skbuff: constify skb_propagate_pfmemalloc() "page" argument
[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
645f0897 74 * IPv6|UDP where the Next Header field in the IPv6
7a6ae71b
TH
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)
29766bcf 369 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
c613c209
WB
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
6e3e939f 433 /* generate wifi status information (where possible) */
62b1a8ab 434 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4 435
e7fd2885
WB
436 /* generate software time stamp when entering packet scheduling */
437 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
438};
439
e1c8a607 440#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 441 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
442#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
443
06b4feb3
JL
444/* Definitions for flags in struct skb_shared_info */
445enum {
446 /* use zcopy routines */
447 SKBFL_ZEROCOPY_ENABLE = BIT(0),
448
449 /* This indicates at least one fragment might be overwritten
450 * (as in vmsplice(), sendfile() ...)
451 * If we need to compute a TX checksum, we'll need to copy
452 * all frags to avoid possible bad checksum
453 */
454 SKBFL_SHARED_FRAG = BIT(1),
455};
456
457#define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
458
a6686f2f
SM
459/*
460 * The callback notifies userspace to release buffers when skb DMA is done in
461 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
462 * The zerocopy_success argument is true if zero copy transmit occurred,
463 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
464 * The ctx field is used to track device context.
465 * The desc field is used to track userspace buffer index.
a6686f2f
SM
466 */
467struct ubuf_info {
36177832
JL
468 void (*callback)(struct sk_buff *, struct ubuf_info *,
469 bool zerocopy_success);
4ab6c99d
WB
470 union {
471 struct {
472 unsigned long desc;
473 void *ctx;
474 };
475 struct {
476 u32 id;
477 u16 len;
478 u16 zerocopy:1;
479 u32 bytelen;
480 };
481 };
c1d1b437 482 refcount_t refcnt;
04c2d33e 483 u8 flags;
a91dbff5
WB
484
485 struct mmpin {
486 struct user_struct *user;
487 unsigned int num_pg;
488 } mmp;
ac45f602
PO
489};
490
52267790
WB
491#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
492
6f89dbce
SV
493int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
494void mm_unaccount_pinned_pages(struct mmpin *mmp);
495
8c793822
JL
496struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size);
497struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
498 struct ubuf_info *uarg);
52267790 499
8c793822 500void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
52267790 501
8c793822
JL
502void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
503 bool success);
52267790 504
b5947e5d 505int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
52267790
WB
506int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
507 struct msghdr *msg, int len,
508 struct ubuf_info *uarg);
509
1da177e4
LT
510/* This data is invariant across clones and lives at
511 * the end of the header data, ie. at skb->end.
512 */
513struct skb_shared_info {
06b4feb3 514 __u8 flags;
de8f3a83
DB
515 __u8 meta_len;
516 __u8 nr_frags;
9f42f126 517 __u8 tx_flags;
7967168c
HX
518 unsigned short gso_size;
519 /* Warning: this field is not always filled in (UFO)! */
520 unsigned short gso_segs;
1da177e4 521 struct sk_buff *frag_list;
ac45f602 522 struct skb_shared_hwtstamps hwtstamps;
7f564528 523 unsigned int gso_type;
09c2d251 524 u32 tskey;
ec7d2f2c
ED
525
526 /*
527 * Warning : all fields before dataref are cleared in __alloc_skb()
528 */
529 atomic_t dataref;
530
69e3c75f
JB
531 /* Intermediate layers must ensure that destructor_arg
532 * remains valid until skb destructor */
533 void * destructor_arg;
a6686f2f 534
fed66381
ED
535 /* must be last field, see pskb_expand_head() */
536 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
537};
538
539/* We divide dataref into two halves. The higher 16 bits hold references
540 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
541 * the entire skb->data. A clone of a headerless skb holds the length of
542 * the header in skb->hdr_len.
1da177e4
LT
543 *
544 * All users must obey the rule that the skb->data reference count must be
545 * greater than or equal to the payload reference count.
546 *
547 * Holding a reference to the payload part means that the user does not
548 * care about modifications to the header part of skb->data.
549 */
550#define SKB_DATAREF_SHIFT 16
551#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
552
d179cd12
DM
553
554enum {
c8753d55
VS
555 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
556 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
557 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
558};
559
7967168c
HX
560enum {
561 SKB_GSO_TCPV4 = 1 << 0,
576a30eb
HX
562
563 /* This indicates the skb is from an untrusted source. */
d9d30adf 564 SKB_GSO_DODGY = 1 << 1,
b0da8537
MC
565
566 /* This indicates the tcp segment has CWR set. */
d9d30adf 567 SKB_GSO_TCP_ECN = 1 << 2,
f83ef8c0 568
d9d30adf 569 SKB_GSO_TCP_FIXEDID = 1 << 3,
01d5b2fc 570
d9d30adf 571 SKB_GSO_TCPV6 = 1 << 4,
68c33163 572
d9d30adf 573 SKB_GSO_FCOE = 1 << 5,
73136267 574
d9d30adf 575 SKB_GSO_GRE = 1 << 6,
0d89d203 576
d9d30adf 577 SKB_GSO_GRE_CSUM = 1 << 7,
cb32f511 578
d9d30adf 579 SKB_GSO_IPXIP4 = 1 << 8,
61c1db7f 580
d9d30adf 581 SKB_GSO_IPXIP6 = 1 << 9,
0f4f4ffa 582
d9d30adf 583 SKB_GSO_UDP_TUNNEL = 1 << 10,
4749c09c 584
d9d30adf 585 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
cbc53e08 586
d9d30adf 587 SKB_GSO_PARTIAL = 1 << 12,
802ab55a 588
d9d30adf 589 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
90017acc 590
d9d30adf 591 SKB_GSO_SCTP = 1 << 14,
c7ef8f0c 592
d9d30adf 593 SKB_GSO_ESP = 1 << 15,
0c19f846
WB
594
595 SKB_GSO_UDP = 1 << 16,
ee80d1eb
WB
596
597 SKB_GSO_UDP_L4 = 1 << 17,
3b335832
SK
598
599 SKB_GSO_FRAGLIST = 1 << 18,
7967168c
HX
600};
601
2e07fa9c
ACM
602#if BITS_PER_LONG > 32
603#define NET_SKBUFF_DATA_USES_OFFSET 1
604#endif
605
606#ifdef NET_SKBUFF_DATA_USES_OFFSET
607typedef unsigned int sk_buff_data_t;
608#else
609typedef unsigned char *sk_buff_data_t;
610#endif
611
161e6137 612/**
1da177e4
LT
613 * struct sk_buff - socket buffer
614 * @next: Next buffer in list
615 * @prev: Previous buffer in list
363ec392 616 * @tstamp: Time we arrived/left
d2f273f0
RD
617 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
618 * for retransmit timer
56b17425 619 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d2f273f0 620 * @list: queue head
d84e0bd7 621 * @sk: Socket we are owned by
d2f273f0
RD
622 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
623 * fragmentation management
1da177e4 624 * @dev: Device we arrived on/are leaving by
d2f273f0 625 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
d84e0bd7 626 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 627 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 628 * @sp: the security path, used for xfrm
1da177e4
LT
629 * @len: Length of actual data
630 * @data_len: Data length
631 * @mac_len: Length of link layer header
334a8132 632 * @hdr_len: writable header length of cloned skb
663ead3b
HX
633 * @csum: Checksum (must include start/offset pair)
634 * @csum_start: Offset from skb->head where checksumming should start
635 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 636 * @priority: Packet queueing priority
60ff7467 637 * @ignore_df: allow local fragmentation
1da177e4 638 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 639 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
640 * @nohdr: Payload reference only, must not modify header
641 * @pkt_type: Packet class
c83c2486 642 * @fclone: skbuff clone status
c83c2486 643 * @ipvs_property: skbuff is owned by ipvs
d2f273f0
RD
644 * @inner_protocol_type: whether the inner protocol is
645 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
646 * @remcsum_offload: remote checksum offload is enabled
875e8939
IS
647 * @offload_fwd_mark: Packet was L2-forwarded in hardware
648 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
e7246e12 649 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 650 * @tc_at_ingress: used within tc_classify to distinguish in/egress
2c64605b
PNA
651 * @redirected: packet was redirected by packet classifier
652 * @from_ingress: packet was redirected from the ingress path
31729363
RD
653 * @peeked: this packet has been seen already, so stats have been
654 * done for it, don't do them again
ba9dda3a 655 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
656 * @protocol: Packet protocol from driver
657 * @destructor: Destruct function
e2080072 658 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
a9e419dc 659 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 660 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 661 * @skb_iif: ifindex of device we arrived on
1da177e4 662 * @tc_index: Traffic control index
61b905da 663 * @hash: the packet hash
d84e0bd7 664 * @queue_mapping: Queue mapping for multiqueue devices
d2f273f0
RD
665 * @head_frag: skb was allocated from page fragments,
666 * not allocated by kmalloc() or vmalloc().
8b700862 667 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
df5042f4 668 * @active_extensions: active extensions (skb_ext_id types)
553a5672 669 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 670 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 671 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 672 * ports.
a3b18ddb 673 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
674 * @wifi_acked_valid: wifi_acked was set
675 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 676 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
d2f273f0
RD
677 * @encapsulation: indicates the inner headers in the skbuff are valid
678 * @encap_hdr_csum: software checksum is needed
679 * @csum_valid: checksum is already valid
dba00306 680 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
d2f273f0
RD
681 * @csum_complete_sw: checksum was completed by software
682 * @csum_level: indicates the number of consecutive checksums found in
683 * the packet minus one that have been verified as
684 * CHECKSUM_UNNECESSARY (max 3)
4ff06203 685 * @dst_pending_confirm: need to confirm neighbour
a48d189e 686 * @decrypted: Decrypted SKB
161e6137 687 * @napi_id: id of the NAPI struct this skb came from
d2f273f0 688 * @sender_cpu: (aka @napi_id) source CPU in XPS
984bc16c 689 * @secmark: security marking
d84e0bd7 690 * @mark: Generic packet mark
d2f273f0
RD
691 * @reserved_tailroom: (aka @mark) number of bytes of free space available
692 * at the tail of an sk_buff
693 * @vlan_present: VLAN tag is present
86a9bad3 694 * @vlan_proto: vlan encapsulation protocol
6aa895b0 695 * @vlan_tci: vlan tag control information
0d89d203 696 * @inner_protocol: Protocol (encapsulation)
d2f273f0
RD
697 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
698 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
6a674e9c
JG
699 * @inner_transport_header: Inner transport layer header (encapsulation)
700 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 701 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
702 * @transport_header: Transport layer header
703 * @network_header: Network layer header
704 * @mac_header: Link layer header
fa69ee5a 705 * @kcov_handle: KCOV remote handle for remote coverage collection
d84e0bd7
DB
706 * @tail: Tail pointer
707 * @end: End pointer
708 * @head: Head of buffer
709 * @data: Data head pointer
710 * @truesize: Buffer size
711 * @users: User count - see {datagram,tcp}.c
df5042f4 712 * @extensions: allocated extensions, valid if active_extensions is nonzero
1da177e4
LT
713 */
714
715struct sk_buff {
363ec392 716 union {
56b17425
ED
717 struct {
718 /* These two members must be first. */
719 struct sk_buff *next;
720 struct sk_buff *prev;
721
722 union {
bffa72cf
ED
723 struct net_device *dev;
724 /* Some protocols might use this space to store information,
725 * while device pointer would be NULL.
726 * UDP receive path is one user.
727 */
728 unsigned long dev_scratch;
56b17425
ED
729 };
730 };
fa0f5273 731 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
d4546c25 732 struct list_head list;
363ec392 733 };
fa0f5273
PO
734
735 union {
736 struct sock *sk;
737 int ip_defrag_offset;
738 };
1da177e4 739
c84d9490 740 union {
bffa72cf 741 ktime_t tstamp;
d3edd06e 742 u64 skb_mstamp_ns; /* earliest departure time */
c84d9490 743 };
1da177e4
LT
744 /*
745 * This is the control buffer. It is free to use for every
746 * layer. Please put your private variables there. If you
747 * want to keep them across layers you have to do a skb_clone()
748 * first. This is owned by whoever has the skb queued ATM.
749 */
da3f5cf1 750 char cb[48] __aligned(8);
1da177e4 751
e2080072
ED
752 union {
753 struct {
754 unsigned long _skb_refdst;
755 void (*destructor)(struct sk_buff *skb);
756 };
757 struct list_head tcp_tsorted_anchor;
758 };
759
b1937227 760#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 761 unsigned long _nfct;
da3f5cf1 762#endif
1da177e4 763 unsigned int len,
334a8132
PM
764 data_len;
765 __u16 mac_len,
766 hdr_len;
b1937227
ED
767
768 /* Following fields are _not_ copied in __copy_skb_header()
769 * Note that queue_mapping is here mostly to fill a hole.
770 */
b1937227 771 __u16 queue_mapping;
36bbef52
DB
772
773/* if you move cloned around you also must adapt those constants */
774#ifdef __BIG_ENDIAN_BITFIELD
775#define CLONED_MASK (1 << 7)
776#else
777#define CLONED_MASK 1
778#endif
779#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
780
d2f273f0 781 /* private: */
36bbef52 782 __u8 __cloned_offset[0];
d2f273f0 783 /* public: */
b1937227 784 __u8 cloned:1,
6869c4d8 785 nohdr:1,
b84f4cc9 786 fclone:2,
a59322be 787 peeked:1,
b1937227 788 head_frag:1,
8b700862 789 pfmemalloc:1;
df5042f4
FW
790#ifdef CONFIG_SKB_EXTENSIONS
791 __u8 active_extensions;
792#endif
b1937227
ED
793 /* fields enclosed in headers_start/headers_end are copied
794 * using a single memcpy() in __copy_skb_header()
795 */
ebcf34f3 796 /* private: */
b1937227 797 __u32 headers_start[0];
ebcf34f3 798 /* public: */
4031ae6e 799
233577a2
HFS
800/* if you move pkt_type around you also must adapt those constants */
801#ifdef __BIG_ENDIAN_BITFIELD
802#define PKT_TYPE_MAX (7 << 5)
803#else
804#define PKT_TYPE_MAX 7
1da177e4 805#endif
233577a2 806#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 807
d2f273f0 808 /* private: */
233577a2 809 __u8 __pkt_type_offset[0];
d2f273f0 810 /* public: */
b1937227 811 __u8 pkt_type:3;
b1937227 812 __u8 ignore_df:1;
b1937227
ED
813 __u8 nf_trace:1;
814 __u8 ip_summed:2;
3853b584 815 __u8 ooo_okay:1;
8b700862 816
61b905da 817 __u8 l4_hash:1;
a3b18ddb 818 __u8 sw_hash:1;
6e3e939f
JB
819 __u8 wifi_acked_valid:1;
820 __u8 wifi_acked:1;
3bdc0eba 821 __u8 no_fcs:1;
77cffe23 822 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 823 __u8 encapsulation:1;
7e2b10c1 824 __u8 encap_hdr_csum:1;
5d0c2b95 825 __u8 csum_valid:1;
8b700862 826
0c4b2d37
MM
827#ifdef __BIG_ENDIAN_BITFIELD
828#define PKT_VLAN_PRESENT_BIT 7
829#else
830#define PKT_VLAN_PRESENT_BIT 0
831#endif
832#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
d2f273f0 833 /* private: */
0c4b2d37 834 __u8 __pkt_vlan_present_offset[0];
d2f273f0 835 /* public: */
0c4b2d37 836 __u8 vlan_present:1;
7e3cead5 837 __u8 csum_complete_sw:1;
b1937227 838 __u8 csum_level:2;
dba00306 839 __u8 csum_not_inet:1;
4ff06203 840 __u8 dst_pending_confirm:1;
b1937227
ED
841#ifdef CONFIG_IPV6_NDISC_NODETYPE
842 __u8 ndisc_nodetype:2;
843#endif
8b700862 844
0c4b2d37 845 __u8 ipvs_property:1;
8bce6d7d 846 __u8 inner_protocol_type:1;
e585f236 847 __u8 remcsum_offload:1;
6bc506b4
IS
848#ifdef CONFIG_NET_SWITCHDEV
849 __u8 offload_fwd_mark:1;
875e8939 850 __u8 offload_l3_fwd_mark:1;
6bc506b4 851#endif
e7246e12
WB
852#ifdef CONFIG_NET_CLS_ACT
853 __u8 tc_skip_classify:1;
8dc07fdb 854 __u8 tc_at_ingress:1;
2c64605b
PNA
855#endif
856#ifdef CONFIG_NET_REDIRECT
857 __u8 redirected:1;
858 __u8 from_ingress:1;
e7246e12 859#endif
a48d189e
SB
860#ifdef CONFIG_TLS_DEVICE
861 __u8 decrypted:1;
862#endif
b1937227
ED
863
864#ifdef CONFIG_NET_SCHED
865 __u16 tc_index; /* traffic control index */
b1937227 866#endif
fe55f6d5 867
b1937227
ED
868 union {
869 __wsum csum;
870 struct {
871 __u16 csum_start;
872 __u16 csum_offset;
873 };
874 };
875 __u32 priority;
876 int skb_iif;
877 __u32 hash;
878 __be16 vlan_proto;
879 __u16 vlan_tci;
2bd82484
ED
880#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
881 union {
882 unsigned int napi_id;
883 unsigned int sender_cpu;
884 };
97fc2f08 885#endif
984bc16c 886#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 887 __u32 secmark;
0c4f691f 888#endif
0c4f691f 889
3b885787
NH
890 union {
891 __u32 mark;
16fad69c 892 __u32 reserved_tailroom;
3b885787 893 };
1da177e4 894
8bce6d7d
TH
895 union {
896 __be16 inner_protocol;
897 __u8 inner_ipproto;
898 };
899
1a37e412
SH
900 __u16 inner_transport_header;
901 __u16 inner_network_header;
902 __u16 inner_mac_header;
b1937227
ED
903
904 __be16 protocol;
1a37e412
SH
905 __u16 transport_header;
906 __u16 network_header;
907 __u16 mac_header;
b1937227 908
fa69ee5a
ME
909#ifdef CONFIG_KCOV
910 u64 kcov_handle;
911#endif
912
ebcf34f3 913 /* private: */
b1937227 914 __u32 headers_end[0];
ebcf34f3 915 /* public: */
b1937227 916
1da177e4 917 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 918 sk_buff_data_t tail;
4305b541 919 sk_buff_data_t end;
1da177e4 920 unsigned char *head,
4305b541 921 *data;
27a884dc 922 unsigned int truesize;
63354797 923 refcount_t users;
df5042f4
FW
924
925#ifdef CONFIG_SKB_EXTENSIONS
926 /* only useable after checking ->active_extensions != 0 */
927 struct skb_ext *extensions;
928#endif
1da177e4
LT
929};
930
931#ifdef __KERNEL__
932/*
933 * Handling routines are only of interest to the kernel
934 */
1da177e4 935
c93bdd0e
MG
936#define SKB_ALLOC_FCLONE 0x01
937#define SKB_ALLOC_RX 0x02
fd11a83d 938#define SKB_ALLOC_NAPI 0x04
c93bdd0e 939
161e6137
PT
940/**
941 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
942 * @skb: buffer
943 */
c93bdd0e
MG
944static inline bool skb_pfmemalloc(const struct sk_buff *skb)
945{
946 return unlikely(skb->pfmemalloc);
947}
948
7fee226a
ED
949/*
950 * skb might have a dst pointer attached, refcounted or not.
