buff *skb);
1273u32 __skb_get_poff(const struct sk_buff *skb, void *data,
72a338bc 1274 const struct flow_keys_basic *keys, int hlen);
e5276937
TH
1275__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1276 void *data, int hlen_proto);
1277
1278static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1279 int thoff, u8 ip_proto)
1280{
1281 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1282}
1283
1284void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1285 const struct flow_dissector_key *key,
1286 unsigned int key_count);
1287
089b19a9
SF
1288struct bpf_flow_dissector;
1289bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
086f9568 1290 __be16 proto, int nhoff, int hlen, unsigned int flags);
089b19a9 1291
3cbf4ffb
SF
1292bool __skb_flow_dissect(const struct net *net,
1293 const struct sk_buff *skb,
e5276937
TH
1294 struct flow_dissector *flow_dissector,
1295 void *target_container,
cd79a238
TH
1296 void *data, __be16 proto, int nhoff, int hlen,
1297 unsigned int flags);
e5276937
TH
1298
1299static inline bool skb_flow_dissect(const struct sk_buff *skb,
1300 struct flow_dissector *flow_dissector,
cd79a238 1301 void *target_container, unsigned int flags)
e5276937 1302{
3cbf4ffb
SF
1303 return __skb_flow_dissect(NULL, skb, flow_dissector,
1304 target_container, NULL, 0, 0, 0, flags);
e5276937
TH
1305}
1306
1307static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1308 struct flow_keys *flow,
1309 unsigned int flags)
e5276937
TH
1310{
1311 memset(flow, 0, sizeof(*flow));
3cbf4ffb
SF
1312 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1313 flow, NULL, 0, 0, 0, flags);
e5276937
TH
1314}
1315
72a338bc 1316static inline bool
3cbf4ffb
SF
1317skb_flow_dissect_flow_keys_basic(const struct net *net,
1318 const struct sk_buff *skb,
72a338bc
PA
1319 struct flow_keys_basic *flow, void *data,
1320 __be16 proto, int nhoff, int hlen,
1321 unsigned int flags)
e5276937
TH
1322{
1323 memset(flow, 0, sizeof(*flow));
3cbf4ffb 1324 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
cd79a238 1325 data, proto, nhoff, hlen, flags);
e5276937
TH
1326}
1327
82828b88
JP
1328void skb_flow_dissect_meta(const struct sk_buff *skb,
1329 struct flow_dissector *flow_dissector,
1330 void *target_container);
1331
75a56758 1332/* Gets a skb connection tracking info, ctinfo map should be a
2ff17117 1333 * map of mapsize to translate enum ip_conntrack_info states
75a56758
PB
1334 * to user states.
1335 */
1336void
1337skb_flow_dissect_ct(const struct sk_buff *skb,
1338 struct flow_dissector *flow_dissector,
1339 void *target_container,
1340 u16 *ctinfo_map,
1341 size_t mapsize);
62b32379
SH
1342void
1343skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1344 struct flow_dissector *flow_dissector,
1345 void *target_container);
1346
0cb09aff
AL
1347void skb_flow_dissect_hash(const struct sk_buff *skb,
1348 struct flow_dissector *flow_dissector,
1349 void *target_container);
1350
3958afa1 1351static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1352{
a3b18ddb 1353 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1354 __skb_get_hash(skb);
bfb564e7 1355
61b905da 1356 return skb->hash;
bfb564e7
KK
1357}
1358
20a17bf6 1359static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1360{
c6cc1ca7
TH
1361 if (!skb->l4_hash && !skb->sw_hash) {
1362 struct flow_keys keys;
de4c1f8b 1363 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1364
de4c1f8b 1365 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1366 }
f70ea018
TH
1367
1368 return skb->hash;
1369}
1370
55667441
ED
1371__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1372 const siphash_key_t *perturb);
50fb7992 1373
57bdf7f4
TH
1374static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1375{
61b905da 1376 return skb->hash;
57bdf7f4
TH
1377}
1378
3df7a74e
TH
1379static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1380{
61b905da 1381 to->hash = from->hash;
a3b18ddb 1382 to->sw_hash = from->sw_hash;
61b905da 1383 to->l4_hash = from->l4_hash;
3df7a74e
TH
1384};
1385
41477662
JK
1386static inline void skb_copy_decrypted(struct sk_buff *to,
1387 const struct sk_buff *from)
1388{
1389#ifdef CONFIG_TLS_DEVICE
1390 to->decrypted = from->decrypted;
1391#endif
1392}
1393
4305b541
ACM
1394#ifdef NET_SKBUFF_DATA_USES_OFFSET
1395static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1396{
1397 return skb->head + skb->end;
1398}
ec47ea82
AD
1399
1400static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1401{
1402 return skb->end;
1403}
4305b541
ACM
1404#else
1405static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1406{
1407 return skb->end;
1408}
ec47ea82
AD
1409
1410static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1411{
1412 return skb->end - skb->head;
1413}
4305b541
ACM
1414#endif
1415
1da177e4 1416/* Internal */
4305b541 1417#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1418
ac45f602
PO
1419static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1420{
1421 return &skb_shinfo(skb)->hwtstamps;
1422}
1423
52267790
WB
1424static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1425{
1426 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1427
1428 return is_zcopy ? skb_uarg(skb) : NULL;
1429}
1430
52900d22
WB
1431static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1432 bool *have_ref)
52267790
WB
1433{
1434 if (skb && uarg && !skb_zcopy(skb)) {
52900d22
WB
1435 if (unlikely(have_ref && *have_ref))
1436 *have_ref = false;
1437 else
1438 sock_zerocopy_get(uarg);
52267790
WB
1439 skb_shinfo(skb)->destructor_arg = uarg;
1440 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1441 }
1442}
1443
5cd8d46e
WB
1444static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1445{
1446 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1447 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1448}
1449
1450static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1451{
1452 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1453}
1454
1455static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1456{
1457 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1458}
1459
52267790
WB
1460/* Release a reference on a zerocopy structure */
1461static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1462{
1463 struct ubuf_info *uarg = skb_zcopy(skb);
1464
1465 if (uarg) {
185ce5c3
WB
1466 if (skb_zcopy_is_nouarg(skb)) {
1467 /* no notification callback */
1468 } else if (uarg->callback == sock_zerocopy_callback) {
0a4a060b
WB
1469 uarg->zerocopy = uarg->zerocopy && zerocopy;
1470 sock_zerocopy_put(uarg);
185ce5c3 1471 } else {
0a4a060b
WB
1472 uarg->callback(uarg, zerocopy);
1473 }
1474
52267790
WB
1475 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1476 }
1477}
1478
1479/* Abort a zerocopy operation and revert zckey on error in send syscall */
1480static inline void skb_zcopy_abort(struct sk_buff *skb)
1481{
1482 struct ubuf_info *uarg = skb_zcopy(skb);
1483
1484 if (uarg) {
52900d22 1485 sock_zerocopy_put_abort(uarg, false);
52267790
WB
1486 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1487 }
1488}
1489
a8305bff
DM
1490static inline void skb_mark_not_on_list(struct sk_buff *skb)
1491{
1492 skb->next = NULL;
1493}
1494
dcfea72e 1495/* Iterate through singly-linked GSO fragments of an skb. */
5eee7bd7
JD
1496#define skb_list_walk_safe(first, skb, next_skb) \
1497 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1498 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
dcfea72e 1499
992cba7e
DM
1500static inline void skb_list_del_init(struct sk_buff *skb)
1501{
1502 __list_del_entry(&skb->list);
1503 skb_mark_not_on_list(skb);
1504}
1505
1da177e4
LT
1506/**
1507 * skb_queue_empty - check if a queue is empty
1508 * @list: queue head
1509 *
1510 * Returns true if the queue is empty, false otherwise.
1511 */
1512static inline int skb_queue_empty(const struct sk_buff_head *list)
1513{
fd44b93c 1514 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1515}
1516
d7d16a89
ED
1517/**
1518 * skb_queue_empty_lockless - check if a queue is empty
1519 * @list: queue head
1520 *
1521 * Returns true if the queue is empty, false otherwise.
1522 * This variant can be used in lockless contexts.
1523 */
1524static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1525{
1526 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1527}
1528
1529
fc7ebb21
DM
1530/**
1531 * skb_queue_is_last - check if skb is the last entry in the queue
1532 * @list: queue head
1533 * @skb: buffer
1534 *
1535 * Returns true if @skb is the last buffer on the list.
1536 */
1537static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1538 const struct sk_buff *skb)
1539{
fd44b93c 1540 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1541}
1542
832d11c5
IJ
1543/**
1544 * skb_queue_is_first - check if skb is the first entry in the queue
1545 * @list: queue head
1546 * @skb: buffer
1547 *
1548 * Returns true if @skb is the first buffer on the list.
1549 */
1550static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1551 const struct sk_buff *skb)
1552{
fd44b93c 1553 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1554}
1555
249c8b42
DM
1556/**
1557 * skb_queue_next - return the next packet in the queue
1558 * @list: queue head
1559 * @skb: current buffer
1560 *
1561 * Return the next packet in @list after @skb. It is only valid to
1562 * call this if skb_queue_is_last() evaluates to false.
1563 */
1564static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1565 const struct sk_buff *skb)
1566{
1567 /* This BUG_ON may seem severe, but if we just return then we
1568 * are going to dereference garbage.
1569 */
1570 BUG_ON(skb_queue_is_last(list, skb));
1571 return skb->next;
1572}
1573
832d11c5
IJ
1574/**
1575 * skb_queue_prev - return the prev packet in the queue
1576 * @list: queue head
1577 * @skb: current buffer
1578 *
1579 * Return the prev packet in @list before @skb. It is only valid to
1580 * call this if skb_queue_is_first() evaluates to false.
1581 */
1582static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1583 const struct sk_buff *skb)
1584{
1585 /* This BUG_ON may seem severe, but if we just return then we
1586 * are going to dereference garbage.
1587 */
1588 BUG_ON(skb_queue_is_first(list, skb));
1589 return skb->prev;
1590}
1591
1da177e4
LT
1592/**
1593 * skb_get - reference buffer
1594 * @skb: buffer to reference
1595 *
1596 * Makes another reference to a socket buffer and returns a pointer
1597 * to the buffer.
1598 */
1599static inline struct sk_buff *skb_get(struct sk_buff *skb)
1600{
63354797 1601 refcount_inc(&skb->users);
1da177e4
LT
1602 return skb;
1603}
1604
1605/*
f8821f96 1606 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1da177e4
LT
1607 */
1608
1da177e4
LT
1609/**
1610 * skb_cloned - is the buffer a clone
1611 * @skb: buffer to check
1612 *
1613 * Returns true if the buffer was generated with skb_clone() and is
1614 * one of multiple shared copies of the buffer. Cloned buffers are
1615 * shared data so must not be written to under normal circumstances.
1616 */
1617static inline int skb_cloned(const struct sk_buff *skb)
1618{
1619 return skb->cloned &&
1620 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1621}
1622
14bbd6a5
PS
1623static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1624{
d0164adc 1625 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1626
1627 if (skb_cloned(skb))
1628 return pskb_expand_head(skb, 0, 0, pri);
1629
1630 return 0;
1631}
1632
1da177e4
LT
1633/**
1634 * skb_header_cloned - is the header a clone
1635 * @skb: buffer to check
1636 *
1637 * Returns true if modifying the header part of the buffer requires
1638 * the data to be copied.
1639 */
1640static inline int skb_header_cloned(const struct sk_buff *skb)
1641{
1642 int dataref;
1643
1644 if (!skb->cloned)
1645 return 0;
1646
1647 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1648 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1649 return dataref != 1;
1650}
1651
9580bf2e
ED
1652static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1653{
1654 might_sleep_if(gfpflags_allow_blocking(pri));
1655
1656 if (skb_header_cloned(skb))
1657 return pskb_expand_head(skb, 0, 0, pri);
1658
1659 return 0;
1660}
1661
f4a775d1
ED
1662/**
1663 * __skb_header_release - release reference to header
1664 * @skb: buffer to operate on
f4a775d1
ED
1665 */
1666static inline void __skb_header_release(struct sk_buff *skb)
1667{
1668 skb->nohdr = 1;
1669 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1670}
1671
1672
1da177e4
LT
1673/**
1674 * skb_shared - is the buffer shared
1675 * @skb: buffer to check
1676 *
1677 * Returns true if more than one person has a reference to this
1678 * buffer.