951 * _skb_refdst low order bit is set if refcount was _not_ taken
952 */
953#define SKB_DST_NOREF 1UL
954#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
955
956/**
957 * skb_dst - returns skb dst_entry
958 * @skb: buffer
959 *
960 * Returns skb dst_entry, regardless of reference taken or not.
961 */
adf30907
ED
962static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
963{
161e6137 964 /* If refdst was not refcounted, check we still are in a
7fee226a
ED
965 * rcu_read_lock section
966 */
967 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
968 !rcu_read_lock_held() &&
969 !rcu_read_lock_bh_held());
970 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
971}
972
7fee226a
ED
973/**
974 * skb_dst_set - sets skb dst
975 * @skb: buffer
976 * @dst: dst entry
977 *
978 * Sets skb dst, assuming a reference was taken on dst and should
979 * be released by skb_dst_drop()
980 */
adf30907
ED
981static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
982{
7fee226a
ED
983 skb->_skb_refdst = (unsigned long)dst;
984}
985
932bc4d7
JA
986/**
987 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
988 * @skb: buffer
989 * @dst: dst entry
990 *
991 * Sets skb dst, assuming a reference was not taken on dst.
992 * If dst entry is cached, we do not take reference and dst_release
993 * will be avoided by refdst_drop. If dst entry is not cached, we take
994 * reference, so that last dst_release can destroy the dst immediately.
995 */
996static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
997{
dbfc4fb7
HFS
998 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
999 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 1000}
7fee226a
ED
1001
1002/**
25985edc 1003 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
1004 * @skb: buffer
1005 */
1006static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1007{
1008 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
1009}
1010
161e6137
PT
1011/**
1012 * skb_rtable - Returns the skb &rtable
1013 * @skb: buffer
1014 */
511c3f92
ED
1015static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1016{
adf30907 1017 return (struct rtable *)skb_dst(skb);
511c3f92
ED
1018}
1019
8b10cab6
JHS
1020/* For mangling skb->pkt_type from user space side from applications
1021 * such as nft, tc, etc, we only allow a conservative subset of
1022 * possible pkt_types to be set.
1023*/
1024static inline bool skb_pkt_type_ok(u32 ptype)
1025{
1026 return ptype <= PACKET_OTHERHOST;
1027}
1028
161e6137
PT
1029/**
1030 * skb_napi_id - Returns the skb's NAPI id
1031 * @skb: buffer
1032 */
90b602f8
ML
1033static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1034{
1035#ifdef CONFIG_NET_RX_BUSY_POLL
1036 return skb->napi_id;
1037#else
1038 return 0;
1039#endif
1040}
1041
161e6137
PT
1042/**
1043 * skb_unref - decrement the skb's reference count
1044 * @skb: buffer
1045 *
1046 * Returns true if we can free the skb.
1047 */
3889a803
PA
1048static inline bool skb_unref(struct sk_buff *skb)
1049{
1050 if (unlikely(!skb))
1051 return false;
63354797 1052 if (likely(refcount_read(&skb->users) == 1))
3889a803 1053 smp_rmb();
63354797 1054 else if (likely(!refcount_dec_and_test(&skb->users)))
3889a803
PA
1055 return false;
1056
1057 return true;
1058}
1059
0a463c78 1060void skb_release_head_state(struct sk_buff *skb);
7965bd4d
JP
1061void kfree_skb(struct sk_buff *skb);
1062void kfree_skb_list(struct sk_buff *segs);
6413139d 1063void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
7965bd4d 1064void skb_tx_error(struct sk_buff *skb);
be769db2
HX
1065
1066#ifdef CONFIG_TRACEPOINTS
7965bd4d 1067void consume_skb(struct sk_buff *skb);
be769db2
HX
1068#else
1069static inline void consume_skb(struct sk_buff *skb)
1070{
1071 return kfree_skb(skb);
1072}
1073#endif
1074
ca2c1418 1075void __consume_stateless_skb(struct sk_buff *skb);
7965bd4d 1076void __kfree_skb(struct sk_buff *skb);
d7e8883c 1077extern struct kmem_cache *skbuff_head_cache;
bad43ca8 1078
7965bd4d
JP
1079void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1080bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1081 bool *fragstolen, int *delta_truesize);
bad43ca8 1082
7965bd4d
JP
1083struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1084 int node);
2ea2f62c 1085struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 1086struct sk_buff *build_skb(void *data, unsigned int frag_size);
ba0509b6
JDB
1087struct sk_buff *build_skb_around(struct sk_buff *skb,
1088 void *data, unsigned int frag_size);
161e6137
PT
1089
1090/**
1091 * alloc_skb - allocate a network buffer
1092 * @size: size to allocate
1093 * @priority: allocation mask
1094 *
1095 * This function is a convenient wrapper around __alloc_skb().
1096 */
d179cd12 1097static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 1098 gfp_t priority)
d179cd12 1099{
564824b0 1100 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
1101}
1102
2e4e4410
ED
1103struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1104 unsigned long data_len,
1105 int max_page_order,
1106 int *errcode,
1107 gfp_t gfp_mask);
da29e4b4 1108struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
2e4e4410 1109
d0bf4a9e
ED
1110/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1111struct sk_buff_fclones {
1112 struct sk_buff skb1;
1113
1114 struct sk_buff skb2;
1115
2638595a 1116 refcount_t fclone_ref;
d0bf4a9e
ED
1117};
1118
1119/**
1120 * skb_fclone_busy - check if fclone is busy
293de7de 1121 * @sk: socket
d0bf4a9e
ED
1122 * @skb: buffer
1123 *
bda13fed 1124 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
1125 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1126 * so we also check that this didnt happen.
d0bf4a9e 1127 */
39bb5e62
ED
1128static inline bool skb_fclone_busy(const struct sock *sk,
1129 const struct sk_buff *skb)
d0bf4a9e
ED
1130{
1131 const struct sk_buff_fclones *fclones;
1132
1133 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1134
1135 return skb->fclone == SKB_FCLONE_ORIG &&
2638595a 1136 refcount_read(&fclones->fclone_ref) > 1 &&
39bb5e62 1137 fclones->skb2.sk == sk;
d0bf4a9e
ED
1138}
1139
161e6137
PT
1140/**
1141 * alloc_skb_fclone - allocate a network buffer from fclone cache
1142 * @size: size to allocate
1143 * @priority: allocation mask
1144 *
1145 * This function is a convenient wrapper around __alloc_skb().
1146 */
d179cd12 1147static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 1148 gfp_t priority)
d179cd12 1149{
c93bdd0e 1150 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
1151}
1152
7965bd4d 1153struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
b0768a86 1154void skb_headers_offset_update(struct sk_buff *skb, int off);
7965bd4d
JP
1155int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1156struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
08303c18 1157void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
7965bd4d 1158struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
1159struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1160 gfp_t gfp_mask, bool fclone);
1161static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1162 gfp_t gfp_mask)
1163{
1164 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1165}
7965bd4d
JP
1166
1167int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1168struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1169 unsigned int headroom);
1170struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1171 int newtailroom, gfp_t priority);
48a1df65
JD
1172int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1173 int offset, int len);
1174int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1175 int offset, int len);
7965bd4d 1176int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
cd0a137a
FF
1177int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1178
1179/**
1180 * skb_pad - zero pad the tail of an skb
1181 * @skb: buffer to pad
1182 * @pad: space to pad
1183 *
1184 * Ensure that a buffer is followed by a padding area that is zero
1185 * filled. Used by network drivers which may DMA or transfer data
1186 * beyond the buffer end onto the wire.
1187 *
1188 * May return error in out of memory cases. The skb is freed on error.
1189 */
1190static inline int skb_pad(struct sk_buff *skb, int pad)
1191{
1192 return __skb_pad(skb, pad, true);
1193}
ead2ceb0 1194#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1195
be12a1fe
HFS
1196int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1197 int offset, size_t size);
1198
d94d9fee 1199struct skb_seq_state {
677e90ed
TG
1200 __u32 lower_offset;
1201 __u32 upper_offset;
1202 __u32 frag_idx;
1203 __u32 stepped_offset;
1204 struct sk_buff *root_skb;
1205 struct sk_buff *cur_skb;
1206 __u8 *frag_data;
97550f6f 1207 __u32 frag_off;
677e90ed
TG
1208};
1209
7965bd4d
JP
1210void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1211 unsigned int to, struct skb_seq_state *st);
1212unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1213 struct skb_seq_state *st);
1214void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1215
7965bd4d 1216unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1217 unsigned int to, struct ts_config *config);
3fc7e8a6 1218
09323cc4
TH
1219/*
1220 * Packet hash types specify the type of hash in skb_set_hash.
1221 *
1222 * Hash types refer to the protocol layer addresses which are used to
1223 * construct a packet's hash. The hashes are used to differentiate or identify
1224 * flows of the protocol layer for the hash type. Hash types are either
1225 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1226 *
1227 * Properties of hashes:
1228 *
1229 * 1) Two packets in different flows have different hash values
1230 * 2) Two packets in the same flow should have the same hash value
1231 *
1232 * A hash at a higher layer is considered to be more specific. A driver should
1233 * set the most specific hash possible.
1234 *
1235 * A driver cannot indicate a more specific hash than the layer at which a hash
1236 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1237 *
1238 * A driver may indicate a hash level which is less specific than the
1239 * actual layer the hash was computed on. For instance, a hash computed
1240 * at L4 may be considered an L3 hash. This should only be done if the
1241 * driver can't unambiguously determine that the HW computed the hash at
1242 * the higher layer. Note that the "should" in the second property above
1243 * permits this.
1244 */
1245enum pkt_hash_types {
1246 PKT_HASH_TYPE_NONE, /* Undefined type */
1247 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1248 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1249 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1250};
1251
bcc83839 1252static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1253{
bcc83839 1254 skb->hash = 0;
a3b18ddb 1255 skb->sw_hash = 0;
bcc83839
TH
1256 skb->l4_hash = 0;
1257}
1258
1259static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1260{
1261 if (!skb->l4_hash)
1262 skb_clear_hash(skb);
1263}
1264
1265static inline void
1266__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1267{
1268 skb->l4_hash = is_l4;
1269 skb->sw_hash = is_sw;
61b905da 1270 skb->hash = hash;
09323cc4
TH
1271}
1272
bcc83839
TH
1273static inline void
1274skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1275{
1276 /* Used by drivers to set hash from HW */
1277 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1278}
1279
1280static inline void
1281__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1282{
1283 __skb_set_hash(skb, hash, true, is_l4);
1284}
1285
e5276937 1286void __skb_get_hash(struct sk_buff *skb);
b917783c 1287u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937
TH
1288u32 skb_get_poff(const struct sk_buff *skb);
1289u32 __skb_get_poff(const struct sk_buff *skb, void *data,
72a338bc 1290 const struct flow_keys_basic *keys, int hlen);
e5276937
TH
1291__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1292 void *data, int hlen_proto);
1293
1294static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1295 int thoff, u8 ip_proto)
1296{
1297 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1298}
1299
1300void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1301 const struct flow_dissector_key *key,
1302 unsigned int key_count);
1303
089b19a9
SF
1304struct bpf_flow_dissector;
1305bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
086f9568 1306 __be16 proto, int nhoff, int hlen, unsigned int flags);
089b19a9 1307
3cbf4ffb
SF
1308bool __skb_flow_dissect(const struct net *net,
1309 const struct sk_buff *skb,
e5276937
TH
1310 struct flow_dissector *flow_dissector,
1311 void *target_container,
cd79a238
TH
1312 void *data, __be16 proto, int nhoff, int hlen,
1313 unsigned int flags);
e5276937
TH
1314
1315static inline bool skb_flow_dissect(const struct sk_buff *skb,
1316 struct flow_dissector *flow_dissector,
cd79a238 1317 void *target_container, unsigned int flags)
e5276937 1318{
3cbf4ffb
SF
1319 return __skb_flow_dissect(NULL, skb, flow_dissector,
1320 target_container, NULL, 0, 0, 0, flags);
e5276937
TH
1321}
1322
1323static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1324 struct flow_keys *flow,
1325 unsigned int flags)
e5276937
TH
1326{
1327 memset(flow, 0, sizeof(*flow));
3cbf4ffb
SF
1328 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1329 flow, NULL, 0, 0, 0, flags);
e5276937
TH
1330}
1331
72a338bc 1332static inline bool
3cbf4ffb
SF
1333skb_flow_dissect_flow_keys_basic(const struct net *net,
1334 const struct sk_buff *skb,
72a338bc
PA
1335 struct flow_keys_basic *flow, void *data,
1336 __be16 proto, int nhoff, int hlen,
1337 unsigned int flags)
e5276937
TH
1338{
1339 memset(flow, 0, sizeof(*flow));
3cbf4ffb 1340 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
cd79a238 1341 data, proto, nhoff, hlen, flags);
e5276937
TH
1342}
1343
82828b88
JP
1344void skb_flow_dissect_meta(const struct sk_buff *skb,
1345 struct flow_dissector *flow_dissector,
1346 void *target_container);
1347
75a56758 1348/* Gets a skb connection tracking info, ctinfo map should be a
2ff17117 1349 * map of mapsize to translate enum ip_conntrack_info states
75a56758
PB
1350 * to user states.
1351 */
1352void
1353skb_flow_dissect_ct(const struct sk_buff *skb,
1354 struct flow_dissector *flow_dissector,
1355 void *target_container,
7baf2429 1356 u16 *ctinfo_map, size_t mapsize,
1357 bool post_ct);
62b32379
SH
1358void
1359skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1360 struct flow_dissector *flow_dissector,
1361 void *target_container);
1362
0cb09aff
AL
1363void skb_flow_dissect_hash(const struct sk_buff *skb,
1364 struct flow_dissector *flow_dissector,
1365 void *target_container);
1366
3958afa1 1367static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1368{
a3b18ddb 1369 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1370 __skb_get_hash(skb);
bfb564e7 1371
61b905da 1372 return skb->hash;
bfb564e7
KK
1373}
1374
20a17bf6 1375static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1376{
c6cc1ca7
TH
1377 if (!skb->l4_hash && !skb->sw_hash) {
1378 struct flow_keys keys;
de4c1f8b 1379 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1380
de4c1f8b 1381 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1382 }
f70ea018
TH
1383
1384 return skb->hash;
1385}
1386
55667441
ED
1387__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1388 const siphash_key_t *perturb);
50fb7992 1389
57bdf7f4
TH
1390static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1391{
61b905da 1392 return skb->hash;
57bdf7f4
TH
1393}
1394
3df7a74e
TH
1395static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1396{
61b905da 1397 to->hash = from->hash;
a3b18ddb 1398 to->sw_hash = from->sw_hash;
61b905da 1399 to->l4_hash = from->l4_hash;
3df7a74e
TH
1400};
1401
41477662
JK
1402static inline void skb_copy_decrypted(struct sk_buff *to,
1403 const struct sk_buff *from)
1404{
1405#ifdef CONFIG_TLS_DEVICE
1406 to->decrypted = from->decrypted;
1407#endif
1408}
1409
4305b541
ACM
1410#ifdef NET_SKBUFF_DATA_USES_OFFSET
1411static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1412{
1413 return skb->head + skb->end;
1414}
ec47ea82
AD
1415
1416static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1417{
1418 return skb->end;
1419}
4305b541
ACM
1420#else
1421static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1422{
1423 return skb->end;
1424}
ec47ea82
AD
1425
1426static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1427{
1428 return skb->end - skb->head;
1429}
4305b541
ACM
1430#endif
1431
1da177e4 1432/* Internal */
4305b541 1433#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1434
ac45f602
PO
1435static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1436{
1437 return &skb_shinfo(skb)->hwtstamps;
1438}
1439
52267790
WB
1440static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1441{
06b4feb3 1442 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
52267790
WB
1443
1444 return is_zcopy ? skb_uarg(skb) : NULL;
1445}
1446
8e044917 1447static inline void net_zcopy_get(struct ubuf_info *uarg)
e76d46cf
JL
1448{
1449 refcount_inc(&uarg->refcnt);
1450}
1451
9ee5e5ad
JL
1452static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1453{
1454 skb_shinfo(skb)->destructor_arg = uarg;
1455 skb_shinfo(skb)->flags |= uarg->flags;
1456}
1457
52900d22
WB
1458static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1459 bool *have_ref)
52267790
WB
1460{
1461 if (skb && uarg && !skb_zcopy(skb)) {
52900d22
WB
1462 if (unlikely(have_ref && *have_ref))
1463 *have_ref = false;
1464 else
8e044917 1465 net_zcopy_get(uarg);
9ee5e5ad 1466 skb_zcopy_init(skb, uarg);
52267790
WB
1467 }
1468}
1469
5cd8d46e
WB
1470static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1471{
1472 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
06b4feb3 1473 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
5cd8d46e
WB
1474}
1475
1476static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1477{
1478 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1479}
1480
1481static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1482{
1483 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1484}
1485
8e044917 1486static inline void net_zcopy_put(struct ubuf_info *uarg)
59776362
JL
1487{
1488 if (uarg)
36177832 1489 uarg->callback(NULL, uarg, true);
59776362
JL
1490}
1491
8e044917 1492static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
236a6b1c
JL
1493{
1494 if (uarg) {
8c793822
JL
1495 if (uarg->callback == msg_zerocopy_callback)
1496 msg_zerocopy_put_abort(uarg, have_uref);
236a6b1c 1497 else if (have_uref)
8e044917 1498 net_zcopy_put(uarg);
236a6b1c
JL
1499 }
1500}
1501
52267790 1502/* Release a reference on a zerocopy structure */
36177832 1503static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
52267790
WB
1504{
1505 struct ubuf_info *uarg = skb_zcopy(skb);
1506
1507 if (uarg) {
36177832
JL
1508 if (!skb_zcopy_is_nouarg(skb))
1509 uarg->callback(skb, uarg, zerocopy_success);
0a4a060b 1510
06b4feb3 1511 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG;
52267790
WB
1512 }
1513}
1514
a8305bff
DM
1515static inline void skb_mark_not_on_list(struct sk_buff *skb)
1516{
1517 skb->next = NULL;
1518}
1519
dcfea72e 1520/* Iterate through singly-linked GSO fragments of an skb. */
5eee7bd7
JD
1521#define skb_list_walk_safe(first, skb, next_skb) \
1522 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1523 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
dcfea72e 1524
992cba7e
DM
1525static inline void skb_list_del_init(struct sk_buff *skb)
1526{
1527 __list_del_entry(&skb->list);
1528 skb_mark_not_on_list(skb);
1529}
1530
1da177e4
LT
1531/**
1532 * skb_queue_empty - check if a queue is empty
1533 * @list: queue head
1534 *
1535 * Returns true if the queue is empty, false otherwise.
1536 */
1537static inline int skb_queue_empty(const struct sk_buff_head *list)
1538{
fd44b93c 1539 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1540}
1541
d7d16a89
ED
1542/**
1543 * skb_queue_empty_lockless - check if a queue is empty
1544 * @list: queue head
1545 *
1546 * Returns true if the queue is empty, false otherwise.
1547 * This variant can be used in lockless contexts.
1548 */
1549static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1550{
1551 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1552}
1553
1554
fc7ebb21
DM
1555/**
1556 * skb_queue_is_last - check if skb is the last entry in the queue
1557 * @list: queue head
1558 * @skb: buffer
1559 *
1560 * Returns true if @skb is the last buffer on the list.
1561 */
1562static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1563 const struct sk_buff *skb)
1564{
fd44b93c 1565 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1566}
1567
832d11c5
IJ
1568/**
1569 * skb_queue_is_first - check if skb is the first entry in the queue
1570 * @list: queue head
1571 * @skb: buffer
1572 *
1573 * Returns true if @skb is the first buffer on the list.
1574 */
1575static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1576 const struct sk_buff *skb)
1577{
fd44b93c 1578 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1579}
1580
249c8b42
DM
1581/**
1582 * skb_queue_next - return the next packet in the queue
1583 * @list: queue head
1584 * @skb: current buffer
1585 *
1586 * Return the next packet in @list after @skb. It is only valid to
1587 * call this if skb_queue_is_last() evaluates to false.
1588 */
1589static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1590 const struct sk_buff *skb)
1591{
1592 /* This BUG_ON may seem severe, but if we just return then we
1593 * are going to dereference garbage.
1594 */
1595 BUG_ON(skb_queue_is_last(list, skb));
1596 return skb->next;
1597}
1598
832d11c5
IJ
1599/**
1600 * skb_queue_prev - return the prev packet in the queue
1601 * @list: queue head
1602 * @skb: current buffer
1603 *
1604 * Return the prev packet in @list before @skb. It is only valid to
1605 * call this if skb_queue_is_first() evaluates to false.
1606 */
1607static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1608 const struct sk_buff *skb)
1609{
1610 /* This BUG_ON may seem severe, but if we just return then we
1611 * are going to dereference garbage.
1612 */
1613 BUG_ON(skb_queue_is_first(list, skb));
1614 return skb->prev;
1615}
1616
1da177e4
LT
1617/**
1618 * skb_get - reference buffer
1619 * @skb: buffer to reference
1620 *
1621 * Makes another reference to a socket buffer and returns a pointer
1622 * to the buffer.
1623 */
1624static inline struct sk_buff *skb_get(struct sk_buff *skb)
1625{
63354797 1626 refcount_inc(&skb->users);
1da177e4
LT
1627 return skb;
1628}
1629
1630/*
f8821f96 1631 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1da177e4
LT
1632 */
1633
1da177e4
LT
1634/**
1635 * skb_cloned - is the buffer a clone
1636 * @skb: buffer to check
1637 *
1638 * Returns true if the buffer was generated with skb_clone() and is
1639 * one of multiple shared copies of the buffer. Cloned buffers are
1640 * shared data so must not be written to under normal circumstances.
1641 */
1642static inline int skb_cloned(const struct sk_buff *skb)
1643{
1644 return skb->cloned &&
1645 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1646}
1647
14bbd6a5
PS
1648static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1649{
d0164adc 1650 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1651
1652 if (skb_cloned(skb))
1653 return pskb_expand_head(skb, 0, 0, pri);
1654
1655 return 0;
1656}
1657
1da177e4
LT
1658/**
1659 * skb_header_cloned - is the header a clone
1660 * @skb: buffer to check
1661 *
1662 * Returns true if modifying the header part of the buffer requires
1663 * the data to be copied.
1664 */
1665static inline int skb_header_cloned(const struct sk_buff *skb)
1666{
1667 int dataref;
1668
1669 if (!skb->cloned)
1670 return 0;
1671
1672 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1673 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1674 return dataref != 1;
1675}
1676
9580bf2e
ED
1677static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1678{
1679 might_sleep_if(gfpflags_allow_blocking(pri));
1680
1681 if (skb_header_cloned(skb))
1682 return pskb_expand_head(skb, 0, 0, pri);
1683
1684 return 0;
1685}
1686
f4a775d1
ED
1687/**
1688 * __skb_header_release - release reference to header
1689 * @skb: buffer to operate on
f4a775d1
ED
1690 */
1691static inline void __skb_header_release(struct sk_buff *skb)
1692{
1693 skb->nohdr = 1;
1694 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1695}
1696
1697
1da177e4
LT
1698/**
1699 * skb_shared - is the buffer shared
1700 * @skb: buffer to check
1701 *
1702 * Returns true if more than one person has a reference to this
1703 * buffer.
1704 */
1705static inline int skb_shared(const struct sk_buff *skb)
1706{
63354797 1707 return refcount_read(&skb->users) != 1;
1da177e4
LT
1708}
1709
1710/**
1711 * skb_share_check - check if buffer is shared and if so clone it
1712 * @skb: buffer to check
1713 * @pri: priority for memory allocation
1714 *
1715 * If the buffer is shared the buffer is cloned and the old copy
1716 * drops a reference. A new clone with a single reference is returned.
1717 * If the buffer is not shared the original buffer is returned. When
1718 * being called from interrupt status or with spinlocks held pri must
1719 * be GFP_ATOMIC.
1720 *
1721 * NULL is returned on a memory allocation failure.
1722 */
47061bc4 1723static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1724{
d0164adc 1725 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1726 if (skb_shared(skb)) {
1727 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1728
1729 if (likely(nskb))
1730 consume_skb(skb);
1731 else
1732 kfree_skb(skb);
1da177e4
LT
1733 skb = nskb;
1734 }
1735 return skb;
1736}
1737
1738/*
1739 * Copy shared buffers into a new sk_buff. We effectively do COW on
1740 * packets to handle cases where we have a local reader and forward
1741 * and a couple of other messy ones. The normal one is tcpdumping
1742 * a packet thats being forwarded.
1743 */
1744
1745/**
1746 * skb_unshare - make a copy of a shared buffer
1747 * @skb: buffer to check
1748 * @pri: priority for memory allocation
1749 *
1750 * If the socket buffer is a clone then this function creates a new
1751 * copy of the data, drops a reference count on the old copy and returns
1752 * the new copy with the reference count at 1. If the buffer is not a clone
1753 * the original buffer is returned. When called with a spinlock held or
1754 * from interrupt state @pri must be %GFP_ATOMIC
1755 *
1756 * %NULL is returned on a memory allocation failure.
1757 */
e2bf521d 1758static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1759 gfp_t pri)
1da177e4 1760{
d0164adc 1761 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1762 if (skb_cloned(skb)) {
1763 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1764
1765 /* Free our shared copy */
1766 if (likely(nskb))
1767 consume_skb(skb);
1768 else
1769 kfree_skb(skb);
1da177e4
LT
1770 skb = nskb;
1771 }
1772 return skb;
1773}
1774
1775/**
1a5778aa 1776 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1777 * @list_: list to peek at
1778 *
1779 * Peek an &sk_buff. Unlike most other operations you _MUST_
1780 * be careful with this one. A peek leaves the buffer on the
1781 * list and someone else may run off with it. You must hold
1782 * the appropriate locks or have a private queue to do this.
1783 *
1784 * Returns %NULL for an empty list or a pointer to the head element.
1785 * The reference count is not incremented and the reference is therefore
1786 * volatile. Use with caution.
1787 */
05bdd2f1 1788static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1789{
18d07000
ED
1790 struct sk_buff *skb = list_->next;
1791
1792 if (skb == (struct sk_buff *)list_)
1793 skb = NULL;
1794 return skb;
1da177e4
LT
1795}
1796
8b69bd7d
DM
1797/**
1798 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1799 * @list_: list to peek at
1800 *
1801 * Like skb_peek(), but the caller knows that the list is not empty.
1802 */
1803static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1804{
1805 return list_->next;
1806}
1807
da5ef6e5
PE
1808/**
1809 * skb_peek_next - peek skb following the given one from a queue
1810 * @skb: skb to start from
1811 * @list_: list to peek at
1812 *
1813 * Returns %NULL when the end of the list is met or a pointer to the
1814 * next element. The reference count is not incremented and the
1815 * reference is therefore volatile. Use with caution.
1816 */
1817static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1818 const struct sk_buff_head *list_)
1819{
1820 struct sk_buff *next = skb->next;
18d07000 1821
da5ef6e5
PE
1822 if (next == (struct sk_buff *)list_)
1823 next = NULL;
1824 return next;
1825}
1826
1da177e4 1827/**
1a5778aa 1828 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1829 * @list_: list to peek at
1830 *
1831 * Peek an &sk_buff. Unlike most other operations you _MUST_
1832 * be careful with this one. A peek leaves the buffer on the
1833 * list and someone else may run off with it. You must hold
1834 * the appropriate locks or have a private queue to do this.
1835 *
1836 * Returns %NULL for an empty list or a pointer to the tail element.
1837 * The reference count is not incremented and the reference is therefore
1838 * volatile. Use with caution.
1839 */
05bdd2f1 1840static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1841{
f8cc62ca 1842 struct sk_buff *skb = READ_ONCE(list_->prev);
18d07000
ED
1843
1844 if (skb == (struct sk_buff *)list_)
1845 skb = NULL;
1846 return skb;
1847
1da177e4
LT
1848}
1849
1850/**
1851 * skb_queue_len - get queue length
1852 * @list_: list to measure
1853 *
1854 * Return the length of an &sk_buff queue.
1855 */
1856static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1857{
1858 return list_->qlen;
1859}
1860
86b18aaa
QC
1861/**
1862 * skb_queue_len_lockless - get queue length
1863 * @list_: list to measure
1864 *
1865 * Return the length of an &sk_buff queue.
1866 * This variant can be used in lockless contexts.
1867 */
1868static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1869{
1870 return READ_ONCE(list_->qlen);
1871}
1872
67fed459
DM
1873/**
1874 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1875 * @list: queue to initialize
1876 *
1877 * This initializes only the list and queue length aspects of
1878 * an sk_buff_head object. This allows to initialize the list
1879 * aspects of an sk_buff_head without reinitializing things like
1880 * the spinlock. It can also be used for on-stack sk_buff_head
1881 * objects where the spinlock is known to not be used.
1882 */
1883static inline void __skb_queue_head_init(struct sk_buff_head *list)
1884{
1885 list->prev = list->next = (struct sk_buff *)list;
1886 list->qlen = 0;
1887}
1888
76f10ad0
AV
1889/*
1890 * This function creates a split out lock class for each invocation;
1891 * this is needed for now since a whole lot of users of the skb-queue
1892 * infrastructure in drivers have different locking usage (in hardirq)
1893 * than the networking core (in softirq only). In the long run either the
1894 * network layer or drivers should need annotation to consolidate the
1895 * main types of usage into 3 classes.
1896 */
1da177e4
LT
1897static inline void skb_queue_head_init(struct sk_buff_head *list)
1898{
1899 spin_lock_init(&list->lock);
67fed459 1900 __skb_queue_head_init(list);
1da177e4
LT
1901}
1902
c2ecba71
PE
1903static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1904 struct lock_class_key *class)
1905{
1906 skb_queue_head_init(list);
1907 lockdep_set_class(&list->lock, class);
1908}
1909
1da177e4 1910/*
bf299275 1911 * Insert an sk_buff on a list.