1679 */
1680static inline int skb_shared(const struct sk_buff *skb)
1681{
63354797 1682 return refcount_read(&skb->users) != 1;
1da177e4
LT
1683}
1684
1685/**
1686 * skb_share_check - check if buffer is shared and if so clone it
1687 * @skb: buffer to check
1688 * @pri: priority for memory allocation
1689 *
1690 * If the buffer is shared the buffer is cloned and the old copy
1691 * drops a reference. A new clone with a single reference is returned.
1692 * If the buffer is not shared the original buffer is returned. When
1693 * being called from interrupt status or with spinlocks held pri must
1694 * be GFP_ATOMIC.
1695 *
1696 * NULL is returned on a memory allocation failure.
1697 */
47061bc4 1698static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1699{
d0164adc 1700 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1701 if (skb_shared(skb)) {
1702 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1703
1704 if (likely(nskb))
1705 consume_skb(skb);
1706 else
1707 kfree_skb(skb);
1da177e4
LT
1708 skb = nskb;
1709 }
1710 return skb;
1711}
1712
1713/*
1714 * Copy shared buffers into a new sk_buff. We effectively do COW on
1715 * packets to handle cases where we have a local reader and forward
1716 * and a couple of other messy ones. The normal one is tcpdumping
1717 * a packet thats being forwarded.
1718 */
1719
1720/**
1721 * skb_unshare - make a copy of a shared buffer
1722 * @skb: buffer to check
1723 * @pri: priority for memory allocation
1724 *
1725 * If the socket buffer is a clone then this function creates a new
1726 * copy of the data, drops a reference count on the old copy and returns
1727 * the new copy with the reference count at 1. If the buffer is not a clone
1728 * the original buffer is returned. When called with a spinlock held or
1729 * from interrupt state @pri must be %GFP_ATOMIC
1730 *
1731 * %NULL is returned on a memory allocation failure.
1732 */
e2bf521d 1733static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1734 gfp_t pri)
1da177e4 1735{
d0164adc 1736 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1737 if (skb_cloned(skb)) {
1738 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1739
1740 /* Free our shared copy */
1741 if (likely(nskb))
1742 consume_skb(skb);
1743 else
1744 kfree_skb(skb);
1da177e4
LT
1745 skb = nskb;
1746 }
1747 return skb;
1748}
1749
1750/**
1a5778aa 1751 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1752 * @list_: list to peek at
1753 *
1754 * Peek an &sk_buff. Unlike most other operations you _MUST_
1755 * be careful with this one. A peek leaves the buffer on the
1756 * list and someone else may run off with it. You must hold
1757 * the appropriate locks or have a private queue to do this.
1758 *
1759 * Returns %NULL for an empty list or a pointer to the head element.
1760 * The reference count is not incremented and the reference is therefore
1761 * volatile. Use with caution.
1762 */
05bdd2f1 1763static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1764{
18d07000
ED
1765 struct sk_buff *skb = list_->next;
1766
1767 if (skb == (struct sk_buff *)list_)
1768 skb = NULL;
1769 return skb;
1da177e4
LT
1770}
1771
8b69bd7d
DM
1772/**
1773 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1774 * @list_: list to peek at
1775 *
1776 * Like skb_peek(), but the caller knows that the list is not empty.
1777 */
1778static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1779{
1780 return list_->next;
1781}
1782
da5ef6e5
PE
1783/**
1784 * skb_peek_next - peek skb following the given one from a queue
1785 * @skb: skb to start from
1786 * @list_: list to peek at
1787 *
1788 * Returns %NULL when the end of the list is met or a pointer to the
1789 * next element. The reference count is not incremented and the
1790 * reference is therefore volatile. Use with caution.
1791 */
1792static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1793 const struct sk_buff_head *list_)
1794{
1795 struct sk_buff *next = skb->next;
18d07000 1796
da5ef6e5
PE
1797 if (next == (struct sk_buff *)list_)
1798 next = NULL;
1799 return next;
1800}
1801
1da177e4 1802/**
1a5778aa 1803 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1804 * @list_: list to peek at
1805 *
1806 * Peek an &sk_buff. Unlike most other operations you _MUST_
1807 * be careful with this one. A peek leaves the buffer on the
1808 * list and someone else may run off with it. You must hold
1809 * the appropriate locks or have a private queue to do this.
1810 *
1811 * Returns %NULL for an empty list or a pointer to the tail element.
1812 * The reference count is not incremented and the reference is therefore
1813 * volatile. Use with caution.
1814 */
05bdd2f1 1815static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1816{
f8cc62ca 1817 struct sk_buff *skb = READ_ONCE(list_->prev);
18d07000
ED
1818
1819 if (skb == (struct sk_buff *)list_)
1820 skb = NULL;
1821 return skb;
1822
1da177e4
LT
1823}
1824
1825/**
1826 * skb_queue_len - get queue length
1827 * @list_: list to measure
1828 *
1829 * Return the length of an &sk_buff queue.
1830 */
1831static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1832{
1833 return list_->qlen;
1834}
1835
86b18aaa
QC
1836/**
1837 * skb_queue_len_lockless - get queue length
1838 * @list_: list to measure
1839 *
1840 * Return the length of an &sk_buff queue.
1841 * This variant can be used in lockless contexts.
1842 */
1843static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1844{
1845 return READ_ONCE(list_->qlen);
1846}
1847
67fed459
DM
1848/**
1849 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1850 * @list: queue to initialize
1851 *
1852 * This initializes only the list and queue length aspects of
1853 * an sk_buff_head object. This allows to initialize the list
1854 * aspects of an sk_buff_head without reinitializing things like
1855 * the spinlock. It can also be used for on-stack sk_buff_head
1856 * objects where the spinlock is known to not be used.
1857 */
1858static inline void __skb_queue_head_init(struct sk_buff_head *list)
1859{
1860 list->prev = list->next = (struct sk_buff *)list;
1861 list->qlen = 0;
1862}
1863
76f10ad0
AV
1864/*
1865 * This function creates a split out lock class for each invocation;
1866 * this is needed for now since a whole lot of users of the skb-queue
1867 * infrastructure in drivers have different locking usage (in hardirq)
1868 * than the networking core (in softirq only). In the long run either the
1869 * network layer or drivers should need annotation to consolidate the
1870 * main types of usage into 3 classes.
1871 */
1da177e4
LT
1872static inline void skb_queue_head_init(struct sk_buff_head *list)
1873{
1874 spin_lock_init(&list->lock);
67fed459 1875 __skb_queue_head_init(list);
1da177e4
LT
1876}
1877
c2ecba71
PE
1878static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1879 struct lock_class_key *class)
1880{
1881 skb_queue_head_init(list);
1882 lockdep_set_class(&list->lock, class);
1883}
1884
1da177e4 1885/*
bf299275 1886 * Insert an sk_buff on a list.
1da177e4
LT
1887 *
1888 * The "__skb_xxxx()" functions are the non-atomic ones that
1889 * can only be called with interrupts disabled.
1890 */
bf299275
GR
1891static inline void __skb_insert(struct sk_buff *newsk,
1892 struct sk_buff *prev, struct sk_buff *next,
1893 struct sk_buff_head *list)
1894{
f8cc62ca
ED
1895 /* See skb_queue_empty_lockless() and skb_peek_tail()
1896 * for the opposite READ_ONCE()
1897 */
d7d16a89
ED
1898 WRITE_ONCE(newsk->next, next);
1899 WRITE_ONCE(newsk->prev, prev);
1900 WRITE_ONCE(next->prev, newsk);
1901 WRITE_ONCE(prev->next, newsk);
bf299275
GR
1902 list->qlen++;
1903}
1da177e4 1904
67fed459
DM
1905static inline void __skb_queue_splice(const struct sk_buff_head *list,
1906 struct sk_buff *prev,
1907 struct sk_buff *next)
1908{
1909 struct sk_buff *first = list->next;
1910 struct sk_buff *last = list->prev;
1911
d7d16a89
ED
1912 WRITE_ONCE(first->prev, prev);
1913 WRITE_ONCE(prev->next, first);
67fed459 1914
d7d16a89
ED
1915 WRITE_ONCE(last->next, next);
1916 WRITE_ONCE(next->prev, last);
67fed459
DM
1917}
1918
1919/**
1920 * skb_queue_splice - join two skb lists, this is designed for stacks
1921 * @list: the new list to add
1922 * @head: the place to add it in the first list
1923 */
1924static inline void skb_queue_splice(const struct sk_buff_head *list,
1925 struct sk_buff_head *head)
1926{
1927 if (!skb_queue_empty(list)) {
1928 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1929 head->qlen += list->qlen;
67fed459
DM
1930 }
1931}
1932
1933/**
d9619496 1934 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1935 * @list: the new list to add
1936 * @head: the place to add it in the first list
1937 *
1938 * The list at @list is reinitialised
1939 */
1940static inline void skb_queue_splice_init(struct sk_buff_head *list,
1941 struct sk_buff_head *head)
1942{
1943 if (!skb_queue_empty(list)) {
1944 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1945 head->qlen += list->qlen;
67fed459
DM
1946 __skb_queue_head_init(list);
1947 }
1948}
1949
1950/**
1951 * skb_queue_splice_tail - join two skb lists, each list being a queue
1952 * @list: the new list to add
1953 * @head: the place to add it in the first list
1954 */
1955static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1956 struct sk_buff_head *head)
1957{
1958 if (!skb_queue_empty(list)) {
1959 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1960 head->qlen += list->qlen;
67fed459
DM
1961 }
1962}
1963
1964/**
d9619496 1965 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1966 * @list: the new list to add
1967 * @head: the place to add it in the first list
1968 *
1969 * Each of the lists is a queue.
1970 * The list at @list is reinitialised
1971 */
1972static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1973 struct sk_buff_head *head)
1974{
1975 if (!skb_queue_empty(list)) {
1976 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1977 head->qlen += list->qlen;
67fed459
DM
1978 __skb_queue_head_init(list);
1979 }
1980}
1981
1da177e4 1982/**
300ce174 1983 * __skb_queue_after - queue a buffer at the list head
1da177e4 1984 * @list: list to use
300ce174 1985 * @prev: place after this buffer
1da177e4
LT
1986 * @newsk: buffer to queue
1987 *
300ce174 1988 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1989 * and you must therefore hold required locks before calling it.
1990 *
1991 * A buffer cannot be placed on two lists at the same time.
1992 */
300ce174
SH
1993static inline void __skb_queue_after(struct sk_buff_head *list,
1994 struct sk_buff *prev,
1995 struct sk_buff *newsk)
1da177e4 1996{
bf299275 1997 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1998}
1999
7965bd4d
JP
2000void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2001 struct sk_buff_head *list);
7de6c033 2002
f5572855
GR
2003static inline void __skb_queue_before(struct sk_buff_head *list,
2004 struct sk_buff *next,
2005 struct sk_buff *newsk)
2006{
2007 __skb_insert(newsk, next->prev, next, list);
2008}
2009
300ce174
SH
2010/**
2011 * __skb_queue_head - queue a buffer at the list head
2012 * @list: list to use
2013 * @newsk: buffer to queue
2014 *
2015 * Queue a buffer at the start of a list. This function takes no locks
2016 * and you must therefore hold required locks before calling it.
2017 *
2018 * A buffer cannot be placed on two lists at the same time.
2019 */
300ce174
SH
2020static inline void __skb_queue_head(struct sk_buff_head *list,
2021 struct sk_buff *newsk)
2022{
2023 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2024}
4ea7b0cf 2025void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174 2026
1da177e4
LT
2027/**
2028 * __skb_queue_tail - queue a buffer at the list tail
2029 * @list: list to use
2030 * @newsk: buffer to queue
2031 *
2032 * Queue a buffer at the end of a list. This function takes no locks
2033 * and you must therefore hold required locks before calling it.
2034 *
2035 * A buffer cannot be placed on two lists at the same time.
2036 */
1da177e4
LT
2037static inline void __skb_queue_tail(struct sk_buff_head *list,
2038 struct sk_buff *newsk)
2039{
f5572855 2040 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4 2041}
4ea7b0cf 2042void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4 2043
1da177e4
LT
2044/*
2045 * remove sk_buff from list. _Must_ be called atomically, and with
2046 * the list known..
2047 */
7965bd4d 2048void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
2049static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2050{
2051 struct sk_buff *next, *prev;
2052
86b18aaa 2053 WRITE_ONCE(list->qlen, list->qlen - 1);
1da177e4
LT
2054 next = skb->next;
2055 prev = skb->prev;
2056 skb->next = skb->prev = NULL;
d7d16a89
ED
2057 WRITE_ONCE(next->prev, prev);
2058 WRITE_ONCE(prev->next, next);
1da177e4
LT
2059}
2060
f525c06d
GR
2061/**
2062 * __skb_dequeue - remove from the head of the queue
2063 * @list: list to dequeue from
2064 *
2065 * Remove the head of the list. This function does not take any locks
2066 * so must be used with appropriate locks held only. The head item is
2067 * returned or %NULL if the list is empty.