1da177e4
LT
1912 *
1913 * The "__skb_xxxx()" functions are the non-atomic ones that
1914 * can only be called with interrupts disabled.
1915 */
bf299275
GR
1916static inline void __skb_insert(struct sk_buff *newsk,
1917 struct sk_buff *prev, struct sk_buff *next,
1918 struct sk_buff_head *list)
1919{
f8cc62ca
ED
1920 /* See skb_queue_empty_lockless() and skb_peek_tail()
1921 * for the opposite READ_ONCE()
1922 */
d7d16a89
ED
1923 WRITE_ONCE(newsk->next, next);
1924 WRITE_ONCE(newsk->prev, prev);
1925 WRITE_ONCE(next->prev, newsk);
1926 WRITE_ONCE(prev->next, newsk);
bf299275
GR
1927 list->qlen++;
1928}
1da177e4 1929
67fed459
DM
1930static inline void __skb_queue_splice(const struct sk_buff_head *list,
1931 struct sk_buff *prev,
1932 struct sk_buff *next)
1933{
1934 struct sk_buff *first = list->next;
1935 struct sk_buff *last = list->prev;
1936
d7d16a89
ED
1937 WRITE_ONCE(first->prev, prev);
1938 WRITE_ONCE(prev->next, first);
67fed459 1939
d7d16a89
ED
1940 WRITE_ONCE(last->next, next);
1941 WRITE_ONCE(next->prev, last);
67fed459
DM
1942}
1943
1944/**
1945 * skb_queue_splice - join two skb lists, this is designed for stacks
1946 * @list: the new list to add
1947 * @head: the place to add it in the first list
1948 */
1949static inline void skb_queue_splice(const struct sk_buff_head *list,
1950 struct sk_buff_head *head)
1951{
1952 if (!skb_queue_empty(list)) {
1953 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1954 head->qlen += list->qlen;
67fed459
DM
1955 }
1956}
1957
1958/**
d9619496 1959 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1960 * @list: the new list to add
1961 * @head: the place to add it in the first list
1962 *
1963 * The list at @list is reinitialised
1964 */
1965static inline void skb_queue_splice_init(struct sk_buff_head *list,
1966 struct sk_buff_head *head)
1967{
1968 if (!skb_queue_empty(list)) {
1969 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1970 head->qlen += list->qlen;
67fed459
DM
1971 __skb_queue_head_init(list);
1972 }
1973}
1974
1975/**
1976 * skb_queue_splice_tail - join two skb lists, each list being a queue
1977 * @list: the new list to add
1978 * @head: the place to add it in the first list
1979 */
1980static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1981 struct sk_buff_head *head)
1982{
1983 if (!skb_queue_empty(list)) {
1984 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1985 head->qlen += list->qlen;
67fed459
DM
1986 }
1987}
1988
1989/**
d9619496 1990 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1991 * @list: the new list to add
1992 * @head: the place to add it in the first list
1993 *
1994 * Each of the lists is a queue.
1995 * The list at @list is reinitialised
1996 */
1997static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1998 struct sk_buff_head *head)
1999{
2000 if (!skb_queue_empty(list)) {
2001 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 2002 head->qlen += list->qlen;
67fed459
DM
2003 __skb_queue_head_init(list);
2004 }
2005}
2006
1da177e4 2007/**
300ce174 2008 * __skb_queue_after - queue a buffer at the list head
1da177e4 2009 * @list: list to use
300ce174 2010 * @prev: place after this buffer
1da177e4
LT
2011 * @newsk: buffer to queue
2012 *
300ce174 2013 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
2014 * and you must therefore hold required locks before calling it.
2015 *
2016 * A buffer cannot be placed on two lists at the same time.
2017 */
300ce174
SH
2018static inline void __skb_queue_after(struct sk_buff_head *list,
2019 struct sk_buff *prev,
2020 struct sk_buff *newsk)
1da177e4 2021{
bf299275 2022 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
2023}
2024
7965bd4d
JP
2025void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2026 struct sk_buff_head *list);
7de6c033 2027
f5572855
GR
2028static inline void __skb_queue_before(struct sk_buff_head *list,
2029 struct sk_buff *next,
2030 struct sk_buff *newsk)
2031{
2032 __skb_insert(newsk, next->prev, next, list);
2033}
2034
300ce174
SH
2035/**
2036 * __skb_queue_head - queue a buffer at the list head
2037 * @list: list to use
2038 * @newsk: buffer to queue
2039 *
2040 * Queue a buffer at the start of a list. This function takes no locks
2041 * and you must therefore hold required locks before calling it.
2042 *
2043 * A buffer cannot be placed on two lists at the same time.
2044 */
300ce174
SH
2045static inline void __skb_queue_head(struct sk_buff_head *list,
2046 struct sk_buff *newsk)
2047{
2048 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2049}
4ea7b0cf 2050void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174 2051
1da177e4
LT
2052/**
2053 * __skb_queue_tail - queue a buffer at the list tail
2054 * @list: list to use
2055 * @newsk: buffer to queue
2056 *
2057 * Queue a buffer at the end of a list. This function takes no locks
2058 * and you must therefore hold required locks before calling it.
2059 *
2060 * A buffer cannot be placed on two lists at the same time.
2061 */
1da177e4
LT
2062static inline void __skb_queue_tail(struct sk_buff_head *list,
2063 struct sk_buff *newsk)
2064{
f5572855 2065 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4 2066}
4ea7b0cf 2067void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4 2068
1da177e4
LT
2069/*
2070 * remove sk_buff from list. _Must_ be called atomically, and with
2071 * the list known..
2072 */
7965bd4d 2073void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
2074static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2075{
2076 struct sk_buff *next, *prev;
2077
86b18aaa 2078 WRITE_ONCE(list->qlen, list->qlen - 1);
1da177e4
LT
2079 next = skb->next;
2080 prev = skb->prev;
2081 skb->next = skb->prev = NULL;
d7d16a89
ED
2082 WRITE_ONCE(next->prev, prev);
2083 WRITE_ONCE(prev->next, next);
1da177e4
LT
2084}
2085
f525c06d
GR
2086/**
2087 * __skb_dequeue - remove from the head of the queue
2088 * @list: list to dequeue from
2089 *
2090 * Remove the head of the list. This function does not take any locks
2091 * so must be used with appropriate locks held only. The head item is
2092 * returned or %NULL if the list is empty.
2093 */
f525c06d
GR
2094static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2095{
2096 struct sk_buff *skb = skb_peek(list);
2097 if (skb)
2098 __skb_unlink(skb, list);
2099 return skb;
2100}
4ea7b0cf 2101struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1da177e4
LT
2102
2103/**
2104 * __skb_dequeue_tail - remove from the tail of the queue
2105 * @list: list to dequeue from
2106 *
2107 * Remove the tail of the list. This function does not take any locks
2108 * so must be used with appropriate locks held only. The tail item is
2109 * returned or %NULL if the list is empty.
2110 */
1da177e4
LT
2111static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2112{
2113 struct sk_buff *skb = skb_peek_tail(list);
2114 if (skb)
2115 __skb_unlink(skb, list);
2116 return skb;
2117}
4ea7b0cf 2118struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
2119
2120
bdcc0924 2121static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
2122{
2123 return skb->data_len;
2124}
2125
2126static inline unsigned int skb_headlen(const struct sk_buff *skb)
2127{
2128 return skb->len - skb->data_len;
2129}
2130
3ece7826 2131static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1da177e4 2132{
c72d8cda 2133 unsigned int i, len = 0;
1da177e4 2134
c72d8cda 2135 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 2136 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
3ece7826
WB
2137 return len;
2138}
2139
2140static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2141{
2142 return skb_headlen(skb) + __skb_pagelen(skb);
1da177e4
LT
2143}
2144
131ea667
IC
2145/**
2146 * __skb_fill_page_desc - initialise a paged fragment in an skb
2147 * @skb: buffer containing fragment to be initialised
2148 * @i: paged fragment index to initialise
2149 * @page: the page to use for this fragment
2150 * @off: the offset to the data with @page
2151 * @size: the length of the data
2152 *
2153 * Initialises the @i'th fragment of @skb to point to &size bytes at
2154 * offset @off within @page.
2155 *
2156 * Does not take any additional reference on the fragment.
2157 */
2158static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2159 struct page *page, int off, int size)
1da177e4
LT
2160{
2161 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2162
c48a11c7 2163 /*
2f064f34
MH
2164 * Propagate page pfmemalloc to the skb if we can. The problem is
2165 * that not all callers have unique ownership of the page but rely
2166 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 2167 */
1dfa5bd3 2168 frag->bv_page = page;
65c84f14 2169 frag->bv_offset = off;
9e903e08 2170 skb_frag_size_set(frag, size);
cca7af38
PE
2171
2172 page = compound_head(page);
2f064f34 2173 if (page_is_pfmemalloc(page))
cca7af38 2174 skb->pfmemalloc = true;
131ea667
IC
2175}
2176
2177/**
2178 * skb_fill_page_desc - initialise a paged fragment in an skb
2179 * @skb: buffer containing fragment to be initialised
2180 * @i: paged fragment index to initialise
2181 * @page: the page to use for this fragment
2182 * @off: the offset to the data with @page
2183 * @size: the length of the data
2184 *
2185 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 2186 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
2187 * addition updates @skb such that @i is the last fragment.
2188 *
2189 * Does not take any additional reference on the fragment.
2190 */
2191static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2192 struct page *page, int off, int size)
2193{
2194 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
2195 skb_shinfo(skb)->nr_frags = i + 1;
2196}
2197
7965bd4d
JP
2198void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2199 int size, unsigned int truesize);
654bed16 2200
f8e617e1
JW
2201void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2202 unsigned int truesize);
2203
1da177e4
LT
2204#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2205
27a884dc
ACM
2206#ifdef NET_SKBUFF_DATA_USES_OFFSET
2207static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2208{
2209 return skb->head + skb->tail;
2210}
2211
2212static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2213{
2214 skb->tail = skb->data - skb->head;
2215}
2216
2217static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2218{
2219 skb_reset_tail_pointer(skb);
2220 skb->tail += offset;
2221}
7cc46190 2222
27a884dc
ACM
2223#else /* NET_SKBUFF_DATA_USES_OFFSET */
2224static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2225{
2226 return skb->tail;
2227}
2228
2229static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2230{
2231 skb->tail = skb->data;
2232}
2233
2234static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2235{
2236 skb->tail = skb->data + offset;
2237}
4305b541 2238
27a884dc
ACM
2239#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2240
1da177e4
LT
2241/*
2242 * Add data to an sk_buff
2243 */
4df864c1
JB
2244void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2245void *skb_put(struct sk_buff *skb, unsigned int len);
2246static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
1da177e4 2247{
4df864c1 2248 void *tmp = skb_tail_pointer(skb);
1da177e4
LT
2249 SKB_LINEAR_ASSERT(skb);
2250 skb->tail += len;
2251 skb->len += len;
2252 return tmp;
2253}
2254
de77b966 2255static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2256{
2257 void *tmp = __skb_put(skb, len);
2258
2259 memset(tmp, 0, len);
2260 return tmp;
2261}
2262
2263static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2264 unsigned int len)
2265{
2266 void *tmp = __skb_put(skb, len);
2267
2268 memcpy(tmp, data, len);
2269 return tmp;
2270}
2271
2272static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2273{
2274 *(u8 *)__skb_put(skb, 1) = val;
2275}
2276
83ad357d 2277static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
e45a79da 2278{
83ad357d 2279 void *tmp = skb_put(skb, len);
e45a79da
JB
2280
2281 memset(tmp, 0, len);
2282
2283 return tmp;
2284}
2285
59ae1d12
JB
2286static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2287 unsigned int len)
2288{
2289 void *tmp = skb_put(skb, len);
2290
2291 memcpy(tmp, data, len);
2292
2293 return tmp;
2294}
2295
634fef61
JB
2296static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2297{
2298 *(u8 *)skb_put(skb, 1) = val;
2299}
2300
d58ff351
JB
2301void *skb_push(struct sk_buff *skb, unsigned int len);
2302static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2303{
2304 skb->data -= len;
2305 skb->len += len;
2306 return skb->data;
2307}
2308
af72868b
JB
2309void *skb_pull(struct sk_buff *skb, unsigned int len);
2310static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2311{
2312 skb->len -= len;
2313 BUG_ON(skb->len < skb->data_len);
2314 return skb->data += len;
2315}
2316
af72868b 2317static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
47d29646
DM
2318{
2319 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2320}
2321
af72868b 2322void *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4 2323
af72868b 2324static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2325{
2326 if (len > skb_headlen(skb) &&
987c402a 2327 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
2328 return NULL;
2329 skb->len -= len;
2330 return skb->data += len;
2331}
2332
af72868b 2333static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2334{
2335 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2336}
2337
b9df4fd7 2338static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2339{
2340 if (likely(len <= skb_headlen(skb)))
b9df4fd7 2341 return true;
1da177e4 2342 if (unlikely(len > skb->len))
b9df4fd7 2343 return false;
987c402a 2344 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
2345}
2346
c8c8b127
ED
2347void skb_condense(struct sk_buff *skb);
2348
1da177e4
LT
2349/**
2350 * skb_headroom - bytes at buffer head
2351 * @skb: buffer to check
2352 *
2353 * Return the number of bytes of free space at the head of an &sk_buff.
2354 */
c2636b4d 2355static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
2356{
2357 return skb->data - skb->head;
2358}
2359
2360/**
2361 * skb_tailroom - bytes at buffer end
2362 * @skb: buffer to check
2363 *
2364 * Return the number of bytes of free space at the tail of an sk_buff
2365 */
2366static inline int skb_tailroom(const struct sk_buff *skb)
2367{
4305b541 2368 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2369}
2370
a21d4572
ED
2371/**
2372 * skb_availroom - bytes at buffer end
2373 * @skb: buffer to check
2374 *
2375 * Return the number of bytes of free space at the tail of an sk_buff
2376 * allocated by sk_stream_alloc()
2377 */
2378static inline int skb_availroom(const struct sk_buff *skb)
2379{
16fad69c
ED
2380 if (skb_is_nonlinear(skb))
2381 return 0;
2382
2383 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2384}
2385
1da177e4
LT
2386/**
2387 * skb_reserve - adjust headroom
2388 * @skb: buffer to alter
2389 * @len: bytes to move
2390 *
2391 * Increase the headroom of an empty &sk_buff by reducing the tail
2392 * room. This is only allowed for an empty buffer.
2393 */
8243126c 2394static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2395{
2396 skb->data += len;
2397 skb->tail += len;
2398}
2399
1837b2e2
BP
2400/**
2401 * skb_tailroom_reserve - adjust reserved_tailroom
2402 * @skb: buffer to alter
2403 * @mtu: maximum amount of headlen permitted
2404 * @needed_tailroom: minimum amount of reserved_tailroom
2405 *
2406 * Set reserved_tailroom so that headlen can be as large as possible but
2407 * not larger than mtu and tailroom cannot be smaller than
2408 * needed_tailroom.
2409 * The required headroom should already have been reserved before using
2410 * this function.