2068 */
f525c06d
GR
2069static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2070{
2071 struct sk_buff *skb = skb_peek(list);
2072 if (skb)
2073 __skb_unlink(skb, list);
2074 return skb;
2075}
4ea7b0cf 2076struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1da177e4
LT
2077
2078/**
2079 * __skb_dequeue_tail - remove from the tail of the queue
2080 * @list: list to dequeue from
2081 *
2082 * Remove the tail of the list. This function does not take any locks
2083 * so must be used with appropriate locks held only. The tail item is
2084 * returned or %NULL if the list is empty.
2085 */
1da177e4
LT
2086static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2087{
2088 struct sk_buff *skb = skb_peek_tail(list);
2089 if (skb)
2090 __skb_unlink(skb, list);
2091 return skb;
2092}
4ea7b0cf 2093struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
2094
2095
bdcc0924 2096static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
2097{
2098 return skb->data_len;
2099}
2100
2101static inline unsigned int skb_headlen(const struct sk_buff *skb)
2102{
2103 return skb->len - skb->data_len;
2104}
2105
3ece7826 2106static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1da177e4 2107{
c72d8cda 2108 unsigned int i, len = 0;
1da177e4 2109
c72d8cda 2110 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 2111 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
3ece7826
WB
2112 return len;
2113}
2114
2115static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2116{
2117 return skb_headlen(skb) + __skb_pagelen(skb);
1da177e4
LT
2118}
2119
131ea667
IC
2120/**
2121 * __skb_fill_page_desc - initialise a paged fragment in an skb
2122 * @skb: buffer containing fragment to be initialised
2123 * @i: paged fragment index to initialise
2124 * @page: the page to use for this fragment
2125 * @off: the offset to the data with @page
2126 * @size: the length of the data
2127 *
2128 * Initialises the @i'th fragment of @skb to point to &size bytes at
2129 * offset @off within @page.
2130 *
2131 * Does not take any additional reference on the fragment.
2132 */
2133static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2134 struct page *page, int off, int size)
1da177e4
LT
2135{
2136 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2137
c48a11c7 2138 /*
2f064f34
MH
2139 * Propagate page pfmemalloc to the skb if we can. The problem is
2140 * that not all callers have unique ownership of the page but rely
2141 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 2142 */
1dfa5bd3 2143 frag->bv_page = page;
65c84f14 2144 frag->bv_offset = off;
9e903e08 2145 skb_frag_size_set(frag, size);
cca7af38
PE
2146
2147 page = compound_head(page);
2f064f34 2148 if (page_is_pfmemalloc(page))
cca7af38 2149 skb->pfmemalloc = true;
131ea667
IC
2150}
2151
2152/**
2153 * skb_fill_page_desc - initialise a paged fragment in an skb
2154 * @skb: buffer containing fragment to be initialised
2155 * @i: paged fragment index to initialise
2156 * @page: the page to use for this fragment
2157 * @off: the offset to the data with @page
2158 * @size: the length of the data
2159 *
2160 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 2161 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
2162 * addition updates @skb such that @i is the last fragment.
2163 *
2164 * Does not take any additional reference on the fragment.
2165 */
2166static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2167 struct page *page, int off, int size)
2168{
2169 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
2170 skb_shinfo(skb)->nr_frags = i + 1;
2171}
2172
7965bd4d
JP
2173void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2174 int size, unsigned int truesize);
654bed16 2175
f8e617e1
JW
2176void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2177 unsigned int truesize);
2178
1da177e4
LT
2179#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2180
27a884dc
ACM
2181#ifdef NET_SKBUFF_DATA_USES_OFFSET
2182static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2183{
2184 return skb->head + skb->tail;
2185}
2186
2187static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2188{
2189 skb->tail = skb->data - skb->head;
2190}
2191
2192static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2193{
2194 skb_reset_tail_pointer(skb);
2195 skb->tail += offset;
2196}
7cc46190 2197
27a884dc
ACM
2198#else /* NET_SKBUFF_DATA_USES_OFFSET */
2199static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2200{
2201 return skb->tail;
2202}
2203
2204static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2205{
2206 skb->tail = skb->data;
2207}
2208
2209static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2210{
2211 skb->tail = skb->data + offset;
2212}
4305b541 2213
27a884dc
ACM
2214#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2215
1da177e4
LT
2216/*
2217 * Add data to an sk_buff
2218 */
4df864c1
JB
2219void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2220void *skb_put(struct sk_buff *skb, unsigned int len);
2221static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
1da177e4 2222{
4df864c1 2223 void *tmp = skb_tail_pointer(skb);
1da177e4
LT
2224 SKB_LINEAR_ASSERT(skb);
2225 skb->tail += len;
2226 skb->len += len;
2227 return tmp;
2228}
2229
de77b966 2230static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2231{
2232 void *tmp = __skb_put(skb, len);
2233
2234 memset(tmp, 0, len);
2235 return tmp;
2236}
2237
2238static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2239 unsigned int len)
2240{
2241 void *tmp = __skb_put(skb, len);
2242
2243 memcpy(tmp, data, len);
2244 return tmp;
2245}
2246
2247static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2248{
2249 *(u8 *)__skb_put(skb, 1) = val;
2250}
2251
83ad357d 2252static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
e45a79da 2253{
83ad357d 2254 void *tmp = skb_put(skb, len);
e45a79da
JB
2255
2256 memset(tmp, 0, len);
2257
2258 return tmp;
2259}
2260
59ae1d12
JB
2261static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2262 unsigned int len)
2263{
2264 void *tmp = skb_put(skb, len);
2265
2266 memcpy(tmp, data, len);
2267
2268 return tmp;
2269}
2270
634fef61
JB
2271static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2272{
2273 *(u8 *)skb_put(skb, 1) = val;
2274}
2275
d58ff351
JB
2276void *skb_push(struct sk_buff *skb, unsigned int len);
2277static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2278{
2279 skb->data -= len;
2280 skb->len += len;
2281 return skb->data;
2282}
2283
af72868b
JB
2284void *skb_pull(struct sk_buff *skb, unsigned int len);
2285static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2286{
2287 skb->len -= len;
2288 BUG_ON(skb->len < skb->data_len);
2289 return skb->data += len;
2290}
2291
af72868b 2292static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
47d29646
DM
2293{
2294 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2295}
2296
af72868b 2297void *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4 2298
af72868b 2299static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2300{
2301 if (len > skb_headlen(skb) &&
987c402a 2302 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
2303 return NULL;
2304 skb->len -= len;
2305 return skb->data += len;
2306}
2307
af72868b 2308static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2309{
2310 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2311}
2312
b9df4fd7 2313static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2314{
2315 if (likely(len <= skb_headlen(skb)))
b9df4fd7 2316 return true;
1da177e4 2317 if (unlikely(len > skb->len))
b9df4fd7 2318 return false;
987c402a 2319 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
2320}
2321
c8c8b127
ED
2322void skb_condense(struct sk_buff *skb);
2323
1da177e4
LT
2324/**
2325 * skb_headroom - bytes at buffer head
2326 * @skb: buffer to check
2327 *
2328 * Return the number of bytes of free space at the head of an &sk_buff.
2329 */
c2636b4d 2330static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
2331{
2332 return skb->data - skb->head;
2333}
2334
2335/**
2336 * skb_tailroom - bytes at buffer end
2337 * @skb: buffer to check
2338 *
2339 * Return the number of bytes of free space at the tail of an sk_buff
2340 */
2341static inline int skb_tailroom(const struct sk_buff *skb)
2342{
4305b541 2343 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2344}
2345
a21d4572
ED
2346/**
2347 * skb_availroom - bytes at buffer end
2348 * @skb: buffer to check
2349 *
2350 * Return the number of bytes of free space at the tail of an sk_buff
2351 * allocated by sk_stream_alloc()
2352 */
2353static inline int skb_availroom(const struct sk_buff *skb)
2354{
16fad69c
ED
2355 if (skb_is_nonlinear(skb))
2356 return 0;
2357
2358 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2359}
2360
1da177e4
LT
2361/**
2362 * skb_reserve - adjust headroom
2363 * @skb: buffer to alter
2364 * @len: bytes to move
2365 *
2366 * Increase the headroom of an empty &sk_buff by reducing the tail
2367 * room. This is only allowed for an empty buffer.
2368 */
8243126c 2369static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2370{
2371 skb->data += len;
2372 skb->tail += len;
2373}
2374
1837b2e2
BP
2375/**
2376 * skb_tailroom_reserve - adjust reserved_tailroom
2377 * @skb: buffer to alter
2378 * @mtu: maximum amount of headlen permitted
2379 * @needed_tailroom: minimum amount of reserved_tailroom
2380 *
2381 * Set reserved_tailroom so that headlen can be as large as possible but
2382 * not larger than mtu and tailroom cannot be smaller than
2383 * needed_tailroom.
2384 * The required headroom should already have been reserved before using
2385 * this function.
2386 */
2387static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2388 unsigned int needed_tailroom)
2389{
2390 SKB_LINEAR_ASSERT(skb);
2391 if (mtu < skb_tailroom(skb) - needed_tailroom)
2392 /* use at most mtu */
2393 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2394 else
2395 /* use up to all available space */
2396 skb->reserved_tailroom = needed_tailroom;
2397}
2398
8bce6d7d
TH
2399#define ENCAP_TYPE_ETHER 0
2400#define ENCAP_TYPE_IPPROTO 1
2401
2402static inline void skb_set_inner_protocol(struct sk_buff *skb,
2403 __be16 protocol)
2404{
2405 skb->inner_protocol = protocol;
2406 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2407}
2408
2409static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2410 __u8 ipproto)
2411{
2412 skb->inner_ipproto = ipproto;
2413 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2414}
2415
6a674e9c
JG
2416static inline void skb_reset_inner_headers(struct sk_buff *skb)
2417{
aefbd2b3 2418 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2419 skb->inner_network_header = skb->network_header;
2420 skb->inner_transport_header = skb->transport_header;
2421}
2422
0b5c9db1
JP
2423static inline void skb_reset_mac_len(struct sk_buff *skb)
2424{
2425 skb->mac_len = skb->network_header - skb->mac_header;
2426}
2427
6a674e9c
JG
2428static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2429 *skb)
2430{
2431 return skb->head + skb->inner_transport_header;
2432}
2433
55dc5a9f
TH
2434static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2435{
2436 return skb_inner_transport_header(skb) - skb->data;
2437}
2438
6a674e9c
JG
2439static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2440{
2441 skb->inner_transport_header = skb->data - skb->head;
2442}
2443
2444static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2445 const int offset)
2446{
2447 skb_reset_inner_transport_header(skb);
2448 skb->inner_transport_header += offset;
2449}
2450
2451static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2452{
2453 return skb->head + skb->inner_network_header;
2454}
2455
2456static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2457{
2458 skb->inner_network_header = skb->data - skb->head;
2459}
2460
2461static inline void skb_set_inner_network_header(struct sk_buff *skb,
2462 const int offset)
2463{
2464 skb_reset_inner_network_header(skb);
2465 skb->inner_network_header += offset;
2466}
2467
aefbd2b3
PS
2468static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2469{
2470 return skb->head + skb->inner_mac_header;
2471}
2472
2473static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2474{
2475 skb->inner_mac_header = skb->data - skb->head;
2476}
2477
2478static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2479 const int offset)
2480{
2481 skb_reset_inner_mac_header(skb);
2482 skb->inner_mac_header += offset;
2483}
fda55eca
ED
2484static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2485{
35d04610 2486 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2487}
2488
9c70220b
ACM
2489static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2490{
2e07fa9c 2491 return skb->head + skb->transport_header;
9c70220b
ACM
2492}
2493
badff6d0
ACM
2494static inline void skb_reset_transport_header(struct sk_buff *skb)
2495{
2e07fa9c 2496 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2497}
2498
967b05f6
ACM
2499static inline void skb_set_transport_header(struct sk_buff *skb,
2500 const int offset)
2501{
2e07fa9c
ACM
2502 skb_reset_transport_header(skb);
2503 skb->transport_header += offset;
ea2ae17d
ACM
2504}
2505
d56f90a7
ACM
2506static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2507{
2e07fa9c 2508 return skb->head + skb->network_header;
d56f90a7
ACM
2509}
2510
c1d2bbe1
ACM
2511static inline void skb_reset_network_header(struct sk_buff *skb)
2512{
2e07fa9c 2513 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2514}
2515
c14d2450
ACM
2516static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2517{
2e07fa9c
ACM
2518 skb_reset_network_header(skb);
2519 skb->network_header += offset;
c14d2450
ACM
2520}
2521
2e07fa9c 2522static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2523{
2e07fa9c 2524 return skb->head + skb->mac_header;
bbe735e4
ACM
2525}
2526
ea6da4fd
AV
2527static inline int skb_mac_offset(const struct sk_buff *skb)
2528{
2529 return skb_mac_header(skb) - skb->data;
2530}
2531
0daf4349
DB
2532static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2533{
2534 return skb->network_header - skb->mac_header;
2535}
2536
2e07fa9c 2537static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2538{
35d04610 2539 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2540}
2541
2542static inline void skb_reset_mac_header(struct sk_buff *skb)
2543{
2544 skb->mac_header = skb->data - skb->head;
2545}
2546
2547static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2548{
2549 skb_reset_mac_header(skb);
2550 skb->mac_header += offset;
2551}
2552
0e3da5bb
TT
2553static inline void skb_pop_mac_header(struct sk_buff *skb)
2554{
2555 skb->mac_header = skb->network_header;
2556}
2557
d2aa125d 2558static inline void skb_probe_transport_header(struct sk_buff *skb)
fbbdb8f0 2559{
72a338bc 2560 struct flow_keys_basic keys;
fbbdb8f0
YX
2561
2562 if (skb_transport_header_was_set(skb))
2563 return;
72a338bc 2564
3cbf4ffb
SF
2565 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2566 NULL, 0, 0, 0, 0))
42aecaa9 2567 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2568}
2569
03606895
ED
2570static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2571{
2572 if (skb_mac_header_was_set(skb)) {
2573 const unsigned char *old_mac = skb_mac_header(skb);
2574
2575 skb_set_mac_header(skb, -skb->mac_len);
2576 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2577 }
2578}
2579
04fb451e
MM
2580static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2581{
2582 return skb->csum_start - skb_headroom(skb);
2583}
2584
08b64fcc
AD
2585static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2586{
2587 return skb->head + skb->csum_start;
2588}
2589
2e07fa9c
ACM
2590static inline int skb_transport_offset(const struct sk_buff *skb)
2591{
2592 return skb_transport_header(skb) - skb->data;
2593}
2594
2595static inline u32 skb_network_header_len(const struct sk_buff *skb)
2596{
2597 return skb->transport_header - skb->network_header;
2598}
2599
6a674e9c
JG
2600static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2601{
2602 return skb->inner_transport_header - skb->inner_network_header;
2603}
2604
2e07fa9c
ACM
2605static inline int skb_network_offset(const struct sk_buff *skb)
2606{
2607 return skb_network_header(skb) - skb->data;
2608}
48d49d0c 2609
6a674e9c
JG
2610static inline int skb_inner_network_offset(const struct sk_buff *skb)
2611{
2612 return skb_inner_network_header(skb) - skb->data;
2613}
2614
f9599ce1
CG
2615static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2616{
2617 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2618}
2619
1da177e4
LT
2620/*
2621 * CPUs often take a performance hit when accessing unaligned memory
2622 * locations. The actual performance hit varies, it can be small if the
2623 * hardware handles it or large if we have to take an exception and fix it
2624 * in software.