2411 */
2412static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2413 unsigned int needed_tailroom)
2414{
2415 SKB_LINEAR_ASSERT(skb);
2416 if (mtu < skb_tailroom(skb) - needed_tailroom)
2417 /* use at most mtu */
2418 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2419 else
2420 /* use up to all available space */
2421 skb->reserved_tailroom = needed_tailroom;
2422}
2423
8bce6d7d
TH
2424#define ENCAP_TYPE_ETHER 0
2425#define ENCAP_TYPE_IPPROTO 1
2426
2427static inline void skb_set_inner_protocol(struct sk_buff *skb,
2428 __be16 protocol)
2429{
2430 skb->inner_protocol = protocol;
2431 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2432}
2433
2434static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2435 __u8 ipproto)
2436{
2437 skb->inner_ipproto = ipproto;
2438 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2439}
2440
6a674e9c
JG
2441static inline void skb_reset_inner_headers(struct sk_buff *skb)
2442{
aefbd2b3 2443 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2444 skb->inner_network_header = skb->network_header;
2445 skb->inner_transport_header = skb->transport_header;
2446}
2447
0b5c9db1
JP
2448static inline void skb_reset_mac_len(struct sk_buff *skb)
2449{
2450 skb->mac_len = skb->network_header - skb->mac_header;
2451}
2452
6a674e9c
JG
2453static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2454 *skb)
2455{
2456 return skb->head + skb->inner_transport_header;
2457}
2458
55dc5a9f
TH
2459static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2460{
2461 return skb_inner_transport_header(skb) - skb->data;
2462}
2463
6a674e9c
JG
2464static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2465{
2466 skb->inner_transport_header = skb->data - skb->head;
2467}
2468
2469static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2470 const int offset)
2471{
2472 skb_reset_inner_transport_header(skb);
2473 skb->inner_transport_header += offset;
2474}
2475
2476static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2477{
2478 return skb->head + skb->inner_network_header;
2479}
2480
2481static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2482{
2483 skb->inner_network_header = skb->data - skb->head;
2484}
2485
2486static inline void skb_set_inner_network_header(struct sk_buff *skb,
2487 const int offset)
2488{
2489 skb_reset_inner_network_header(skb);
2490 skb->inner_network_header += offset;
2491}
2492
aefbd2b3
PS
2493static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2494{
2495 return skb->head + skb->inner_mac_header;
2496}
2497
2498static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2499{
2500 skb->inner_mac_header = skb->data - skb->head;
2501}
2502
2503static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2504 const int offset)
2505{
2506 skb_reset_inner_mac_header(skb);
2507 skb->inner_mac_header += offset;
2508}
fda55eca
ED
2509static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2510{
35d04610 2511 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2512}
2513
9c70220b
ACM
2514static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2515{
2e07fa9c 2516 return skb->head + skb->transport_header;
9c70220b
ACM
2517}
2518
badff6d0
ACM
2519static inline void skb_reset_transport_header(struct sk_buff *skb)
2520{
2e07fa9c 2521 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2522}
2523
967b05f6
ACM
2524static inline void skb_set_transport_header(struct sk_buff *skb,
2525 const int offset)
2526{
2e07fa9c
ACM
2527 skb_reset_transport_header(skb);
2528 skb->transport_header += offset;
ea2ae17d
ACM
2529}
2530
d56f90a7
ACM
2531static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2532{
2e07fa9c 2533 return skb->head + skb->network_header;
d56f90a7
ACM
2534}
2535
c1d2bbe1
ACM
2536static inline void skb_reset_network_header(struct sk_buff *skb)
2537{
2e07fa9c 2538 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2539}
2540
c14d2450
ACM
2541static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2542{
2e07fa9c
ACM
2543 skb_reset_network_header(skb);
2544 skb->network_header += offset;
c14d2450
ACM
2545}
2546
2e07fa9c 2547static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2548{
2e07fa9c 2549 return skb->head + skb->mac_header;
bbe735e4
ACM
2550}
2551
ea6da4fd
AV
2552static inline int skb_mac_offset(const struct sk_buff *skb)
2553{
2554 return skb_mac_header(skb) - skb->data;
2555}
2556
0daf4349
DB
2557static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2558{
2559 return skb->network_header - skb->mac_header;
2560}
2561
2e07fa9c 2562static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2563{
35d04610 2564 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2565}
2566
b4ab3141
DB
2567static inline void skb_unset_mac_header(struct sk_buff *skb)
2568{
2569 skb->mac_header = (typeof(skb->mac_header))~0U;
2570}
2571
2e07fa9c
ACM
2572static inline void skb_reset_mac_header(struct sk_buff *skb)
2573{
2574 skb->mac_header = skb->data - skb->head;
2575}
2576
2577static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2578{
2579 skb_reset_mac_header(skb);
2580 skb->mac_header += offset;
2581}
2582
0e3da5bb
TT
2583static inline void skb_pop_mac_header(struct sk_buff *skb)
2584{
2585 skb->mac_header = skb->network_header;
2586}
2587
d2aa125d 2588static inline void skb_probe_transport_header(struct sk_buff *skb)
fbbdb8f0 2589{
72a338bc 2590 struct flow_keys_basic keys;
fbbdb8f0
YX
2591
2592 if (skb_transport_header_was_set(skb))
2593 return;
72a338bc 2594
3cbf4ffb
SF
2595 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2596 NULL, 0, 0, 0, 0))
42aecaa9 2597 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2598}
2599
03606895
ED
2600static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2601{
2602 if (skb_mac_header_was_set(skb)) {
2603 const unsigned char *old_mac = skb_mac_header(skb);
2604
2605 skb_set_mac_header(skb, -skb->mac_len);
2606 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2607 }
2608}
2609
04fb451e
MM
2610static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2611{
2612 return skb->csum_start - skb_headroom(skb);
2613}
2614
08b64fcc
AD
2615static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2616{
2617 return skb->head + skb->csum_start;
2618}
2619
2e07fa9c
ACM
2620static inline int skb_transport_offset(const struct sk_buff *skb)
2621{
2622 return skb_transport_header(skb) - skb->data;
2623}
2624
2625static inline u32 skb_network_header_len(const struct sk_buff *skb)
2626{
2627 return skb->transport_header - skb->network_header;
2628}
2629
6a674e9c
JG
2630static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2631{
2632 return skb->inner_transport_header - skb->inner_network_header;
2633}
2634
2e07fa9c
ACM
2635static inline int skb_network_offset(const struct sk_buff *skb)
2636{
2637 return skb_network_header(skb) - skb->data;
2638}
48d49d0c 2639
6a674e9c
JG
2640static inline int skb_inner_network_offset(const struct sk_buff *skb)
2641{
2642 return skb_inner_network_header(skb) - skb->data;
2643}
2644
f9599ce1
CG
2645static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2646{
2647 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2648}
2649
1da177e4
LT
2650/*
2651 * CPUs often take a performance hit when accessing unaligned memory
2652 * locations. The actual performance hit varies, it can be small if the
2653 * hardware handles it or large if we have to take an exception and fix it
2654 * in software.
2655 *
2656 * Since an ethernet header is 14 bytes network drivers often end up with
2657 * the IP header at an unaligned offset. The IP header can be aligned by
2658 * shifting the start of the packet by 2 bytes. Drivers should do this
2659 * with:
2660 *
8660c124 2661 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2662 *
2663 * The downside to this alignment of the IP header is that the DMA is now
2664 * unaligned. On some architectures the cost of an unaligned DMA is high
2665 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2666 *
1da177e4
LT
2667 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2668 * to be overridden.
2669 */
2670#ifndef NET_IP_ALIGN
2671#define NET_IP_ALIGN 2
2672#endif
2673
025be81e
AB
2674/*
2675 * The networking layer reserves some headroom in skb data (via
2676 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2677 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2678 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2679 *
2680 * Unfortunately this headroom changes the DMA alignment of the resulting
2681 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2682 * on some architectures. An architecture can override this value,
2683 * perhaps setting it to a cacheline in size (since that will maintain
2684 * cacheline alignment of the DMA). It must be a power of 2.
2685 *
d6301d3d 2686 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2687 * headroom, you should not reduce this.
5933dd2f
ED
2688 *
2689 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2690 * to reduce average number of cache lines per packet.
645f0897 2691 * get_rps_cpu() for example only access one 64 bytes aligned block :
18e8c134 2692 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2693 */
2694#ifndef NET_SKB_PAD
5933dd2f 2695#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2696#endif
2697
7965bd4d 2698int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2699
5293efe6 2700static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2701{
5e1abdc3 2702 if (WARN_ON(skb_is_nonlinear(skb)))
3cc0e873 2703 return;
27a884dc
ACM
2704 skb->len = len;
2705 skb_set_tail_pointer(skb, len);
1da177e4
LT
2706}
2707
5293efe6
DB
2708static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2709{
2710 __skb_set_length(skb, len);
2711}
2712
7965bd4d 2713void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2714
2715static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2716{
3cc0e873
HX
2717 if (skb->data_len)
2718 return ___pskb_trim(skb, len);
2719 __skb_trim(skb, len);
2720 return 0;
1da177e4
LT
2721}
2722
2723static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2724{
2725 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2726}
2727
e9fa4f7b
HX
2728/**
2729 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2730 * @skb: buffer to alter
2731 * @len: new length
2732 *
2733 * This is identical to pskb_trim except that the caller knows that
2734 * the skb is not cloned so we should never get an error due to out-
2735 * of-memory.
2736 */
2737static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2738{
2739 int err = pskb_trim(skb, len);
2740 BUG_ON(err);
2741}
2742
5293efe6
DB
2743static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2744{
2745 unsigned int diff = len - skb->len;
2746
2747 if (skb_tailroom(skb) < diff) {
2748 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2749 GFP_ATOMIC);
2750 if (ret)
2751 return ret;
2752 }
2753 __skb_set_length(skb, len);
2754 return 0;
2755}
2756
1da177e4
LT
2757/**
2758 * skb_orphan - orphan a buffer
2759 * @skb: buffer to orphan
2760 *
2761 * If a buffer currently has an owner then we call the owner's
2762 * destructor function and make the @skb unowned. The buffer continues
2763 * to exist but is no longer charged to its former owner.
2764 */
2765static inline void skb_orphan(struct sk_buff *skb)
2766{
c34a7612 2767 if (skb->destructor) {
1da177e4 2768 skb->destructor(skb);
c34a7612
ED
2769 skb->destructor = NULL;
2770 skb->sk = NULL;
376c7311
ED
2771 } else {
2772 BUG_ON(skb->sk);
c34a7612 2773 }
1da177e4
LT
2774}
2775
a353e0ce
MT
2776/**
2777 * skb_orphan_frags - orphan the frags contained in a buffer
2778 * @skb: buffer to orphan frags from
2779 * @gfp_mask: allocation mask for replacement pages
2780 *
2781 * For each frag in the SKB which needs a destructor (i.e. has an
2782 * owner) create a copy of that frag and release the original
2783 * page by calling the destructor.
2784 */
2785static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2786{
1f8b977a
WB
2787 if (likely(!skb_zcopy(skb)))
2788 return 0;
185ce5c3 2789 if (!skb_zcopy_is_nouarg(skb) &&
8c793822 2790 skb_uarg(skb)->callback == msg_zerocopy_callback)
1f8b977a
WB
2791 return 0;
2792 return skb_copy_ubufs(skb, gfp_mask);
2793}
2794
2795/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2796static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2797{
2798 if (likely(!skb_zcopy(skb)))
a353e0ce
MT
2799 return 0;
2800 return skb_copy_ubufs(skb, gfp_mask);
2801}
2802
1da177e4
LT
2803/**
2804 * __skb_queue_purge - empty a list
2805 * @list: list to empty
2806 *
2807 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2808 * the list and one reference dropped. This function does not take the
2809 * list lock and the caller must hold the relevant locks to use it.
2810 */
1da177e4
LT
2811static inline void __skb_queue_purge(struct sk_buff_head *list)
2812{
2813 struct sk_buff *skb;
2814 while ((skb = __skb_dequeue(list)) != NULL)
2815 kfree_skb(skb);
2816}
4ea7b0cf 2817void skb_queue_purge(struct sk_buff_head *list);
1da177e4 2818
385114de 2819unsigned int skb_rbtree_purge(struct rb_root *root);
9f5afeae 2820
7965bd4d 2821void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2822
7965bd4d
JP
2823struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2824 gfp_t gfp_mask);
8af27456
CH
2825
2826/**
2827 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2828 * @dev: network device to receive on
2829 * @length: length to allocate
2830 *
2831 * Allocate a new &sk_buff and assign it a usage count of one. The
2832 * buffer has unspecified headroom built in. Users should allocate
2833 * the headroom they think they need without accounting for the
2834 * built in space. The built in space is used for optimisations.
2835 *
2836 * %NULL is returned if there is no free memory. Although this function
2837 * allocates memory it can be called from an interrupt.
2838 */
2839static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2840 unsigned int length)
8af27456
CH
2841{
2842 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2843}
2844
6f532612
ED
2845/* legacy helper around __netdev_alloc_skb() */
2846static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2847 gfp_t gfp_mask)
2848{
2849 return __netdev_alloc_skb(NULL, length, gfp_mask);
2850}
2851
2852/* legacy helper around netdev_alloc_skb() */
2853static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2854{
2855 return netdev_alloc_skb(NULL, length);
2856}
2857
2858
4915a0de
ED
2859static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2860 unsigned int length, gfp_t gfp)
61321bbd 2861{
4915a0de 2862 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2863
2864 if (NET_IP_ALIGN && skb)
2865 skb_reserve(skb, NET_IP_ALIGN);
2866 return skb;
2867}
2868
4915a0de
ED
2869static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2870 unsigned int length)
2871{
2872 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2873}
2874
181edb2b
AD
2875static inline void skb_free_frag(void *addr)
2876{
8c2dd3e4 2877 page_frag_free(addr);
181edb2b
AD
2878}
2879
ffde7328 2880void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2881struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2882 unsigned int length, gfp_t gfp_mask);
2883static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2884 unsigned int length)
2885{
2886 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2887}
795bb1c0
JDB
2888void napi_consume_skb(struct sk_buff *skb, int budget);
2889
2890void __kfree_skb_flush(void);
15fad714 2891void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2892
71dfda58
AD
2893/**
2894 * __dev_alloc_pages - allocate page for network Rx
2895 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2896 * @order: size of the allocation
2897 *
2898 * Allocate a new page.
2899 *
2900 * %NULL is returned if there is no free memory.
2901*/
2902static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2903 unsigned int order)
2904{
2905 /* This piece of code contains several assumptions.
2906 * 1. This is for device Rx, therefor a cold page is preferred.
2907 * 2. The expectation is the user wants a compound page.
2908 * 3. If requesting a order 0 page it will not be compound
2909 * due to the check to see if order has a value in prep_new_page
2910 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2911 * code in gfp_to_alloc_flags that should be enforcing this.
2912 */
453f85d4 2913 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
71dfda58
AD
2914
2915 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2916}
2917
2918static inline struct page *dev_alloc_pages(unsigned int order)
2919{
95829b3a 2920 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2921}
2922
2923/**
2924 * __dev_alloc_page - allocate a page for network Rx
2925 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2926 *
2927 * Allocate a new page.
2928 *
2929 * %NULL is returned if there is no free memory.
2930 */
2931static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2932{
2933 return __dev_alloc_pages(gfp_mask, 0);
2934}
2935
2936static inline struct page *dev_alloc_page(void)
2937{
95829b3a 2938 return dev_alloc_pages(0);
71dfda58
AD
2939}
2940
0614002b
MG
2941/**
2942 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2943 * @page: The page that was allocated from skb_alloc_page
2944 * @skb: The skb that may need pfmemalloc set
2945 */
48f971c9
AL
2946static inline void skb_propagate_pfmemalloc(const struct page *page,
2947 struct sk_buff *skb)
0614002b 2948{
2f064f34 2949 if (page_is_pfmemalloc(page))
0614002b
MG
2950 skb->pfmemalloc = true;
2951}
2952
7240b60c
JL
2953/**
2954 * skb_frag_off() - Returns the offset of a skb fragment
2955 * @frag: the paged fragment
2956 */
2957static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2958{
65c84f14 2959 return frag->bv_offset;
7240b60c
JL
2960}
2961
2962/**
2963 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2964 * @frag: skb fragment
2965 * @delta: value to add
2966 */
2967static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2968{
65c84f14 2969 frag->bv_offset += delta;
7240b60c
JL
2970}
2971
2972/**
2973 * skb_frag_off_set() - Sets the offset of a skb fragment
2974 * @frag: skb fragment
2975 * @offset: offset of fragment
2976 */
2977static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2978{
65c84f14 2979 frag->bv_offset = offset;
7240b60c
JL
2980}
2981
2982/**
2983 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2984 * @fragto: skb fragment where offset is set
2985 * @fragfrom: skb fragment offset is copied from
2986 */
2987static inline void skb_frag_off_copy(skb_frag_t *fragto,
2988 const skb_frag_t *fragfrom)
2989{
65c84f14 2990 fragto->bv_offset = fragfrom->bv_offset;
7240b60c
JL
2991}
2992
131ea667 2993/**
e227867f 2994 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2995 * @frag: the paged fragment
2996 *
2997 * Returns the &struct page associated with @frag.
2998 */
2999static inline struct page *skb_frag_page(const skb_frag_t *frag)
3000{
1dfa5bd3 3001 return frag->bv_page;
131ea667
IC
3002}
3003
3004/**
3005 * __skb_frag_ref - take an addition reference on a paged fragment.
3006 * @frag: the paged fragment
3007 *
3008 * Takes an additional reference on the paged fragment @frag.
3009 */
3010static inline void __skb_frag_ref(skb_frag_t *frag)
3011{
3012 get_page(skb_frag_page(frag));
3013}
3014
3015/**
3016 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3017 * @skb: the buffer
3018 * @f: the fragment offset.
3019 *
3020 * Takes an additional reference on the @f'th paged fragment of @skb.
3021 */
3022static inline void skb_frag_ref(struct sk_buff *skb, int f)
3023{
3024 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3025}
3026
3027/**
3028 * __skb_frag_unref - release a reference on a paged fragment.
3029 * @frag: the paged fragment
3030 *
3031 * Releases a reference on the paged fragment @frag.
3032 */
3033static inline void __skb_frag_unref(skb_frag_t *frag)
3034{
3035 put_page(skb_frag_page(frag));
3036}
3037
3038/**
3039 * skb_frag_unref - release a reference on a paged fragment of an skb.
3040 * @skb: the buffer
3041 * @f: the fragment offset
3042 *
3043 * Releases a reference on the @f'th paged fragment of @skb.
3044 */
3045static inline void skb_frag_unref(struct sk_buff *skb, int f)
3046{
3047 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
3048}
3049
3050/**
3051 * skb_frag_address - gets the address of the data contained in a paged fragment
3052 * @frag: the paged fragment buffer
3053 *
3054 * Returns the address of the data within @frag. The page must already
3055 * be mapped.
3056 */
3057static inline void *skb_frag_address(const skb_frag_t *frag)
3058{
7240b60c 3059 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
131ea667
IC
3060}
3061
3062/**
3063 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3064 * @frag: the paged fragment buffer
3065 *
3066 * Returns the address of the data within @frag. Checks that the page
3067 * is mapped and returns %NULL otherwise.
3068 */
3069static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3070{
3071 void *ptr = page_address(skb_frag_page(frag));
3072 if (unlikely(!ptr))
3073 return NULL;
3074
7240b60c
JL
3075 return ptr + skb_frag_off(frag);
3076}
3077
3078/**
3079 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3080 * @fragto: skb fragment where page is set
3081 * @fragfrom: skb fragment page is copied from
3082 */
3083static inline void skb_frag_page_copy(skb_frag_t *fragto,
3084 const skb_frag_t *fragfrom)
3085{
3086 fragto->bv_page = fragfrom->bv_page;
131ea667
IC
3087}
3088
3089/**
3090 * __skb_frag_set_page - sets the page contained in a paged fragment
3091 * @frag: the paged fragment
3092 * @page: the page to set
3093 *
3094 * Sets the fragment @frag to contain @page.
3095 */
3096static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3097{
1dfa5bd3 3098 frag->bv_page = page;
131ea667
IC
3099}
3100
3101/**
3102 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3103 * @skb: the buffer
3104 * @f: the fragment offset
3105 * @page: the page to set
3106 *
3107 * Sets the @f'th fragment of @skb to contain @page.
3108 */
3109static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3110 struct page *page)
3111{
3112 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3113}
3114
400dfd3a
ED
3115bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3116
131ea667
IC
3117/**
3118 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 3119 * @dev: the device to map the fragment to
131ea667
IC
3120 * @frag: the paged fragment to map
3121 * @offset: the offset within the fragment (starting at the
3122 * fragment's own offset)
3123 * @size: the number of bytes to map
771b00a8 3124 * @dir: the direction of the mapping (``PCI_DMA_*``)
131ea667
IC
3125 *
3126 * Maps the page associated with @frag to @device.
3127 */
3128static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3129 const skb_frag_t *frag,
3130 size_t offset, size_t size,
3131 enum dma_data_direction dir)
3132{
3133 return dma_map_page(dev, skb_frag_page(frag),
7240b60c 3134 skb_frag_off(frag) + offset, size, dir);
131ea667
IC
3135}
3136
117632e6
ED
3137static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3138 gfp_t gfp_mask)
3139{
3140 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3141}
3142
bad93e9d
OP
3143
3144static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3145 gfp_t gfp_mask)
3146{
3147 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3148}
3149
3150
334a8132
PM
3151/**
3152 * skb_clone_writable - is the header of a clone writable
3153 * @skb: buffer to check
3154 * @len: length up to which to write
3155 *
3156 * Returns true if modifying the header part of the cloned buffer
3157 * does not requires the data to be copied.
3158 */
05bdd2f1 3159static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
3160{
3161 return !skb_header_cloned(skb) &&
3162 skb_headroom(skb) + len <= skb->hdr_len;
3163}
3164
3697649f
DB
3165static inline int skb_try_make_writable(struct sk_buff *skb,
3166 unsigned int write_len)
3167{
3168 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3169 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3170}
3171
d9cc2048
HX
3172static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3173 int cloned)
3174{
3175 int delta = 0;
3176
d9cc2048
HX
3177 if (headroom > skb_headroom(skb))
3178 delta = headroom - skb_headroom(skb);
3179
3180 if (delta || cloned)
3181 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3182 GFP_ATOMIC);
3183 return 0;
3184}
3185
1da177e4
LT
3186/**
3187 * skb_cow - copy header of skb when it is required
3188 * @skb: buffer to cow
3189 * @headroom: needed headroom
3190 *
3191 * If the skb passed lacks sufficient headroom or its data part
3192 * is shared, data is reallocated. If reallocation fails, an error
3193 * is returned and original skb is not changed.
3194 *
3195 * The result is skb with writable area skb->head...skb->tail
3196 * and at least @headroom of space at head.
3197 */
3198static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3199{
d9cc2048
HX
3200 return __skb_cow(skb, headroom, skb_cloned(skb));
3201}
1da177e4 3202
d9cc2048
HX
3203/**
3204 * skb_cow_head - skb_cow but only making the head writable
3205 * @skb: buffer to cow
3206 * @headroom: needed headroom
3207 *
3208 * This function is identical to skb_cow except that we replace the
3209 * skb_cloned check by skb_header_cloned. It should be used when
3210 * you only need to push on some header and do not need to modify
3211 * the data.
3212 */
3213static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3214{
3215 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
3216}
3217
3218/**
3219 * skb_padto - pad an skbuff up to a minimal size
3220 * @skb: buffer to pad
3221 * @len: minimal length
3222 *
3223 * Pads up a buffer to ensure the trailing bytes exist and are
3224 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
3225 * is untouched. Otherwise it is extended. Returns zero on
3226 * success. The skb is freed on error.
1da177e4 3227 */
5b057c6b 3228static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
3229{
3230 unsigned int size = skb->len;
3231 if (likely(size >= len))
5b057c6b 3232 return 0;
987c402a 3233 return skb_pad(skb, len - size);
1da177e4
LT
3234}
3235
9c0c1124 3236/**
4ea7b0cf 3237 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
9c0c1124
AD
3238 * @skb: buffer to pad
3239 * @len: minimal length
cd0a137a 3240 * @free_on_error: free buffer on error
9c0c1124
AD
3241 *
3242 * Pads up a buffer to ensure the trailing bytes exist and are
3243 * blanked. If the buffer already contains sufficient data it
3244 * is untouched. Otherwise it is extended. Returns zero on
cd0a137a 3245 * success. The skb is freed on error if @free_on_error is true.
9c0c1124 3246 */
4a009cb0
ED
3247static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3248 unsigned int len,
3249 bool free_on_error)
9c0c1124
AD
3250{
3251 unsigned int size = skb->len;
3252
3253 if (unlikely(size < len)) {
3254 len -= size;
cd0a137a 3255 if (__skb_pad(skb, len, free_on_error))
9c0c1124
AD
3256 return -ENOMEM;
3257 __skb_put(skb, len);
3258 }
3259 return 0;
3260}
3261
cd0a137a
FF
3262/**
3263 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3264 * @skb: buffer to pad
3265 * @len: minimal length
3266 *
3267 * Pads up a buffer to ensure the trailing bytes exist and are
3268 * blanked. If the buffer already contains sufficient data it
3269 * is untouched. Otherwise it is extended. Returns zero on
3270 * success. The skb is freed on error.
3271 */
4a009cb0 3272static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
cd0a137a
FF
3273{
3274 return __skb_put_padto(skb, len, true);
3275}
3276
1da177e4 3277static inline int skb_add_data(struct sk_buff *skb,
af2b040e 3278 struct iov_iter *from, int copy)
1da177e4
LT
3279{
3280 const int off = skb->len;
3281
3282 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 3283 __wsum csum = 0;
15e6cb46
AV
3284 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3285 &csum, from)) {
1da177e4
LT
3286 skb->csum = csum_block_add(skb->csum, csum, off);
3287 return 0;
3288 }
15e6cb46 3289 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
3290 return 0;
3291
3292 __skb_trim(skb, off);
3293 return -EFAULT;
3294}
3295
38ba0a65
ED
3296static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3297 const struct page *page, int off)
1da177e4 3298{
1f8b977a
WB
3299 if (skb_zcopy(skb))
3300 return false;
1da177e4 3301 if (i) {
d8e18a51 3302 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 3303
ea2ab693 3304 return page == skb_frag_page(frag) &&
7240b60c 3305 off == skb_frag_off(frag) + skb_frag_size(frag);
1da177e4 3306 }
38ba0a65 3307 return false;
1da177e4
LT
3308}
3309
364c6bad
HX
3310static inline int __skb_linearize(struct sk_buff *skb)
3311{
3312 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3313}
3314
1da177e4
LT
3315/**
3316 * skb_linearize - convert paged skb to linear one
3317 * @skb: buffer to linarize
1da177e4
LT
3318 *
3319 * If there is no free memory -ENOMEM is returned, otherwise zero
3320 * is returned and the old skb data released.
3321 */
364c6bad
HX
3322static inline int skb_linearize(struct sk_buff *skb)
3323{
3324 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3325}
3326
cef401de
ED
3327/**
3328 * skb_has_shared_frag - can any frag be overwritten
3329 * @skb: buffer to test
3330 *
3331 * Return true if the skb has at least one frag that might be modified
3332 * by an external entity (as in vmsplice()/sendfile())
3333 */
3334static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3335{
c9af6db4 3336 return skb_is_nonlinear(skb) &&
06b4feb3 3337 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
cef401de
ED
3338}
3339
364c6bad
HX
3340/**
3341 * skb_linearize_cow - make sure skb is linear and writable
3342 * @skb: buffer to process
3343 *
3344 * If there is no free memory -ENOMEM is returned, otherwise zero
3345 * is returned and the old skb data released.
3346 */
3347static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 3348{
364c6bad
HX
3349 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3350 __skb_linearize(skb) : 0;
1da177e4
LT
3351}
3352
479ffccc
DB
3353static __always_inline void
3354__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3355 unsigned int off)
3356{
3357 if (skb->ip_summed == CHECKSUM_COMPLETE)
3358 skb->csum = csum_block_sub(skb->csum,
3359 csum_partial(start, len, 0), off);
3360 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3361 skb_checksum_start_offset(skb) < 0)
3362 skb->ip_summed = CHECKSUM_NONE;
3363}
3364
1da177e4
LT
3365/**
3366 * skb_postpull_rcsum - update checksum for received skb after pull
3367 * @skb: buffer to update
3368 * @start: start of data before pull
3369 * @len: length of data pulled
3370 *
3371 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
3372 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3373 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 3374 */
1da177e4 3375static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 3376 const void *start, unsigned int len)
1da177e4 3377{
479ffccc 3378 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
3379}
3380
479ffccc
DB
3381static __always_inline void
3382__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3383 unsigned int off)
3384{
3385 if (skb->ip_summed == CHECKSUM_COMPLETE)
3386 skb->csum = csum_block_add(skb->csum,
3387 csum_partial(start, len, 0), off);
3388}
cbb042f9 3389
479ffccc
DB
3390/**
3391 * skb_postpush_rcsum - update checksum for received skb after push
3392 * @skb: buffer to update
3393 * @start: start of data after push
3394 * @len: length of data pushed
3395 *
3396 * After doing a push on a received packet, you need to call this to
3397 * update the CHECKSUM_COMPLETE checksum.
3398 */
f8ffad69
DB
3399static inline void skb_postpush_rcsum(struct sk_buff *skb,
3400 const void *start, unsigned int len)
3401{
479ffccc 3402 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
3403}
3404
af72868b 3405void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
479ffccc 3406
82a31b92
WC
3407/**
3408 * skb_push_rcsum - push skb and update receive checksum
3409 * @skb: buffer to update
3410 * @len: length of data pulled
3411 *
3412 * This function performs an skb_push on the packet and updates
3413 * the CHECKSUM_COMPLETE checksum. It should be used on
3414 * receive path processing instead of skb_push unless you know
3415 * that the checksum difference is zero (e.g., a valid IP header)
3416 * or you are setting ip_summed to CHECKSUM_NONE.
3417 */
d58ff351 3418static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
82a31b92
WC
3419{
3420 skb_push(skb, len);
3421 skb_postpush_rcsum(skb, skb->data, len);
3422 return skb->data;
3423}
3424
88078d98 3425int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
7ce5a27f
DM
3426/**
3427 * pskb_trim_rcsum - trim received skb and update checksum
3428 * @skb: buffer to trim
3429 * @len: new length
3430 *
3431 * This is exactly the same as pskb_trim except that it ensures the
3432 * checksum of received packets are still valid after the operation.
6c57f045 3433 * It can change skb pointers.