2625 *
2626 * Since an ethernet header is 14 bytes network drivers often end up with
2627 * the IP header at an unaligned offset. The IP header can be aligned by
2628 * shifting the start of the packet by 2 bytes. Drivers should do this
2629 * with:
2630 *
8660c124 2631 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2632 *
2633 * The downside to this alignment of the IP header is that the DMA is now
2634 * unaligned. On some architectures the cost of an unaligned DMA is high
2635 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2636 *
1da177e4
LT
2637 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2638 * to be overridden.
2639 */
2640#ifndef NET_IP_ALIGN
2641#define NET_IP_ALIGN 2
2642#endif
2643
025be81e
AB
2644/*
2645 * The networking layer reserves some headroom in skb data (via
2646 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2647 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2648 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2649 *
2650 * Unfortunately this headroom changes the DMA alignment of the resulting
2651 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2652 * on some architectures. An architecture can override this value,
2653 * perhaps setting it to a cacheline in size (since that will maintain
2654 * cacheline alignment of the DMA). It must be a power of 2.
2655 *
d6301d3d 2656 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2657 * headroom, you should not reduce this.
5933dd2f
ED
2658 *
2659 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2660 * to reduce average number of cache lines per packet.
2661 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2662 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2663 */
2664#ifndef NET_SKB_PAD
5933dd2f 2665#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2666#endif
2667
7965bd4d 2668int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2669
5293efe6 2670static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2671{
5e1abdc3 2672 if (WARN_ON(skb_is_nonlinear(skb)))
3cc0e873 2673 return;
27a884dc
ACM
2674 skb->len = len;
2675 skb_set_tail_pointer(skb, len);
1da177e4
LT
2676}
2677
5293efe6
DB
2678static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2679{
2680 __skb_set_length(skb, len);
2681}
2682
7965bd4d 2683void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2684
2685static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2686{
3cc0e873
HX
2687 if (skb->data_len)
2688 return ___pskb_trim(skb, len);
2689 __skb_trim(skb, len);
2690 return 0;
1da177e4
LT
2691}
2692
2693static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2694{
2695 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2696}
2697
e9fa4f7b
HX
2698/**
2699 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2700 * @skb: buffer to alter
2701 * @len: new length
2702 *
2703 * This is identical to pskb_trim except that the caller knows that
2704 * the skb is not cloned so we should never get an error due to out-
2705 * of-memory.
2706 */
2707static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2708{
2709 int err = pskb_trim(skb, len);
2710 BUG_ON(err);
2711}
2712
5293efe6
DB
2713static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2714{
2715 unsigned int diff = len - skb->len;
2716
2717 if (skb_tailroom(skb) < diff) {
2718 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2719 GFP_ATOMIC);
2720 if (ret)
2721 return ret;
2722 }
2723 __skb_set_length(skb, len);
2724 return 0;
2725}
2726
1da177e4
LT
2727/**
2728 * skb_orphan - orphan a buffer
2729 * @skb: buffer to orphan
2730 *
2731 * If a buffer currently has an owner then we call the owner's
2732 * destructor function and make the @skb unowned. The buffer continues
2733 * to exist but is no longer charged to its former owner.
2734 */
2735static inline void skb_orphan(struct sk_buff *skb)
2736{
c34a7612 2737 if (skb->destructor) {
1da177e4 2738 skb->destructor(skb);
c34a7612
ED
2739 skb->destructor = NULL;
2740 skb->sk = NULL;
376c7311
ED
2741 } else {
2742 BUG_ON(skb->sk);
c34a7612 2743 }
1da177e4
LT
2744}
2745
a353e0ce
MT
2746/**
2747 * skb_orphan_frags - orphan the frags contained in a buffer
2748 * @skb: buffer to orphan frags from
2749 * @gfp_mask: allocation mask for replacement pages
2750 *
2751 * For each frag in the SKB which needs a destructor (i.e. has an
2752 * owner) create a copy of that frag and release the original
2753 * page by calling the destructor.
2754 */
2755static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2756{
1f8b977a
WB
2757 if (likely(!skb_zcopy(skb)))
2758 return 0;
185ce5c3
WB
2759 if (!skb_zcopy_is_nouarg(skb) &&
2760 skb_uarg(skb)->callback == sock_zerocopy_callback)
1f8b977a
WB
2761 return 0;
2762 return skb_copy_ubufs(skb, gfp_mask);
2763}
2764
2765/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2766static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2767{
2768 if (likely(!skb_zcopy(skb)))
a353e0ce
MT
2769 return 0;
2770 return skb_copy_ubufs(skb, gfp_mask);
2771}
2772
1da177e4
LT
2773/**
2774 * __skb_queue_purge - empty a list
2775 * @list: list to empty
2776 *
2777 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2778 * the list and one reference dropped. This function does not take the
2779 * list lock and the caller must hold the relevant locks to use it.
2780 */
1da177e4
LT
2781static inline void __skb_queue_purge(struct sk_buff_head *list)
2782{
2783 struct sk_buff *skb;
2784 while ((skb = __skb_dequeue(list)) != NULL)
2785 kfree_skb(skb);
2786}
4ea7b0cf 2787void skb_queue_purge(struct sk_buff_head *list);
1da177e4 2788
385114de 2789unsigned int skb_rbtree_purge(struct rb_root *root);
9f5afeae 2790
7965bd4d 2791void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2792
7965bd4d
JP
2793struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2794 gfp_t gfp_mask);
8af27456
CH
2795
2796/**
2797 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2798 * @dev: network device to receive on
2799 * @length: length to allocate
2800 *
2801 * Allocate a new &sk_buff and assign it a usage count of one. The
2802 * buffer has unspecified headroom built in. Users should allocate
2803 * the headroom they think they need without accounting for the
2804 * built in space. The built in space is used for optimisations.
2805 *
2806 * %NULL is returned if there is no free memory. Although this function
2807 * allocates memory it can be called from an interrupt.
2808 */
2809static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2810 unsigned int length)
8af27456
CH
2811{
2812 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2813}
2814
6f532612
ED
2815/* legacy helper around __netdev_alloc_skb() */
2816static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2817 gfp_t gfp_mask)
2818{
2819 return __netdev_alloc_skb(NULL, length, gfp_mask);
2820}
2821
2822/* legacy helper around netdev_alloc_skb() */
2823static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2824{
2825 return netdev_alloc_skb(NULL, length);
2826}
2827
2828
4915a0de
ED
2829static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2830 unsigned int length, gfp_t gfp)
61321bbd 2831{
4915a0de 2832 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2833
2834 if (NET_IP_ALIGN && skb)
2835 skb_reserve(skb, NET_IP_ALIGN);
2836 return skb;
2837}
2838
4915a0de
ED
2839static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2840 unsigned int length)
2841{
2842 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2843}
2844
181edb2b
AD
2845static inline void skb_free_frag(void *addr)
2846{
8c2dd3e4 2847 page_frag_free(addr);
181edb2b
AD
2848}
2849
ffde7328 2850void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2851struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2852 unsigned int length, gfp_t gfp_mask);
2853static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2854 unsigned int length)
2855{
2856 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2857}
795bb1c0
JDB
2858void napi_consume_skb(struct sk_buff *skb, int budget);
2859
2860void __kfree_skb_flush(void);
15fad714 2861void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2862
71dfda58
AD
2863/**
2864 * __dev_alloc_pages - allocate page for network Rx
2865 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2866 * @order: size of the allocation
2867 *
2868 * Allocate a new page.
2869 *
2870 * %NULL is returned if there is no free memory.
2871*/
2872static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2873 unsigned int order)
2874{
2875 /* This piece of code contains several assumptions.
2876 * 1. This is for device Rx, therefor a cold page is preferred.
2877 * 2. The expectation is the user wants a compound page.
2878 * 3. If requesting a order 0 page it will not be compound
2879 * due to the check to see if order has a value in prep_new_page
2880 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2881 * code in gfp_to_alloc_flags that should be enforcing this.
2882 */
453f85d4 2883 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
71dfda58
AD
2884
2885 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2886}
2887
2888static inline struct page *dev_alloc_pages(unsigned int order)
2889{
95829b3a 2890 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2891}
2892
2893/**
2894 * __dev_alloc_page - allocate a page for network Rx
2895 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2896 *
2897 * Allocate a new page.
2898 *
2899 * %NULL is returned if there is no free memory.
2900 */
2901static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2902{
2903 return __dev_alloc_pages(gfp_mask, 0);
2904}
2905
2906static inline struct page *dev_alloc_page(void)
2907{
95829b3a 2908 return dev_alloc_pages(0);
71dfda58
AD
2909}
2910
0614002b
MG
2911/**
2912 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2913 * @page: The page that was allocated from skb_alloc_page
2914 * @skb: The skb that may need pfmemalloc set
2915 */
2916static inline void skb_propagate_pfmemalloc(struct page *page,
2917 struct sk_buff *skb)
2918{
2f064f34 2919 if (page_is_pfmemalloc(page))
0614002b
MG
2920 skb->pfmemalloc = true;
2921}
2922
7240b60c
JL
2923/**
2924 * skb_frag_off() - Returns the offset of a skb fragment
2925 * @frag: the paged fragment
2926 */
2927static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2928{
65c84f14 2929 return frag->bv_offset;
7240b60c
JL
2930}
2931
2932/**
2933 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2934 * @frag: skb fragment
2935 * @delta: value to add
2936 */
2937static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2938{
65c84f14 2939 frag->bv_offset += delta;
7240b60c
JL
2940}
2941
2942/**
2943 * skb_frag_off_set() - Sets the offset of a skb fragment
2944 * @frag: skb fragment
2945 * @offset: offset of fragment
2946 */
2947static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2948{
65c84f14 2949 frag->bv_offset = offset;
7240b60c
JL
2950}
2951
2952/**
2953 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2954 * @fragto: skb fragment where offset is set
2955 * @fragfrom: skb fragment offset is copied from
2956 */
2957static inline void skb_frag_off_copy(skb_frag_t *fragto,
2958 const skb_frag_t *fragfrom)
2959{
65c84f14 2960 fragto->bv_offset = fragfrom->bv_offset;
7240b60c
JL
2961}
2962
131ea667 2963/**
e227867f 2964 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2965 * @frag: the paged fragment
2966 *
2967 * Returns the &struct page associated with @frag.