7ce5a27f
DM
3434 */
3435
3436static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3437{
3438 if (likely(len >= skb->len))
3439 return 0;
88078d98 3440 return pskb_trim_rcsum_slow(skb, len);
7ce5a27f
DM
3441}
3442
5293efe6
DB
3443static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3444{
3445 if (skb->ip_summed == CHECKSUM_COMPLETE)
3446 skb->ip_summed = CHECKSUM_NONE;
3447 __skb_trim(skb, len);
3448 return 0;
3449}
3450
3451static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3452{
3453 if (skb->ip_summed == CHECKSUM_COMPLETE)
3454 skb->ip_summed = CHECKSUM_NONE;
3455 return __skb_grow(skb, len);
3456}
3457
18a4c0ea
ED
3458#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3459#define skb_rb_first(root) rb_to_skb(rb_first(root))
3460#define skb_rb_last(root) rb_to_skb(rb_last(root))
3461#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3462#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3463
1da177e4
LT
3464#define skb_queue_walk(queue, skb) \
3465 for (skb = (queue)->next; \
a1e4891f 3466 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3467 skb = skb->next)
3468
46f8914e
JC
3469#define skb_queue_walk_safe(queue, skb, tmp) \
3470 for (skb = (queue)->next, tmp = skb->next; \
3471 skb != (struct sk_buff *)(queue); \
3472 skb = tmp, tmp = skb->next)
3473
1164f52a 3474#define skb_queue_walk_from(queue, skb) \
a1e4891f 3475 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3476 skb = skb->next)
3477
18a4c0ea
ED
3478#define skb_rbtree_walk(skb, root) \
3479 for (skb = skb_rb_first(root); skb != NULL; \
3480 skb = skb_rb_next(skb))
3481
3482#define skb_rbtree_walk_from(skb) \
3483 for (; skb != NULL; \
3484 skb = skb_rb_next(skb))
3485
3486#define skb_rbtree_walk_from_safe(skb, tmp) \
3487 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3488 skb = tmp)
3489
1164f52a
DM
3490#define skb_queue_walk_from_safe(queue, skb, tmp) \
3491 for (tmp = skb->next; \
3492 skb != (struct sk_buff *)(queue); \
3493 skb = tmp, tmp = skb->next)
3494
300ce174
SH
3495#define skb_queue_reverse_walk(queue, skb) \
3496 for (skb = (queue)->prev; \
a1e4891f 3497 skb != (struct sk_buff *)(queue); \
300ce174
SH
3498 skb = skb->prev)
3499
686a2955
DM
3500#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3501 for (skb = (queue)->prev, tmp = skb->prev; \
3502 skb != (struct sk_buff *)(queue); \
3503 skb = tmp, tmp = skb->prev)
3504
3505#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3506 for (tmp = skb->prev; \
3507 skb != (struct sk_buff *)(queue); \
3508 skb = tmp, tmp = skb->prev)
1da177e4 3509
21dc3301 3510static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3511{
3512 return skb_shinfo(skb)->frag_list != NULL;
3513}
3514
3515static inline void skb_frag_list_init(struct sk_buff *skb)
3516{
3517 skb_shinfo(skb)->frag_list = NULL;
3518}
3519
ee039871
DM
3520#define skb_walk_frags(skb, iter) \
3521 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3522
ea3793ee 3523
b50b0580
SD
3524int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3525 int *err, long *timeo_p,
ea3793ee 3526 const struct sk_buff *skb);
65101aec
PA
3527struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3528 struct sk_buff_head *queue,
3529 unsigned int flags,
fd69c399 3530 int *off, int *err,
65101aec 3531 struct sk_buff **last);
b50b0580
SD
3532struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3533 struct sk_buff_head *queue,
e427cad6 3534 unsigned int flags, int *off, int *err,
ea3793ee 3535 struct sk_buff **last);
b50b0580
SD
3536struct sk_buff *__skb_recv_datagram(struct sock *sk,
3537 struct sk_buff_head *sk_queue,
e427cad6 3538 unsigned int flags, int *off, int *err);
7965bd4d
JP
3539struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3540 int *err);
a11e1d43
LT
3541__poll_t datagram_poll(struct file *file, struct socket *sock,
3542 struct poll_table_struct *wait);
c0371da6
AV
3543int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3544 struct iov_iter *to, int size);
51f3d02b
DM
3545static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3546 struct msghdr *msg, int size)
3547{
e5a4b0bb 3548 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3549}
e5a4b0bb
AV
3550int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3551 struct msghdr *msg);
65d69e25
SG
3552int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3553 struct iov_iter *to, int len,
3554 struct ahash_request *hash);
3a654f97
AV
3555int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3556 struct iov_iter *from, int len);
3a654f97 3557int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3558void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3559void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3560static inline void skb_free_datagram_locked(struct sock *sk,
3561 struct sk_buff *skb)
3562{
3563 __skb_free_datagram_locked(sk, skb, 0);
3564}
7965bd4d 3565int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3566int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3567int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3568__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
8d5930df 3569 int len);
a60e3cc7 3570int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3571 struct pipe_inode_info *pipe, unsigned int len,
25869262 3572 unsigned int flags);
20bf50de
TH
3573int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3574 int len);
7965bd4d 3575void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3576unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3577int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3578 int len, int hlen);
7965bd4d
JP
3579void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3580int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3581void skb_scrub_packet(struct sk_buff *skb, bool xnet);
779b7931 3582bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
2b16f048 3583bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
7965bd4d 3584struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3a1296a3
SK
3585struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3586 unsigned int offset);
0d5501c1 3587struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3588int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3589int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3590int skb_vlan_pop(struct sk_buff *skb);
3591int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
19fbcb36
GN
3592int skb_eth_pop(struct sk_buff *skb);
3593int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3594 const unsigned char *src);
fa4e0f88 3595int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
d04ac224 3596 int mac_len, bool ethernet);
040b5cfb
MV
3597int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3598 bool ethernet);
d27cf5c5 3599int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
2a2ea508 3600int skb_mpls_dec_ttl(struct sk_buff *skb);
6fa01ccd
SV
3601struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3602 gfp_t gfp);
20380731 3603
6ce8e9ce
AV
3604static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3605{
3073f070 3606 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3607}
3608
7eab8d9e
AV
3609static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3610{
e5a4b0bb 3611 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3612}
3613
2817a336
DB
3614struct skb_checksum_ops {
3615 __wsum (*update)(const void *mem, int len, __wsum wsum);
3616 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3617};
3618
9617813d
DC
3619extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3620
2817a336
DB
3621__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3622 __wsum csum, const struct skb_checksum_ops *ops);
3623__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3624 __wsum csum);
3625
1e98a0f0
ED
3626static inline void * __must_check
3627__skb_header_pointer(const struct sk_buff *skb, int offset,
3628 int len, void *data, int hlen, void *buffer)
1da177e4 3629{
55820ee2 3630 if (hlen - offset >= len)
690e36e7 3631 return data + offset;
1da177e4 3632
690e36e7
DM
3633 if (!skb ||
3634 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3635 return NULL;
3636
3637 return buffer;
3638}
3639
1e98a0f0
ED
3640static inline void * __must_check
3641skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3642{
3643 return __skb_header_pointer(skb, offset, len, skb->data,
3644 skb_headlen(skb), buffer);
3645}
3646
4262e5cc
DB
3647/**
3648 * skb_needs_linearize - check if we need to linearize a given skb
3649 * depending on the given device features.
3650 * @skb: socket buffer to check
3651 * @features: net device features
3652 *
3653 * Returns true if either:
3654 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3655 * 2. skb is fragmented and the device does not support SG.
3656 */
3657static inline bool skb_needs_linearize(struct sk_buff *skb,
3658 netdev_features_t features)
3659{
3660 return skb_is_nonlinear(skb) &&
3661 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3662 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3663}
3664
d626f62b
ACM
3665static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3666 void *to,
3667 const unsigned int len)
3668{
3669 memcpy(to, skb->data, len);
3670}
3671
3672static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3673 const int offset, void *to,
3674 const unsigned int len)
3675{
3676 memcpy(to, skb->data + offset, len);
3677}
3678
27d7ff46
ACM
3679static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3680 const void *from,
3681 const unsigned int len)
3682{
3683 memcpy(skb->data, from, len);
3684}
3685
3686static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3687 const int offset,
3688 const void *from,
3689 const unsigned int len)
3690{
3691 memcpy(skb->data + offset, from, len);
3692}
3693
7965bd4d 3694void skb_init(void);
1da177e4 3695
ac45f602
PO
3696static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3697{
3698 return skb->tstamp;
3699}
3700
a61bbcf2
PM
3701/**
3702 * skb_get_timestamp - get timestamp from a skb
3703 * @skb: skb to get stamp from
13c6ee2a 3704 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
a61bbcf2
PM
3705 *
3706 * Timestamps are stored in the skb as offsets to a base timestamp.
3707 * This function converts the offset back to a struct timeval and stores
3708 * it in stamp.
3709 */
ac45f602 3710static inline void skb_get_timestamp(const struct sk_buff *skb,
13c6ee2a 3711 struct __kernel_old_timeval *stamp)
a61bbcf2 3712{
13c6ee2a 3713 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
a61bbcf2
PM
3714}
3715
887feae3
DD
3716static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3717 struct __kernel_sock_timeval *stamp)
3718{
3719 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3720
3721 stamp->tv_sec = ts.tv_sec;
3722 stamp->tv_usec = ts.tv_nsec / 1000;
3723}
3724
ac45f602 3725static inline void skb_get_timestampns(const struct sk_buff *skb,
df1b4ba9 3726 struct __kernel_old_timespec *stamp)
ac45f602 3727{
df1b4ba9
AB
3728 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3729
3730 stamp->tv_sec = ts.tv_sec;
3731 stamp->tv_nsec = ts.tv_nsec;
ac45f602
PO
3732}
3733
887feae3
DD
3734static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3735 struct __kernel_timespec *stamp)
3736{
3737 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3738
3739 stamp->tv_sec = ts.tv_sec;
3740 stamp->tv_nsec = ts.tv_nsec;
3741}
3742
b7aa0bf7 3743static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3744{
b7aa0bf7 3745 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3746}
3747
164891aa
SH
3748static inline ktime_t net_timedelta(ktime_t t)
3749{
3750 return ktime_sub(ktime_get_real(), t);
3751}
3752
b9ce204f
IJ
3753static inline ktime_t net_invalid_timestamp(void)
3754{
8b0e1953 3755 return 0;
b9ce204f 3756}
a61bbcf2 3757
de8f3a83
DB
3758static inline u8 skb_metadata_len(const struct sk_buff *skb)
3759{
3760 return skb_shinfo(skb)->meta_len;
3761}
3762
3763static inline void *skb_metadata_end(const struct sk_buff *skb)
3764{
3765 return skb_mac_header(skb);
3766}
3767
3768static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3769 const struct sk_buff *skb_b,
3770 u8 meta_len)
3771{
3772 const void *a = skb_metadata_end(skb_a);
3773 const void *b = skb_metadata_end(skb_b);
3774 /* Using more efficient varaiant than plain call to memcmp(). */
3775#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3776 u64 diffs = 0;
3777
3778 switch (meta_len) {
3779#define __it(x, op) (x -= sizeof(u##op))
3780#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3781 case 32: diffs |= __it_diff(a, b, 64);
df561f66 3782 fallthrough;
de8f3a83 3783 case 24: diffs |= __it_diff(a, b, 64);
df561f66 3784 fallthrough;
de8f3a83 3785 case 16: diffs |= __it_diff(a, b, 64);
df561f66 3786 fallthrough;
de8f3a83
DB
3787 case 8: diffs |= __it_diff(a, b, 64);
3788 break;
3789 case 28: diffs |= __it_diff(a, b, 64);
df561f66 3790 fallthrough;
de8f3a83 3791 case 20: diffs |= __it_diff(a, b, 64);
df561f66 3792 fallthrough;
de8f3a83 3793 case 12: diffs |= __it_diff(a, b, 64);
df561f66 3794 fallthrough;
de8f3a83
DB
3795 case 4: diffs |= __it_diff(a, b, 32);
3796 break;
3797 }
3798 return diffs;
3799#else
3800 return memcmp(a - meta_len, b - meta_len, meta_len);
3801#endif
3802}
3803
3804static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3805 const struct sk_buff *skb_b)
3806{
3807 u8 len_a = skb_metadata_len(skb_a);
3808 u8 len_b = skb_metadata_len(skb_b);
3809
3810 if (!(len_a | len_b))
3811 return false;
3812
3813 return len_a != len_b ?
3814 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3815}
3816
3817static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3818{
3819 skb_shinfo(skb)->meta_len = meta_len;
3820}
3821
3822static inline void skb_metadata_clear(struct sk_buff *skb)
3823{
3824 skb_metadata_set(skb, 0);
3825}
3826
62bccb8c
AD
3827struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3828
c1f19b51
RC
3829#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3830
7965bd4d
JP
3831void skb_clone_tx_timestamp(struct sk_buff *skb);
3832bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3833
3834#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3835
3836static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3837{
3838}
3839
3840static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3841{
3842 return false;
3843}
3844
3845#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3846
3847/**
3848 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3849 *
da92b194
RC
3850 * PHY drivers may accept clones of transmitted packets for
3851 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3852 * must call this function to return the skb back to the stack with a
3853 * timestamp.
da92b194 3854 *
2ff17117 3855 * @skb: clone of the original outgoing packet
7a76a021 3856 * @hwtstamps: hardware time stamps
c1f19b51
RC
3857 *
3858 */
3859void skb_complete_tx_timestamp(struct sk_buff *skb,
3860 struct skb_shared_hwtstamps *hwtstamps);
3861
e7ed11ee 3862void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
e7fd2885
WB
3863 struct skb_shared_hwtstamps *hwtstamps,
3864 struct sock *sk, int tstype);
3865
ac45f602
PO
3866/**
3867 * skb_tstamp_tx - queue clone of skb with send time stamps
3868 * @orig_skb: the original outgoing packet
3869 * @hwtstamps: hardware time stamps, may be NULL if not available
3870 *
3871 * If the skb has a socket associated, then this function clones the
3872 * skb (thus sharing the actual data and optional structures), stores
3873 * the optional hardware time stamping information (if non NULL) or
3874 * generates a software time stamp (otherwise), then queues the clone
3875 * to the error queue of the socket. Errors are silently ignored.
3876 */
7965bd4d
JP
3877void skb_tstamp_tx(struct sk_buff *orig_skb,
3878 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3879
4507a715
RC
3880/**
3881 * skb_tx_timestamp() - Driver hook for transmit timestamping
3882 *
3883 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3884 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3885 *
73409f3b
DM
3886 * Specifically, one should make absolutely sure that this function is
3887 * called before TX completion of this packet can trigger. Otherwise
3888 * the packet could potentially already be freed.
3889 *
4507a715
RC
3890 * @skb: A socket buffer.
3891 */
3892static inline void skb_tx_timestamp(struct sk_buff *skb)
3893{
c1f19b51 3894 skb_clone_tx_timestamp(skb);
b50a5c70
ML
3895 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3896 skb_tstamp_tx(skb, NULL);
4507a715
RC
3897}
3898
6e3e939f
JB
3899/**
3900 * skb_complete_wifi_ack - deliver skb with wifi status
3901 *
3902 * @skb: the original outgoing packet
3903 * @acked: ack status
3904 *
3905 */
3906void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3907
7965bd4d
JP
3908__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3909__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3910
60476372
HX
3911static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3912{
6edec0e6
TH
3913 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3914 skb->csum_valid ||
3915 (skb->ip_summed == CHECKSUM_PARTIAL &&
3916 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3917}
3918
fb286bb2
HX
3919/**
3920 * skb_checksum_complete - Calculate checksum of an entire packet
3921 * @skb: packet to process
3922 *
3923 * This function calculates the checksum over the entire packet plus
3924 * the value of skb->csum. The latter can be used to supply the
3925 * checksum of a pseudo header as used by TCP/UDP. It returns the
3926 * checksum.
3927 *
3928 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3929 * this function can be used to verify that checksum on received
3930 * packets. In that case the function should return zero if the
3931 * checksum is correct. In particular, this function will return zero
3932 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3933 * hardware has already verified the correctness of the checksum.
3934 */
4381ca3c 3935static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3936{
60476372
HX
3937 return skb_csum_unnecessary(skb) ?
3938 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3939}
3940
77cffe23
TH
3941static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3942{
3943 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3944 if (skb->csum_level == 0)
3945 skb->ip_summed = CHECKSUM_NONE;
3946 else
3947 skb->csum_level--;
3948 }
3949}
3950
3951static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3952{
3953 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3954 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3955 skb->csum_level++;
3956 } else if (skb->ip_summed == CHECKSUM_NONE) {
3957 skb->ip_summed = CHECKSUM_UNNECESSARY;
3958 skb->csum_level = 0;
3959 }
3960}
3961
836e66c2
DB
3962static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
3963{
3964 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3965 skb->ip_summed = CHECKSUM_NONE;
3966 skb->csum_level = 0;
3967 }
3968}
3969
76ba0aae
TH
3970/* Check if we need to perform checksum complete validation.
3971 *
3972 * Returns true if checksum complete is needed, false otherwise
3973 * (either checksum is unnecessary or zero checksum is allowed).
3974 */
3975static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3976 bool zero_okay,
3977 __sum16 check)
3978{
5d0c2b95
TH
3979 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3980 skb->csum_valid = 1;
77cffe23 3981 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3982 return false;
3983 }
3984
3985 return true;
3986}
3987
da279887 3988/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
76ba0aae
TH
3989 * in checksum_init.
3990 */
3991#define CHECKSUM_BREAK 76
3992
4e18b9ad
TH
3993/* Unset checksum-complete
3994 *
3995 * Unset checksum complete can be done when packet is being modified
3996 * (uncompressed for instance) and checksum-complete value is
3997 * invalidated.
3998 */
3999static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4000{
4001 if (skb->ip_summed == CHECKSUM_COMPLETE)
4002 skb->ip_summed = CHECKSUM_NONE;
4003}
4004
76ba0aae
TH
4005/* Validate (init) checksum based on checksum complete.
4006 *
4007 * Return values:
4008 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4009 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4010 * checksum is stored in skb->csum for use in __skb_checksum_complete
4011 * non-zero: value of invalid checksum
4012 *
4013 */
4014static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4015 bool complete,
4016 __wsum psum)
4017{
4018 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4019 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 4020 skb->csum_valid = 1;
76ba0aae
TH
4021 return 0;
4022 }
4023 }
4024
4025 skb->csum = psum;
4026
5d0c2b95
TH
4027 if (complete || skb->len <= CHECKSUM_BREAK) {
4028 __sum16 csum;
4029
4030 csum = __skb_checksum_complete(skb);
4031 skb->csum_valid = !csum;
4032 return csum;
4033 }
76ba0aae
TH
4034
4035 return 0;
4036}
4037
4038static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4039{
4040 return 0;
4041}
4042
4043/* Perform checksum validate (init). Note that this is a macro since we only
4044 * want to calculate the pseudo header which is an input function if necessary.
4045 * First we try to validate without any computation (checksum unnecessary) and
4046 * then calculate based on checksum complete calling the function to compute
4047 * pseudo header.
4048 *
4049 * Return values:
4050 * 0: checksum is validated or try to in skb_checksum_complete
4051 * non-zero: value of invalid checksum
4052 */
4053#define __skb_checksum_validate(skb, proto, complete, \
4054 zero_okay, check, compute_pseudo) \
4055({ \
4056 __sum16 __ret = 0; \
5d0c2b95 4057 skb->csum_valid = 0; \
76ba0aae
TH
4058 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4059 __ret = __skb_checksum_validate_complete(skb, \
4060 complete, compute_pseudo(skb, proto)); \
4061 __ret; \
4062})
4063
4064#define skb_checksum_init(skb, proto, compute_pseudo) \
4065 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4066
4067#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4068 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4069
4070#define skb_checksum_validate(skb, proto, compute_pseudo) \
4071 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4072
4073#define skb_checksum_validate_zero_check(skb, proto, check, \
4074 compute_pseudo) \
096a4cfa 4075 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
4076
4077#define skb_checksum_simple_validate(skb) \
4078 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4079
d96535a1
TH
4080static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4081{
219f1d79 4082 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
4083}
4084
e4aa33ad 4085static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
d96535a1
TH
4086{
4087 skb->csum = ~pseudo;
4088 skb->ip_summed = CHECKSUM_COMPLETE;
4089}
4090
e4aa33ad 4091#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
4092do { \
4093 if (__skb_checksum_convert_check(skb)) \
e4aa33ad 4094 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
d96535a1
TH
4095} while (0)
4096
15e2396d
TH
4097static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4098 u16 start, u16 offset)
4099{
4100 skb->ip_summed = CHECKSUM_PARTIAL;
4101 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4102 skb->csum_offset = offset - start;
4103}
4104
dcdc8994
TH
4105/* Update skbuf and packet to reflect the remote checksum offload operation.