2968 */
2969static inline struct page *skb_frag_page(const skb_frag_t *frag)
2970{
1dfa5bd3 2971 return frag->bv_page;
131ea667
IC
2972}
2973
2974/**
2975 * __skb_frag_ref - take an addition reference on a paged fragment.
2976 * @frag: the paged fragment
2977 *
2978 * Takes an additional reference on the paged fragment @frag.
2979 */
2980static inline void __skb_frag_ref(skb_frag_t *frag)
2981{
2982 get_page(skb_frag_page(frag));
2983}
2984
2985/**
2986 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2987 * @skb: the buffer
2988 * @f: the fragment offset.
2989 *
2990 * Takes an additional reference on the @f'th paged fragment of @skb.
2991 */
2992static inline void skb_frag_ref(struct sk_buff *skb, int f)
2993{
2994 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2995}
2996
2997/**
2998 * __skb_frag_unref - release a reference on a paged fragment.
2999 * @frag: the paged fragment
3000 *
3001 * Releases a reference on the paged fragment @frag.
3002 */
3003static inline void __skb_frag_unref(skb_frag_t *frag)
3004{
3005 put_page(skb_frag_page(frag));
3006}
3007
3008/**
3009 * skb_frag_unref - release a reference on a paged fragment of an skb.
3010 * @skb: the buffer
3011 * @f: the fragment offset
3012 *
3013 * Releases a reference on the @f'th paged fragment of @skb.
3014 */
3015static inline void skb_frag_unref(struct sk_buff *skb, int f)
3016{
3017 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
3018}
3019
3020/**
3021 * skb_frag_address - gets the address of the data contained in a paged fragment
3022 * @frag: the paged fragment buffer
3023 *
3024 * Returns the address of the data within @frag. The page must already
3025 * be mapped.
3026 */
3027static inline void *skb_frag_address(const skb_frag_t *frag)
3028{
7240b60c 3029 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
131ea667
IC
3030}
3031
3032/**
3033 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3034 * @frag: the paged fragment buffer
3035 *
3036 * Returns the address of the data within @frag. Checks that the page
3037 * is mapped and returns %NULL otherwise.
3038 */
3039static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3040{
3041 void *ptr = page_address(skb_frag_page(frag));
3042 if (unlikely(!ptr))
3043 return NULL;
3044
7240b60c
JL
3045 return ptr + skb_frag_off(frag);
3046}
3047
3048/**
3049 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3050 * @fragto: skb fragment where page is set
3051 * @fragfrom: skb fragment page is copied from
3052 */
3053static inline void skb_frag_page_copy(skb_frag_t *fragto,
3054 const skb_frag_t *fragfrom)
3055{
3056 fragto->bv_page = fragfrom->bv_page;
131ea667
IC
3057}
3058
3059/**
3060 * __skb_frag_set_page - sets the page contained in a paged fragment
3061 * @frag: the paged fragment
3062 * @page: the page to set
3063 *
3064 * Sets the fragment @frag to contain @page.
3065 */
3066static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3067{
1dfa5bd3 3068 frag->bv_page = page;
131ea667
IC
3069}
3070
3071/**
3072 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3073 * @skb: the buffer
3074 * @f: the fragment offset
3075 * @page: the page to set
3076 *
3077 * Sets the @f'th fragment of @skb to contain @page.
3078 */
3079static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3080 struct page *page)
3081{
3082 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3083}
3084
400dfd3a
ED
3085bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3086
131ea667
IC
3087/**
3088 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 3089 * @dev: the device to map the fragment to
131ea667
IC
3090 * @frag: the paged fragment to map
3091 * @offset: the offset within the fragment (starting at the
3092 * fragment's own offset)
3093 * @size: the number of bytes to map
771b00a8 3094 * @dir: the direction of the mapping (``PCI_DMA_*``)
131ea667
IC
3095 *
3096 * Maps the page associated with @frag to @device.
3097 */
3098static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3099 const skb_frag_t *frag,
3100 size_t offset, size_t size,
3101 enum dma_data_direction dir)
3102{
3103 return dma_map_page(dev, skb_frag_page(frag),
7240b60c 3104 skb_frag_off(frag) + offset, size, dir);
131ea667
IC
3105}
3106
117632e6
ED
3107static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3108 gfp_t gfp_mask)
3109{
3110 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3111}
3112
bad93e9d
OP
3113
3114static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3115 gfp_t gfp_mask)
3116{
3117 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3118}
3119
3120
334a8132
PM
3121/**
3122 * skb_clone_writable - is the header of a clone writable
3123 * @skb: buffer to check
3124 * @len: length up to which to write
3125 *
3126 * Returns true if modifying the header part of the cloned buffer
3127 * does not requires the data to be copied.
3128 */
05bdd2f1 3129static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
3130{
3131 return !skb_header_cloned(skb) &&
3132 skb_headroom(skb) + len <= skb->hdr_len;
3133}
3134
3697649f
DB
3135static inline int skb_try_make_writable(struct sk_buff *skb,
3136 unsigned int write_len)
3137{
3138 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3139 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3140}
3141
d9cc2048
HX
3142static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3143 int cloned)
3144{
3145 int delta = 0;
3146
d9cc2048
HX
3147 if (headroom > skb_headroom(skb))
3148 delta = headroom - skb_headroom(skb);
3149
3150 if (delta || cloned)
3151 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3152 GFP_ATOMIC);
3153 return 0;
3154}
3155
1da177e4
LT
3156/**
3157 * skb_cow - copy header of skb when it is required
3158 * @skb: buffer to cow
3159 * @headroom: needed headroom
3160 *
3161 * If the skb passed lacks sufficient headroom or its data part
3162 * is shared, data is reallocated. If reallocation fails, an error
3163 * is returned and original skb is not changed.
3164 *
3165 * The result is skb with writable area skb->head...skb->tail
3166 * and at least @headroom of space at head.
3167 */
3168static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3169{
d9cc2048
HX
3170 return __skb_cow(skb, headroom, skb_cloned(skb));
3171}
1da177e4 3172
d9cc2048
HX
3173/**
3174 * skb_cow_head - skb_cow but only making the head writable
3175 * @skb: buffer to cow
3176 * @headroom: needed headroom
3177 *
3178 * This function is identical to skb_cow except that we replace the
3179 * skb_cloned check by skb_header_cloned. It should be used when
3180 * you only need to push on some header and do not need to modify
3181 * the data.
3182 */
3183static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3184{
3185 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
3186}
3187
3188/**
3189 * skb_padto - pad an skbuff up to a minimal size
3190 * @skb: buffer to pad
3191 * @len: minimal length
3192 *
3193 * Pads up a buffer to ensure the trailing bytes exist and are
3194 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
3195 * is untouched. Otherwise it is extended. Returns zero on
3196 * success. The skb is freed on error.
1da177e4 3197 */
5b057c6b 3198static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
3199{
3200 unsigned int size = skb->len;
3201 if (likely(size >= len))
5b057c6b 3202 return 0;
987c402a 3203 return skb_pad(skb, len - size);
1da177e4
LT
3204}
3205
9c0c1124 3206/**
4ea7b0cf 3207 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
9c0c1124
AD
3208 * @skb: buffer to pad
3209 * @len: minimal length
cd0a137a 3210 * @free_on_error: free buffer on error
9c0c1124
AD
3211 *
3212 * Pads up a buffer to ensure the trailing bytes exist and are
3213 * blanked. If the buffer already contains sufficient data it
3214 * is untouched. Otherwise it is extended. Returns zero on
cd0a137a 3215 * success. The skb is freed on error if @free_on_error is true.
9c0c1124 3216 */
cd0a137a
FF
3217static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
3218 bool free_on_error)
9c0c1124
AD
3219{
3220 unsigned int size = skb->len;
3221
3222 if (unlikely(size < len)) {
3223 len -= size;
cd0a137a 3224 if (__skb_pad(skb, len, free_on_error))
9c0c1124
AD
3225 return -ENOMEM;
3226 __skb_put(skb, len);
3227 }
3228 return 0;
3229}
3230
cd0a137a
FF
3231/**
3232 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3233 * @skb: buffer to pad
3234 * @len: minimal length
3235 *
3236 * Pads up a buffer to ensure the trailing bytes exist and are
3237 * blanked. If the buffer already contains sufficient data it
3238 * is untouched. Otherwise it is extended. Returns zero on
3239 * success. The skb is freed on error.
3240 */
3241static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
3242{
3243 return __skb_put_padto(skb, len, true);
3244}
3245
1da177e4 3246static inline int skb_add_data(struct sk_buff *skb,
af2b040e 3247 struct iov_iter *from, int copy)
1da177e4
LT
3248{
3249 const int off = skb->len;
3250
3251 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 3252 __wsum csum = 0;
15e6cb46
AV
3253 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3254 &csum, from)) {
1da177e4
LT
3255 skb->csum = csum_block_add(skb->csum, csum, off);
3256 return 0;
3257 }
15e6cb46 3258 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
3259 return 0;
3260
3261 __skb_trim(skb, off);
3262 return -EFAULT;
3263}
3264
38ba0a65
ED
3265static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3266 const struct page *page, int off)
1da177e4 3267{
1f8b977a
WB
3268 if (skb_zcopy(skb))
3269 return false;
1da177e4 3270 if (i) {
d8e18a51 3271 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 3272
ea2ab693 3273 return page == skb_frag_page(frag) &&
7240b60c 3274 off == skb_frag_off(frag) + skb_frag_size(frag);
1da177e4 3275 }
38ba0a65 3276 return false;
1da177e4
LT
3277}
3278
364c6bad
HX
3279static inline int __skb_linearize(struct sk_buff *skb)
3280{
3281 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3282}
3283
1da177e4
LT
3284/**
3285 * skb_linearize - convert paged skb to linear one
3286 * @skb: buffer to linarize
1da177e4
LT
3287 *
3288 * If there is no free memory -ENOMEM is returned, otherwise zero
3289 * is returned and the old skb data released.
3290 */
364c6bad
HX
3291static inline int skb_linearize(struct sk_buff *skb)
3292{
3293 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3294}
3295
cef401de
ED
3296/**
3297 * skb_has_shared_frag - can any frag be overwritten
3298 * @skb: buffer to test
3299 *
3300 * Return true if the skb has at least one frag that might be modified
3301 * by an external entity (as in vmsplice()/sendfile())
3302 */
3303static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3304{
c9af6db4
PS
3305 return skb_is_nonlinear(skb) &&
3306 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
3307}
3308
364c6bad
HX
3309/**
3310 * skb_linearize_cow - make sure skb is linear and writable
3311 * @skb: buffer to process
3312 *
3313 * If there is no free memory -ENOMEM is returned, otherwise zero
3314 * is returned and the old skb data released.
3315 */
3316static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 3317{
364c6bad
HX
3318 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3319 __skb_linearize(skb) : 0;
1da177e4
LT
3320}
3321
479ffccc
DB
3322static __always_inline void
3323__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3324 unsigned int off)
3325{
3326 if (skb->ip_summed == CHECKSUM_COMPLETE)
3327 skb->csum = csum_block_sub(skb->csum,
3328 csum_partial(start, len, 0), off);
3329 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3330 skb_checksum_start_offset(skb) < 0)
3331 skb->ip_summed = CHECKSUM_NONE;
3332}
3333
1da177e4
LT
3334/**
3335 * skb_postpull_rcsum - update checksum for received skb after pull
3336 * @skb: buffer to update
3337 * @start: start of data before pull
3338 * @len: length of data pulled
3339 *
3340 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
3341 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3342 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 3343 */
1da177e4 3344static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 3345 const void *start, unsigned int len)
1da177e4 3346{
479ffccc 3347 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
3348}
3349
479ffccc
DB
3350static __always_inline void
3351__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3352 unsigned int off)
3353{
3354 if (skb->ip_summed == CHECKSUM_COMPLETE)
3355 skb->csum = csum_block_add(skb->csum,
3356 csum_partial(start, len, 0), off);
3357}
cbb042f9 3358
479ffccc
DB
3359/**
3360 * skb_postpush_rcsum - update checksum for received skb after push
3361 * @skb: buffer to update
3362 * @start: start of data after push
3363 * @len: length of data pushed
3364 *
3365 * After doing a push on a received packet, you need to call this to
3366 * update the CHECKSUM_COMPLETE checksum.
3367 */
f8ffad69
DB
3368static inline void skb_postpush_rcsum(struct sk_buff *skb,
3369 const void *start, unsigned int len)
3370{
479ffccc 3371 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
3372}
3373
af72868b 3374void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
479ffccc 3375
82a31b92
WC
3376/**
3377 * skb_push_rcsum - push skb and update receive checksum
3378 * @skb: buffer to update
3379 * @len: length of data pulled
3380 *
3381 * This function performs an skb_push on the packet and updates
3382 * the CHECKSUM_COMPLETE checksum. It should be used on
3383 * receive path processing instead of skb_push unless you know
3384 * that the checksum difference is zero (e.g., a valid IP header)
3385 * or you are setting ip_summed to CHECKSUM_NONE.
3386 */
d58ff351 3387static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
82a31b92
WC
3388{
3389 skb_push(skb, len);
3390 skb_postpush_rcsum(skb, skb->data, len);
3391 return skb->data;
3392}
3393
88078d98 3394int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
7ce5a27f
DM
3395/**
3396 * pskb_trim_rcsum - trim received skb and update checksum
3397 * @skb: buffer to trim
3398 * @len: new length
3399 *
3400 * This is exactly the same as pskb_trim except that it ensures the
3401 * checksum of received packets are still valid after the operation.
6c57f045 3402 * It can change skb pointers.
7ce5a27f
DM
3403 */
3404
3405static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3406{
3407 if (likely(len >= skb->len))
3408 return 0;
88078d98 3409 return pskb_trim_rcsum_slow(skb, len);
7ce5a27f
DM
3410}
3411
5293efe6
DB
3412static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3413{
3414 if (skb->ip_summed == CHECKSUM_COMPLETE)
3415 skb->ip_summed = CHECKSUM_NONE;
3416 __skb_trim(skb, len);
3417 return 0;
3418}
3419
3420static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3421{
3422 if (skb->ip_summed == CHECKSUM_COMPLETE)
3423 skb->ip_summed = CHECKSUM_NONE;
3424 return __skb_grow(skb, len);
3425}
3426
18a4c0ea
ED
3427#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3428#define skb_rb_first(root) rb_to_skb(rb_first(root))
3429#define skb_rb_last(root) rb_to_skb(rb_last(root))
3430#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3431#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3432
1da177e4
LT
3433#define skb_queue_walk(queue, skb) \
3434 for (skb = (queue)->next; \
a1e4891f 3435 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3436 skb = skb->next)
3437
46f8914e
JC
3438#define skb_queue_walk_safe(queue, skb, tmp) \
3439 for (skb = (queue)->next, tmp = skb->next; \
3440 skb != (struct sk_buff *)(queue); \
3441 skb = tmp, tmp = skb->next)
3442
1164f52a 3443#define skb_queue_walk_from(queue, skb) \
a1e4891f 3444 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3445 skb = skb->next)
3446
18a4c0ea
ED
3447#define skb_rbtree_walk(skb, root) \
3448 for (skb = skb_rb_first(root); skb != NULL; \
3449 skb = skb_rb_next(skb))
3450
3451#define skb_rbtree_walk_from(skb) \
3452 for (; skb != NULL; \
3453 skb = skb_rb_next(skb))
3454
3455#define skb_rbtree_walk_from_safe(skb, tmp) \
3456 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3457 skb = tmp)
3458
1164f52a
DM
3459#define skb_queue_walk_from_safe(queue, skb, tmp) \
3460 for (tmp = skb->next; \
3461 skb != (struct sk_buff *)(queue); \
3462 skb = tmp, tmp = skb->next)
3463
300ce174
SH
3464#define skb_queue_reverse_walk(queue, skb) \
3465 for (skb = (queue)->prev; \
a1e4891f 3466 skb != (struct sk_buff *)(queue); \
300ce174
SH
3467 skb = skb->prev)
3468
686a2955
DM
3469#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3470 for (skb = (queue)->prev, tmp = skb->prev; \
3471 skb != (struct sk_buff *)(queue); \
3472 skb = tmp, tmp = skb->prev)
3473
3474#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3475 for (tmp = skb->prev; \
3476 skb != (struct sk_buff *)(queue); \
3477 skb = tmp, tmp = skb->prev)
1da177e4 3478
21dc3301 3479static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3480{
3481 return skb_shinfo(skb)->frag_list != NULL;
3482}
3483
3484static inline void skb_frag_list_init(struct sk_buff *skb)
3485{
3486 skb_shinfo(skb)->frag_list = NULL;
3487}
3488
ee039871
DM
3489#define skb_walk_frags(skb, iter) \
3490 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3491
ea3793ee 3492
b50b0580
SD
3493int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3494 int *err, long *timeo_p,
ea3793ee 3495 const struct sk_buff *skb);
65101aec
PA
3496struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3497 struct sk_buff_head *queue,
3498 unsigned int flags,
fd69c399 3499 int *off, int *err,
65101aec 3500 struct sk_buff **last);
b50b0580
SD
3501struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3502 struct sk_buff_head *queue,
e427cad6 3503 unsigned int flags, int *off, int *err,
ea3793ee 3504 struct sk_buff **last);
b50b0580
SD
3505struct sk_buff *__skb_recv_datagram(struct sock *sk,
3506 struct sk_buff_head *sk_queue,
e427cad6 3507 unsigned int flags, int *off, int *err);
7965bd4d
JP
3508struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3509 int *err);
a11e1d43
LT
3510__poll_t datagram_poll(struct file *file, struct socket *sock,
3511 struct poll_table_struct *wait);
c0371da6
AV
3512int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3513 struct iov_iter *to, int size);
51f3d02b
DM
3514static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3515 struct msghdr *msg, int size)
3516{
e5a4b0bb 3517 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3518}
e5a4b0bb
AV
3519int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3520 struct msghdr *msg);
65d69e25
SG
3521int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3522 struct iov_iter *to, int len,
3523 struct ahash_request *hash);
3a654f97
AV
3524int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3525 struct iov_iter *from, int len);
3a654f97 3526int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3527void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3528void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3529static inline void skb_free_datagram_locked(struct sock *sk,
3530 struct sk_buff *skb)
3531{
3532 __skb_free_datagram_locked(sk, skb, 0);
3533}
7965bd4d 3534int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3535int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3536int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3537__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3538 int len, __wsum csum);
a60e3cc7 3539int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3540 struct pipe_inode_info *pipe, unsigned int len,
25869262 3541 unsigned int flags);
20bf50de
TH
3542int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3543 int len);
7965bd4d 3544void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3545unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3546int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3547 int len, int hlen);
7965bd4d
JP
3548void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3549int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3550void skb_scrub_packet(struct sk_buff *skb, bool xnet);
779b7931 3551bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
2b16f048 3552bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
7965bd4d 3553struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3a1296a3
SK
3554struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3555 unsigned int offset);
0d5501c1 3556struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3557int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3558int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3559int skb_vlan_pop(struct sk_buff *skb);
3560int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
fa4e0f88 3561int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
d04ac224 3562 int mac_len, bool ethernet);
040b5cfb
MV
3563int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3564 bool ethernet);
d27cf5c5 3565int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
2a2ea508 3566int skb_mpls_dec_ttl(struct sk_buff *skb);
6fa01ccd
SV
3567struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3568 gfp_t gfp);
20380731 3569
6ce8e9ce
AV
3570static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3571{
3073f070 3572 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3573}
3574
7eab8d9e
AV
3575static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3576{
e5a4b0bb 3577 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3578}
3579
2817a336
DB
3580struct skb_checksum_ops {
3581 __wsum (*update)(const void *mem, int len, __wsum wsum);
3582 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3583};
3584
9617813d
DC
3585extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3586
2817a336
DB
3587__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3588 __wsum csum, const struct skb_checksum_ops *ops);
3589__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3590 __wsum csum);
3591
1e98a0f0
ED
3592static inline void * __must_check
3593__skb_header_pointer(const struct sk_buff *skb, int offset,
3594 int len, void *data, int hlen, void *buffer)
1da177e4 3595{
55820ee2 3596 if (hlen - offset >= len)
690e36e7 3597 return data + offset;
1da177e4 3598
690e36e7
DM
3599 if (!skb ||
3600 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3601 return NULL;
3602
3603 return buffer;
3604}
3605
1e98a0f0
ED
3606static inline void * __must_check
3607skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3608{
3609 return __skb_header_pointer(skb, offset, len, skb->data,
3610 skb_headlen(skb), buffer);
3611}
3612
4262e5cc
DB
3613/**
3614 * skb_needs_linearize - check if we need to linearize a given skb
3615 * depending on the given device features.
3616 * @skb: socket buffer to check
3617 * @features: net device features
3618 *
3619 * Returns true if either:
3620 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3621 * 2. skb is fragmented and the device does not support SG.
3622 */
3623static inline bool skb_needs_linearize(struct sk_buff *skb,
3624 netdev_features_t features)
3625{
3626 return skb_is_nonlinear(skb) &&
3627 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3628 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3629}
3630
d626f62b
ACM
3631static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3632 void *to,
3633 const unsigned int len)
3634{
3635 memcpy(to, skb->data, len);
3636}
3637
3638static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3639 const int offset, void *to,
3640 const unsigned int len)
3641{
3642 memcpy(to, skb->data + offset, len);
3643}
3644
27d7ff46
ACM
3645static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3646 const void *from,
3647 const unsigned int len)
3648{
3649 memcpy(skb->data, from, len);
3650}
3651
3652static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3653 const int offset,
3654 const void *from,
3655 const unsigned int len)
3656{
3657 memcpy(skb->data + offset, from, len);
3658}
3659
7965bd4d 3660void skb_init(void);
1da177e4 3661
ac45f602
PO
3662static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3663{
3664 return skb->tstamp;
3665}
3666
a61bbcf2
PM
3667/**
3668 * skb_get_timestamp - get timestamp from a skb
3669 * @skb: skb to get stamp from
13c6ee2a 3670 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
a61bbcf2
PM
3671 *
3672 * Timestamps are stored in the skb as offsets to a base timestamp.
3673 * This function converts the offset back to a struct timeval and stores
3674 * it in stamp.
3675 */
ac45f602 3676static inline void skb_get_timestamp(const struct sk_buff *skb,
13c6ee2a 3677 struct __kernel_old_timeval *stamp)
a61bbcf2 3678{
13c6ee2a 3679 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
a61bbcf2
PM
3680}
3681
887feae3
DD
3682static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3683 struct __kernel_sock_timeval *stamp)
3684{
3685 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3686
3687 stamp->tv_sec = ts.tv_sec;
3688 stamp->tv_usec = ts.tv_nsec / 1000;
3689}
3690
ac45f602 3691static inline void skb_get_timestampns(const struct sk_buff *skb,
df1b4ba9 3692 struct __kernel_old_timespec *stamp)
ac45f602 3693{
df1b4ba9
AB
3694 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3695
3696 stamp->tv_sec = ts.tv_sec;
3697 stamp->tv_nsec = ts.tv_nsec;
ac45f602
PO
3698}
3699
887feae3
DD
3700static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3701 struct __kernel_timespec *stamp)
3702{
3703 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3704
3705 stamp->tv_sec = ts.tv_sec;
3706 stamp->tv_nsec = ts.tv_nsec;
3707}
3708
b7aa0bf7 3709static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3710{
b7aa0bf7 3711 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3712}
3713
164891aa
SH
3714static inline ktime_t net_timedelta(ktime_t t)
3715{
3716 return ktime_sub(ktime_get_real(), t);
3717}
3718
b9ce204f
IJ
3719static inline ktime_t net_invalid_timestamp(void)
3720{
8b0e1953 3721 return 0;
b9ce204f 3722}
a61bbcf2 3723
de8f3a83
DB
3724static inline u8 skb_metadata_len(const struct sk_buff *skb)
3725{
3726 return skb_shinfo(skb)->meta_len;
3727}
3728
3729static inline void *skb_metadata_end(const struct sk_buff *skb)
3730{
3731 return skb_mac_header(skb);
3732}
3733
3734static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3735 const struct sk_buff *skb_b,
3736 u8 meta_len)
3737{
3738 const void *a = skb_metadata_end(skb_a);
3739 const void *b = skb_metadata_end(skb_b);
3740 /* Using more efficient varaiant than plain call to memcmp(). */
3741#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3742 u64 diffs = 0;
3743
3744 switch (meta_len) {
3745#define __it(x, op) (x -= sizeof(u##op))
3746#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3747 case 32: diffs |= __it_diff(a, b, 64);
df561f66 3748 fallthrough;
de8f3a83 3749 case 24: diffs |= __it_diff(a, b, 64);
df561f66 3750 fallthrough;
de8f3a83 3751 case 16: diffs |= __it_diff(a, b, 64);
df561f66 3752 fallthrough;
de8f3a83
DB
3753 case 8: diffs |= __it_diff(a, b, 64);
3754 break;
3755 case 28: diffs |= __it_diff(a, b, 64);
df561f66 3756 fallthrough;
de8f3a83 3757 case 20: diffs |= __it_diff(a, b, 64);
df561f66 3758 fallthrough;
de8f3a83 3759 case 12: diffs |= __it_diff(a, b, 64);
df561f66 3760 fallthrough;
de8f3a83
DB
3761 case 4: diffs |= __it_diff(a, b, 32);
3762 break;
3763 }
3764 return diffs;
3765#else
3766 return memcmp(a - meta_len, b - meta_len, meta_len);
3767#endif
3768}
3769
3770static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3771 const struct sk_buff *skb_b)
3772{
3773 u8 len_a = skb_metadata_len(skb_a);
3774 u8 len_b = skb_metadata_len(skb_b);
3775
3776 if (!(len_a | len_b))
3777 return false;
3778
3779 return len_a != len_b ?