4106 * When called, ptr indicates the starting point for skb->csum when
4107 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4108 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4109 */
4110static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 4111 int start, int offset, bool nopartial)
dcdc8994
TH
4112{
4113 __wsum delta;
4114
15e2396d
TH
4115 if (!nopartial) {
4116 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4117 return;
4118 }
4119
dcdc8994
TH
4120 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4121 __skb_checksum_complete(skb);
4122 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4123 }
4124
4125 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4126
4127 /* Adjust skb->csum since we changed the packet */
4128 skb->csum = csum_add(skb->csum, delta);
4129}
4130
cb9c6836
FW
4131static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4132{
4133#if IS_ENABLED(CONFIG_NF_CONNTRACK)
261db6c2 4134 return (void *)(skb->_nfct & NFCT_PTRMASK);
cb9c6836
FW
4135#else
4136 return NULL;
4137#endif
4138}
4139
261db6c2 4140static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
1da177e4 4141{
261db6c2
JS
4142#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4143 return skb->_nfct;
4144#else
4145 return 0UL;
4146#endif
1da177e4 4147}
261db6c2
JS
4148
4149static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
1da177e4 4150{
261db6c2
JS
4151#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4152 skb->_nfct = nfct;
2fc72c7b 4153#endif
261db6c2 4154}
df5042f4
FW
4155
4156#ifdef CONFIG_SKB_EXTENSIONS
4157enum skb_ext_id {
4158#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4159 SKB_EXT_BRIDGE_NF,
4165079b
FW
4160#endif
4161#ifdef CONFIG_XFRM
4162 SKB_EXT_SEC_PATH,
95a7233c
PB
4163#endif
4164#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4165 TC_SKB_EXT,
3ee17bc7
MM
4166#endif
4167#if IS_ENABLED(CONFIG_MPTCP)
4168 SKB_EXT_MPTCP,
df5042f4
FW
4169#endif
4170 SKB_EXT_NUM, /* must be last */
4171};
4172
4173/**
4174 * struct skb_ext - sk_buff extensions
4175 * @refcnt: 1 on allocation, deallocated on 0
4176 * @offset: offset to add to @data to obtain extension address
4177 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4178 * @data: start of extension data, variable sized
4179 *
4180 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4181 * to use 'u8' types while allowing up to 2kb worth of extension data.
4182 */
4183struct skb_ext {
4184 refcount_t refcnt;
4185 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4186 u8 chunks; /* same */
5c91aa1d 4187 char data[] __aligned(8);
df5042f4
FW
4188};
4189
4930f483 4190struct skb_ext *__skb_ext_alloc(gfp_t flags);
8b69a803
PA
4191void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4192 struct skb_ext *ext);
df5042f4
FW
4193void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4194void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4195void __skb_ext_put(struct skb_ext *ext);
4196
4197static inline void skb_ext_put(struct sk_buff *skb)
4198{
4199 if (skb->active_extensions)
4200 __skb_ext_put(skb->extensions);
4201}
4202
df5042f4
FW
4203static inline void __skb_ext_copy(struct sk_buff *dst,
4204 const struct sk_buff *src)
4205{
4206 dst->active_extensions = src->active_extensions;
4207
4208 if (src->active_extensions) {
4209 struct skb_ext *ext = src->extensions;
4210
4211 refcount_inc(&ext->refcnt);
4212 dst->extensions = ext;
4213 }
4214}
4215
4216static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4217{
4218 skb_ext_put(dst);
4219 __skb_ext_copy(dst, src);
4220}
4221
4222static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4223{
4224 return !!ext->offset[i];
4225}
4226
4227static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4228{
4229 return skb->active_extensions & (1 << id);
4230}
4231
4232static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4233{
4234 if (skb_ext_exist(skb, id))
4235 __skb_ext_del(skb, id);
4236}
4237
4238static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4239{
4240 if (skb_ext_exist(skb, id)) {
4241 struct skb_ext *ext = skb->extensions;
4242
4243 return (void *)ext + (ext->offset[id] << 3);
4244 }
4245
4246 return NULL;
4247}
174e2381
FW
4248
4249static inline void skb_ext_reset(struct sk_buff *skb)
4250{
4251 if (unlikely(skb->active_extensions)) {
4252 __skb_ext_put(skb->extensions);
4253 skb->active_extensions = 0;
4254 }
4255}
677bf08c
FW
4256
4257static inline bool skb_has_extensions(struct sk_buff *skb)
4258{
4259 return unlikely(skb->active_extensions);
4260}
df5042f4
FW
4261#else
4262static inline void skb_ext_put(struct sk_buff *skb) {}
174e2381 4263static inline void skb_ext_reset(struct sk_buff *skb) {}
df5042f4
FW
4264static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4265static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4266static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
677bf08c 4267static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
df5042f4
FW
4268#endif /* CONFIG_SKB_EXTENSIONS */
4269
895b5c9f 4270static inline void nf_reset_ct(struct sk_buff *skb)
a193a4ab 4271{
5f79e0f9 4272#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4273 nf_conntrack_put(skb_nfct(skb));
4274 skb->_nfct = 0;
2fc72c7b 4275#endif
a193a4ab
PM
4276}
4277
124dff01
PM
4278static inline void nf_reset_trace(struct sk_buff *skb)
4279{
478b360a 4280#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
4281 skb->nf_trace = 0;
4282#endif
a193a4ab
PM
4283}
4284
2b5ec1a5
YY
4285static inline void ipvs_reset(struct sk_buff *skb)
4286{
4287#if IS_ENABLED(CONFIG_IP_VS)
4288 skb->ipvs_property = 0;
4289#endif
4290}
4291
de8bda1d 4292/* Note: This doesn't put any conntrack info in dst. */
b1937227
ED
4293static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4294 bool copy)
edda553c 4295{
5f79e0f9 4296#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4297 dst->_nfct = src->_nfct;
4298 nf_conntrack_get(skb_nfct(src));
2fc72c7b 4299#endif
478b360a 4300#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
4301 if (copy)
4302 dst->nf_trace = src->nf_trace;
478b360a 4303#endif
edda553c
YK
4304}
4305
e7ac05f3
YK
4306static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4307{
e7ac05f3 4308#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 4309 nf_conntrack_put(skb_nfct(dst));
e7ac05f3 4310#endif
b1937227 4311 __nf_copy(dst, src, true);
e7ac05f3
YK
4312}
4313
984bc16c
JM
4314#ifdef CONFIG_NETWORK_SECMARK
4315static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4316{
4317 to->secmark = from->secmark;
4318}
4319
4320static inline void skb_init_secmark(struct sk_buff *skb)
4321{
4322 skb->secmark = 0;
4323}
4324#else
4325static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4326{ }
4327
4328static inline void skb_init_secmark(struct sk_buff *skb)
4329{ }
4330#endif
4331
7af8f4ca
FW
4332static inline int secpath_exists(const struct sk_buff *skb)
4333{
4334#ifdef CONFIG_XFRM
4165079b 4335 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
7af8f4ca
FW
4336#else
4337 return 0;
4338#endif
4339}
4340
574f7194
EB
4341static inline bool skb_irq_freeable(const struct sk_buff *skb)
4342{
4343 return !skb->destructor &&
7af8f4ca 4344 !secpath_exists(skb) &&
cb9c6836 4345 !skb_nfct(skb) &&
574f7194
EB
4346 !skb->_skb_refdst &&
4347 !skb_has_frag_list(skb);
4348}
4349
f25f4e44
PWJ
4350static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4351{
f25f4e44 4352 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
4353}
4354
9247744e 4355static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 4356{
4e3ab47a 4357 return skb->queue_mapping;
4e3ab47a
PE
4358}
4359
f25f4e44
PWJ
4360static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4361{
f25f4e44 4362 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
4363}
4364
d5a9e24a
DM
4365static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4366{
4367 skb->queue_mapping = rx_queue + 1;
4368}
4369
9247744e 4370static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
4371{
4372 return skb->queue_mapping - 1;
4373}
4374
9247744e 4375static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 4376{
a02cec21 4377 return skb->queue_mapping != 0;
d5a9e24a
DM
4378}
4379
4ff06203
JA
4380static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4381{
4382 skb->dst_pending_confirm = val;
4383}
4384
4385static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4386{
4387 return skb->dst_pending_confirm != 0;
4388}
4389
2294be0f 4390static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
def8b4fa 4391{
0b3d8e08 4392#ifdef CONFIG_XFRM
4165079b 4393 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
def8b4fa 4394#else
def8b4fa 4395 return NULL;
def8b4fa 4396#endif
0b3d8e08 4397}
def8b4fa 4398
68c33163
PS
4399/* Keeps track of mac header offset relative to skb->head.
4400 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4401 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
4402 * tunnel skb it points to outer mac header.
4403 * Keeps track of level of encapsulation of network headers.
4404 */
68c33163 4405struct skb_gso_cb {
802ab55a
AD
4406 union {
4407 int mac_offset;
4408 int data_offset;
4409 };
3347c960 4410 int encap_level;
76443456 4411 __wsum csum;
7e2b10c1 4412 __u16 csum_start;
68c33163 4413};
a08e7fd9
CZ
4414#define SKB_GSO_CB_OFFSET 32
4415#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
68c33163
PS
4416
4417static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4418{
4419 return (skb_mac_header(inner_skb) - inner_skb->head) -
4420 SKB_GSO_CB(inner_skb)->mac_offset;
4421}
4422
1e2bd517
PS
4423static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4424{
4425 int new_headroom, headroom;
4426 int ret;
4427
4428 headroom = skb_headroom(skb);
4429 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4430 if (ret)
4431 return ret;
4432
4433 new_headroom = skb_headroom(skb);
4434 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4435 return 0;
4436}
4437
08b64fcc
AD
4438static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4439{
4440 /* Do not update partial checksums if remote checksum is enabled. */
4441 if (skb->remcsum_offload)
4442 return;
4443
4444 SKB_GSO_CB(skb)->csum = res;
4445 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4446}
4447
7e2b10c1
TH
4448/* Compute the checksum for a gso segment. First compute the checksum value
4449 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4450 * then add in skb->csum (checksum from csum_start to end of packet).
4451 * skb->csum and csum_start are then updated to reflect the checksum of the
4452 * resultant packet starting from the transport header-- the resultant checksum
4453 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4454 * header.
4455 */
4456static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4457{
76443456
AD
4458 unsigned char *csum_start = skb_transport_header(skb);
4459 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4460 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 4461
76443456
AD
4462 SKB_GSO_CB(skb)->csum = res;
4463 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 4464
76443456 4465 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
4466}
4467
bdcc0924 4468static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
4469{
4470 return skb_shinfo(skb)->gso_size;
4471}
4472
36a8f39e 4473/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 4474static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
4475{
4476 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4477}
4478
d02f51cb
DA
4479/* Note: Should be called only if skb_is_gso(skb) is true */
4480static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4481{
4482 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4483}
4484
4c3024de 4485/* Note: Should be called only if skb_is_gso(skb) is true */
b90efd22
WB
4486static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4487{
4c3024de 4488 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
b90efd22
WB
4489}
4490
5293efe6
DB
4491static inline void skb_gso_reset(struct sk_buff *skb)
4492{
4493 skb_shinfo(skb)->gso_size = 0;
4494 skb_shinfo(skb)->gso_segs = 0;
4495 skb_shinfo(skb)->gso_type = 0;
4496}
4497
d02f51cb
DA
4498static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4499 u16 increment)
4500{
4501 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4502 return;
4503 shinfo->gso_size += increment;
4504}
4505
4506static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4507 u16 decrement)
4508{
4509 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4510 return;
4511 shinfo->gso_size -= decrement;
4512}
4513
7965bd4d 4514void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
4515
4516static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4517{
4518 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4519 * wanted then gso_type will be set. */
05bdd2f1
ED
4520 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4521
b78462eb
AD
4522 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4523 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
4524 __skb_warn_lro_forwarding(skb);
4525 return true;
4526 }
4527 return false;
4528}
4529
35fc92a9
HX
4530static inline void skb_forward_csum(struct sk_buff *skb)
4531{
4532 /* Unfortunately we don't support this one. Any brave souls? */
4533 if (skb->ip_summed == CHECKSUM_COMPLETE)
4534 skb->ip_summed = CHECKSUM_NONE;
4535}
4536
bc8acf2c
ED
4537/**
4538 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4539 * @skb: skb to check
4540 *
4541 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4542 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4543 * use this helper, to document places where we make this assertion.
4544 */
05bdd2f1 4545static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
4546{
4547#ifdef DEBUG
4548 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4549#endif
4550}
4551
f35d9d8a 4552bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 4553
ed1f50c3 4554int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
4555struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4556 unsigned int transport_len,
4557 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 4558
3a7c1ee4
AD
4559/**
4560 * skb_head_is_locked - Determine if the skb->head is locked down
4561 * @skb: skb to check
4562 *
4563 * The head on skbs build around a head frag can be removed if they are
4564 * not cloned. This function returns true if the skb head is locked down
4565 * due to either being allocated via kmalloc, or by being a clone with
4566 * multiple references to the head.
4567 */
4568static inline bool skb_head_is_locked(const struct sk_buff *skb)
4569{
4570 return !skb->head_frag || skb_cloned(skb);
4571}
fe6cc55f 4572
179bc67f
EC
4573/* Local Checksum Offload.
4574 * Compute outer checksum based on the assumption that the
4575 * inner checksum will be offloaded later.
d0dcde64 4576 * See Documentation/networking/checksum-offloads.rst for
e8ae7b00 4577 * explanation of how this works.
179bc67f
EC
4578 * Fill in outer checksum adjustment (e.g. with sum of outer
4579 * pseudo-header) before calling.
4580 * Also ensure that inner checksum is in linear data area.
4581 */
4582static inline __wsum lco_csum(struct sk_buff *skb)
4583{
9e74a6da
AD
4584 unsigned char *csum_start = skb_checksum_start(skb);
4585 unsigned char *l4_hdr = skb_transport_header(skb);
4586 __wsum partial;
179bc67f
EC
4587
4588 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
4589 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4590 skb->csum_offset));
4591
179bc67f 4592 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 4593 * adjustment filled in by caller) and return result.
179bc67f 4594 */
9e74a6da 4595 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
4596}
4597
2c64605b
PNA
4598static inline bool skb_is_redirected(const struct sk_buff *skb)
4599{
4600#ifdef CONFIG_NET_REDIRECT
4601 return skb->redirected;
4602#else
4603 return false;
4604#endif
4605}
4606
4607static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4608{
4609#ifdef CONFIG_NET_REDIRECT
4610 skb->redirected = 1;
4611 skb->from_ingress = from_ingress;
4612 if (skb->from_ingress)
4613 skb->tstamp = 0;
4614#endif
4615}
4616
4617static inline void skb_reset_redirect(struct sk_buff *skb)
4618{
4619#ifdef CONFIG_NET_REDIRECT
4620 skb->redirected = 0;
4621#endif
4622}
4623
fa821170
XL
4624static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4625{
4626 return skb->csum_not_inet;
4627}
4628
6370cc3b
AN
4629static inline void skb_set_kcov_handle(struct sk_buff *skb,
4630 const u64 kcov_handle)
4631{
fa69ee5a
ME
4632#ifdef CONFIG_KCOV
4633 skb->kcov_handle = kcov_handle;
4634#endif
6370cc3b
AN
4635}
4636
4637static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4638{
fa69ee5a
ME
4639#ifdef CONFIG_KCOV
4640 return skb->kcov_handle;
6370cc3b 4641#else
fa69ee5a
ME
4642 return 0;
4643#endif
4644}
6370cc3b 4645
1da177e4
LT
4646#endif /* __KERNEL__ */
4647#endif /* _LINUX_SKBUFF_H */