3780 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3781}
3782
3783static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3784{
3785 skb_shinfo(skb)->meta_len = meta_len;
3786}
3787
3788static inline void skb_metadata_clear(struct sk_buff *skb)
3789{
3790 skb_metadata_set(skb, 0);
3791}
3792
62bccb8c
AD
3793struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3794
c1f19b51
RC
3795#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3796
7965bd4d
JP
3797void skb_clone_tx_timestamp(struct sk_buff *skb);
3798bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3799
3800#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3801
3802static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3803{
3804}
3805
3806static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3807{
3808 return false;
3809}
3810
3811#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3812
3813/**
3814 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3815 *
da92b194
RC
3816 * PHY drivers may accept clones of transmitted packets for
3817 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3818 * must call this function to return the skb back to the stack with a
3819 * timestamp.
da92b194 3820 *
2ff17117 3821 * @skb: clone of the original outgoing packet
7a76a021 3822 * @hwtstamps: hardware time stamps
c1f19b51
RC
3823 *
3824 */
3825void skb_complete_tx_timestamp(struct sk_buff *skb,
3826 struct skb_shared_hwtstamps *hwtstamps);
3827
e7fd2885
WB
3828void __skb_tstamp_tx(struct sk_buff *orig_skb,
3829 struct skb_shared_hwtstamps *hwtstamps,
3830 struct sock *sk, int tstype);
3831
ac45f602
PO
3832/**
3833 * skb_tstamp_tx - queue clone of skb with send time stamps
3834 * @orig_skb: the original outgoing packet
3835 * @hwtstamps: hardware time stamps, may be NULL if not available
3836 *
3837 * If the skb has a socket associated, then this function clones the
3838 * skb (thus sharing the actual data and optional structures), stores
3839 * the optional hardware time stamping information (if non NULL) or
3840 * generates a software time stamp (otherwise), then queues the clone
3841 * to the error queue of the socket. Errors are silently ignored.
3842 */
7965bd4d
JP
3843void skb_tstamp_tx(struct sk_buff *orig_skb,
3844 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3845
4507a715
RC
3846/**
3847 * skb_tx_timestamp() - Driver hook for transmit timestamping
3848 *
3849 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3850 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3851 *
73409f3b
DM
3852 * Specifically, one should make absolutely sure that this function is
3853 * called before TX completion of this packet can trigger. Otherwise
3854 * the packet could potentially already be freed.
3855 *
4507a715
RC
3856 * @skb: A socket buffer.
3857 */
3858static inline void skb_tx_timestamp(struct sk_buff *skb)
3859{
c1f19b51 3860 skb_clone_tx_timestamp(skb);
b50a5c70
ML
3861 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3862 skb_tstamp_tx(skb, NULL);
4507a715
RC
3863}
3864
6e3e939f
JB
3865/**
3866 * skb_complete_wifi_ack - deliver skb with wifi status
3867 *
3868 * @skb: the original outgoing packet
3869 * @acked: ack status
3870 *
3871 */
3872void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3873
7965bd4d
JP
3874__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3875__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3876
60476372
HX
3877static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3878{
6edec0e6
TH
3879 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3880 skb->csum_valid ||
3881 (skb->ip_summed == CHECKSUM_PARTIAL &&
3882 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3883}
3884
fb286bb2
HX
3885/**
3886 * skb_checksum_complete - Calculate checksum of an entire packet
3887 * @skb: packet to process
3888 *
3889 * This function calculates the checksum over the entire packet plus
3890 * the value of skb->csum. The latter can be used to supply the
3891 * checksum of a pseudo header as used by TCP/UDP. It returns the
3892 * checksum.
3893 *
3894 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3895 * this function can be used to verify that checksum on received
3896 * packets. In that case the function should return zero if the
3897 * checksum is correct. In particular, this function will return zero
3898 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3899 * hardware has already verified the correctness of the checksum.
3900 */
4381ca3c 3901static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3902{
60476372
HX
3903 return skb_csum_unnecessary(skb) ?
3904 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3905}
3906
77cffe23
TH
3907static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3908{
3909 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3910 if (skb->csum_level == 0)
3911 skb->ip_summed = CHECKSUM_NONE;
3912 else
3913 skb->csum_level--;
3914 }
3915}
3916
3917static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3918{
3919 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3920 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3921 skb->csum_level++;
3922 } else if (skb->ip_summed == CHECKSUM_NONE) {
3923 skb->ip_summed = CHECKSUM_UNNECESSARY;
3924 skb->csum_level = 0;
3925 }
3926}
3927
836e66c2
DB
3928static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
3929{
3930 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3931 skb->ip_summed = CHECKSUM_NONE;
3932 skb->csum_level = 0;
3933 }
3934}
3935
76ba0aae
TH
3936/* Check if we need to perform checksum complete validation.
3937 *
3938 * Returns true if checksum complete is needed, false otherwise
3939 * (either checksum is unnecessary or zero checksum is allowed).
3940 */
3941static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3942 bool zero_okay,
3943 __sum16 check)
3944{
5d0c2b95
TH
3945 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3946 skb->csum_valid = 1;
77cffe23 3947 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3948 return false;
3949 }
3950
3951 return true;
3952}
3953
da279887 3954/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
76ba0aae
TH
3955 * in checksum_init.
3956 */
3957#define CHECKSUM_BREAK 76
3958
4e18b9ad
TH
3959/* Unset checksum-complete
3960 *
3961 * Unset checksum complete can be done when packet is being modified
3962 * (uncompressed for instance) and checksum-complete value is
3963 * invalidated.
3964 */
3965static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3966{
3967 if (skb->ip_summed == CHECKSUM_COMPLETE)
3968 skb->ip_summed = CHECKSUM_NONE;
3969}
3970
76ba0aae
TH
3971/* Validate (init) checksum based on checksum complete.
3972 *
3973 * Return values:
3974 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3975 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3976 * checksum is stored in skb->csum for use in __skb_checksum_complete
3977 * non-zero: value of invalid checksum
3978 *
3979 */
3980static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3981 bool complete,
3982 __wsum psum)
3983{
3984 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3985 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3986 skb->csum_valid = 1;
76ba0aae
TH
3987 return 0;
3988 }
3989 }
3990
3991 skb->csum = psum;
3992
5d0c2b95
TH
3993 if (complete || skb->len <= CHECKSUM_BREAK) {
3994 __sum16 csum;
3995
3996 csum = __skb_checksum_complete(skb);
3997 skb->csum_valid = !csum;
3998 return csum;
3999 }
76ba0aae
TH
4000
4001 return 0;
4002}
4003
4004static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4005{
4006 return 0;
4007}
4008
4009/* Perform checksum validate (init). Note that this is a macro since we only
4010 * want to calculate the pseudo header which is an input function if necessary.
4011 * First we try to validate without any computation (checksum unnecessary) and
4012 * then calculate based on checksum complete calling the function to compute
4013 * pseudo header.
4014 *
4015 * Return values:
4016 * 0: checksum is validated or try to in skb_checksum_complete
4017 * non-zero: value of invalid checksum
4018 */
4019#define __skb_checksum_validate(skb, proto, complete, \
4020 zero_okay, check, compute_pseudo) \
4021({ \
4022 __sum16 __ret = 0; \
5d0c2b95 4023 skb->csum_valid = 0; \
76ba0aae
TH
4024 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4025 __ret = __skb_checksum_validate_complete(skb, \
4026 complete, compute_pseudo(skb, proto)); \
4027 __ret; \
4028})
4029
4030#define skb_checksum_init(skb, proto, compute_pseudo) \
4031 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4032
4033#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4034 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4035
4036#define skb_checksum_validate(skb, proto, compute_pseudo) \
4037 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4038
4039#define skb_checksum_validate_zero_check(skb, proto, check, \
4040 compute_pseudo) \
096a4cfa 4041 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
4042
4043#define skb_checksum_simple_validate(skb) \
4044 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4045
d96535a1
TH
4046static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4047{
219f1d79 4048 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
4049}
4050
e4aa33ad 4051static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
d96535a1
TH
4052{
4053 skb->csum = ~pseudo;
4054 skb->ip_summed = CHECKSUM_COMPLETE;
4055}
4056
e4aa33ad 4057#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
4058do { \
4059 if (__skb_checksum_convert_check(skb)) \
e4aa33ad 4060 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
d96535a1
TH
4061} while (0)
4062
15e2396d
TH
4063static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4064 u16 start, u16 offset)
4065{
4066 skb->ip_summed = CHECKSUM_PARTIAL;
4067 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4068 skb->csum_offset = offset - start;
4069}
4070
dcdc8994
TH
4071/* Update skbuf and packet to reflect the remote checksum offload operation.
4072 * When called, ptr indicates the starting point for skb->csum when
4073 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4074 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4075 */
4076static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 4077 int start, int offset, bool nopartial)
dcdc8994
TH
4078{
4079 __wsum delta;
4080
15e2396d
TH
4081 if (!nopartial) {
4082 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4083 return;
4084 }
4085
dcdc8994
TH
4086 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4087 __skb_checksum_complete(skb);
4088 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4089 }
4090
4091 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4092
4093 /* Adjust skb->csum since we changed the packet */
4094 skb->csum = csum_add(skb->csum, delta);
4095}
4096
cb9c6836
FW
4097static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4098{
4099#if IS_ENABLED(CONFIG_NF_CONNTRACK)
261db6c2 4100 return (void *)(skb->_nfct & NFCT_PTRMASK);
cb9c6836
FW
4101#else
4102 return NULL;
4103#endif
4104}
4105
261db6c2 4106static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
1da177e4 4107{
261db6c2
JS
4108#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4109 return skb->_nfct;
4110#else
4111 return 0UL;
4112#endif
1da177e4 4113}
261db6c2
JS
4114
4115static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
1da177e4 4116{
261db6c2
JS
4117#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4118 skb->_nfct = nfct;
2fc72c7b 4119#endif
261db6c2 4120}
df5042f4
FW
4121
4122#ifdef CONFIG_SKB_EXTENSIONS
4123enum skb_ext_id {
4124#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4125 SKB_EXT_BRIDGE_NF,
4165079b
FW
4126#endif
4127#ifdef CONFIG_XFRM
4128 SKB_EXT_SEC_PATH,
95a7233c
PB
4129#endif
4130#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4131 TC_SKB_EXT,
3ee17bc7
MM
4132#endif
4133#if IS_ENABLED(CONFIG_MPTCP)
4134 SKB_EXT_MPTCP,
df5042f4
FW
4135#endif
4136 SKB_EXT_NUM, /* must be last */
4137};
4138
4139/**
4140 * struct skb_ext - sk_buff extensions
4141 * @refcnt: 1 on allocation, deallocated on 0
4142 * @offset: offset to add to @data to obtain extension address
4143 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4144 * @data: start of extension data, variable sized
4145 *
4146 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4147 * to use 'u8' types while allowing up to 2kb worth of extension data.
4148 */
4149struct skb_ext {
4150 refcount_t refcnt;
4151 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4152 u8 chunks; /* same */
5c91aa1d 4153 char data[] __aligned(8);
df5042f4
FW
4154};
4155
4930f483 4156struct skb_ext *__skb_ext_alloc(gfp_t flags);
8b69a803
PA
4157void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4158 struct skb_ext *ext);
df5042f4
FW
4159void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4160void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4161void __skb_ext_put(struct skb_ext *ext);
4162
4163static inline void skb_ext_put(struct sk_buff *skb)
4164{
4165 if (skb->active_extensions)
4166 __skb_ext_put(skb->extensions);
4167}
4168
df5042f4
FW
4169static inline void __skb_ext_copy(struct sk_buff *dst,
4170 const struct sk_buff *src)
4171{
4172 dst->active_extensions = src->active_extensions;
4173
4174 if (src->active_extensions) {
4175 struct skb_ext *ext = src->extensions;
4176
4177 refcount_inc(&ext->refcnt);
4178 dst->extensions = ext;
4179 }
4180}
4181
4182static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4183{
4184 skb_ext_put(dst);
4185 __skb_ext_copy(dst, src);
4186}
4187
4188static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4189{
4190 return !!ext->offset[i];
4191}
4192
4193static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4194{
4195 return skb->active_extensions & (1 << id);
4196}
4197
4198static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4199{
4200 if (skb_ext_exist(skb, id))
4201 __skb_ext_del(skb, id);
4202}
4203
4204static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4205{
4206 if (skb_ext_exist(skb, id)) {
4207 struct skb_ext *ext = skb->extensions;
4208
4209 return (void *)ext + (ext->offset[id] << 3);
4210 }
4211
4212 return NULL;
4213}
174e2381
FW
4214
4215static inline void skb_ext_reset(struct sk_buff *skb)
4216{
4217 if (unlikely(skb->active_extensions)) {
4218 __skb_ext_put(skb->extensions);
4219 skb->active_extensions = 0;
4220 }
4221}
677bf08c
FW
4222
4223static inline bool skb_has_extensions(struct sk_buff *skb)
4224{
4225 return unlikely(skb->active_extensions);
4226}
df5042f4
FW
4227#else
4228static inline void skb_ext_put(struct sk_buff *skb) {}
174e2381 4229static inline void skb_ext_reset(struct sk_buff *skb) {}
df5042f4
FW
4230static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4231static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4232static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
677bf08c 4233static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
df5042f4
FW
4234#endif /* CONFIG_SKB_EXTENSIONS */
4235
895b5c9f 4236static inline void nf_reset_ct(struct sk_buff *skb)
a193a4ab 4237{
5f79e0f9 4238#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4239 nf_conntrack_put(skb_nfct(skb));
4240 skb->_nfct = 0;
2fc72c7b 4241#endif
a193a4ab
PM
4242}
4243
124dff01
PM
4244static inline void nf_reset_trace(struct sk_buff *skb)
4245{
478b360a 4246#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
4247 skb->nf_trace = 0;
4248#endif
a193a4ab
PM
4249}
4250
2b5ec1a5
YY
4251static inline void ipvs_reset(struct sk_buff *skb)
4252{
4253#if IS_ENABLED(CONFIG_IP_VS)
4254 skb->ipvs_property = 0;
4255#endif
4256}
4257
de8bda1d 4258/* Note: This doesn't put any conntrack info in dst. */
b1937227
ED
4259static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4260 bool copy)
edda553c 4261{
5f79e0f9 4262#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4263 dst->_nfct = src->_nfct;
4264 nf_conntrack_get(skb_nfct(src));
2fc72c7b 4265#endif
478b360a 4266#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
4267 if (copy)
4268 dst->nf_trace = src->nf_trace;
478b360a 4269#endif
edda553c
YK
4270}
4271
e7ac05f3
YK
4272static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4273{
e7ac05f3 4274#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 4275 nf_conntrack_put(skb_nfct(dst));
e7ac05f3 4276#endif
b1937227 4277 __nf_copy(dst, src, true);
e7ac05f3
YK
4278}
4279
984bc16c
JM
4280#ifdef CONFIG_NETWORK_SECMARK
4281static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4282{
4283 to->secmark = from->secmark;
4284}
4285
4286static inline void skb_init_secmark(struct sk_buff *skb)
4287{
4288 skb->secmark = 0;
4289}
4290#else
4291static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4292{ }
4293
4294static inline void skb_init_secmark(struct sk_buff *skb)
4295{ }
4296#endif
4297
7af8f4ca
FW
4298static inline int secpath_exists(const struct sk_buff *skb)
4299{
4300#ifdef CONFIG_XFRM
4165079b 4301 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
7af8f4ca
FW
4302#else
4303 return 0;
4304#endif
4305}
4306
574f7194
EB
4307static inline bool skb_irq_freeable(const struct sk_buff *skb)
4308{
4309 return !skb->destructor &&
7af8f4ca 4310 !secpath_exists(skb) &&
cb9c6836 4311 !skb_nfct(skb) &&
574f7194
EB
4312 !skb->_skb_refdst &&
4313 !skb_has_frag_list(skb);
4314}
4315
f25f4e44
PWJ
4316static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4317{
f25f4e44 4318 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
4319}
4320
9247744e 4321static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 4322{
4e3ab47a 4323 return skb->queue_mapping;
4e3ab47a
PE
4324}
4325
f25f4e44
PWJ
4326static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4327{
f25f4e44 4328 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
4329}
4330
d5a9e24a
DM
4331static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4332{
4333 skb->queue_mapping = rx_queue + 1;
4334}
4335
9247744e 4336static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
4337{
4338 return skb->queue_mapping - 1;
4339}
4340
9247744e 4341static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 4342{
a02cec21 4343 return skb->queue_mapping != 0;
d5a9e24a
DM
4344}
4345
4ff06203
JA
4346static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4347{
4348 skb->dst_pending_confirm = val;
4349}
4350
4351static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4352{
4353 return skb->dst_pending_confirm != 0;
4354}
4355
2294be0f 4356static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
def8b4fa 4357{
0b3d8e08 4358#ifdef CONFIG_XFRM
4165079b 4359 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
def8b4fa 4360#else
def8b4fa 4361 return NULL;
def8b4fa 4362#endif
0b3d8e08 4363}
def8b4fa 4364
68c33163
PS
4365/* Keeps track of mac header offset relative to skb->head.
4366 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4367 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
4368 * tunnel skb it points to outer mac header.
4369 * Keeps track of level of encapsulation of network headers.
4370 */
68c33163 4371struct skb_gso_cb {
802ab55a
AD
4372 union {
4373 int mac_offset;
4374 int data_offset;
4375 };
3347c960 4376 int encap_level;
76443456 4377 __wsum csum;
7e2b10c1 4378 __u16 csum_start;
68c33163 4379};
a08e7fd9
CZ
4380#define SKB_GSO_CB_OFFSET 32
4381#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
68c33163
PS
4382
4383static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4384{
4385 return (skb_mac_header(inner_skb) - inner_skb->head) -
4386 SKB_GSO_CB(inner_skb)->mac_offset;
4387}
4388
1e2bd517
PS
4389static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4390{
4391 int new_headroom, headroom;
4392 int ret;
4393
4394 headroom = skb_headroom(skb);
4395 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4396 if (ret)
4397 return ret;
4398
4399 new_headroom = skb_headroom(skb);
4400 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4401 return 0;
4402}
4403
08b64fcc
AD
4404static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4405{
4406 /* Do not update partial checksums if remote checksum is enabled. */
4407 if (skb->remcsum_offload)
4408 return;
4409
4410 SKB_GSO_CB(skb)->csum = res;
4411 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4412}
4413
7e2b10c1
TH
4414/* Compute the checksum for a gso segment. First compute the checksum value
4415 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4416 * then add in skb->csum (checksum from csum_start to end of packet).
4417 * skb->csum and csum_start are then updated to reflect the checksum of the
4418 * resultant packet starting from the transport header-- the resultant checksum
4419 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4420 * header.
4421 */
4422static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4423{
76443456
AD
4424 unsigned char *csum_start = skb_transport_header(skb);
4425 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4426 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 4427
76443456
AD
4428 SKB_GSO_CB(skb)->csum = res;
4429 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 4430
76443456 4431 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
4432}
4433
bdcc0924 4434static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
4435{
4436 return skb_shinfo(skb)->gso_size;
4437}
4438
36a8f39e 4439/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 4440static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
4441{
4442 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4443}
4444
d02f51cb
DA
4445/* Note: Should be called only if skb_is_gso(skb) is true */
4446static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4447{
4448 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4449}
4450
4c3024de 4451/* Note: Should be called only if skb_is_gso(skb) is true */
b90efd22
WB
4452static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4453{
4c3024de 4454 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
b90efd22
WB
4455}
4456
5293efe6
DB
4457static inline void skb_gso_reset(struct sk_buff *skb)
4458{
4459 skb_shinfo(skb)->gso_size = 0;
4460 skb_shinfo(skb)->gso_segs = 0;
4461 skb_shinfo(skb)->gso_type = 0;
4462}
4463
d02f51cb
DA
4464static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4465 u16 increment)
4466{
4467 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4468 return;
4469 shinfo->gso_size += increment;
4470}
4471
4472static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4473 u16 decrement)
4474{
4475 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4476 return;
4477 shinfo->gso_size -= decrement;
4478}
4479
7965bd4d 4480void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
4481
4482static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4483{
4484 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4485 * wanted then gso_type will be set. */
05bdd2f1
ED
4486 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4487
b78462eb
AD
4488 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4489 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
4490 __skb_warn_lro_forwarding(skb);
4491 return true;
4492 }
4493 return false;
4494}
4495
35fc92a9
HX
4496static inline void skb_forward_csum(struct sk_buff *skb)
4497{
4498 /* Unfortunately we don't support this one. Any brave souls? */
4499 if (skb->ip_summed == CHECKSUM_COMPLETE)
4500 skb->ip_summed = CHECKSUM_NONE;
4501}
4502
bc8acf2c
ED
4503/**
4504 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4505 * @skb: skb to check
4506 *
4507 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4508 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4509 * use this helper, to document places where we make this assertion.
4510 */
05bdd2f1 4511static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
4512{
4513#ifdef DEBUG
4514 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4515#endif
4516}
4517
f35d9d8a 4518bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 4519
ed1f50c3 4520int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
4521struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4522 unsigned int transport_len,
4523 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 4524
3a7c1ee4
AD
4525/**
4526 * skb_head_is_locked - Determine if the skb->head is locked down
4527 * @skb: skb to check
4528 *
4529 * The head on skbs build around a head frag can be removed if they are
4530 * not cloned. This function returns true if the skb head is locked down
4531 * due to either being allocated via kmalloc, or by being a clone with
4532 * multiple references to the head.
4533 */
4534static inline bool skb_head_is_locked(const struct sk_buff *skb)
4535{
4536 return !skb->head_frag || skb_cloned(skb);
4537}
fe6cc55f 4538
179bc67f
EC
4539/* Local Checksum Offload.
4540 * Compute outer checksum based on the assumption that the
4541 * inner checksum will be offloaded later.
d0dcde64 4542 * See Documentation/networking/checksum-offloads.rst for
e8ae7b00 4543 * explanation of how this works.
179bc67f
EC
4544 * Fill in outer checksum adjustment (e.g. with sum of outer
4545 * pseudo-header) before calling.
4546 * Also ensure that inner checksum is in linear data area.
4547 */
4548static inline __wsum lco_csum(struct sk_buff *skb)
4549{
9e74a6da
AD
4550 unsigned char *csum_start = skb_checksum_start(skb);
4551 unsigned char *l4_hdr = skb_transport_header(skb);
4552 __wsum partial;
179bc67f
EC
4553
4554 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
4555 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4556 skb->csum_offset));
4557
179bc67f 4558 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 4559 * adjustment filled in by caller) and return result.
179bc67f 4560 */
9e74a6da 4561 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
4562}
4563
2c64605b
PNA
4564static inline bool skb_is_redirected(const struct sk_buff *skb)
4565{
4566#ifdef CONFIG_NET_REDIRECT
4567 return skb->redirected;
4568#else
4569 return false;
4570#endif
4571}
4572
4573static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4574{
4575#ifdef CONFIG_NET_REDIRECT
4576 skb->redirected = 1;
4577 skb->from_ingress = from_ingress;
4578 if (skb->from_ingress)
4579 skb->tstamp = 0;
4580#endif
4581}
4582
4583static inline void skb_reset_redirect(struct sk_buff *skb)
4584{
4585#ifdef CONFIG_NET_REDIRECT
4586 skb->redirected = 0;
4587#endif
4588}
4589
1da177e4
LT
4590#endif /* __KERNEL__ */
4591#endif /* _LINUX_SKBUFF_H */